JP3926641B2 - Multi-input multi-output turbo receiver - Google Patents

Description

【００２１】
ここで、ｈ_1,1（ｋ，０）は、１番目の送信アンテナＡＮＳ-１から１番目の受信アンテナＡＮＲ-１への伝送路における、時点ｋのインパルスレスポンスの０シンボル遅延の値を表す。同様にして、ｈ_1,1（ｋ，Ｑ−１）は、１番目の送信アンテナＡＮＳ-１から１番目の受信アンテナＡＮＲ-１に至る伝送路における、時点ｋのインパルスレスポンスの（Ｑ−１）シンボル遅延の値を表し、ｈ_1,N（ｋ，０）は、１番目の送信アンテナＡＮＳ-１からＮ番目の受信アンテナＡＮＲ-Ｎに至る伝送路における、時点ｋのインパルスレスポンスの０シンボル遅延の値を表し、ｈ_1,N（ｋ，Ｑ−１）は、１番目の送信アンテナＡＮＳ-１からＮ番目の受信アンテナＡＮＲ-Ｎに至る伝送路における、時点ｋのインパルスレスポンスの（Ｑ−１）シンボル遅延の値を表す。
【００２２】
同様にして２…，Ｍ番目の送信アンテナＡＮＳ-２〜ＡＮＳ-Ｍからそれぞれ送信された信号の各受信信号アンテナＡＮＲ-１〜ＡＮＲ-Ｎに至るインパルスレスポンスを用いて、行列Ｈ ₂（ｋ），…，Ｈ _M（ｋ）を定義する。この場合、Ｍ番目の送信アンテナＡＮＳ-Ｍと各受信信号アンテナＡＮＲ-１〜ＡＮＲ-Ｎ間のインパルスレスポンスを表す行列Ｈ _M（ｋ）は以下の式で表される。
【００２３】
【数２】

【００２４】
次に、１番目の送信アンテナＡＮＳ-１から送信された第１ストリームの信号の時点ｋにおけるビットの尤度値λ₁（ｋ）を以下の式で定義する。
λ₁（ｋ）＝ｌｏｇ［Ｐ[ｂ（ｋ）＝０]／Ｐ[ｂ（ｋ）＝１]］ （５）
この尤度値を用いて、１番目のストリームの尤度値ベクトルｂ ₁（ｋ）を以下の式で定義する。

ここでtanhはハイバポリックタンジェントを表す。−∞から＋∞の値を取る尤度値λが、tanh（λ／２）とすることにより、−１から１の値を取ることとなる。同様にして、２，…，Ｍ番目の送信アンテナＡＮＳ-２，…，ＡＮＳ-Ｍから送信された信号の時点ｋにおけるビットの尤度値をλ₂（ｋ），…，λ_M（ｋ）として、第２，…，Ｍストリームの尤度値ベクトルをｂ ₂（ｋ），…，ｂ _M（ｋ）とする。このとき、ｂ _M（ｋ）は以下の式で表される。

以上、定義された変数を用いて図２について説明する。はじめに、１回目の第１ストリーム用等化器２１-１における処理について説明する。伝送路推定器１３によりインパルスレスポンスを推定して、式（３）で定義される行列Ｈ ₁（ｋ）を生成する。ここでｋは時点を表す。この段階では、他のストリーム用等化器２１-２〜２１-Ｍにおいて処理が終了していないために、干渉成分生成器２３には端子２２-１〜２２-Ｍよりの各ストリームの尤度値出力は入力されないため、ｍ＝１，…，Ｍの全ての尤度値λ_m（ｋ）はゼロとなり、尤度値ベクトルｂ _m（ｋ）もゼロベクトルとなる。したがって、干渉成分生成器２３の出力もゼロとなり、受信信号ベクトルｒ（ｋ）がＭＭＳＥフィルタ１６に入力される。ＭＭＳＥフィルタ１６のフィルタ係数ベクトルをＷ ₁（ｋ）とするとＷ ₁（ｋ）は、行列Ｈ ₁（ｋ）を用いて以下の式で表される。
Ｗ ₁（ｋ）＝｛Ｈ ₁（ｋ）Ｈ ₁ ^H（ｋ）＋σ² Ｉ｝^-1 Ｈ ₁（ｋ）ｅ _Q（８）
ここで、σ²は平均雑音電力を表し、Ｉは単位行列を表す。また、添字の^Hは複素共役転置を表す。ｅ _Qは、中央に位置する要素が１で、それ以外の要素が０となる全要素の数が２Ｑ−１の列ベクトルであり、以下の式で定義される。
【００２５】
ｅ _Q＝［０…０１０…０］^T （９）
このとき、ＭＭＳＥフィルタ１６の出力ｚ₁（ｋ）は以下の式で表される。

この、ＭＭＳＥフィルタ１６の出力ｚ₁（ｋ）とインパルスレスポンスを用いて生成された行列Ｈ ₁（ｋ）を用いて、尤度値計算器２４により尤度値λ₁（ｋ）が計算される［文献５］。
λ₁（ｋ）＝４×Ｒｅ｛ｚ₁（ｋ）｝／（１−μ₁（ｋ）） （１１）
ここで、Ｒｅ｛ｚ（ｋ）｝はｚ（ｋ）の実数部の値を取ることを表し、μ₁（ｋ）は以下の式で定義される。
【００２６】

以上が、１回目の第１ストリーム用等化器２１-１の処理であり、尤度値λ₁（ｋ）が出力される。
次に、１回目の第２ストリーム用等化器２１-２の処理について説明する。第２ストリームの信号検出においては、第１ストリームの信号成分は干渉成分となるので、受信信号ベクトルから干渉成分を引いた信号ベクトルがＭＭＳＥフィルタ１６にて処理される。以下、第２ストリーム用等化器２１-２における処理を順に述べる。はじめに伝送路推定器１３により推定されたインパルスレスポンスを用いて、行列Ｈ ₂（ｋ）を生成する。干渉成分生成器２３では、第１ストリーム用等化器２１-１の尤度値λ₁（ｋ）の出力を用いて、式（６）で定義される尤度値ベクトルｂ ₁（ｋ）を生成し、これと行列Ｈ ₁（ｋ）とから第１ストリーム信号成分Ｈ ₁（ｋ）ｂ ₁（ｋ）を干渉成分として生成する。式（１３）に示すように受信信号ベクトルから第１ストリームによる干渉成分が差し引かれ、ＭＭＳＥフィルタ１６に入力される。
【００２７】
ｒ ₁′（ｋ）＝ｒ（ｋ）−Ｈ ₁（ｋ）ｂ ₁（ｋ） （１３）
なお、受信信号ｒ₁（ｋ），…，ｒ_N（ｋ）からＨ ₁（ｋ）ｂ ₁（ｋ）の対応する成分が引算部１５-１〜１５-Ｎでそれぞれ差し引かれる。
ＭＭＳＥにフィルタ１６では、信号ベクトルｒ′（ｋ）と、インパレスレスポンスから得られる行列Ｈ ₁（ｋ），Ｈ ₂（ｋ）と、尤度値ベクトルｂ ₁（ｋ）を用いてフィルタ処理が行われる。このとき、フィルタ出力ｚ₂（ｋ）は以下の式で表される。

ここで、Λ₁（ｋ）はｂ ₁（ｋ）の要素を用いて以下の形で定義される対角行列である。
【００２８】
【数３】

【００２９】
最後に、尤度値計算器２４では、フィルタ１６の出力と尤度値ベクトルｂ ₁
（ｋ）とインパレスレスポンスから得られる行列Ｈ ₁（ｋ），Ｈ ₂（ｋ）を用いて、以下の式で表される尤度値出力λ₂（ｋ）が計算される。
λ₂（ｋ）＝４×Ｒｅ｛ｚ₂（ｋ）｝／（１−μ₂（ｋ）） （１６）
ここで、μ₂（ｋ）は以下の式で定義される。

同様にして、１回目の第Ｍストリーム用等化器２１-Ｍにおける処理では、既に処理が終了している尤度値λ₁（ｋ），…，λ_M-1（ｋ）から得られる尤度値ベクトル（ｂ ₁（ｋ），…，ｂ _M-1（ｋ）を用いて、受信信号ベクトルｒ（ｋ）から他のストリームの信号成分が干渉成分として式（１８）に示すように差し引かれ、

この信号ベクトルｒ′（ｋ）がＭＭＳＥフィルタ１６で処理され、フィルタ出力ｚ_M（ｋ）が式（１９）に示すように得られる。
【００３０】

尤度値計算器２４では、フィルタ出力ｚ_M（ｋ）、既に処理が終了している尤度値λ₁（ｋ），…，λ_M-1（ｋ）、推定されたインパルスレスポンスから得られる行列Ｈ ₁（ｋ），…，Ｈ _M（ｋ）を用いて、尤度値出力λ_M（ｋ）が計算される。
λ_M（ｋ）＝４×Ｒｅ｛ｚ_M（ｋ）｝／（１−μ_M（ｋ）） （２０）
ここで、μ_M（ｋ）は以下の式で定義される。
【００３１】

以上が、１回目の各ストリーム用等化器における処理となる。続いて、２回目における各ストリーム用等化器における処理を説明する。はじめに、２回目の第１ストリーム用等化器２１-１における処理について述べる。１回目の処理とは異なり、尤度値λ₁（ｋ），…，λ_M（ｋ）を利用して処理が行われる。受信信号ベクトルから第２ストリームから第Ｍストリームの信号成分が干渉成分として差し引かれる。
【００３２】

ここで、ｂ ₁′（ｋ）は式（６）におけるｂ ₁（ｋ）の中央に位置する要素をゼロとしたベクトルであり、以下の形で定義される。
ｂ ₁′（ｋ）＝［tanh(λ₁(k+(Q-1))/2…tanh(λ₁(k+1)/2 ０ tanh(λ₁(k-1))/2…tanh(λ₁(k-(Q-1))/2]^T （２３）
これは、第１ストリームの時点ｋの信号成分を残し、その前後の符号間干渉成分を除去するように作用する。
【００３３】
干渉成分が差し引かれた式（２２）で与えられる信号ベクトルｒ ₁′（ｋ）がＭＭＳＥフィルタ１６にて処理され、２回目のフィルタ出力ｚ₁′（ｋ）が式（２４）により計算される。

ここで、Λ₁′（ｋ）は式（１５）で定義されるΛ₁（ｋ）の中央に位置する要素を１とした対角行列であり、以下の形で表される。
【００３４】
【数４】

【００３５】
中央の要素のみを１とすることは、ｂ ₁′（ｋ）の場合と同様に、時点ｋの
信号成分を残し、その前後の符号間干渉成分を除去するように作用する。最後に、尤度値計算器２４により尤度値出力λ₁′（ｋ）が計算される。
λ₁′（ｋ）＝４×Ｒｅ｛ｚ₁′（ｋ）｝／（１−μ₁′（ｋ）） （２６）
ここで、μ₁′（ｋ）は以下の式で定義される。
μ₁′(k)＝ｅ _Q ^T Ｈ ₁ ^H(k)｛Ｈ ₁(k)Λ₁′(k)Ｈ ₁ ^H(k)＋Ｈ ₂(k)Λ₂(k)Ｈ ₂ ^H(k)＋…＋Ｈ _M(k)Λ_M Ｈ _M ^H(k)＋σ² Ｉ｝^-1 Ｈ ₁(k)ｅ _Q (２７）
同様にして、第２ストリーム用等化器２１-２の２回目の処理においても、はじめに受信信号ベクトルから干渉成分を引いた信号ベクトルを以下の式により求める。
【００３６】

ここで、ｂ ₁(ｋ）は、第１ストリーム用等化器２１-１において２回目の等化処理により導出された尤度値出力λ₁′(ｋ）を用いて生成される尤度値ベクトルであり、ｂ ₂′(ｋ）は式（２３）の場合と同様に、中央の要素を０として、残りは１回目の第２ストリーム用等化器２１-２から出力された尤度値λ₂(ｋ）を用いて生成される尤度値ベクトルである。この差ベクトルｒ ₂′(ｋ）がＭＭＳＥフィルタ１６に入力されてフィルタ処理される。このとき、フィルタ出力ｚ₂′(ｋ）は以下の式で表される。
【００３７】
ｚ₂′(k)＝ｅ _Q ^T Ｈ ₂ ^H(k)｛Ｈ ₁(k)Λ₁(k)Ｈ ₁ ^H(k)＋Ｈ ₂(k)Λ₂′(k)Ｈ ₂ ^H(k)＋…＋Ｈ _M(k)Λ_M(ｋ）Ｈ _M ^H(k)＋σ² Ｉ｝^-1 ｒ ₂′(k) (２９)
ここで、Λ₂′(ｋ）はΛ₂(ｋ）の中央に位置する要素を１とした対角行列である。このフィルタ出力を用いて、尤度値計算器２４により尤度値出力λ₂′(ｋ）が計算される。
λ₂′(k）＝４×Ｒｅ｛ｚ₂′(k)｝／（１−μ₂′（ｋ）） （３０）
ここで、μ₂′（ｋ）は以下の式で定義される。
【００３８】
μ₂′(k)＝ｅ _Q ^T Ｈ ₂ ^H(k)｛Ｈ ₁(k)Λ₁(k)Ｈ ₁ ^H(k)＋Ｈ ₂(k)Λ₂′(k)Ｈ ₂ ^H(k)＋Ｈ ₃(k)Λ₃(ｋ） Ｈ ₃ ^H(k)＋…＋Ｈ _M(k)Λ_M(ｋ） Ｈ _M ^H(k)＋σ² Ｉ｝^-1 Ｈ ₂(k)ｅ _Q (３１）
同様にして、第ｍストリーム用等化器２１-ｍのａ回目（ａは１以上の任意の整数）の処理においては、ａ−１回目の等化処理における第ｍストリーム用等化器２１-ｍから第Ｍストリーム等化器２１-Ｍの尤度値出力と、ａ回目の等化処理における第１ストリーム用等化器２１-１から第ｍ−１ストリーム用等化器２１-ｍ−１の尤度値出力を用いて、はじめに受信信号から干渉成分を次式により差し引く。
【００３９】
ｒ _m′(k）＝ｒ（ｋ）−Ｈ ₁(ｋ）ｂ ₁(ｋ）−…−Ｈ _m-1(ｋ）ｂ _m-1(ｋ）−Ｈ _m(ｋ）ｂ _m′(ｋ）−Ｈ _m+1(ｋ）ｂ _m+1(ｋ）−…−Ｈ _M(ｋ）ｂ _M(ｋ） （３２）
ここで、ｂ ₁(ｋ），…，ｂ _m-1(ｋ）はそれぞれ第１ストリーム用等化器２１-１から第ｍ−１ストリーム用等化器２１-ｍ−１のａ回目の等化処理により得られた尤度値出力より生成される尤度値ベクトルを表し、ｂ _m′(ｋ），ｂ _m+1(ｋ），…，ｂ _M(ｋ）は第ｍストリーム用等化器２１-ｍから第Ｍストリーム用等化器２１-Ｍのａ−１回目の等化処理により得られた尤度値出力より生成される尤度値ベクトルを表す（ａ＝１である１回目の等化処理においては、ｂ ₁(ｋ），…，ｂ _M(ｋ）は全てゼロにより構成されるベクトルとなる）。また、尤度値ベクトルｂ _k′（ｋ）の中央に位置する要素はゼロとする。この信号ベクトルｒ′（ｋ）がＭＭＳＥフィルタ１６により処理されて、以下の式で示されるフィルタ出力ｚ_m(ｋ）を得る。
【００４０】
ｚ_m(k)＝ｅ _Q ^T Ｈ _m ^H(k)｛Ｈ ₁(k)Λ₁(k)Ｈ ₁ ^H(k)＋…＋Ｈ _m-1(k)Λ_m-1(ｋ）Ｈ _m-1 ^H(k)＋Ｈ _m(k)Λ_m′(ｋ）Ｈ _m ^H(k)＋Ｈ _m+1(k)Λ_m+1(ｋ）Ｈ _m+1 ^H(k)＋…＋Ｈ _M(k)Λ_M(ｋ）Ｈ _M ^H(k)＋σ² Ｉ｝^-1ｒ_m′(k) (３３)
ここで、Λ₁(ｋ），…，Λ_m-1(ｋ）は、それぞれ第１ストリーム用等化器２１-１から第ｍ−１ストリーム用等化器２１-ｍ−１のａ回目の等化処理により得られた尤度値出力により生成される対角行列であり、Λ_m′（ｋ），Λ_m+1(ｋ），…，Λ_M(ｋ）は、それぞれ第ｍストリーム用等化器２１-ｍから第Ｍストリーム用等化器２１-Ｍのａ−１回目の等化処理により得られた尤度値出力により生成される対角行列である。また、Λ_m′（ｋ）の中央に位置する要素は１とする。最後にフィルタ出力を用いて、尤度値計算器２４により尤度値出力λ_m(ｋ）が得られる。
【００４１】
λ_m(ｋ）＝４×Ｒｅ｛ｚ_m(ｋ）｝／（１−μ_m(ｋ）） （３４）
ここで、μ_m(ｋ）は以下の式で定義される。
μ_m(k)＝ｅ _Q ^T Ｈ _m ^H(k)｛Ｈ ₁(k)Λ₁(k)Ｈ ₁ ^H(k)＋…＋Ｈ _m-1(k)Λ_m-1(k)Ｈ _m-1 ^H(k)＋Ｈ _m(k)Λ_m′(ｋ） Ｈ _m ^H(k)＋Ｈ _m+1(k)Λ_m+1(ｋ） Ｈ _m+1 ^H(k)＋…＋Ｈ _M(k)Λ_M(ｋ） Ｈ _M ^H(k)＋σ² Ｉ｝^-1 Ｈ _m(k)ｅ _Q (３５）
以上の処理を繰り返し行うことで、各ストリーム用等化器において、より確からしい尤度値出力を得ることができるために、最終的に得られた尤度値λ₁(ｋ）〜λ_M(ｋ）の出力を判定器１７−１〜１７−Ｍに通して得られるデータ出力として誤りの少ないデータを得ることができる。なお、第ｍストリーム用等化器２１-ｍにおける干渉成分生成器２３の機能構成は図３Ａに示すようになる。端子２２-１〜２２-ｍ−１よりの今回（ａ回目）の等化処理で得られた尤度値λ₁（ｋ）〜λ_m-1（ｋ）が尤度ベクトル生成部２６-１〜２６-ｍ−１に入力されて、それぞれ式（６）中の添字「１」をそれぞれ「１」，〜，「ｍ−１」とした尤度ベクトルｂ ₁（ｋ）〜ｂ _m-1（ｋ）が生成され、端子２２-ｍ〜２２-Ｍより前回（ａ−１回目）の等化処理で得られた尤度値λ_m（ｋ）〜λ_M（ｋ）が尤度ベクトル生成部２６-ｍ〜２６-Ｍに入力され、尤度ベクトル生成部２６-ｍ＋１〜２６-Ｍでは式（６）中の添字「１」をそれぞれ「ｍ＋１」，〜，「Ｍ」とした尤度ベクトルｂ _m+1（ｋ）〜ｂ _M（ｋ）が生成され、尤度ベクトル生成部２６-ｍでは式（２３）中の添字「１」を「ｍ」とした尤度ベクトルｂ _m′（ｋ）が生成される。
【００４２】
これら尤度ベクトルｂ ₁（ｋ），ｂ _m-1（ｋ），ｂ _m′（ｋ），ｂ _m+1（ｋ），…ｂ _M（ｋ）と伝送路推定器１３よりのインパルスレスポンス行列Ｈ ₁（ｋ），…，Ｈ _m-1（ｋ），Ｈ _m（ｋ），Ｈ _m+1（ｋ），…，Ｈ _M（ｋ）とがそれぞれ乗算部２７-１〜２７-Ｍで乗算されて、第１〜第Ｍストリームの第ｍストリームに対する各干渉成分が生成される。
またフィルタ１６の機能構成は図３Ｂに示すように、対角行列生成部２８に今回（ａ回目）の等化処理で得られた尤度値λ₁（ｋ）〜λ_m-1（ｋ）と前回（ａ−１回目）の等化処理で得られた尤度値λ_m+1（ｋ）〜λ_M（ｋ）が入力され、式（１５）中の添字「１」をそれぞれ「１」〜「ｍ−１」，「ｍ＋１」〜「Ｍ」とした対角行列Λ₁（ｋ）〜Λ_m-1（ｋ），Λ_m+1（ｋ）〜Λ_M（ｋ）が生成され、また前回（ａ−１回目）の等化処理で得られた尤度値λ_m（ｋ）が対角行列生成部２８′に入力され、式（２５）中の添字「１」を「ｍ」とした対角行列Λ_m′（ｋ）が生成される。これら対角行列Λ₁（ｋ）〜Λ_m-1（ｋ），Λ_m+1（ｋ）〜Λ_M（ｋ）及びΛ_m′（ｋ）と、インパルスレスポンスＨ ₁（ｋ）〜Ｈ _M（ｋ）と平均雑音電力σ²とが行列演算部２９に入力され、式（３３）の右辺中の差ベクトルｒ _m′（ｋ）を除いた式が演算されてフィルタ係数Ｗ _m（ｋ）が計算される。対角行列生成部２８及び２８′と行列演算部２９によりフィルタ係数計算部３１が構成される。受信ベクトルｒ（ｋ）から干渉成分が差し引かれた差ベクトルｒ _m′（ｋ）がフィルタ係数Ｗ _m（ｋ）とによりフィルタ処理部３２でフィルタ演算処理され、つまりＷ _m（ｋ）ｒ _m′（ｋ）が演算されてフィルタ出力Ｚ_m（ｋ）が得られる。
【００４３】
尤度値計算部２４ではフィルタ出力ｚ_m（ｋ）と、フィルタ係数Ｗ _m（ｋ）と、インパルスレスポンスＨ _m（ｋ）とが入力されて、μ_m（ｋ）＝Ｗ _m（ｋ）Ｈ _m（ｋ）ｅ _Q、つまり式（３５）が演算部２４ａで計算され、更に演算部２４ｂでμ_m（ｋ）とｚ_m（ｋ）とから式（３４）が計算されて尤度値λ_m（ｋ）が得られる。この尤度値λ_m（ｋ）は判定器１７-ｍに入力され、２値判定が行われる。前述したように尤度値λ_m（ｋ）はtanh（λ_m（ｋ）／２）と正規化されて各種演算に利用されるため、各ストリーム用等化器２１-ｍ（ｍ＝１，・・・，Ｍ）に共通の正規化部により各tanh（λ_m（ｋ）／２）の演算を行うことが好ましい。
【００４４】
以上の等化処理の手順の例を図４に示す。まずインパルスレスポンスＨ ₁（ｋ）〜Ｈ _M（ｋ）を求め（Ｓ０）、パラメータｍを１とし（Ｓ１）、今回求めた尤度値λ₁（ｋ）〜λ_m-1（ｋ）（又はtanh（λ₁（ｋ）／２）〜tanh（λ_m-1（ｋ）／２）以下同様）と前回求めた尤度値λ_m（ｋ）〜λ_M（ｋ）（又はtanh（λ_m（ｋ）／２）〜tanh（λ_M（ｋ）／２）以下同様）により尤度ベクトル生成部２６−１〜２６−Ｍで尤度ベクトルΛ₁（ｋ），…，Λ_m-1（ｋ），Λ_m′（ｋ），Λ_m+1（ｋ），…，Λ_M（ｋ）を生成する（Ｓ２）。これら尤度ベクトルとインパルスレスポンスＨ ₁（ｋ）〜Ｈ _M（ｋ）とを用いて干渉成分を乗算部２７-１〜２７-Ｍで計算する（Ｓ３）。各尤度ベクトルを生成するごとに対応する干渉生成を計算してもよい。
【００４５】
受信ベクトルｒ（ｋ）からこれら干渉成分を除去して差ベクトルｒ _m′（ｋ）を求める（Ｓ４）。また今回求めた尤度値λ₁（ｋ）〜λ_m-1（ｋ）と前回求めた尤度値λ_m（ｋ）〜λ_M（ｋ）とインパルスレスポンスＨ ₁（ｋ）〜Ｈ _M（ｋ）と平均雑音電力σ²とを用いてフィルタ係数計算部３１でフィルタ係数Ｗ _m（ｋ）を計算する（Ｓ５）。フィルタ係数Ｗ _m（ｋ）で差ベクトルｒ _m′（ｋ）をフィルタ処理してフィルタ出力ｚ_m（ｋ）を求める（Ｓ６）。フィルタ出力ｚ_m（ｋ）、フィルタ係数Ｗ _m（ｋ）、インパルスレスポンスＨ _m（ｋ）により尤度値計算器２４により尤度値λ_m（ｋ）（又はtanh（λ_m（ｋ）／２））を計算する（Ｓ７）。
【００４６】
ｍがＭになったかを調べ（Ｓ８）、Ｍになっていなければｍを＋１してステップＳ２に戻る（Ｓ９）。ｍがＭになったら、等化処理を所定回数行ったかを調べ（Ｓ１０）、所定回数行っていなければステップＳ１に戻り、所定回数行ったならば、この時の尤度値λ_m（ｋ）（又はtanh（λ_m（ｋ）／２））を判定器１７で判定して出力する（Ｓ１１）。ステップＳ８でｍ＝Ｍであったら、破線枠Ｓ１２で示すように尤度値λ_m（ｋ）の判定器１７による判定を行って、途中結果として出力するようにしてもよい。ステップＳ５のフィルタ係数計算を、ステップＳ２〜Ｓ４の前に行ってもよい。
【００４７】
計算機シミュレーションにより、実施例１による受信機と、文献４による受信機の特性比較を行った結果を図５に示す。計算機シミュレーションの諸元として、送信アンテナ数Ｍ及び受信アンテナ数Ｎはそれぞれ５として変調方式はＢＰＳＫとした。伝搬路は１波レイリーフェージング伝搬路として、送信信号の１フレーム内で時間変動はしないこととし、伝送路（インパルスレスポンス）推定は理想的に行われているものとし、各フレームは情報２５６シンボルにより構成されることとした。受信機ではインパルスレスポンスにより各ストリームの受信信号電力を比較して、最も受信電力の強いストリームを第１ストリームとして、受信電力の大きい順に各ストリーム用等化器における処理を行うこととした。
【００４８】
図５において破線３５は文献４の受信機構成において第１ストリーム用等化器１２-１から第Ｍストリーム用等化器１２-Ｍまでの処理を行った後の、各ストリームのデータ出力の平均ビット誤り率を示す。破線３６は文献４の受信機構成において第１ストリーム用等化器１２-１から第Ｍストリーム用等化器１２-Ｍまでの処理を行った後、得られたデータ出力を用いて、さらに第１ストリーム用等化器１２-１から第Ｍストリーム用等化器１２-Ｍまでの処理を行った後の各ストリームのデータ出力の平均ビット誤り率を示す。実線３７は実施例１の受信機構成を用いて第１ストリーム用等化器２１-１から第Ｍストリーム用等化器２１-Ｍまでの処理が終わった後での平均ビット誤り率を示し、実線３８は実施例１の受信機構成を用いて第１ストリーム用等化器２１-１から第Ｍストリーム用等化器２１-Ｍまでの処理が終わった後、さらに再び第１ストリーム用等化器２１-１から第Ｍストリーム用等化器２１-Ｍまでの処理を行った後での平均ビット誤り率を示す。つまり、実線３８は、各ストリーム用等化器における処理を２回繰り返した後の特性となる。
【００４９】
Ｅ_bはビットエネルギー、Ｎ₀ は雑音電力を示す。
図５に示す結果より、文献４の方法を用いた場合は処理を繰り返しても殆ど特性が改善されないが、実施例１の構成では、処理を繰り返すことにより特性が改善するとともに、各ストリーム用等化器の処理を１回だけ行った場合の結果においても、文献４の受信機構成による特性よりも優れていることが確認された。
次に、３波レイリーフェージング伝搬路（先行波、１シンボル遅延波、２シンボル遅延波）における特性を図６に示す。ここで、等化器が考慮する最大遅延シンボル数は２とした。破線４１，４２，４３は文献４の方法を用いてそれぞれ１回、２回、３回処理した場合であり、実線４４，４５，４６は実施例１の構成によりそれぞれ１回、２回、３回処理した場合である。この場合は文献４の方法も処理を繰返し行うことにより特性が改善しているが、実施例１の構成の特性の方が優れていることが確認できる。文献４の方法では遅延波が考慮されていないが、図６より遅延波を考慮した場合においても、実施例１は文献４の方法に比べて特性を改善することができる。
【００５０】
図１に示した構成中の判定器１７-１〜１７-Ｍを除いた部分は多入力多出力等化器を構成している。
実施例２
続いて、この発明の実施例２を図７に基づいて説明する。図７においては図１に示したストリーム用等化器２１-１〜２１-Ｍの代りにストリーム用等化器４７-１〜４７-Ｍを用い、そのストリーム用等化器の尤度値出力の受け渡し方が実施例１と異なる。実施例１では、第ｍストリーム用等化器２１-ｍの処理により得られた尤度値出力を第ｍ＋１ストリーム用等化器２１-ｍ＋１の処理において用いていたが、実施例２では第ｍストリーム用等化器４７-ｍよりの尤度値出力を第ｍ＋１ストリーム用等化器４７-ｍ＋１の処理に用いない。
【００５１】
したがって、実施例２では始めに第１ストリーム用等化器４７-１から第Ｍストリーム用等化器４７-Ｍは受信信号ベクトルのみを用いて等化処理を行う。全ストリーム用等化器４７-１〜４７-Ｍにおける１回目の処理が終了した後、得られた各尤度値出力と同一の受信信号ベクトルを用いて、再び各ストリーム用等化器４７-１〜４７-Ｍにおいて処理が行われる。この構成では、各ストリーム用等化器４７-１〜４７-Ｍの処理を同時に行うことができるので、全体の処理に必要な時間を削減できる。
この場合の処理手順の例を図８を参照して簡単に説明する。まずインパルスレスポンスＨ ₁（ｋ）〜Ｈ _M（ｋ）を推定し（Ｓ０）、これらを用いて以下の処理をｍ＝１，…，Ｍについて行う。まずフィルタ係数Ｗ _m（ｋ）を、式（８）の添字の「１」を「ｍ」とすることにより計算し（Ｓ１）、これら各フィルタ係数Ｗ _m（ｋ）により受信ベクトルｒ（ｋ）をフィルタ処理し（Ｓ２）、各フィルタ処理結果ｚ_m（ｋ）とフィルタ係数Ｗ _m（ｋ）、インパルスレスポンスＨ _m（ｋ）を用いて尤度値λ_m（ｋ）（又はtanh（λ_m／２））を求める（Ｓ３）。これまでが１回目の処理である。
【００５２】
次にｍ＝１，…，Ｍについて以下の処理を行う。まず前回の尤度値λ₁（ｋ）〜λ_M（ｋ）を用いて尤度ベクトルを生成する（Ｓ４）、この尤度ベクトルの生成は図４中のステップＳ２と同様であるが、ただ尤度値として前回処理により求めたものを全て用いる点が異なる。次に各ｍ＝１，…，Ｍストリームに対する干渉成分をそれぞれ生成し（Ｓ５）、これら干渉成分を、受信ベクトルｒ（ｋ）からそれぞれ除去してｒ ₁′（ｋ）〜ｒ _M′（ｋ）を求め（Ｓ６）、またフィルタ係数Ｗ ₁（ｋ）〜Ｗ _M（ｋ）を計算する（Ｓ７）。この計算は図４中のステップＳ５と同様であるが、ただ尤度値として前回処理により求めたものを全て用いる点が異なる。
【００５３】
これらフィルタ係数Ｗ ₁（ｋ）〜Ｗ _M（ｋ）によりｒ ₁′（ｋ）〜ｒ _M′（ｋ）をそれぞれフィルタ処理し（Ｓ８）、これらフィルタ処理結果ｚ₁（ｋ）〜ｚ_M（ｋ）とフィルタ係数Ｗ ₁（ｋ）〜Ｗ _M（ｋ）と、インパルスレスポンスＨ ₁（ｋ）〜Ｈ _M（ｋ）をそれぞれ用いて、図４中のステップＳ７と同様な計算により尤度値λ₁（ｋ）〜λ_M（ｋ）をそれぞれ計算する（Ｓ９）。処理回数が所定回数になったかを調べ（Ｓ１０）、なっていなければステップＳ４に戻り、なっていれば尤度値λ₁（ｋ）〜λ_M（ｋ）をそれぞれ判定器で判定して出力する（Ｓ１１）。ステップＳ９の後に、尤度値λ₁（ｋ）〜λ_M（ｋ）を判定して途中結果として出力してもよい（Ｓ１２）。
【００５４】
この、実施例１による受信機構成と実施例１による受信機構成の各特性を計算機シミュレーションにより比較した。その結果を図９に示す。計算機シミュレーション諸元は図６の場合と同等であり、図９における実線５１，５２，５３は実施例２の受信機構成における１回、２回、３回の処理の特性を示し、破線５４，５５，５６は実施例１における受信機構成の１回、２回、３回の処理の特性を示す。実施例２の受信機構成における１回目、２回目の処理が終了した後の特性は、実施例１の受信機構成における特性の対応するものよりも劣化しているが、３回目の処理が終了した後の特性は、ほぼ同等である。したがって、３回繰返し処理を行う場合は、実施例２の受信機構成を用いることにより、実施例１の受信機構成を用いる場合よりも、処理時間を削減することができる。
【００５５】
図７中の判定器１７-１〜１７-Ｍを除いた部分は多入力多出力等化器を構成している。
実施例３
次に、この発明の実施例３について説明する。文献５ではＭＩＭＯチャネル信号伝送に関して、等化と復号を繰返し処理する受信機構成が提案されている。図２２は、文献５における送受信機構成を示す。第１〜第Ｍストリームのデータは符号器６１-１〜６１-Ｍでそれぞれ符号化され、これら符号化データはインターリーバ６２-１〜６２-Ｍでフレーム間の符号化系列の並ぶ順序が変えられ、これらデータにより、送信機６３-１〜６３-Ｍで搬送波が変調され、送信アンテナＡＮＳ-１〜ＡＮＳ-Ｎに供給される。受信アンテナＡＮＲ-１〜ＡＮＲ-Ｍに受信された信号はＭＩＭＯ等化器６４に入力され、その等化出力（尤度値出力）はデインターリーバ６５-１〜６５-Ｍでそれぞれ並びかえが戻され、復号器６６-１〜６６-Ｍで復号される。これら復号器６６-１〜６６-Ｍより各尤度値出力はインターリーバ６７-１〜６７-Ｍでそれぞれ並びかえられてＭＩＭＯ等化器６４に入力され、先の受信信号の等化処理に利用される。
【００５６】
ＭＩＭＯ等化器６４は図２３に示すように各アンテナＡＮＲ-１〜ＡＮＲ-Ｎからの受信信号サンプル値は第１〜第Ｍストリーム用等化器６８-１〜６８-Ｍにそれぞれ入力され、これら等化器６８-１〜６８-Ｍには復号器６６-１〜６６-Ｍからの各尤度値がインターリーバをそれぞれ介して入力される。各ストリーム用等化器６８-ｍにおいて尤度値出力を導出して、その値をＳＩＳＯ（Soft-Input Soft-Output）復号器６６-ｍにて復号を行う。復号器６６-ｍから得られた尤度値出力は、インターリーバ６７-ｍを介して再び各ストリーム用等化器６８-ｍにおいて用いられ処理が行われる。以上の処理を繰返し行い、最終的に復号器６６-１〜６６-Ｍから第１〜第Ｍストリームのデータ出力が得られる。図２０に示した構成では、各ストリーム用等化器６８-１〜６８-Ｍの処理を同時に行うことができる。
【００５７】
実施例３は、各ストリーム用等化器は処理を順次行い、復号器よりの尤度値と今回の処理におけるストリーム用等化器の出力尤度値も利用される。つまり図１０に示すように第１〜第Ｍストリーム用等化器７１-１〜７１-Ｍにおいて、これらは処理を順次行い、第ｍストリーム用等化器７１-１には第１〜第ｍ−１ストリーム用等化器７１-１〜７１-ｍ−１よりの尤度値出力と、復号器６６-ｍ〜６６-Ｍよりの尤度値出力とが入力される。このように第ｍストリーム用等化器７１-ｍにおいて得られた尤度値出力を第ｍ＋１ストリーム用等化器７１-ｍ＋１において用いることで、より確からしい尤度値を用いて処理を行うことができるために、特性を改善することができる。
実施例４
次に、この発明の実施例４について図１１に基づいて説明する。実施例４は実施例１の方法をマルチキャリア型として遅延波の影響軽減を行い、各ストリーム用等化器において等化器が到来波を一波のみ考慮する構成として、各ストリーム用等化器の演算量を削減することを特徴とする。マルチキャリア方式は、高速伝送を行う上で問題となる遅延波の影響を軽減するために複数のキャリア（搬送波）を用いて信号を伝送する方式であり、デジタル放送などにおいて注目されている技術である。マルチキャリア方式は、一つのキャリアを用いて信号伝送を行う場合と同じ帯域幅の周波数を複数のサブキャリアに分けて信号伝送を行うことにより、各サブキャリアにおける情報伝送速度を下げることができるために、遅延波の影響を軽減することができる。また、ガードインターバルを設けることにより、効率的に遅延波の影響を軽減することができる。このマルチキャリア方式は、送受信機をキャリアの数だけ用いることにより実現することが可能であるが、ＯＦＤＭ（Orthogonal Frequency Division Multiplexing）と呼ばれる高速フーリエ変換を用いてマルチキャリア多重化を行う方法が一般的である。
【００５８】
図１１を参照してマルチキャリア方式について説明する。送信側では第１〜第Ｍストリームにおいて、変調器７３-１〜７３-ＭによりＢＰＳＫやＱＰＳＫなどの変調を行い、その後で直列−並列変換器７４-１〜７４-Ｍにより直並列変換を行い、更にマルチキャリア化を行う。マルチキャリア化の一手段として、図１１では高速逆フーリエ変換（Inverse Fast Fourier Transform）器７５-１〜７５-Ｍを用い、各並列出力ごとに高速逆フーリエ変換を行ってマルチキャリア化を行っている。図１１では略されているが、高速逆フーリエ変換された信号はガードインターバルが付加された後で送信される。
【００５９】
受信側では、各アンテナＡＮＲ-１〜ＡＮＲ-Ｎにおいて受信されたマルチキャリア多重化された信号を各サブキャリアごとの信号に戻す。図１１ではマルチキャリア多重化された信号を各サブキャリアごとの信号に戻す一手段として高速フーリエ変換（Fast Fourier Transform）器７６-１〜７６-Ｎを用いている。高速フーリエ変換器７６-１〜７６-Ｎで高速フーリエ変換された各周波数成分、つまり各サブキャリアは、サブキャリアごとにＭＩＭＯ等化器７７-１〜７７-Ｊにより、複数の送信アンテナから送信された信号の分離を行う。つまりアンテナＡＮＲ-１〜ＡＮＲ-Ｎより得られた各サブキャリアｆ_j(ｊ＝１，…，Ｊ）は同一のＭＩＭＯ等化器７７-ｊに入力される。
【００６０】
各ＭＩＭＯ等化器７７-ｊの構成は図１２に示す受信機構成となる。同一サブキャリアｆ_jの各アンテナ対応の受信信号サンプル値が端子１１-１〜１１-Ｎに入力される。これらは第１〜第Ｍストリーム用等化器７９-１〜７９-Ｍに入力される。端子８１-１〜８１-Ｍに後で述べる並列直列変換器７８-１〜７８-Ｍよりの第１〜第Ｍストリームの尤度値が入力される。第ｍストリーム用等化器７９-ｍには第１〜第ｍ−１ストリーム用等化器７９-１〜７９-ｍ−１の出力尤度値と、端子８１-ｍ〜８１-Ｍよりの尤度値が入力されて、等化処理に用いられる。つまりこれ等の等化処理動作は図１、図１０と同様である。
【００６１】
先に述べたように、直並列変換することにより、更に必要に応じてガードインターバルを設けることにより、遅延波の影響を軽減することができるため、各ストリーム用等化器７９-Ｍの構成は図２に示した構成において、到来波を１波のみ考慮して遅延波を考慮しない場合の構成となる。各ＭＩＭＯ等化器７７-１〜７７-Ｊの各第１〜第Ｍストリーム対応尤度値出力はそれぞれ第１〜第Ｍ並列−直列変換器７８-１〜７８-Ｍでそれぞれ直列の尤度値出力とされて、これら第１〜第Ｍ直列尤度値出力はそれぞれ判定器１７-１〜１７-Ｍへ供給されると共に図１２中の端子８１-１〜８１-Ｍへ供給される。
【００６２】
各ストリーム用等化器７９-ｍは到来波を１波のみ考慮し、遅延を考慮しない構成であるため、その結果、マルチキャリア型としない場合は受信アンテナ数Ｎ、等化器２１-ｍが考慮する最大遅延シンボル数をＱ−１とすると、式（３５）の逆行列演算のために演算量は（Ｎ×Ｑ）³ のオーダーで増加してしまうのに対して、実施例４では等化器７９-ｍは遅延波を考慮しないために、演算量はＮ³ のオーダーでの増加に抑えることができる。つまり、実施例４により演算量を従来の方式と比較して１／Ｑ³に削減することができる。
なお、高速フーリエ変換器７６-ｎ（ｎ＝１，…，Ｎ）の代りに複数の周波数発振器に周波数変換器の組合せを用いてもよい、つまり各受信アンテナＡＮＲ-ｎの受信信号対応は、その受信信号をサブキャリアごとに分離する分離部を設ければよい。
実施例５
この発明の実施例５について説明する。図１１に示したマルチキャリア型ＭＩＭＯチャネル信号伝送において、実施例４では、ＭＩＭＯ等化器７７-ｊの構成は図１２に示した構成としたが、この実施例５ではＭＩＭＯ等化器７７-ｊにおけるストリーム用等化器７９-１〜７９-Ｍの接続構成を図１３に示すように図７に示したと同様の接続構成とする。この構成により各ストリーム用等化器７９-ｍの処理を同時に行うことができる。したがって、図１２に示した構成と比較して、処理にかかる時間を削減することができる。
実施例６
この発明の実施例６について図１４及び図１５を参照して説明する。図１４及び図１５に示すマルチキャリア型ＭＩＭＯチャネル信号伝送の送受信構成は、図１１に示したマルチキャリア型ＭＩＭＯチヤネル信号伝送に、図２２に示した符号化を加えた構成であって、対応する部分に同一参照符号を付けてある。図１４において、第１ストリームから第Ｍストリームの各送信アンテナから送信される情報は、始めに符号器６１-ｍによりそれぞれ符号化が行われ、インターリーバ６２-ｍによってフレーム内の符号化系列が並ぶ順序が変えられる。ここで、図では各送信アンテナにおける符号化を別々に表しているが、例えば一人のユーザが複数のアンテナＡＮＳ-１〜ＡＮＳ-Ｍを用いて信号伝送を行う場合は、各アンテナＡＮＳ-１〜ＡＮＳ-Ｍに対応する符号化を協調して行うことにより、受信品質の特性が改善できる。これ以降における送信側での処理は、図１１に示した処理と同等となる。
【００６３】
受信側では、図１５に示すように、マルチキャリア多重化された信号を各キャリアごとの信号に戻して、図１０に示した構成のＭＩＭＯ等化器７７-ｊを用いて信号分離を行い、並列−直列変換器７８-ｍにより並直列変換を行う。その後に、デインターリーバ６５-ｍを介して復号器６６-ｍにより復号を行う。ここで、送信側において各送信アンテナにおける符号化を協調して行った場合は、復号についても一括で処理を行うこととなる。復号器６６-ｍとしてはＳＩＳＯ型の復号器を用いて、符号化系列の尤度値出力を導出して、その結果を、インターリーバ６７-ｍを介してＭＩＭＯ等化器７７-１〜７７-Ｊに戻す。このように、信号分離と復号の処理を繰返し行い、復号器６６-ｍから出力されるデータ出力が最終的な判定結果となる。
【００６４】
図１５に示したようにマルチキャリア化することにより、各ストリーム用等化器における演算量を削減することができる。具体的には、マルチキャリア型としない場合は受信アンテナ数Ｎ、等化器が考慮する最大遅延シンボル数をＱ−１とすると、式（３５）の逆行列演算のために各ストリーム用等化器の演算量は（Ｎ×Ｑ）³ のオーダーで増加してしまうのに対して、実施例６では等化器は遅延波を考慮しないために、演算量はＮ³ のオーダーでの増加に抑えることができる。つまり、この実施例６により演算量を従来の方式と比較して１／Ｑ³ に削減することができる。
実施例７
図１５に示した実施例６における送受信機構成において、ＭＩＭＯ等化器７７−ｊの構成を図２３に示した構成とすることにより、各ストリーム用等化器の処理を同時に行うことができる。したがって、図１０に示した構成をとる実施例６の方法と比較して、処理に要する時間を削減することができる。
実施例８
図２０及び図２１に示したＶ−ＢＬＡＳＴ法の構成では、発明が解決しようとする課題の項で述べたように、あるストリームの等化処理により得られた信号判定結果が間違っていた場合、その次に行われるストリームの等化処理において、受信信号から正しく他のストリームの信号成分を差し引くことができないために、さらに誤りが生じてしまう。これは誤り伝播と呼ばれる。Ｖ−ＢＬＡＳＴ型等化器においてより確からしい信号判定結果を用いる方法として、送信側で誤り訂正符号化を行い、受信側で復号器から得られる結果を用いてＶ−ＢＬＡＳＴ型等化器において処理を行い、等化と復号の処理を繰返し行うという方式が考えられる。しかし、この場合においても誤り伝播の影響による特性劣化が起きてしまう。そこでこの発明の実施例８では、等化と復号の処理を繰返し行うこととして、１回目の等化処理は図２０に示した構成により行い、２回目以降の等化処理は図１６に示す構成により前回の復号結果を用い、各ストリーム用等化器１２-ｍとしては尤度値の導出を必要としない図２１に示した構成を用いる。
【００６５】
つまり例えば図１７に示すように、伝送路推定器１３によりインパルスレスポンス行列Ｈ ₁(ｋ)〜Ｈ _M(ｋ)を推定し（Ｓ１）、ｍを１に初期化する（Ｓ２）。第ｍストリーム等化器１２-ｍにおいて、既に１回目の等化処理を済した第１〜第ｍ−１ストリーム等化器１２-１〜１２-ｍ−１よりの判定器１７-１〜１７-ｍ−１の出力データｂ₁(ｋ)〜ｂ_m-1(ｋ)と、インパルスレスポンス行列Ｈ ₁(ｋ)〜Ｈ _m-1(ｋ)を用いて、受信ベクトルｒ（ｋ）中の第１〜第ｍ−１ストリームの信号による干渉成分を、干渉成分生成器１４で生成する（Ｓ３）。式（６）中の尤度値ｔａｎｈ（λ₁(ｋ)）の代りにｂ₁(ｋ)を用いてｂ ₁(ｋ)と対応する干渉シンボルベクトル
ｂ″₁(k)＝［ｂ₁(ｋ＋Ｑ−１)…ｂ₁(ｋ)…ｂ₁(ｋ−Ｑ＋１)］^T （36）
を作り、同様にｂ₂(ｋ)〜ｂ_m-1(ｋ)についても干渉シンボルベクトルｂ″₂(k)〜ｂ″_m-1(k)を作り、これらと対応するインパルスレスポンスＨ ₁(ｋ)〜Ｈ _m-1(ｋ)との積により干渉成分を生成する。
【００６６】
これら干渉成分を受信ベクトルｒ（ｋ）から差し引き、差ベクトルｒ′_m(ｋ）を求める（Ｓ４）。またＨ ₁(ｋ)〜Ｈ _m(ｋ)よりフィルタ係数ｗ _m(ｋ)を計算し（Ｓ５）、このフィルタ係数ｗ _m(ｋ)で差ベクトルｒ′_m(ｋ）を線形フィルタ１６でフィルタ処理し（Ｓ６）、そのフィルタ処理結果ｚ_m(ｋ)を判定器１７-ｍで判定し、第ｍストリームデータ出力を得る（Ｓ７）。この第ｍストリームデータ出力を図１６に示すように復号器９１-ｍで誤り訂正復号する（Ｓ８）。送信側ではデータを誤り訂正符号方法により符号化している。
次にｍがＭになったかを調べ（Ｓ９）、Ｍになっていなければｍを＋１してステップＳ３に戻り次のストリーム等化器の処理に移る（Ｓ１０）。ステップＳ９でｍ＝Ｍであれば、第１〜第Ｍストリーム等化器１２-１〜１２-Ｍにおける１回目の等化処理が終了する。要するにこの１回目の等化処理は実施例１において尤度値ｔａｎｈ（λ_m(ｋ)）の代りに既に処理された判定結果ｂ_m(ｋ)のみを用いたものであり、文献４に示すＶ−ＢＬＡＳＴ法と同一である。
【００６７】
２回目以後の等化処理においては、前回の等化処理、誤り訂正復号された出力（復号結果）ｂ′₁(ｋ)〜ｂ′_M(ｋ)を、端子９２-１〜９２-Ｍより全ストリーム等化器１２-１〜１２-Ｍに入力して、その干渉成分生成器１４においてインパルスレスポンス行列Ｈ ₁(ｋ)〜Ｈ _m(ｋ)を用い、各ストリームの干渉成分を生成する（Ｓ１１）。この場合第ｍストリーム等化器１２-ｍにおいては干渉ベクトルｂ″₁(ｋ)〜ｂ″_m-1(ｋ)，ｂ″_m+1(ｋ)〜ｂ″_M(ｋ)は式（３６）の添字「１」を対応した添字と置きかえたものであるが、ｂ″_m(ｋ)は式（３６）の添字「１」を「ｍ」とし、干渉ベクトルｂ″_m(ｋ)の中央に位置する要素をゼロとしたものである。
【００６８】
各第ｍストリーム等化器１２-ｍにおいて、同一受信ベクトルｒ（ｋ）から対応する干渉成分を除去してそれぞれ差ベクトルｒ′_m(ｋ)を求める（Ｓ１２）。各第ｍストリーム等化器１２-ｍの線形フィルタ１６でフィルタ係数ｗ _m(ｋ)を計算する（Ｓ１３）。このフィルタ係数ｗ _m(ｋ)は実施例１における式（３３）中のフィルタ係数の項のｔａｎｈ（λ_m(ｋ)／２）、（ｍ＝１，…，ｎ）の代りに前回の復号結果ｂ′_m(ｋ)を用いて計算すればよい。そのフィルタ係数ｗ _m(ｋ)により差ベクトルｒ′_m(ｋ)をフィルタ処理する（Ｓ１４）。その各フィルタ処理結果ｚ_m(ｋ)をそれぞれ判定器１７-ｍで判定し（Ｓ１５）、その判定結果を復号器９１-ｍでそれぞれ誤り訂正復号する（Ｓ１６）。
【００６９】
この前回の復号結果ｂ′₁(ｋ)〜ｂ′_M(ｋ)を用いた等化処理が所定回数になっていなければ（Ｓ１７）、ステップＳ１１に戻り、所定回数になっていれば、終了する。
以上述べたようにこの実施例８では２回目以降の等化処理においては前回の誤り訂正復号結果を用い、処理が終了したストリームの判定結果を次のストリームにおける処理において用いない。その結果、誤り伝播に起因する平均誤り率特性の劣化を抑えることができる。これにより、２回目以降の等化処理においては、各ストリーム用等化器１２-１〜１２-Ｍにおける処理を同時に行うことができるので、処理遅延を低減することができる。
【００７０】
この実施例８では等化器１２-１〜１２-Ｍでは判定結果及び復号結果を用いて処理を行い、また復号器９１-１〜９１-Ｍではフィルタ１６の出力を判定器１７で判定された結果を用いて処理を行うために、等化器１２-１〜１２-Ｍにおいて尤度値の計算を行う尤度値計算器２４は必要としない。また、復号器９１-１〜９１-Ｍにおいても尤度値を導出するための装置を必要としない。さらに、この実施例８はマルチキャリア型ＭＩＭＯチャネル信号伝送の受信機においても、同様に用いることができる。つまり図１５において全てのデインターリーバ６５-ｍ及び全てのインターリーバ６７を省略し、復号器６６-Ｍとして復号器９１-ｍを用い、各多入力多出力等化器７７−ｊを、図２０、図２１及び図１６に示した構成とし、図１７に示した手順で処理すればよい。
【００７１】
図１８にこの実施例８の受信機構成における平均誤り率（ＢＥＲ）特性の計算機シミュレーション結果を示す。シミュレーション諸元として、送信アンテナ数及び受信アンテナ数はそれぞれ２として、変調方式はＢＰＳＫとして、伝搬路は遅延時間差が０，１，２，３，４シンボルの５パスレイリーフェージング伝搬路を用いた。遅延波による特性劣化を抑えるためにサブキャリア数１６、ガードインターバル４シンボルのＯＦＤＭ（マルチキャリア方式）を用いた図中破線９３と破線９４はＶ−ＢＬＡＳＴ型等化器と復号器を組み合わせた従来の受信機構成における結果を示し、実線９５と９６は実施例８の受信機構成における結果を示す。また、破線９３と実線９５は１回目の等化、復号の結果を示し、破線９４及び実線９６は等化→復号→等化→復号を行った後、つまり２回処理した結果を示す。これらの結果より、等化、復号を１回行った後の特性は同等であるのに対して、等化、復号を２回行った後の特性は、実施例８の特性が従来のＶ−ＢＬＡＳＴ法と比較して改善されている。これは、誤り伝播の影響を軽減できているためである。
【００７２】
上述において、第１〜第Ｍストリーム用等化器における伝送路推定器１３は共通に用いることができる。また上述において同一マルチキャリア内の各サブキャリアは同一伝搬路を通るから、同様のことが云えるから１つの多入力多出力等化器７７-ｊにおける１つのストリーム用等化器７９の伝送路推定器を全ての多入力多出力等化器７７-ｊに共通に利用することもできる。図１５及び図２２に示した実施例において、送信側でインターリーバを省略している場合は、受信側でデインターリーバ及びインターリーバを省略する。しかし、これらインターリーバ、デインターリーバを用いた方が、特性がよいものとなる。
【００７３】
【発明の効果】
この発明の実施例１により、文献４の受信機構成と比較して、誤り率特性を改善することができる。
この発明の実施例２により、文献４の受信機構成と比較して、処理に要する時間を削減し、誤り率特性を改善することができる。
この発明の実施例３により、文献５の方法と比較して、誤り率特性を改善することができる。
この発明の実施例４により、実施例１と比較して演算量を削減することができる。
【００７４】
この発明の実施例５により、実施例４と比較して、処理に要する時間を削減することができる。
この発明の実施例６により、実施例３と比較して、演算量を削減することができる。
この発明の実施例７により、実施例６と比較して、処理に要する時間を削減することができる。
この発明の実施例８により、文献４記載の受信機構成と復号器を用いた場合と比較して、誤り率特性を改善することができる。
【図面の簡単な説明】
【図１】実施例１の機能構成を示す図。
【図２】図１中の各ストリーム用等化器２１-ｍの機能構成例を示す図。
【図３】Ａは図２中の干渉成分生成器２３の機能構成例を示す図、Ｂは図２中のフィルタ１６及び尤度値計算器２４の機能構成例を示す図である。
【図４】図１に示す構成の処理手順の例を示す流れ図。
【図５】実施例１と文献４における受信機構成との１波レイリーフェージング伝搬路における平均ＢＥＲ特性を示す図。
【図６】実施例１と、文献４における受信機構成の３波レイリーフェージング伝搬路における平均ＢＥＲ特性を示す図。
【図７】実施例２の受信機構成を示す図。
【図８】図７に示す受信機の処理手順の例を示す流れ図。
【図９】実施例２の平均ＢＥＲ特性と、実施例１の平均ＢＥＲ特性を示す図。
【図１０】実施例３におけるＭＩＭＯ等化器６４の構成を示す図。
【図１１】マルチキャリア型ＭＩＭＯチャネル信号伝送の送受信構成を示す図。
【図１２】実施例４の各ＭＩＭＯ等化器７７-ｊの構成を示す図。
【図１３】実施例５の各ＭＩＭＯ等化器７７-ｊの受信機構成を示す図。
【図１４】実施例６における受信機構成と対応する送信機構成を示す図。
【図１５】実施例６の受信機構成を示す図。
【図１６】実施例８における２回目以後の処理での受信機構成を示す図。
【図１７】実施例８の処理手順の例を示す流れ図。
【図１８】実施例８における受信機構成と文献４における受信機構成の平均ＢＥＲ特性を示す図。
【図１９】ＭＩＭＯチャネル信号伝送における送受信構成を示す図。
【図２０】文献４に示す受信機構成を示す図。
【図２１】図２０中のストリーム用等化器１２-ｍの構成を示す図。
【図２２】文献５における送受信機構成を示す図。
【図２３】図２２中のＭＩＭＯ等化器６４の構成を示す図。
［参考文献］
１．冨里繁，浅井孝浩，松本正，“移動通信用ＭＩＭＯチャネル信号伝送における無線信号処理，”信学技報，RCS2001-136, pp.43-48, Oct,2001
２．G.J. Foschini and M.J. Gans,“On limits of wireless communications in a fading environment when using multiple antennas,”Wireless Personal Communication,vol.6,no.3,pp.315-335,March,1998.
３．H.Yoshino, K.Fukawa, H.Suzuki,“Interference canceling equalizer(ICE) for mobile radio communication,”IEEE Trans, on VT,vol.46,no.4,pp.849-861,Nov.1997.
４．P.W.Wolniansky, G.J.Foschini, G.D.Golden, and R.A.Valenzuela,“V-blast:An architecture for realizing very high data rates over the rich-scattering wireless channel.”URSI International Symposium on Signals, Systems, and Electronics Conference Proceedings, pages 295-300,1998.
５．阿部哲士，松本正，“周波数選択性ＭＩＭＯチャネルにおける時空ターボ等化器，”信学技法，RCS2000-256, pp.75-80, March,2001[0001]
BACKGROUND OF THE INVENTION
The present invention relates to a method of receiving signals transmitted from a plurality of transmitting antennas by a plurality of antennas, for example, a so-called MIMO (Multiple Input Multiple Output) channel signal receiving method and an apparatus thereof.
[0002]
[Prior art]
A MIMO (Multiple Input Multiple Output) channel signal transmission scheme aimed at increasing the transmission speed and increasing the system capacity in mobile communication has been studied [Reference 1]. FIG. 19 shows an example of a transmission / reception system for MIMO channel signal transmission. Here, the number of transmission antennas ANS is M, and the number of reception antennas ANR is N. In MIMO channel signal transmission, different information is transmitted as a signal using the same frequency using a plurality of transmitters TX-1 to TX-M and antennas ANS-1 to ANS-M on the transmission side, and on the reception side. This is a method of performing reception using a plurality of antennas ANR-1 to ANR-N and one receiver RX, and it has been shown that the communication capacity can be increased [Reference 2]. For this MIMO channel signal transmission, a single user (user) performs signal transmission using a plurality of transmission / reception antennas, or a plurality of users use the same frequency by using one or more transmission antennas from the mobile device. A method of use in which transmission is performed and reception is performed using a plurality of antennas on the base station side is conceivable.
[0003]
In this MIMO channel signal transmission, since different information is transmitted using the same frequency by a plurality of antennas on the transmission side, these transmission signals are added and received. Therefore, it is necessary to perform signal processing for separating the signals transmitted from the respective transmission antennas ANS-1 to ANS-M from the received signals. With respect to this signal processing, optimum characteristics can be obtained by using the method of Document 3. However, this method increases the amount of computation in the receiver as the number of transmitting antennas increases. In view of this, a method called V-BLAST (Vertical Bell Laboratories Layered Space Time Architecture) that can perform processing with a small amount of calculation although optimum characteristics cannot be obtained has been studied [Reference 4].
[0004]
A receiver configuration of the V-BLAST method is shown in FIG. In the figure, received signals from the antennas ANR-1 to ANR-N are used as baseband signals, and further, received signal sample value sequences sampled at a frequency equal to or higher than the symbol frequency of transmission data and converted into digital values are input. The signals are respectively input to the first to Mth stream equalizers 12-1 to 12-M from terminals 11-1 to 11-N. The first stream equalizer 12-1 represents an equalizer for processing the data sequence transmitted from the first transmission antenna ANS-1, and similarly, the M-th stream equalizer 12-M , Represents an equalizer for processing a data sequence transmitted from the Mth transmitting antenna ANS-M.
[0005]
First, the first stream equalizer 12-1 performs processing using signals received by the plurality of receiving antennas ANR-1 to ANR-N. Next, the second stream equalizer 12-2 performs processing using the first stream data output from the first stream equalizer 12-1 and all the received signals. Thus, the stream data output of each equalizer obtained so far and processing using all received signals are performed up to the M-th stream equalizer 12-M. That is, the M-th stream equalizer 12-M equalizes the received signal and each data output of the M-1st stream equalizers 12-1 to 12-M-1 from the first stream. Process.
[0006]
FIG. 21 shows the configuration of each stream equalizer 12-m (m = 1, 2,..., M). The transmission path estimator 13 estimates the impulse response between the transmitting and receiving antennas using all received signals. The interference component generator 14 generates the received signal component by using the estimated impulse response and the data output of the stream that has already been processed, and uses these as the interference component for the m-th stream. The components corresponding to the transmission path from the transmission antenna ANS-m to the reception antennas ANR-1 to ANR-N are subtracted from the reception signal by the subtraction units 15-1 to 15-N, and the result is input to the linear filter 16. . Document 4 introduces a linear filter based on ZF (Zero Forcing) and MMSF (Minimum Mean Squared Error) as the linear filter 16. Finally, the output of the linear filter 16 is determined by the determiner 17, whereby the data output of the m-th stream can be obtained at the output terminal 18-m.
[0007]
[Problems to be solved by the invention]
In the V-BLAST method, when the signal determination result obtained by equalization processing of a certain stream, for example, the m-th stream is incorrect, the interference component is correctly detected from the received signal in the subsequent m + 1-th stream equalization processing. It is possible to further subtract the error because it cannot be deducted. An object of the present invention is to provide a multi-input multi-output turbo receiver capable of suppressing such a phenomenon that an error propagates, improving an error rate characteristic, and reducing a calculation amount.
[0008]
[Means for Solving the Problems]
In order to solve the above problems, the invention uses the following means. In the first embodiment of the present invention, a MIMO equalizer including a plurality of stream equalizers and a determiner are used, and the likelihood value output of each stream equalizer is used for a stream to be processed thereafter. Is input to the generator. Each stream equalizer consists of a transmission path estimator, an interference component signal generator, an MMSE filter, and a likelihood value calculator. Received signal sample values received by multiple antennas are input to the transmission path estimator. The The impulse response estimated by the transmission path estimator and the likelihood value output of each stream are input to an interference component signal generator. The signal obtained by subtracting the interference component signal from the received signal, the likelihood value output of each stream, and the estimated impulse response are input to the MMSE filter. The likelihood value calculator receives the output of the MMSE filter and the estimated impulse response of each stream and the estimated impulse response.
[0009]
In the second embodiment of the present invention, the configuration is the same as the receiver according to the first embodiment, but the likelihood value output used in the signal generator of the interference component of each stream equalizer is different. In the a-th process (a is an arbitrary integer equal to or greater than 1), each stream equalizer uses the likelihood value output derived in each a-1th stream equalizer.
In the third embodiment of the present invention, the configuration of each stream equalizer including a combination of a MIMO equalizer including a plurality of stream equalizers and a determinator, and a deinterleaver and a decoder is the same as that of the first embodiment. The likelihood value output of each stream equalizer is input to a stream equalizer to be processed thereafter. After the processing in the a-th equalizer for all streams is completed, the obtained likelihood value output is input to the deinterleaver and input to the decoder. The likelihood value output output from the decoder is input to each stream equalizer via an interleaver, and the (a + 1) th equalization process is performed.
[0010]
The fourth embodiment of the present invention comprises a combination of a fast Fourier transformer, a parallel-serial converter, and a MIMO equalizer according to the first embodiment for returning a signal multiplexed and multiplexed. Here, instead of using the fast Fourier transformer, a general Fourier transformer or a plurality of frequency oscillators can be used.
The fifth embodiment of the present invention comprises a combination of a fast Fourier transformer, a parallel-serial converter, and a MIMO equalizer according to the second embodiment for returning a signal multiplexed and multiplexed. Here, instead of using the fast Fourier transformer, a general Fourier transformer or a plurality of frequency oscillators can be used.
[0011]
The sixth embodiment of the present invention comprises a combination of a fast Fourier transformer, a parallel-serial converter, and a MIMO equalizer according to the third embodiment for returning the multiplexed and multiplexed signal. Here, instead of using the fast Fourier transformer, a general Fourier transformer or a plurality of frequency oscillators can be used.
The seventh embodiment of the present invention is the same as the configuration in the sixth embodiment, but in the MIMO equalizer, the likelihood value output in each stream equalizer of the a-th (a is an arbitrary integer of 1 or more) is: It is not input to the other stream equalizers, but is input to the decoder via parallel-serial conversion and deinterleaver. The likelihood value output from the decoder is input to each stream equalizer of each MIMO equalizer via an interleaver, and the (a + 1) th processing is performed.
[0012]
According to the eighth aspect of the present invention, the multi-input multi-output equalizer including a plurality of stream equalizers including a determiner and a decoder are combined. Each stream equalizer includes an interference component generator, an MMSE filter, and a determiner. In the first equalization process, the determination output of each stream equalizer is a stream that is processed thereafter. Input to the equalizer. Then, after the first equalization process for all streams is completed, decoding is performed in the decoder, and the equalization process is performed again using the encoded sequence obtained as a result. In the second and subsequent equalization processes, the encoding sequence obtained from the decoder is used in the interference component generator in each stream equalizer. That is, unlike the equalization process in the first time, the determination result in a certain stream equalizer is not input to the stream equalizer to be processed next. The previous decoding result is used for equalization processing, and all the stream equalizers are processed simultaneously.
[0013]
DETAILED DESCRIPTION OF THE INVENTION
Embodiments of the present invention will be described with reference to examples.
Example 1
FIG. 1 shows the configuration of Embodiment 1 of the present invention. The difference between this configuration and the V-BLAST method shown in FIG. 20 is that a likelihood value is output from each of the stream equalizers 21-1 to 21-M, and the likelihood value has not yet been equalized. Is input to the equalizer of each stream that has not been used, and is used for equalization processing in that stream, and each process of the first stream equalizer 21-1 to the Mth stream equalizer 21-M is completed. After that, the process from the first stream equalizer 21-1 to the M-th stream equalizer 21-M is repeated using the obtained likelihood value output. By doing so, the characteristics can be improved. Here, the likelihood value is a log likelihood ratio of the probability that each bit (symbol) is 0 and the probability that it is 1, and is defined by the following equation (1).
[0014]
log [P [b (i) = 0] / P [b (i) = 1]] (1)
The likelihood value may be defined as log [P [b (i) = 1] / P [b (i) = 0]] by reversing the denominator and numerator of equation (1). In this equation, P [b (i) = 0] represents the probability that the bit b (i) at the time point i is 0, and P [b (i) = 1] is the bit b (i) at the time point i being 1. Represents the probability of The likelihood value defined by the equation (1) takes a value from −∞ to + ∞, and becomes a positive value as the possibility that the focused bit is 0 is high, and the focused bit is 1 The higher the probability of, the larger the negative value. When the likelihood value is 0, the probability that the focused bit is 0 is equal to the probability that it is 1.
[0015]
A processing flow of the first embodiment will be described. First, processing is performed in the first stream equalizer 21-1 using the received signal, and the likelihood value output λ of the first stream is obtained.₁Is derived. Next, the derived likelihood value output λ of the first stream₁The second stream equalizer 21-2 is processed using the received signal. Similarly, the equalizer for the M-th stream 21-M outputs the first stream likelihood value output λ derived earlier.₁To (M-1) th stream likelihood value output λ_M-1And processing is performed using the received signal. The above processing is the first equalization processing from the first stream equalizer 21-1 to the Mth stream equalizer 21-M, and the likelihood value λ of each stream₁~ Λ_MAre determined by the determiners 17-1 to 17-M, respectively, thereby obtaining a data output in each stream.
[0016]
In the first embodiment, equalization processing in each stream is further repeated for the same received signal. That is, after the first equalization process from the first stream to the Mth stream is finished, the likelihood value output λ from each stream₁~ Λ_MAnd the first stream equalizer 21-1 again using the received signal and the likelihood value λ₁Is output. As a result, the more likely likelihood value λ₁Can be obtained. In the first equalization process in the first stream, only the received signal can be used, whereas in the second equalization process in the first stream, the likelihood of the received signal and each stream is increased. Degree value λ₁~ Λ_MThis is because the output of can be used.
[0017]
Similarly, the second stream equalizer 21-2 for the second time outputs the likelihood value output λ obtained by the second process of the equalizer 21-1 for the first stream.₁And each likelihood value λ of the second to Mth streams obtained by the first processing₂~ Λ_MThe equalization process is performed using the output and the received signal. It is a feature of the first embodiment that the above processing is repeated, and by performing the processing repeatedly, a more reliable likelihood value can be obtained, so that characteristics can be improved. In the V-BLAST method shown in FIG. 20, each stream equalizer can obtain only a discrete value data output instead of a continuous value likelihood value output. Even if the treatment is repeated, the characteristics are hardly improved.
[0018]
Next, the configuration of each stream equalizer 21-m in FIG. 1 will be described with reference to FIG. The transmission path estimator 13 estimates an impulse response between the transmission antenna ANS-m and each of the reception antennas ANR-1 to ANR-N using the reception signal in the same manner as in FIG. In the first equalization process, the estimated impulse response and the likelihood value λ of the first to (m-1) th streams that have already been processed from the terminals 22-1 to 22-m-1.₁~ Λ_m-1The interference component generator 23 generates signal components of the first to (m-1) th streams that have already been processed, and subtracts them from the received signal as interference components for the mth stream. The results are subtracted by the units 15-1 to 15-N, and the result is input to the linear filter 16. The linear filter 16 performs filter processing based on MMSE. Finally, the output of the linear filter 16, the impulse response, and the likelihood value λ of each stream that has already been processed.₁~ Λ_m-1The likelihood value output λ_mIs calculated by the likelihood value calculator 24. A specific method for calculating the likelihood value will be described below. First received signal vector at time kr(K) is defined in the following form.
[0019]
r(k) = [r₁(k + Q-1)] ... r_N(k + Q-1) r_l(k + Q-2) ... r_N(k + Q-2) ... r_l(k) ... r_N(k)]^T          (2)
Here, N is the number of receiving antennas, Q is the maximum number of delay symbols considered by the equalizer 21-m + 1, and the superscript^TRepresents transposition. R₁(K) represents a received signal sample value at a time point k from the first receiving antenna ANR-1, r_N(K) represents a received signal sample value at a time point k from the Nth receiving antenna ANR-N. Next, the following matrix is defined using the estimated impulse response.
[0020]
[Expression 1]

[0021]
Where h_1,1(K, 0) represents the value of the 0 symbol delay of the impulse response at time k in the transmission path from the first transmitting antenna ANS-1 to the first receiving antenna ANR-1. Similarly, h_1,1(K, Q-1) represents the value of the (Q-1) symbol delay of the impulse response at time k in the transmission path from the first transmitting antenna ANS-1 to the first receiving antenna ANR-1. h_{1, N}(K, 0) represents the value of the 0 symbol delay of the impulse response at time k in the transmission path from the first transmitting antenna ANS-1 to the Nth receiving antenna ANR-N, h_{1, N}(K, Q-1) represents the value of the (Q-1) symbol delay of the impulse response at time k in the transmission path from the first transmitting antenna ANS-1 to the Nth receiving antenna ANR-N.
[0022]
Similarly, using the impulse responses from the 2nd and Mth transmitting antennas ANS-2 to ANS-M to the received signal antennas ANR-1 to ANR-N, the matrixH ₂(K), ...,H _MDefine (k). In this case, a matrix representing an impulse response between the Mth transmission antenna ANS-M and each of the reception signal antennas ANR-1 to ANR-N.H _M(K) is represented by the following equation.
[0023]
[Expression 2]

[0024]
Next, the bit likelihood value λ at time k of the signal of the first stream transmitted from the first transmitting antenna ANS-1₁(K) is defined by the following equation.
λ₁(K) = log [P [b (k) = 0] / P [b (k) = 1]] (5)
Using this likelihood value, the likelihood value vector of the first streamb ₁(K) is defined by the following equation.

Here, tanh represents a hiberpolik tangent. When the likelihood value λ taking a value from −∞ to + ∞ is tanh (λ / 2), a value from −1 to 1 is taken. Similarly, the likelihood value of the bit at the time point k of the signal transmitted from the 2,..., M-th transmitting antenna ANS-2,.₂(K), ..., λ_MAs (k), the likelihood value vector of the second,.b ₂(K), ...,b _M(K). At this time,b _M(K) is represented by the following equation.

2 will be described using the defined variables. First, the first process in the first stream equalizer 21-1 will be described. A matrix defined by the equation (3) by estimating the impulse response by the transmission path estimator 13H ₁(K) is generated. Here, k represents a time point. At this stage, since the processing is not completed in the other stream equalizers 21-2 to 21-M, the interference component generator 23 has the likelihood of each stream from the terminals 22-1 to 22-M. Since no value output is input, all likelihood values λ of m = 1,._m(K) is zero, likelihood value vectorb _m(K) is also a zero vector. Therefore, the output of the interference component generator 23 is also zero, and the received signal vectorr(K) is input to the MMSE filter 16. The filter coefficient vector of the MMSE filter 16 isW ₁(K)W ₁(K) is a matrixH ₁It is expressed by the following formula using (k).
W ₁(K) = {H ₁(K)H ₁ ^H(K) + σ² I}^-1 H ₁(K)e _Q(8)
Where σ²Represents the average noise power,IRepresents an identity matrix. Also, subscript^HRepresents a complex conjugate transpose.e _QIs a column vector in which the number of all elements in which the element located at the center is 1 and the other elements are 0 is 2Q-1, and is defined by the following equation.
[0025]
e _Q= [0 ... 010 ... 0]^T          (9)
At this time, the output z of the MMSE filter 16₁(K) is represented by the following equation.

The output z of this MMSE filter 16₁Matrix generated using (k) and impulse responseH ₁(K), the likelihood value calculator 24 uses the likelihood value λ.₁(K) is calculated [5].
λ₁(K) = 4 × Re {z₁(K)} / (1-μ₁(K)) (11)
Here, Re {z (k)} represents taking the value of the real part of z (k), and μ₁(K) is defined by the following equation.
[0026]

The above is the first processing of the first stream equalizer 21-1, and the likelihood value λ₁(K) is output.
Next, the first processing of the second stream equalizer 21-2 will be described. In signal detection of the second stream, since the signal component of the first stream becomes an interference component, a signal vector obtained by subtracting the interference component from the received signal vector is processed by the MMSE filter 16. Hereinafter, the processing in the second stream equalizer 21-2 will be described in order. First, using the impulse response estimated by the transmission path estimator 13, the matrixH ₂(K) is generated. In the interference component generator 23, the likelihood value λ of the first stream equalizer 21-1 is used.₁Likelihood value vector defined by equation (6) using the output of (k)b ₁(K) is generated and this is a matrixH ₁The first stream signal component from (k)H ₁(K)b ₁(K) is generated as an interference component. As shown in Expression (13), the interference component due to the first stream is subtracted from the received signal vector and input to the MMSE filter 16.
[0027]
r ₁′ (K) =r(K)-H ₁(K)b ₁(K) (13)
The received signal r₁(K), ..., r_NFrom (k)H ₁(K)b ₁The corresponding components of (k) are subtracted by the subtraction units 15-1 to 15-N, respectively.
In MMSE to filter 16, the signal vectorr'(K) and matrix obtained from imperial responseH ₁(K),H ₂(K) and likelihood value vectorb ₁Filter processing is performed using (k). At this time, the filter output z₂(K) is represented by the following equation.

Where Λ₁(K) isb ₁It is a diagonal matrix defined in the following form using the element of (k).
[0028]
[Equation 3]

[0029]
Finally, the likelihood value calculator 24 outputs the output of the filter 16 and the likelihood value vector.b ₁
(K) and the matrix obtained from the imperial responseH ₁(K),H ₂Using (k), the likelihood value output λ expressed by the following equation:₂(K) is calculated.
λ₂(K) = 4 × Re {z₂(K)} / (1-μ₂(K)) (16)
Where μ₂(K) is defined by the following equation.

Similarly, in the first processing in the equalizer for the Mth stream 21-M, the likelihood value λ that has already been processed.₁(K), ..., λ_M-1Likelihood value vector obtained from (k) (b ₁(K), ...,b _M-1Using (k), the received signal vectorrThe signal component of the other stream is subtracted from (k) as an interference component as shown in Equation (18),

This signal vectorr′ (K) is processed by the MMSE filter 16 and the filter output z_M(K) is obtained as shown in equation (19).
[0030]

In the likelihood value calculator 24, the filter output z_M(K) Likelihood value λ that has already been processed₁(K), ..., λ_M-1(K), a matrix obtained from the estimated impulse responseH ₁(K), ...,H _MUsing (k), the likelihood value output λ_M(K) is calculated.
λ_M(K) = 4 × Re {z_M(K)} / (1-μ_M(K)) (20)
Where μ_M(K) is defined by the following equation.
[0031]

The above is the first process in each stream equalizer. Next, the processing in each stream equalizer in the second time will be described. First, the second processing in the first stream equalizer 21-1 will be described. Unlike the first process, the likelihood value λ₁(K), ..., λ_MProcessing is performed using (k). The signal components of the second stream to the Mth stream are subtracted as interference components from the received signal vector.
[0032]

here,b ₁′ (K) in equation (6)b ₁This is a vector in which the element located at the center of (k) is zero, and is defined in the following form.
b ₁′ (K) = [tanh (λ₁(k + (Q-1)) / 2 ... tanh (λ₁(k + 1) / 2 0 tanh (λ₁(k-1)) / 2 ... tanh (λ₁(k- (Q-1)) / 2]^T          (23)
This leaves the signal component at the time point k of the first stream, and acts to remove the intersymbol interference components before and after that.
[0033]
Signal vector given by equation (22) from which the interference component is subtractedr ₁′ (K) is processed by the MMSE filter 16 and the second filter output z₁'(K) is calculated according to equation (24).

Where Λ₁′ (K) is defined by equation (15).₁It is a diagonal matrix with the element located at the center of (k) as 1, and is expressed in the following form.
[0034]
[Expression 4]

[0035]
Setting only the central element to 1 meansb ₁As in the case of ′ (k),
The signal component is left, and the intersymbol interference component before and after the signal component is removed. Finally, the likelihood value calculator 24 outputs the likelihood value output λ.₁'(K) is calculated.
λ₁′ (K) = 4 × Re {z₁′ (K)} / (1-μ₁'(K)) (26)
Where μ₁′ (K) is defined by the following equation.
μ₁′ (K) ＝e _Q ^T H ₁ ^H(k) {H ₁(k) Λ₁′ (K)H ₁ ^H(k) +H ₂(k) Λ₂(k)H ₂ ^H(k) + ... +H _M(k) Λ_M H _M ^H(k) + σ² I}^-1 H ₁(k)e _Q  (27)
Similarly, in the second processing of the second stream equalizer 21-2, first, a signal vector obtained by subtracting the interference component from the received signal vector is obtained by the following equation.
[0036]

here,b ₁(k) is a likelihood value output λ derived by the second equalization process in the first stream equalizer 21-1.₁A likelihood value vector generated using ′ (k),b ₂As in the case of Expression (23), ′ (k) sets the central element to 0, and the remaining likelihood value λ output from the first-time second stream equalizer 21-2.₂A likelihood value vector generated using (k). This difference vectorr ₂′ (K) is input to the MMSE filter 16 and filtered. At this time, the filter output z₂′ (K) is represented by the following equation.
[0037]
z₂′ (K) ＝e _Q ^T H ₂ ^H(k) {H ₁(k) Λ₁(k)H ₁ ^H(k) +H ₂(k) Λ₂′ (K)H ₂ ^H(k) + ... +H _M(k) Λ_M(k)H _M ^H(k) + σ² I}^-1 r ₂′ (K) (29)
Where Λ₂′ (K) is Λ₂It is a diagonal matrix with the element located at the center of (k) as 1. Using this filter output, the likelihood value calculator 24 uses the likelihood value output λ.₂′ (K) is calculated.
λ₂′ (K) = 4 × Re {z₂′ (K)} / (1-μ₂'(K)) (30)
Where μ₂′ (K) is defined by the following equation.
[0038]
μ₂′ (K) ＝e _Q ^T H ₂ ^H(k) {H ₁(k) Λ₁(k)H ₁ ^H(k) +H ₂(k) Λ₂′ (K)H ₂ ^H(k) +H _Three(k) Λ_Three(k) H _Three ^H(k) + ... +H _M(k) Λ_M(k) H _M ^H(k) + σ² I}^-1 H ₂(k)e _Q      (31)
Similarly, in the a-th process (a is an arbitrary integer equal to or greater than 1) of the m-th stream equalizer 21-m, the m-th stream equalizer 21- in the a-1th equalization process. Likelihood value output of the Mth stream equalizer 21-M from m and the first stream equalizer 21-1 to m-1th stream equalizer 21-m-1 in the a-th equalization process First, the interference component is subtracted from the received signal by the following equation using the likelihood value output of
[0039]
r _m′ (K) =r(K)-H ₁(k)b ₁(k) -...-H _m-1(k)b _m-1(k) −H _m(k)b _m′ (K) −H _{m + 1}(k)b _{m + 1}(k) -...-H _M(k)b _M(k) (32)
here,b ₁(k), ...,b _m-1(k) is a likelihood generated from the likelihood value output obtained by the a-th equalization process from the first stream equalizer 21-1 to the (m-1) th stream equalizer 21-m-1. Represents a vector of degree values,b _m′ (K),b _{m + 1}(k), ...,b _M(k) is a likelihood value vector generated from the likelihood value output obtained by the a-1th equalization process from the m-th stream equalizer 21-m to the M-th stream equalizer 21-M. (In the first equalization process where a = 1,b ₁(k), ...,b _M(k) is a vector composed of all zeros). Likelihood value vectorb _kThe element located at the center of ′ (k) is zero. This signal vectorr′ (K) is processed by the MMSE filter 16 and the filter output z shown by the following equation:_m(k) is obtained.
[0040]
z_m(k) =e _Q ^T H _m ^H(k) {H ₁(k) Λ₁(k)H ₁ ^H(k) + ... +H _m-1(k) Λ_m-1(k)H _m-1 ^H(k) +H _m(k) Λ_m′ (K)H _m ^H(k) +H _{m + 1}(k) Λ_{m + 1}(k)H _{m + 1} ^H(k) + ... +H _M(k) Λ_M(k)H _M ^H(k) + σ² I}^-1r_m′ (K)  (33)
Where Λ₁(k), ..., Λ_m-1(k) is generated by the likelihood value output obtained by the a-th equalization process from the first stream equalizer 21-1 to the (m-1) th stream equalizer 21-m-1. Diagonal matrix, Λ_m′ (K), Λ_{m + 1}(k), ..., Λ_M(k) is a diagonal generated from the likelihood value output obtained by the a-1th equalization process from the m-th stream equalizer 21-m to the M-th stream equalizer 21-M. It is a matrix. Also, Λ_mThe element located at the center of ′ (k) is 1. Finally, using the filter output, the likelihood value calculator 24 uses the likelihood value output λ._m(k) is obtained.
[0041]
λ_m(k) = 4 × Re {z_m(k)} / (1-μ_m(k)) (34)
Where μ_m(k) is defined by the following equation.
μ_m(k) =e _Q ^T H _m ^H(k) {H ₁(k) Λ₁(k)H ₁ ^H(k) + ... +H _m-1(k) Λ_m-1(k)H _m-1 ^H(k) +H _m(k) Λ_m′ (K) H _m ^H(k) +H _{m + 1}(k) Λ_{m + 1}(k) H _{m + 1} ^H(k) + ... +H _M(k) Λ_M(k) H _M ^H(k) + σ² I}^-1 H _m(k)e _Q                                                          (35)
By repeatedly performing the above processing, each stream equalizer can obtain a more likely likelihood value output, and thus the finally obtained likelihood value λ₁(k) to λ_MData with few errors can be obtained as a data output obtained by passing the output of (k) through the determiners 17-1 to 17-M. The functional configuration of the interference component generator 23 in the m-th stream equalizer 21-m is as shown in FIG. 3A. Likelihood value λ obtained by the present (a-th) equalization processing from terminals 22-1 to 22-m-1₁(K) to λ_m-1(K) is input to the likelihood vector generators 26-1 to 26 -m−1, and the likelihood “1” in equation (6) is set to “1”, “−”, “m−1”, respectively. Degree vectorb ₁(K) ~b _m-1(K) is generated, and the likelihood value λ obtained by the previous (a-1) equalization processing from the terminals 22-m to 22-M._m(K) to λ_M(K) is input to the likelihood vector generators 26-m to 26-M, and in the likelihood vector generators 26-m + 1 to 26-M, the subscript “1” in the equation (6) is changed to “m + 1”,. , “M” likelihood vectorb _{m + 1}(K) ~b _M(K) is generated, and the likelihood vector generation unit 26-m uses the likelihood vector in which the subscript “1” in Expression (23) is “m”.b _m'(K) is generated.
[0042]
These likelihood vectorsb ₁(K),b _m-1(K),b _m′ (K),b _{m + 1}(K), ...b _M(K) and impulse response matrix from the transmission path estimator 13H ₁(K), ...,H _m-1(K),H _m(K),H _{m + 1}(K), ...,H _M(K) is multiplied by multipliers 27-1 to 27 -M, respectively, and interference components for the m-th stream of the first to M-th streams are generated.
Further, as shown in FIG. 3B, the functional configuration of the filter 16 is the likelihood value λ obtained by the diagonal matrix generation unit 28 in the current (a-th) equalization process.₁(K) to λ_m-1Likelihood value λ obtained by equalization processing of (k) and the previous time (a-1)_{m + 1}(K) to λ_M(K) is input, and the diagonal matrix Λ in which the subscript “1” in equation (15) is “1” to “m−1” and “m + 1” to “M”, respectively.₁(K) to Λ_m-1(K), Λ_{m + 1}(K) to Λ_M(K) is generated, and the likelihood value λ obtained by the previous (a-1th) equalization process_m(K) is input to the diagonal matrix generation unit 28 ′, and the diagonal matrix Λ with the subscript “1” in Expression (25) as “m”._m'(K) is generated. These diagonal matrices Λ₁(K) to Λ_m-1(K), Λ_{m + 1}(K) to Λ_M(K) and Λ_m'(K) and impulse responseH ₁(K) ~H _M(K) and average noise power σ²Are input to the matrix calculation unit 29, and the difference vector in the right side of Expression (33)r _mThe filter coefficient is calculated by calculating the expression excluding ′ (k)W _m(K) is calculated. The diagonal matrix generation units 28 and 28 ′ and the matrix calculation unit 29 constitute a filter coefficient calculation unit 31. Receive vectorrDifference vector with interference component subtracted from (k)r _m′ (K) is the filter coefficientW _m(K) is subjected to filter calculation processing by the filter processing unit 32, that is,W _m(K)r _m'(K) is calculated and the filter output Z_m(K) is obtained.
[0043]
In the likelihood value calculation unit 24, the filter output z_m(K) and filter coefficientsW _m(K) and impulse responseH _m(K) is input and μ_m(K) =W _m(K)H _m(K)e _QThat is, the equation (35) is calculated by the calculation unit 24a, and the calculation unit 24b further calculates μ._m(K) and z_mEquation (34) is calculated from (k) and the likelihood value λ_m(K) is obtained. This likelihood value λ_m(K) is input to the determiner 17-m, and binary determination is performed. Likelihood value λ as described above_m(K) is tanh (λ_mSince (k) / 2) is normalized and used for various operations, each stream equalizer 21-m (m = 1,._mIt is preferable to perform the calculation of (k) / 2).
[0044]
An example of the above equalization processing procedure is shown in FIG. First impulse responseH ₁(K) ~H _M(K) is obtained (S0), parameter m is set to 1 (S1), and the likelihood value λ obtained this time₁(K) to λ_m-1(K) (or tanh (λ₁(K) / 2) to tanh (λ_m-1(K) / 2) and so on) and likelihood value λ obtained last time_m(K) to λ_M(K) (or tanh (λ_m(K) / 2) to tanh (λ_M(K) / 2) The same applies to the likelihood vector generators 26-1 to 26-M.₁(K), ..., Λ_m-1(K), Λ_m′ (K), Λ_{m + 1}(K), ..., Λ_M(K) is generated (S2). These likelihood vectors and impulse responsesH ₁(K) ~H _MThe interference components are calculated by the multipliers 27-1 to 27 -M using (k) (S 3). A corresponding interference generation may be calculated for each likelihood vector generated.
[0045]
Receive vectorrThe difference vector is obtained by removing these interference components from (k).r _m'(K) is obtained (S4). The likelihood value λ obtained this time₁(K) to λ_m-1(K) and the previously obtained likelihood value λ_m(K) to λ_M(K) and impulse responseH ₁(K) ~H _M(K) and average noise power σ²And the filter coefficient calculation unit 31 uses the filter coefficientW _m(K) is calculated (S5). Filter coefficientW _m(K) the difference vectorr _mFilter output z by filtering ′ (k)_m(K) is obtained (S6). Filter output z_m(K), filter coefficientW _m(K) Impulse responseH _mThe likelihood value λ is calculated by the likelihood value calculator 24 according to (k)._m(K) (or tanh (λ_m(K) / 2)) is calculated (S7).
[0046]
It is checked whether m is M (S8). If M is not M, m is incremented by 1 and the process returns to step S2 (S9). When m reaches M, it is checked whether equalization processing has been performed a predetermined number of times (S10). If not, the process returns to step S1, and if it has been performed a predetermined number of times, the likelihood value λ at this time_m(K) (or tanh (λ_m(K) / 2)) is determined by the determiner 17 and output (S11). If m = M in step S8, the likelihood value λ as shown by the broken line frame S12._mThe determination by the determiner 17 in (k) may be performed and output as an intermediate result. The filter coefficient calculation in step S5 may be performed before steps S2 to S4.
[0047]
FIG. 5 shows the result of comparison of the characteristics of the receiver according to the first embodiment and the receiver according to Reference 4 by computer simulation. As specifications of the computer simulation, the number M of transmitting antennas and the number N of receiving antennas are 5 and the modulation method is BPSK. The propagation path is a one-wave Rayleigh fading propagation path, and there is no time variation within one frame of the transmission signal. The transmission path (impulse response) is estimated ideally, and each frame is based on 256 symbols of information. It was decided to be composed. In the receiver, the received signal power of each stream is compared by an impulse response, the stream having the strongest received power is set as the first stream, and the processing in each stream equalizer is performed in descending order of the received power.
[0048]
In FIG. 5, the broken line 35 indicates the average of the data output of each stream after the processing from the first stream equalizer 12-1 to the M-th stream equalizer 12 -M is performed in the receiver configuration of Document 4. Indicates the bit error rate. A broken line 36 indicates that the processing from the first stream equalizer 12-1 to the Mth stream equalizer 12-M in the receiver configuration of Document 4 is further performed using the data output obtained. The average bit error rate of the data output of each stream after the processing from the equalizer for 1 stream 12-1 to the equalizer for the Mth stream 12-M is shown. A solid line 37 indicates an average bit error rate after the processing from the first stream equalizer 21-1 to the M-th stream equalizer 21-M is completed using the receiver configuration of the first embodiment. The solid line 38 is equalized again for the first stream after the processing from the first stream equalizer 21-1 to the Mth stream equalizer 21-M is completed using the receiver configuration of the first embodiment. The average bit error rate after processing from the equalizer 21-1 to the M-th stream equalizer 21-M is shown. That is, the solid line 38 shows the characteristic after the process in each stream equalizer is repeated twice.
[0049]
E_bIs bit energy, N₀Indicates noise power.
From the results shown in FIG. 5, when the method of Document 4 is used, the characteristics are hardly improved even if the process is repeated. However, in the configuration of the first embodiment, the characteristics are improved by repeating the process, and for each stream, etc. Even in the case where the processing of the generator was performed only once, it was confirmed that it was superior to the characteristics of the receiver configuration of Document 4.
Next, FIG. 6 shows characteristics in a three-wave Rayleigh fading channel (preceding wave, one symbol delay wave, and two symbol delay wave). Here, the maximum number of delay symbols considered by the equalizer is 2. Dashed lines 41, 42, and 43 indicate cases where the process of Reference 4 is performed once, twice, and three times, respectively, and solid lines 44, 45, and 46 indicate once, twice, and three according to the configuration of the first embodiment. This is the case of processing once. In this case, the method of Reference 4 also improves the characteristics by repeatedly performing the process, but it can be confirmed that the characteristics of the configuration of Example 1 are superior. Although the delayed wave is not considered in the method of Literature 4, even when the delayed wave is taken into consideration from FIG. 6, the characteristics of Example 1 can be improved as compared with the method of Literature 4.
[0050]
The portion excluding the determiners 17-1 to 17-M in the configuration shown in FIG. 1 constitutes a multi-input multi-output equalizer.
Example 2
Next, a second embodiment of the present invention will be described with reference to FIG. In FIG. 7, instead of the stream equalizers 21-1 to 21-M shown in FIG. 1, stream equalizers 47-1 to 47-M are used, and the likelihood value output of the stream equalizer is output. Is different from the first embodiment. In the first embodiment, the likelihood value output obtained by the process of the m-th stream equalizer 21-m is used in the process of the m + 1-th stream equalizer 21-m + 1. The likelihood value output from the stream equalizer 47-m is not used for the processing of the (m + 1) th stream equalizer 47-m + 1.
[0051]
Therefore, in the second embodiment, first, the first stream equalizer 47-1 to the Mth stream equalizer 47-M perform equalization processing using only the received signal vector. After the first processing in all stream equalizers 47-1 to 47-M is completed, each stream equalizer 47- is again used by using the same received signal vector as the obtained likelihood value output. Processing is performed at 1-47-M. In this configuration, the processing of each of the stream equalizers 47-1 to 47-M can be performed at the same time, so that the time required for the entire processing can be reduced.
An example of the processing procedure in this case will be briefly described with reference to FIG. First impulse responseH ₁(K) ~H _M(K) is estimated (S0), and the following processing is performed for m = 1,. First, the filter coefficientW _m(K) is calculated by substituting “1” of the subscript of equation (8) as “m” (S1), and each of these filter coefficientsW _mReceive vector by (k)r(K) is filtered (S2), and each filter processing result z_m(K) and filter coefficientW _m(K) Impulse responseH _mThe likelihood value λ using (k)_m(K) (or tanh (λ_m/ 2)) is obtained (S3). This is the first process.
[0052]
Next, the following processing is performed for m = 1,. First, the previous likelihood value λ₁(K) to λ_M(K) is used to generate a likelihood vector (S4). This likelihood vector is generated in the same manner as in step S2 in FIG. 4, except that all likelihood values obtained by the previous processing are used. Is different. Next, interference components for m = 1,..., M streams are respectively generated (S5), and these interference components are converted into received vectors.rRemove each from (k)r ₁'(K) ~r _M'(K) is obtained (S6) and the filter coefficientW ₁(K) ~W _M(K) is calculated (S7). This calculation is the same as step S5 in FIG. 4, except that all the values obtained by the previous process are used as likelihood values.
[0053]
These filter coefficientsW ₁(K) ~W _MBy (k)r ₁'(K) ~r _M′ (K) is filtered (S8), and the result z₁(K) -z_M(K) and filter coefficientW ₁(K) ~W _M(K) and impulse responseH ₁(K) ~H _MUsing (k), the likelihood value λ is calculated by the same calculation as in step S7 in FIG.₁(K) to λ_M(K) is calculated (S9). It is checked whether the number of processing times has reached a predetermined number (S10). If not, the process returns to step S4, and if not, the likelihood value λ₁(K) to λ_M(K) is determined by the determiner and output (S11). After step S9, the likelihood value λ₁(K) to λ_M(K) may be determined and output as an intermediate result (S12).
[0054]
The characteristics of the receiver configuration according to the first embodiment and the receiver configuration according to the first embodiment were compared by computer simulation. The result is shown in FIG. The computer simulation specifications are the same as those in FIG. 6, and solid lines 51, 52, and 53 in FIG. 9 indicate the characteristics of processing once, twice, and three times in the receiver configuration of the second embodiment. Reference numerals 55 and 56 indicate the characteristics of one, two, and three processes of the receiver configuration in the first embodiment. The characteristics after the first and second processing in the receiver configuration of the second embodiment are degraded from the corresponding characteristics of the receiver configuration of the first embodiment, but the third processing is completed. The characteristics after the are almost the same. Therefore, when the process is repeated three times, the processing time can be reduced by using the receiver configuration of the second embodiment as compared with the case of using the receiver configuration of the first embodiment.
[0055]
The portion excluding the determiners 17-1 to 17-M in FIG. 7 constitutes a multi-input multi-output equalizer.
Example 3
Next, a third embodiment of the present invention will be described. Document 5 proposes a receiver configuration that repeatedly performs equalization and decoding for MIMO channel signal transmission. FIG. 22 shows a transceiver configuration in Document 5. The data of the first to Mth streams are encoded by the encoders 61-1 to 61-M, respectively, and these encoded data are changed in the order in which the encoded sequences are arranged between the frames by the interleavers 62-1 to 62-M. With these data, the carrier waves are modulated by the transmitters 63-1 to 63-M and supplied to the transmission antennas ANS-1 to ANS-N. Signals received by the receiving antennas ANR-1 to ANR-M are input to the MIMO equalizer 64, and the equalized outputs (likelihood value outputs) are rearranged by the deinterleavers 65-1 to 65-M, respectively. Returned and decoded by the decoders 66-1 to 66 -M. The likelihood value outputs from the decoders 66-1 to 66-M are rearranged by the interleavers 67-1 to 67-M and input to the MIMO equalizer 64 for equalization processing of the previous received signal. Used.
[0056]
As shown in FIG. 23, the MIMO equalizer 64 receives received signal sample values from the antennas ANR-1 to ANR-N as input to first to Mth stream equalizers 68-1 to 68-M, The likelihood values from the decoders 66-1 to 66-M are input to these equalizers 68-1 to 68-M via interleavers, respectively. The likelihood value output is derived in each stream equalizer 68-m, and the value is decoded in a SISO (Soft-Input Soft-Output) decoder 66-m. The likelihood value output obtained from the decoder 66-m is again used in the stream equalizer 68-m via the interleaver 67-m and processed. The above processing is repeated, and the first to Mth stream data outputs are finally obtained from the decoders 66-1 to 66-M. In the configuration shown in FIG. 20, the processing of each stream equalizer 68-1 to 68-M can be performed simultaneously.
[0057]
In the third embodiment, each stream equalizer sequentially performs processing, and the likelihood value from the decoder and the output likelihood value of the stream equalizer in the current processing are also used. That is, as shown in FIG. 10, in the first to Mth stream equalizers 71-1 to 71-M, these are sequentially processed, and the mth stream equalizer 71-1 has the first to mth streams. Likelihood value outputs from the -1 stream equalizers 71-1 to 71-m-1 and likelihood value outputs from the decoders 66-m to 66-M are input. By using the likelihood value output obtained in the m-th stream equalizer 71-m in this way in the m + 1-th stream equalizer 71-m + 1, processing is performed using more likely likelihood values. Therefore, the characteristics can be improved.
Example 4
Next, a fourth embodiment of the present invention will be described with reference to FIG. In the fourth embodiment, the effects of delayed waves are reduced by using the method of the first embodiment as a multi-carrier type, and each stream equalizer is configured so that the equalizer considers only one incoming wave in each stream equalizer. It is characterized by reducing the amount of computation. The multi-carrier method is a method for transmitting signals using a plurality of carriers (carrier waves) in order to reduce the influence of delayed waves, which is a problem in high-speed transmission. is there. The multi-carrier method can reduce the information transmission rate in each subcarrier by dividing the frequency of the same bandwidth as when performing signal transmission using one carrier into multiple subcarriers and performing signal transmission. In addition, the influence of the delayed wave can be reduced. Further, by providing the guard interval, it is possible to efficiently reduce the influence of the delayed wave. This multi-carrier scheme can be realized by using as many transceivers as the number of carriers. However, it is common to perform multi-carrier multiplexing using fast Fourier transform called OFDM (Orthogonal Frequency Division Multiplexing). It is.
[0058]
The multi-carrier scheme will be described with reference to FIG. On the transmission side, in the first to Mth streams, modulation such as BPSK and QPSK is performed by the modulators 73-1 to 73-M, and then serial / parallel conversion is performed by the serial-parallel converters 74-1 to 74-M. In addition, multi-carrier is performed. As a means for multicarrier, in FIG. 11, a fast inverse Fourier transform (Inverse Fast Fourier Transform) devices 75-1 to 75-M are used to perform multicarrier conversion by performing fast inverse Fourier transform for each parallel output. Yes. Although omitted in FIG. 11, the signal subjected to the fast inverse Fourier transform is transmitted after the guard interval is added.
[0059]
On the receiving side, the multicarrier multiplexed signal received at each of the antennas ANR-1 to ANR-N is returned to the signal for each subcarrier. In FIG. 11, Fast Fourier Transformers 76-1 to 76-N are used as one means for returning a multicarrier multiplexed signal to a signal for each subcarrier. Each frequency component, that is, each subcarrier subjected to the fast Fourier transform by the fast Fourier transformers 76-1 to 76 -N, is transmitted from a plurality of transmission antennas by the MIMO equalizers 77-1 to 77 -J for each subcarrier. Signal separation is performed. That is, each subcarrier f obtained from the antennas ANR-1 to ANR-N._j(j = 1,..., J) are input to the same MIMO equalizer 77-j.
[0060]
The configuration of each MIMO equalizer 77-j is the receiver configuration shown in FIG. Same subcarrier f_jThe received signal sample values corresponding to the antennas are input to the terminals 11-1 to 11-N. These are input to first to Mth stream equalizers 79-1 to 79-M. The likelihood values of the first to Mth streams from the parallel-serial converters 78-1 to 78-M described later are input to the terminals 81-1 to 81-M. The m-th stream equalizer 79-m includes output likelihood values of the first to (m-1) th stream equalizers 79-1 to 79-m-1, and terminals 81-m to 81-M. A likelihood value is input and used for equalization processing. That is, these equalization processing operations are the same as those shown in FIGS.
[0061]
As described above, since the influence of the delayed wave can be reduced by performing the serial-parallel conversion and further providing the guard interval as necessary, the configuration of each stream equalizer 79-M is as follows. The configuration shown in FIG. 2 is a configuration in which only one incoming wave is considered and no delayed wave is considered. The first to Mth stream-corresponding likelihood value outputs of the MIMO equalizers 77-1 to 77-J are serial likelihoods of the first to Mth parallel-serial converters 78-1 to 78-M, respectively. The first to Mth series likelihood value outputs are supplied to the determiners 17-1 to 17-M and supplied to the terminals 81-1 to 81-M in FIG.
[0062]
Since each stream equalizer 79-m has a configuration in which only one incoming wave is taken into account and no delay is taken into account, as a result, the number of reception antennas N and the equalizer 21-m in the case where the multi-carrier type is not used. Assuming that the maximum number of delay symbols to be considered is Q−1, the amount of computation is (N × Q) for the inverse matrix computation of Equation (35).^ThreeOn the other hand, in the fourth embodiment, the equalizer 79-m does not consider the delayed wave, so the amount of calculation is N^ThreeIt can be suppressed to an increase in order. That is, according to the fourth embodiment, the calculation amount is 1 / Q compared with the conventional method.^ThreeCan be reduced.
Note that a combination of frequency converters may be used for a plurality of frequency oscillators instead of the fast Fourier transformers 76-n (n = 1,..., N), that is, the reception signal correspondence of each reception antenna ANR-n is A separation unit that separates the received signal for each subcarrier may be provided.
Example 5
Embodiment 5 of the present invention will be described. In the multicarrier MIMO channel signal transmission shown in FIG. 11, in the fourth embodiment, the configuration of the MIMO equalizer 77-j is the same as that shown in FIG. 12, but in the fifth embodiment, the MIMO equalizer 77-j is configured. As shown in FIG. 13, the connection configuration of the stream equalizers 79-1 to 79-M in j is the same as that shown in FIG. With this configuration, the processing of each stream equalizer 79-m can be performed simultaneously. Therefore, the time required for processing can be reduced as compared with the configuration shown in FIG.
Example 6
Embodiment 6 of the present invention will be described with reference to FIGS. The transmission / reception configuration of the multicarrier MIMO channel signal transmission shown in FIGS. 14 and 15 is a configuration in which the encoding shown in FIG. 22 is added to the multicarrier MIMO channel signal transmission shown in FIG. Parts are given the same reference numerals. In FIG. 14, the information transmitted from the transmission antennas of the first stream to the Mth stream is first encoded by the encoder 61-m, and the encoded sequence in the frame is converted by the interleaver 62-m. The ordering can be changed. Here, although the encoding in each transmitting antenna is represented separately in the figure, for example, when one user performs signal transmission using a plurality of antennas ANS-1 to ANS-M, each antenna ANS-1 to ANS-1 By performing the encoding corresponding to ANS-M in cooperation, the reception quality characteristic can be improved. The subsequent processing on the transmission side is equivalent to the processing shown in FIG.
[0063]
On the receiving side, as shown in FIG. 15, the multi-carrier multiplexed signal is returned to the signal for each carrier, and signal separation is performed using the MIMO equalizer 77-j having the configuration shown in FIG. Parallel-serial conversion is performed by the parallel-serial converter 78-m. Thereafter, decoding is performed by the decoder 66-m via the deinterleaver 65-m. Here, when encoding on each transmitting antenna is performed cooperatively on the transmission side, decoding is also performed collectively. As the decoder 66-m, a SISO type decoder is used to derive the likelihood value output of the encoded sequence, and the result is sent to the MIMO equalizers 77-1 to 77-77 via the interleaver 67-m. -Return to J. In this way, the signal separation and decoding processes are repeated, and the data output output from the decoder 66-m becomes the final determination result.
[0064]
As shown in FIG. 15, the amount of calculation in each stream equalizer can be reduced by using multi-carrier. Specifically, if the multi-carrier type is not used, the number of reception antennas is N, and the maximum number of delay symbols considered by the equalizer is Q-1, so that equalization for each stream is performed for the inverse matrix operation of Equation (35). The calculation amount of the instrument is (N × Q)^ThreeIn the sixth embodiment, since the equalizer does not consider the delayed wave, the amount of calculation is N.^ThreeIt can be suppressed to an increase in order. That is, according to the sixth embodiment, the calculation amount is 1 / Q compared with the conventional method.^ThreeCan be reduced.
Example 7
In the transmitter / receiver configuration of the sixth embodiment illustrated in FIG. 15, the configuration of the MIMO equalizer 77-j is configured as illustrated in FIG. 23, whereby the processing of each stream equalizer can be performed simultaneously. Therefore, the time required for processing can be reduced as compared with the method of the sixth embodiment having the configuration shown in FIG.
Example 8
In the configuration of the V-BLAST method shown in FIG. 20 and FIG. 21, as described in the section of the problem to be solved by the invention, when the signal determination result obtained by equalization processing of a certain stream is incorrect, In the subsequent stream equalization processing, the signal component of the other stream cannot be correctly subtracted from the received signal, so that further errors occur. This is called error propagation. As a method of using a more reliable signal determination result in the V-BLAST type equalizer, error correction coding is performed on the transmission side, and processing is performed in the V-BLAST type equalizer using the result obtained from the decoder on the reception side. It is conceivable to perform the equalization and decoding processes repeatedly. However, even in this case, characteristic degradation occurs due to the effect of error propagation. Therefore, in the eighth embodiment of the present invention, the equalization and decoding processes are repeatedly performed, and the first equalization process is performed according to the configuration shown in FIG. 20, and the second and subsequent equalization processes are configured as shown in FIG. Thus, the previous decoding result is used, and the configuration shown in FIG. 21 that does not require the derivation of the likelihood value is used as each stream equalizer 12-m.
[0065]
That is, for example, as shown in FIG.H ₁(k) ~H _M(k) is estimated (S1), and m is initialized to 1 (S2). In the m-th stream equalizer 12-m, the determiners 17-1 to 17-1 from the first to m-1th stream equalizers 12-1 to 12-m-1 that have already been subjected to the first equalization process. -m-1 output data b₁(k) -b_m-1(k) and impulse response matrixH ₁(k) ~H _m-1Using (k), the received vectorrAn interference component generated by the signals of the first to (m-1) th streams in (k) is generated by the interference component generator 14 (S3). Likelihood value tanh (λ in equation (6)₁(k)) instead of b₁Using (k)b ₁Interference symbol vector corresponding to (k)
b″₁(k) = [b₁(k + Q-1) ... b₁(k) ... b₁(k−Q + 1)]^T  (36)
Make b₂(k) -b_m-1Interference symbol vector for (k)b″₂(k) ~b″_m-1Create (k) and corresponding impulse responseH ₁(k) ~H _m-1An interference component is generated by the product of (k).
[0066]
These interference components are received vectorsrSubtract from (k), difference vectorr′_m(k) is obtained (S4). AlsoH ₁(k) ~H _mFilter coefficient from (k)w _m(k) is calculated (S5), and this filter coefficientw _mDifference vector at (k)r′_m(k) is filtered by the linear filter 16 (S6), and the filtered result z_m(k) is determined by the determiner 17-m, and the m-th stream data output is obtained (S7). The m-th stream data output is subjected to error correction decoding by the decoder 91-m as shown in FIG. 16 (S8). On the transmission side, data is encoded by an error correction code method.
Next, it is checked whether m has become M (S9). If M has not been reached, m is incremented by 1 and the process returns to step S3 to proceed to the next stream equalizer process (S10). If m = M in step S9, the first equalization process in the first to Mth stream equalizers 12-1 to 12-M is completed. In short, this first equalization processing is performed by the likelihood value tanh (λ_m(k)) instead of already processed decision result b_mOnly (k) is used, which is the same as the V-BLAST method shown in Document 4.
[0067]
In the equalization processing after the second time, the previous equalization processing, error-corrected decoded output (decoding result) b ′₁(k) -b '_M(k) is input to all stream equalizers 12-1 to 12-M from terminals 92-1 to 92-M, and an impulse response matrix is generated in the interference component generator 14.H ₁(k) ~H _mUsing (k), an interference component of each stream is generated (S11). In this case, in the m-th stream equalizer 12-m, the interference vectorb″₁(k) ~b″_m-1(k),b″_{m + 1}(k) ~b″_M(k) is obtained by replacing the subscript “1” in the expression (36) with the corresponding subscript.b″_m(k) indicates that the subscript “1” in equation (36) is “m” and the interference vectorb″_mThe element located at the center of (k) is zero.
[0068]
In each m-th stream equalizer 12-m, the same reception vectorrRemove corresponding interference components from (k)r′_m(k) is obtained (S12). Filter coefficients in the linear filter 16 of each m-th stream equalizer 12-mw _m(k) is calculated (S13). This filter coefficientw _m(k) is tanh (λ of the term of the filter coefficient in the equation (33) in the first embodiment._m(k) / 2), the previous decoding result b ′ instead of (m = 1,..., n)_mWhat is necessary is just to calculate using (k). Its filter coefficientsw _mThe difference vector by (k)r′_m(k) is filtered (S14). Each filter processing result z_m(k) is determined by the determiner 17-m (S15), and the determination result is error-corrected and decoded by the decoder 91-m (S16).
[0069]
This previous decryption result b ′₁(k) -b '_MIf the equalization process using (k) has not reached the predetermined number (S17), the process returns to step S11, and if it has reached the predetermined number, the process ends.
As described above, in the eighth embodiment, the previous error correction decoding result is used in the second and subsequent equalization processing, and the determination result of the stream for which processing has been completed is not used in the processing in the next stream. As a result, it is possible to suppress deterioration of the average error rate characteristic due to error propagation. Thereby, in the second and subsequent equalization processes, the processes in the respective stream equalizers 12-1 to 12-M can be performed at the same time, so that the processing delay can be reduced.
[0070]
In the eighth embodiment, the equalizers 12-1 to 12-M perform processing using the determination result and the decoding result. In the decoders 91-1 to 91-M, the output of the filter 16 is determined by the determiner 17. In order to perform processing using the obtained results, the likelihood value calculator 24 for calculating likelihood values in the equalizers 12-1 to 12-M is not required. Also, the decoders 91-1 to 91-M do not require a device for deriving likelihood values. Further, the eighth embodiment can be similarly used in a receiver for multicarrier MIMO channel signal transmission. That is, in FIG. 15, all the deinterleavers 65-m and all the interleavers 67 are omitted, the decoder 91-m is used as the decoder 66-M, and each multi-input multi-output equalizer 77-j is shown in FIG. 20, the configuration shown in FIG. 21 and FIG. 16, and the processing shown in FIG.
[0071]
FIG. 18 shows a computer simulation result of average error rate (BER) characteristics in the receiver configuration of the eighth embodiment. As simulation specifications, the number of transmission antennas and the number of reception antennas were each 2, the modulation method was BPSK, and the propagation path was a 5-path Rayleigh fading propagation path with delay time differences of 0, 1, 2, 3, and 4 symbols. In the figure, a broken line 93 and a broken line 94 using OFDM (multicarrier system) with 16 subcarriers and 4 symbols of guard intervals in order to suppress characteristic deterioration due to delayed waves are the conventional combinations of a V-BLAST type equalizer and a decoder. The solid lines 95 and 96 show the results in the receiver configuration of the eighth embodiment. A broken line 93 and a solid line 95 indicate the result of the first equalization and decoding, and a broken line 94 and a solid line 96 indicate the result of performing the equalization → decoding → equalization → decoding, that is, the result of the second processing. From these results, the characteristics after the equalization and decoding are performed once are equivalent, while the characteristics after the equalization and decoding are performed twice are the same as those in the eighth embodiment. It is improved compared to the BLAST method. This is because the influence of error propagation can be reduced.
[0072]
In the above description, the transmission path estimator 13 in the first to Mth stream equalizers can be used in common. In addition, in the above description, since each subcarrier in the same multicarrier passes through the same propagation path, the same can be said. Therefore, the transmission path of one stream equalizer 79 in one multi-input multi-output equalizer 77-j. The estimator can be commonly used for all the multi-input multi-output equalizers 77-j. In the embodiment shown in FIGS. 15 and 22, when the interleaver is omitted on the transmission side, the deinterleaver and interleaver are omitted on the reception side. However, the characteristics are better when these interleavers and deinterleavers are used.
[0073]
【The invention's effect】
According to the first embodiment of the present invention, the error rate characteristic can be improved as compared with the receiver configuration of Document 4.
According to the second embodiment of the present invention, the time required for processing can be reduced and the error rate characteristic can be improved as compared with the receiver configuration of Document 4.
According to the third embodiment of the present invention, the error rate characteristics can be improved as compared with the method of Document 5.
According to the fourth embodiment of the present invention, the amount of calculation can be reduced as compared with the first embodiment.
[0074]
According to the fifth embodiment of the present invention, the time required for processing can be reduced as compared with the fourth embodiment.
According to the sixth embodiment of the present invention, the amount of calculation can be reduced as compared with the third embodiment.
According to the seventh embodiment of the present invention, the time required for processing can be reduced as compared with the sixth embodiment.
According to the eighth embodiment of the present invention, the error rate characteristics can be improved as compared with the case where the receiver configuration and the decoder described in Document 4 are used.
[Brief description of the drawings]
FIG. 1 is a diagram illustrating a functional configuration of a first embodiment.
FIG. 2 is a diagram illustrating an example of a functional configuration of each stream equalizer 21-m in FIG. 1;
3 is a diagram illustrating an example of a functional configuration of an interference component generator 23 in FIG. 2, and FIG. 3B is a diagram illustrating an example of a functional configuration of a filter 16 and a likelihood value calculator 24 in FIG.
4 is a flowchart showing an example of a processing procedure of the configuration shown in FIG. 1;
FIG. 5 is a diagram showing an average BER characteristic in a one-wave Rayleigh fading channel between the first embodiment and the receiver configuration in Document 4.
6 is a graph showing average BER characteristics in a three-wave Rayleigh fading propagation path of the receiver configuration in Embodiment 1 and Reference 4. FIG.
7 is a diagram showing a receiver configuration of Embodiment 2. FIG.
8 is a flowchart showing an example of a processing procedure of the receiver shown in FIG.
9 is a graph showing an average BER characteristic of Example 2 and an average BER characteristic of Example 1. FIG.
10 is a diagram showing a configuration of a MIMO equalizer 64 in Embodiment 3. FIG.
FIG. 11 is a diagram showing a transmission / reception configuration of multicarrier MIMO channel signal transmission.
FIG. 12 is a diagram illustrating a configuration of each MIMO equalizer 77-j according to the fourth embodiment.
FIG. 13 is a diagram illustrating a receiver configuration of each MIMO equalizer 77-j according to the fifth embodiment.
14 is a diagram showing a transmitter configuration corresponding to the receiver configuration in Embodiment 6. FIG.
15 is a diagram showing a receiver configuration of Embodiment 6. FIG.
FIG. 16 is a diagram illustrating a receiver configuration in the second and subsequent processes in the eighth embodiment.
FIG. 17 is a flowchart illustrating an example of a processing procedure according to the eighth embodiment.
18 is a diagram illustrating average BER characteristics of the receiver configuration in Example 8 and the receiver configuration in Literature 4. FIG.
FIG. 19 is a diagram showing a transmission / reception configuration in MIMO channel signal transmission.
FIG. 20 is a diagram showing a receiver configuration described in Document 4.
FIG. 21 is a diagram showing a configuration of a stream equalizer 12-m in FIG.
22 is a diagram showing a transceiver configuration in Document 5. FIG.
23 is a diagram showing a configuration of a MIMO equalizer 64 in FIG.
[References]
1. Shigeru Akari, Takahiro Asai and Tadashi Matsumoto, “Radio signal processing in MIMO channel signal transmission for mobile communications,” IEICE Technical Report, RCS2001-136, pp.43-48, Oct, 2001
2. G.J.Foschini and M.J.Gans, “On limits of wireless communications in a fading environment when using multiple antennas,” Wireless Personal Communication, vol.6, no.3, pp.315-335, March, 1998.
3. H. Yoshino, K. Fukawa, H. Suzuki, “Interference canceling equalizer (ICE) for mobile radio communication,” IEEE Trans, on VT, vol. 46, no. 4, pp. 849-861, Nov. 1997.
4). PWWolniansky, GJFoschini, GDGolden, and RAValenzuela, “V-blast: An architecture for realizing very high data rates over the rich-scattering wireless channel.” URSI International Symposium on Signals, Systems, and Electronics Conference Proceedings, pages 295 -300,1998.
5. Tetsushi Abe and Tadashi Matsumoto, “Space-Time Turbo Equalizer in Frequency Selective MIMO Channel,” IEICE Tech., RCS2000-256, pp.75-80, March, 2001

Claims (15)

In a receiver that receives signals of a plurality of streams transmitted from a plurality of transmission antennas,
A multi-input multi-output equalizer having a plurality of stream equalizers provided in correspondence with the respective streams and in which the processing order is determined ;
Each of the above stream equalizers is
A transmission path estimator that receives the received signal and estimates the impulse response of the signal propagation path of each stream;
The likelihood value output from the stream equalizer that has been processed this time, the previous likelihood value of the stream equalizer that has not been processed this time, and the impulse response are input and processed. An interference component generator that generates, as an interference component, a stream reception signal corresponding to the stream equalizer that has been
A signal from a received signal by subtracting the interference components, and the impulse response, the last processing of the stream equalizer likelihood value and the process above process is outputted from the equalizer for stream dispensed is not dispensed A filter that receives the input likelihood value , filters the received signal from which the interference component is subtracted, and outputs a signal corresponding to the stream equalizer;
The filter output, the likelihood value output from the stream equalizer that has been subjected to the above processing, the likelihood value that was previously processed by the stream equalizer that has not been subjected to the above processing, and the impulse response are input. is, multiple input multiple output turbo receiver; and a likelihood value calculator that outputs a likelihood value. In a receiver that receives signals of a plurality of streams transmitted from a plurality of transmission antennas,
Comprising a multiple-input multiple-output equalizer having a plurality of streams for the equalizer provided in correspondence with the respective stream,
Each stream for the equalizer,
A transmission path estimator that receives the received signal and estimates the impulse response of the signal propagation path of each stream;
An likelihood component from each stream equalizer output by the previous process, an impulse response, and an interference component generator that generates an interference component;
A signal obtained by subtracting the interference component from the received signal, an impulse response, the likelihood value, and a filter that outputs a signal corresponding to the stream equalizer;
And the filter output, and the likelihood value, the impulse response is input, multiple-input multiple-output turbo receiver; and a likelihood value calculator that outputs a likelihood value. Each stream signal is serial-to-parallel converted, and each parallel signal is a multicarrier multiplexed signal on a subcarrier,
A plurality of separation units for separating the received signal from each receiving antenna into a signal of each subcarrier;
A plurality of multi-input multi-output equalizers according to claim 1 for each subcarrier to which a reception signal from the same subcarrier separated by the separation unit is input.
A plurality of parallel values are input to each stream corresponding likelihood values from these multi-input multi-output equalizers, output serial likelihood values corresponding to the streams, and input to the multi-input multi-output equalizers. The multi-input multi-output turbo receiver according to claim 1, further comprising a serial converter. Each stream signal is serial-to-parallel converted, and each parallel signal is a multicarrier multiplexed signal on a subcarrier,
A plurality of separation units for separating the received signal from each receiving antenna into a signal of each subcarrier;
A plurality of multi-input multi-output equalizers according to claim 2, for each subcarrier to which a reception signal from the same subcarrier separated by the separation unit is input,
A plurality of parallel values are input to each stream corresponding likelihood values from these multi-input multi-output equalizers, output serial likelihood values corresponding to the streams, and input to the multi-input multi-output equalizers. The multi-input multi-output turbo receiver according to claim 2, further comprising a serial converter. In a receiver that receives signals of a plurality of streams transmitted from a plurality of transmission antennas,
A multi-input multi-output equalizer that is provided corresponding to each of the transmission antennas, includes a plurality of stream equalizers that output a first likelihood value, the processing order being determined;
A plurality of decoders each supplied with a first likelihood value output from each stream equalizer and outputting a second likelihood value;
Comprising
Each stream for the equalizer,
A transmission path estimator that receives the received signal and estimates the impulse response of the signal propagation path of each stream;
The first likelihood value from the stream equalizer that has been processed, the second likelihood value from the decoder that corresponds to the stream equalizer that has not been processed, and the impulse response are input. An interference component generator for generating interference components;
A signal obtained by subtracting the interference component from the received signal, and the impulse response, and the first likelihood value than the equalizer for processing has been dispensed stream corresponding to the stream equalizer processing has not been dispensed and the second the likelihood value from the decoder is input, a filter for outputting a signal corresponding to the stream for the equalizer a received signal interference component is pulled to filter,
A likelihood value calculator that receives the filter output, the first likelihood value, the second likelihood value, and the impulse response, and outputs the first likelihood value;
Multiple input multiple output turbo receiver, characterized in that it comprises a. Each stream signal is serial-to-parallel converted, and each parallel signal is a multicarrier multiplexed signal on a subcarrier,
A plurality of separation units for separating the received signal from each receiving antenna into a signal of each subcarrier;
A plurality of multi-input multi-output equalizers according to claim 5 for each subcarrier into which received signals from the same subcarrier separated by the separation unit are input,
A first likelihood value corresponding to each stream is input from each of these multi-input multi-output equalizers, a plurality of parallel first-likelihood values corresponding to streams are output and input to the corresponding decoders. 6. The multi-input multi-output turbo receiver according to claim 5, further comprising a serial converter. In a receiver that receives signals transmitted from a plurality of transmitting antennas, a multi-carrier multiplexed signal in which each stream signal is serial-parallel converted and each parallel signal is placed on a subcarrier.
A plurality of separation units for separating the received signal from each receiving antenna into a signal of each subcarrier;
A plurality of multi-input multi-output equalizers for each subcarrier to which a reception signal from the same subcarrier separated by the separation unit is input;
A plurality of parallel-to-serial converters, each of which receives a first likelihood value corresponding to each stream from the multi-input multi-output equalizer and outputs a first serial likelihood value corresponding to the stream;
A plurality of decoders each receiving a corresponding one of the first likelihood values in series and supplying a second likelihood value to the multiple multi-input multi-output equalizer;
Each of the multi-input multi-output equalizers includes a plurality of stream equalizers provided corresponding to the streams.
Each of the above stream equalizers is
A transmission path estimator that receives the received signal and estimates the impulse response of the signal propagation path of each stream;
An interference component generator that receives the second likelihood value from each decoder output by the previous process and an impulse response and generates an interference component;
A signal obtained by subtracting the interference component from the received signal, and the impulse response, the second likelihood value is input, a filter for outputting a signal corresponding to the equalizer for the stream,
A multi-input multi-characteristic device comprising: a likelihood value calculator that receives the filter output, the second likelihood value, and the impulse response and outputs the first likelihood value. Output turbo receiver. In a receiver that receives signals of a plurality of streams transmitted from a plurality of transmission antennas,
A multi-input multi-output equalizer comprising a plurality of stream equalizers provided in correspondence with the respective streams;
A determination output of each stream equalizer is input, and a plurality of decoders for performing a decoding process,
Each stream for the equalizer,
A transmission path estimator that receives the received signal and estimates the impulse response of the signal propagation path of each stream;
An interference component generator that receives the previous decoding results and impulse responses from all the decoders and generates an interference component;
A signal obtained by subtracting the interference component from the received signal, an impulse response, and the previous decoding result, and a filter that outputs a signal corresponding to the stream equalizer;
A multi-input multi-output turbo receiver, comprising: a determination unit that receives the filter output, performs binary determination, and uses the determination result as an output of a stream equalizer. Each stream signal is serial-to-parallel converted, and each parallel signal is a multicarrier multiplexed signal on a subcarrier,
A plurality of separation units for separating the received signal from each receiving antenna into a signal of each subcarrier;
A plurality of multi-input multi-output equalizers according to claim 8 for each subcarrier to which a reception signal from the same subcarrier separated by the separation unit is input.
A plurality of parallel-to-serial converters for inputting the stream-corresponding determination results from the multi-input / multi-output equalizers, outputting the stream-corresponding serial determination results and inputting them to the corresponding ones of the decoder; The multi-input multi-output turbo receiver according to claim 8, comprising: Receive M streams of signals transmitted from integer M antennas greater than or equal to 2 ,
Each of M stream equalizers
Estimate the impulse response of the signal propagation path of each stream from the received signal,
Generate an interference component corresponding to each stream using the likelihood value and impulse response of each stream,
Subtract the interference component corresponding to each stream from the received signal,
In the reception method of filtering the subtracted signal using the likelihood value and impulse response of each stream,
The M stream equalizers are processed in the order of the first to Mth stream equalizers,
In an arbitrary integer a-th process, the m th (m = 1,..., M) stream equalizer uses the likelihood values from the first to (m-1) th stream equalizers as the likelihood values, and Using the likelihood value from the equalizer for the m-th to M-th streams obtained in the a-1th process,
Each of the equalizers for the m-th stream has its filter processing output, the likelihood value from the a-th first to (m-1) th stream equalizers, and the (m-1) th to (m) th times obtained in the (a-1) th process. A multi-input multi-output turbo reception method, wherein a likelihood value is calculated using a likelihood value and an impulse response from an equalizer for M streams. Receive M streams of signals transmitted from integer M antennas greater than or equal to 2,
Each the M streams for the equalizer
Estimate the impulse response of the signal propagation path of each stream from the received signal,
Generate an interference component corresponding to each stream using the likelihood value and impulse response of each stream,
Subtract the interference component corresponding to each stream from the received signal,
In the reception method of filtering the subtracted signal using the likelihood value and impulse response of each stream,
First, the received signal is filtered using an impulse response to obtain a signal for each stream,
For each stream, calculate the likelihood value for that stream using the filtered signal and the impulse response,
In the subsequent processing, each stream processing generates an interference component corresponding to each stream using the M stream likelihood values obtained in the previous processing and the impulse response,
A multi-input multi-output turbo reception method characterized in that M stream likelihood values obtained in the previous process are also used for the filtering process and the likelihood value calculation, respectively. The likelihood value used for generating the interference component and calculating the likelihood value is obtained by a decoding process corresponding to the stream,
The multi-input multi-output according to claim 10, wherein the likelihood value used for generating the interference component and calculating the likelihood value is decoded by decoding the calculated likelihood value corresponding to each stream. Turbo reception method. The signal of the transmitted stream is a multi-carrier signal in which the parallel signal is serial-parallel converted, the parallel signal is placed on different subcarriers, and these subcarriers are multiplexed.
Separate the received signal for each receiving antenna into subcarriers,
For each of these subcarriers, the generation of the interference component corresponding to each stream, the filtering process, and the calculation of the likelihood value are performed to obtain M likelihood values,
13. The stream corresponding to the M likelihood values obtained for each subcarrier is parallel-serial converted to obtain the likelihood value of each stream. The multi-input multi-output turbo reception method described in 1. Receives M stream signals transmitted from integer M transmission antennas of 2 or more, estimates propagation path impulse response of reception signals from each transmission antenna, and equalizes them by M stream equalizers In the receiving method of processing and decoding the equalization processing result by the decoder,
As the first reception process,
The M stream equalizers are processed in the order of the first to Mth stream equalizers,
The m-th (m = 1,..., M) stream equalizer generates an interference component corresponding to each stream using the determination result and the impulse response from the first to (m-1) th stream equalizers.
Subtract the interference component corresponding to each stream from the received signal,
The subtracted signal is filtered using the determination result from the first to (m-1) th stream equalizers and the impulse response,
The filter processing result is subjected to determination processing, and the determination processing result is supplied to the decoder for the m-th stream,
As the second and subsequent reception processing,
Each stream equalizer generates an interference component corresponding to each stream using the decoding results and impulse responses of all previous streams,
Subtract the interference component corresponding to each stream from the received signal,
Filter the subtracted signal using the decoding results and impulse responses of all streams,
A multi-input multi-output turbo reception method characterized in that, for each stream, the filtered signal is subjected to binary determination, and each determination result is decoded. The signal of the transmitted stream is serial-parallel converted,
The parallel signal is put on different subcarriers, and these subcarriers are multiplexed multicarrier signals.
Separate the received signal for each receiving antenna into subcarriers,
For each of these subcarriers, the generation of the interference component corresponding to each stream, the filtering process, and the determination process are performed to obtain M determination results,
15. The multi-input multi-output turbo according to claim 14, wherein the streams corresponding to the M determination results obtained for each subcarrier are parallel-serial converted to obtain the determination results for each stream. Receiving method.

Images