200813977 九、發明說明： 【發明所屬之技術領域】 本發明係涉及一種音樂伴奏裝置之自動跟 系統之創新設計者。 乃法以及 【先前技術】 下唱歌 音準， ’便造 的致果 士按一般人在有音樂伴奏（如伴唱機）的情形 日守’在往會因為伴奏音樂的音調過高或過低而失去 如此當演唱者的歌聲音調跟不上伴奏音樂之音調時 成歌聲與音樂聲不協調的情形’進而大幅影響歌唱 夠嚷立:伴^目前雖有相關業界針對上揭問題研發出能 :伴奏裝置隨著演唱者歌聲音調而改變伴奏音樂音 :二:；惟，其採用之技術特點，係必須測量演唱者 預定時間週期，復藉由運算獲取此時間週期之 】：：」，再透過該「平均音調」與一匹配伴奏音樂 疋* $考日调加以比齡以接祉 樂之音調配訊號，進以轉變伴奏音 於_曰^ ’此種習知音樂伴奏之自動跟調方法，i對 瞀而猫π 調之计π疋取一段時間週期之音調加以計 斤獲仔一平均值（即平均音 θ 甘—a 個平均值的時間點門f ①、、母:人獲得- L占間隔（如5秒）已然造成比對上產生明 二;遲的情形，若再加上計算、比對所需的時間，則延遲 的障形將更加明顯，田 法於實際應用姑驗中發現’：音樂伴奏之自動跟調方 、，二鲦中發現，其改變伴奏音樂音調以符合演 5 200813977 =者音調的過程並無法達到較佳之即時性，往往必須等待 决唱者唱完—句歌詞之後，於下-句歌詞進行時才會達成 音樂音：之轉變；且上揭習知方法係為兩個固定點數值的 比對’貫難以獲得較高準確性之移調值，“匕顯然不足因 應使用者最期盼的需求’實有再加以改善突破之空間存在 〇 是以，針對上述習知音樂伴奏裝置之自動跟調方法使 用上所存在之問題點，如何 蚀处私s _ U — I出一種能夠更具理想實用 性之創新構造，實有待相關章界 Μ月灯仰關糸介再加以、思索突破之目標及 方向者。 有鑑於此，發明人本於多年從事相關產品之製造開發 與設計經驗，針對上述之目#，詳加設計與審慎評估後， 終得一確具實用性之本發明。 【發明内容】 本發明之主要目的，係在提供一種音樂伴奏裝置之自 動跟調方法’令人們唱歌時，能根據演唱者的音調，即時 且自動地調整伴奏歌曲的音調，以減少歌聲失去音準的機 會，並適合自己的歌唱的音域；其技術特點包括：提供一 音調估測器，以將使用I的聲音每隔一段預定時間計算盆 基本頻率（Fundamental Frequency);在一音階序列記錄哭中了 將使用者聲音的基本頻率轉為使用者音階參數序列了並記 錄之；藉-音階匹配器以㈣使用者音階參數序列盘主: 律音階序列之差異性；復利用移調判斷器決定是否須移= 及何時移調，若需移調，則自動調整音樂合成器中之音階 6 200813977 麥數’最後音樂合m將讀位記狀樂譜合成實際的 音樂波形二然後與使用者歌聲-起經混音器(mixer)、：叭 輸出者；藉此創新獨特設計，使用者之歌聲音調計算方式 係採用每-小段時間計算—次之方式，復藉該音階:列^ 錄器將使用者聲音的基本頻率料使用者音階參數序列； 亦即’本發明係、比對主旋律音階序列及使用㈣聲音階序 列，所述音階匹配器係比對一段音階序列的匹配程度:屬 於動態比對音階曲線之模式，並輸出其最佳匹配時的音階 差，因為該音階匹配器是動態比對一段時間的音階序= 線，故所得之移調值具有較高之準確性，能夠獲得最佳之 音調調整效果，更符合使用者之所需。 又，本發明之技術特點係直接取得已記錄歌曲之主旋 律，並無需複雜的運算過程，具有較低之系統運算量，佔 用較低的系統資源，故整體具有較高之運作效率和即時性 ’可大幅改善習知系統延遲之問題點，達到更加實用好用 之進步性者。 【實施方式】 請參閱第1 、2、3圖所示，係本發明音樂伴奏裝置 之自動跟調方法之較佳實施例，惟此等實施例僅供說明之 用，在專利申請上並不受此結構之限制；所述自動跟調方 法係包括： 如第1圖所示，首先將使用者的聲音每隔一小段時間 (約0.1秒），經由一音調估測器11計算其基本頻率（Fund 200813977 amental Frequency)，該音調估測器u乃是計算此段聲音之基 本週期或頻率，它可由計瞀异去倍 A 干匕」田<斤取大值之自相關函數（如仂⑺汀 e⑽1〇n function)獲得或在頻譜上由I值（驗㈣之相關 位置及間距而得到，基本週期與頻率的關係是： 基本頻率=取樣頻率/基本週期 該取樣頻率是聲音每秒取樣的樣本點數，接著在一音 階序列記錄器12中，將一連串使用去* & 疋甲仗用考輸入聲音的基本頻率 轉為使用者音階序列121 ，並加以圮钸，立帆咖此方^ 儿刀μ 口匕綠，音階與頻率的關 係如下： 音階Α4的頻率是440 Ηζ，每升高半個音階頻率升高 %倍，每降低半個音階頻率降低％倍，所以每升高12 個音階頻率共升高2倍，每降低12個音階頻率亦降低7倍 ，然後，透過一音階匹配器13計算比對使用者音階序列i 21與主旋律音階序列14之差異性，在此，該主旋律音階序 列14是預先儲存於音樂的樂譜15中，例如··在midi (唧 al instmment digital interface)檔案中即可同時記錄這些樂 譜的資訊，該音階匹配器13則是利用類似動態時間扭曲（ Dynamic Time Warping)或校準的方式，比對使用者音階序列 與主旋律音階序列14的差異性，其具體作法如丁： 假4:使用者音階序列121為nl，n2, ···，nj，代表每隔固 定一小段時間（如〇· 1秒）所連續計算出的使用者（或歌 者）的音階（或音調），並假設主旋律音階序列14為ml，m 2,…，mj，代表每隔固定一小段時間（如0.1秒）的主旋律 音階，在此音階是以數字1〜255表示，音階C3的代表數值為 60，音階D3的代表數值為61，音階B3的代表數值為59，以 8 200813977 此類推，因為在唱歌時，歌唱 與背景音樂播放的拍點位置相 做動態時間校準，以產生較正 者的聲音拍點位置不見得會 同’所以在音階比對時必須 確的比對結果’如下示意圖 nl n2 n3 n4 n5 n6200813977 IX. Description of the Invention: [Technical Field of the Invention] The present invention relates to an innovative designer of an automatic tracking system for a musical accompaniment apparatus. Naifa and [previous technique] sing the pitch, 'The result is that the average person is in the situation of having a music accompaniment (such as a phonograph). In the past, the pitch of the accompaniment music is too high or too low. When the singer's voice can't keep up with the pitch of the accompaniment music, the situation that the song and the music are not coordinated will further greatly affect the singing. The singer's voice is changed to change the accompaniment music sound: 2:; However, the technical characteristics of the singer's use must measure the singer's predetermined time period, and use the operation to obtain the time period::", and then through the "average Tone and a matching accompaniment music 疋* $ 考 调 调 调 调 调 调 调 调 调 调 调 调 调 调 调 调 调 调 调 调 调 调 调 调 调 调 调 调 调 调 调 调 调 调 调 调 调 调 调 调 调 调 调 调 调 调 调And the cat π 调 计 疋 疋 一段时间 一段时间 一段时间 一段时间 一段时间 一段时间 一段时间 一段时间 一段时间 一段时间 一段时间 一段时间 一段时间 一段时间 一段时间 一段时间 一段时间 一段时间 一段时间 一段时间 一段时间 一段时间 一段时间 一段时间 一段时间 一段时间 一段时间 一段时间 一段时间 一段时间 一段时间 一段时间 一段时间 一段时间 一段时间 一段时间 一段时间 一段时间 一段时间 一段时间 一段时间 一段时间 一段时间Such as 5 Seconds) already caused the second on the comparison; in the late case, if the time required for calculation and comparison is added, the delay of the delay will be more obvious, and Tian Fa finds in the actual application: music accompaniment The automatic follow-up, and found in the second singer, the process of changing the accompaniment music to meet the tone of the 5 200813977 = can not achieve better immediacy, often have to wait for the singer to sing the sentence - after the lyrics, under - When the lyrics are played, the music sound is changed: the change is based on the comparison of the two fixed point values. It is difficult to obtain a transposition value with higher accuracy. The expectation of demand 'has actually improved the space for breakthroughs. Therefore, the problem of using the automatic tune method of the above-mentioned conventional music accompaniment device is how to erode the private s _ U - I The innovative structure with more ideal and practicality needs to be related to the chapters, the moonlights, and the target and direction of the breakthrough. In view of this, the inventors have been engaged in related products for many years. The development and design experience, in view of the above-mentioned purpose #, detailed design and prudent evaluation, the invention has finally become practical. [Summary of the Invention] The main object of the present invention is to provide an automatic music accompaniment device. The method of adjusting the singer can adjust the pitch of the accompaniment song instantly and automatically according to the vocal tone of the singer, so as to reduce the chance that the vocal song loses the pitch and suit the vocal range of the singing; its technical features include: providing a tone The estimator calculates the fundamental frequency of the sound using the sound of I every predetermined period of time; in the sequence of recording a cry, the basic frequency of the user's voice is converted into a sequence of user scale parameters and recorded. The borrowing-scale matching device uses (4) user scale parameter sequence disc master: the difference of the rhythm sequence; the complex shifting determiner determines whether to shift = and when to transpose, if the transposition is required, the scale in the music synthesizer is automatically adjusted. 6 200813977 Mai number 'Last music combined with m will be read into the score sheet to synthesize the actual music waveform 2 and then with the user's voice - Through the mixer, the output of the speaker; with this innovative and unique design, the user's song tone calculation method is calculated every time - the second time, the scale is borrowed: the column player will be the user The basic frequency of the sound is the user's scale parameter sequence; that is, 'the invention system, the main melody scale sequence, and the (4) sound order sequence, the scale matcher matches the degree of the scale sequence: belonging to the dynamic alignment scale The mode of the curve, and output the scale difference of the best match, because the scale matcher dynamically compares the scale order = line for a period of time, so the resulting transposition value has higher accuracy and can obtain the best pitch. Adjust the effect to better suit the needs of the user. Moreover, the technical feature of the present invention directly obtains the main melody of the recorded song, does not require a complicated operation process, has a low system operation amount, and occupies low system resources, so the overall operation efficiency and immediacy are high. It can greatly improve the problem of the delay of the conventional system and achieve more practical and useful progress. [Embodiment] Please refer to Figures 1, 2, and 3, which are preferred embodiments of the automatic tracking method of the musical accompaniment apparatus of the present invention, but the embodiments are for illustrative purposes only, and are not used in patent applications. The automatic tracking method includes: as shown in FIG. 1, the user's voice is first calculated by a tone estimator 11 every short period of time (about 0.1 second). (Fund 200813977 amental Frequency), the tone estimator u is the basic period or frequency for calculating the sound of this segment, which can be used to calculate the autocorrelation function of the large value. (7) Ting e(10)1〇n function) is obtained or spectrally obtained from the I value (test (4) related position and spacing. The relationship between the fundamental period and frequency is: Basic frequency = sampling frequency / basic period The sampling frequency is sound sampling per second. The number of sample points is then converted into a user scale sequence 121 by a series of using the frequency of the input sounds in a scale sequence recorder 12, and then smashed. ^ knife knife mouth匕 Green, the relationship between scale and frequency is as follows: The frequency of scale Α4 is 440 Ηζ, the frequency of each half of the scale is increased by a factor of 5%, and the frequency of each half of the scale is reduced by a factor of 5%, so the height of each scale is increased by 12 scales. 2 times higher, the frequency is reduced by 7 times for every 12 scales, and then the difference between the user scale sequence i 21 and the main melody scale sequence 14 is calculated by a scale matching unit 13, where the main melody scale sequence 14 is Pre-stored in the music score 15 of the music, for example, the information of the scores can be simultaneously recorded in the midi (唧al instmment digital interface) file, and the scale matcher 13 utilizes similar dynamic time warping ( The method of calibration compares the difference between the user scale sequence and the main melody scale sequence 14, and the specific method is as follows: false 4: the user scale sequence 121 is nl, n2, ···, nj, representing a fixed segment every time. The scale (or pitch) of the user (or singer) that is continuously calculated for the time (eg 〇·1 second), and assumes that the main melody scale sequence 14 is ml, m 2,...,mj, representing every fixed one The main melody scale for a short period of time (such as 0.1 second), the scale is represented by the numbers 1~255, the representative value of the scale C3 is 60, the representative value of the scale D3 is 61, and the representative value of the scale B3 is 59, to 8 200813977 By analogy, because when singing, the singing and the background music play the position of the beat to do dynamic time calibration, in order to produce the correcter's sound beat position does not necessarily be the same 'so the comparison result must be confirmed when the scale is compared' Nl n2 n3 n4 n5 n6
ml m2 m3 m4 m5 m6Ml m2 m3 m4 m5 m6
上揭貫施例中，就時間的角度而言，該n2、n3 (即使 用者音階序列）係隨著m2 (即主旋律音階序列）作校準的 狀態，以使背景音樂播放的拍點位置盥歌哐去沾& + 、艾曰有的聲音拍點 位置能夠在正確相對應之拍點位置作匹· %比對，而進行移 調時’則是令主旋律音階序列隨著使用者音階序列作移調 的動作。 9 200813977 假5又dist(ni，mk)表示音階ni，mk的誤差值，acu_dist(ni ，mk)表示過去最佳路徑到音階ni，mk的累積誤差值，則上圖 匹配之每個節點的最小累積誤差為： acu_dist(ni,mk) = dist(ni,mk) + min{ acujist(ni-l,mk), acu 一dist(ni，mk-1)，acu—dist(ni-1，mk-1)，···} 其中，min{···}表示取最小值，在{···}中的範圍則由經 驗決定，通常取時間校準值—2〜+2的範圍内，因此最後 匹配的結果誤差為acu—dist(nj，mj·)，]·為此次比對的最後一 個時間點，其大小由實驗決定通常大於40(4秒）小於100( 10秒）而其中隶佳路控是指累積誤差最小的路徑，實作 上並不需求出。 通常設K1二一 6 基於上述做法，我們可用以算出主旋律需移調多少， 如第2圖所示，首先設定主旋律音階移調值為s = ki ，$ j表不音階升高-個半音，s = _丨表示音符降低一個半 音，然後，利用上述做法比對使用者音階序列及移調s後 的主旋律音階序列，並記錄其最後匹配結果的累積誤差值 •s(s)，接著令s=s+1，再次計算Dis(s)直到㈣2，最後找 出移調值s = Smin使得Dis (Smn)是最小值，其中Ki<= s〈嘟 K2 接著，再藉由一移調判斷器16決定是否須移調及何 移調，該移調判斷器16是根據誤差值Dis(Smn)小於某—絰^ 值常數D時才做移調處理，移調處理是將主旋律音符撿 二個半音，在調整時，並應考慮音樂的和諧及自然，4 = :隔-段時間調整一次，且通常在主旋律音長較長時：： 10 200813977 -其1，該音樂合成器17是將以數位記錄的毕级15人成 貫際的曰樂波形’最後與使用者的歌聲一二二 18(mixer)輪出，若雲敕 此曰為 …且： 周”則調整音樂合成器17中之音 二體做法係可將樂譜15中所有音符音 咼或降低幾個音階數，此 升 车立、， 吓4曰F白數通常小於或等於6個 曰，仁亚不限定’因為相差12個半音(8度音）等於 剛好相差兩倍，頻率差兩彳立對 、，… 古 貝手差兩乜對音感而言似乎是相同的，而 鬲於6個半音以上可以用降含來 川|牛㊈术處理，低於6個丰 則可用升調來處理。 以下茲提出一實作範例： 、當播放背景音樂時’即開始進行錄音，並設定聲音格 式為單音16 bits ’取樣頻率為麵Hz，每次錄製^度 :〇.1秒’接著在步驟音調估測器^中，計算歌唱者的^ 音基本頻率，其方法如下，假設所錄製的聲音為： I· 計真自相關函數/•/尤），其中： n(k) = Σ n x(n)x(n-k)，n=0，l，2,…，N - 1 ， k=22, 23, Η …，674 或々值的範圍代表所欲谓測的頻率範圍： 44100/22〜44100/674 = 2004.54 〜65·43Ηζ 尋我Ls^argOnaxCnOO)h)，k·代表鹜n(k)有i欠饭碍的 女值。 基本頻率 然後，將該基本頻率轉換為 音卩皆代號，假設若基本頻率=440 Hz則將之換為音产匕 A4(中音La)，音階代號為69，每差半音，頻率差 2. 200813977 ^倍，音符代號差1 ，由音階序列記錄器12記錄之 ，並在主旋律音階序列14記錄主旋律音階代號，在音 階匹配器13中，首先設定K1二一6 ，K2 = 6 ，妙1 π 定其音階代號序列長度為4秒（j=40)，每錄音〇丨秒 計算一次，4秒共40次，假設記錄主旋律音階序列14 為mi，i = 0，1，2，…，39，使用者歌聲音階序列為& i = 〇，1，2，…，39，移調為s，並假設音階代號mi nk 的差值為 dist (mi, nk)，令其中 dist (mi，他）>=〇， 且令當mi，nk相差一個八度音（12個半音）時，其古十 算誤差值結果相等，即： dist (mi, nk) = dist (mi+12*N, nk), 其中N為整數，並設時間校準值-1〜+〇的範圍内 ，則音階匹配器13之做法如下， 1. 設定s=Kl 2. 設定，且設定累積誤差值陣列 acu—dist[0〜39][0〜39] 初值為一個非常大的值 1000000 3. 計算 acu_dist [0][0]= dist (mO+s，n0) 4. 設定 j = i 一 1 5. 若 j >= 40則跳到步驟8 6. acu一dist [i][j] = min{ acu—dist [i-l][j—1]， acu一dist [i-l][j]， acu一dist [i][j-1]} + dist (mi+s， nj) 7. j = j +1若 j O i +1 則回到步驟5 8. i = i +1 若 i < 40 則回到步驟4 9. Dis(s)二 dtw 一dist[39][39] 12 200813977 則回到步驟 10· S = S+1 11·若 s〈二 K2 12. 結束。 然後’在移調判斷器16中，若Dis㈤<=4〇(4〇係為經 驗值），i目前播放之主旋律音符音長〉=1秒時，則將主 旋律移調^個半音並間隔4秒（4秒係為經驗值）以上，才 進仃下-次之移調；最後音樂合成器17是將以數位記錄的 ㈣合成實㈣音樂波形，然後與使用者歌聲—起經混音 杰18(mixer)、喇口八19輸出者。 【本發明之優點】 [、本發明所揭音樂伴奏裝置之自動跟調方法，其使用者 之歌聲音調計算方式並非取其平均值，而是採用每— :段時間計算-次（如0-1秒）之方式，復藉由所述 “皆序列記錄器12將使用者聲音的基本頻率轉為使用 者音階參數序列121 ·女日η 1 1Z1 ，亦即，本發明係比對主旋律立 階序列14及使用者歌聲音階序列m ,而非比對盆平 均音調，所述之音階匹配器13是比對-段音階序列的 匹配程度，此屬於一種動態比對音階曲線之模式’並 輸出其取佳匹配時的音階差，因為該音階匹配器 動態比對-段時間的音階序列曲線，而非僅—段時門 2音調平均值’故所得之移調值具有較高之準確性， 此夠獲侍取佳之音調調整效果，進而更符合使 所需。 、 2 本發明之技術特點 …、係直接取侍已記錄歌曲之主旋律， 13 200813977 並不需要一堆複雜的運算過程，具有較低之系統運算 里，佔用較低的系統資源，故整體具有較高之運作效 率和即時性，確可大幅改善習知系統延遲之問題點， 達到更加實用好用之進步性者。 f 上述實施例所揭示者係藉以具體說明本發明，且文中 雖透過特定的術語進行說明，#不“此限定本發明之專 利犯圍m此項技術領域之人士t可在瞭解本發明之精 神與原則後對其進行變更與修改而達到等效之目的，而此 等變更與修改，皆應涵蓋於如后所述 定範疇中。 疋之申清專利範圍所界 14 200813977 【圖式簡單說明】 第1圖：係本發明音樂伴 塊圖。 第2圖：係本發明之音階 【主要元件符號說明】 音調估測器 音階序列記錄器 使用者音階序列 音階匹配器 主旋律音階序列 樂譜 移調判斷器 音樂合成器 混音器 口刺口八 置之自動跟調方法之系統方 器動作流程之方塊示意圖。 15In the above embodiment, in terms of time, the n2 and n3 (ie, the user scale sequence) are in a state of being calibrated with m2 (ie, the main melody scale sequence) so that the position of the background music is played. The songs are smeared & +, Ai's voice position can be matched at the correct corresponding position, and when transposed, the main melody sequence is made with the user's scale sequence. Transposed action. 9 200813977 False 5 and dist(ni,mk) represent the error value of the scale ni,mk, acu_dist(ni ,mk) represents the cumulative error value of the past best path to the scale ni,mk, then the upper graph matches each node's The minimum cumulative error is: acu_dist(ni,mk) = dist(ni,mk) + min{ acujist(ni-l,mk), acu a dist(ni,mk-1),acu-dist(ni-1,mk -1),···} where min{···} indicates the minimum value, and the range in {···} is determined by experience, usually taking the time calibration value—2 to +2, so The final matching result error is acu_dist(nj,mj·),]· is the last time point of the comparison, and its size is determined by experiments, usually greater than 40 (4 seconds) and less than 100 (10 seconds). Good road control refers to the path with the smallest accumulated error, which is not required in practice. Usually set K1 2:6 Based on the above method, we can use to calculate how much the main melody needs to be transferred. As shown in Fig. 2, first set the main melody scale shift value to s = ki, and $ j to increase the scale - semitone, s = _丨 indicates that the note is reduced by one semitone. Then, the above method is used to compare the user's scale sequence and the main melody scale sequence after transposition s, and record the cumulative error value of the last matching result • s(s), and then let s=s+ 1. Calculate Dis(s) again until (4) 2, and finally find the transposition value s = Smin such that Dis (Smn) is the minimum value, where Ki <= s < toss K2. Then, by a transposition determiner 16, it is decided whether to transpose And transposition, the transposition determiner 16 performs the transposition processing according to the error value Dis(Smn) being less than a certain value - D value constant D, and the transposition processing is to modulate the main melody note by two semitones, and when adjusting, the music should be considered. Harmony and nature, 4 = : interval-time adjustment, and usually when the main melody has a long length:: 10 200813977 -1, the music synthesizer 17 is a 15th grader who will record in digital format The music waveform 'final with the user's The song is one or two of the 18 (mixer) rounds, if the cloud is this... and: "week" adjusts the sound of the music synthesizer 17 in the two-body practice can be used to reduce all the notes in the score 15 or reduce the number of scales This rises the car, and scares 4曰F white number is usually less than or equal to 6 曰, Renya is not limited to 'because the difference of 12 semitones (8 octaves) is equal to exactly twice the difference, the frequency difference is two pairs, ... Gubei hand difference seems to be the same for the sound, and 鬲 6 6 6 6 6 6 6 6 6 6 6 6 6 6 6 6 6 6 6 6 6 6 6 6 6 6 6 6 6 6 6 6 6 6 6 6 6 6 6 6 6 6 6 6 6 6 6 6 6 6 6 6 6 6 6 6 6 6 6 6 6 6 6 6 6 6 6 6 6 6 6 6 6 6 6 6 6 A practical example: When playing background music, 'start recording, and set the sound format to single tone 16 bits' sampling frequency is face Hz, each recording ^ degree: 〇.1 second' then estimate the pitch in steps In the device ^, calculate the basic frequency of the vocalist's sound, as follows, assuming that the recorded sound is: I· Measure the true autocorrelation function /•/ especially), where: n(k) = Σ nx(n)x (nk), n=0,l,2,...,N - 1 , k=22, 23, Η ..., 674 or the range of 々 values represent the frequency of the desired measurement Rate range: 44100/22~44100/674 = 2004.54 ~65·43Ηζ Find me Ls^argOnaxCnOO)h), k· stands for 鹜n(k) has the female value of i owe the rice. Basic frequency Then, the basic frequency Convert to the sound 卩 code, suppose if the basic frequency = 440 Hz, then change it to sound 匕 A4 (middle La), the scale code is 69, each semitone, frequency difference 2. 200813977 ^ times, note code difference 1 Recorded by the scale sequence recorder 12, and record the main melody scale code in the main melody scale sequence 14, in the scale matcher 13, first set K1 two-six, K2 = 6, and the magic 1 π determines the length of the scale code sequence to be 4 Seconds (j=40), calculated once per recording leap second, 4 times in 4 seconds, assuming that the main melody scale sequence 14 is recorded as mi, i = 0, 1, 2, ..., 39, the user's song sequence is & i = 〇,1,2,...,39, transpose to s, and assume that the difference between the scale code mi nk is dist (mi, nk), let dist (mi, he)>=〇, and let mi When nk differs by one octave (12 semitones), the error value of the ancient ten arithmetic is equal, namely: dist (mi, nk) = dist (mi+12*N, nk), When N is an integer and the time calibration value is within the range of -1 to +〇, the scale matching unit 13 is as follows: 1. Set s=Kl 2. Set, and set the cumulative error value array acu_dist[0~ 39][0~39] The initial value is a very large value of 1000000 3. Calculate acu_dist [0][0]= dist (mO+s,n0) 4. Set j = i to 1 5. If j >= 40 then jump to step 8 6. acu-dist [i][j] = min{ acu-dist [il][j-1], acu-dist [il][j], acu-dist [i][j -1]} + dist (mi+s, nj) 7. j = j +1 If j O i +1 then go back to step 5. 8. i = i +1 If i < 40 then go back to step 4. Dis(s) two dtw a dist[39][39] 12 200813977 Then go back to step 10 · S = S+1 11 · If s < two K2 12. End. Then, in the transposition judger 16, if Dis(5)<=4〇(4〇 is the empirical value), i currently plays the main melody note length > =1 seconds, then the main melody is shifted by a semitone and separated by 4 seconds ( 4 seconds is the experience value) above, only to enter the next-second shift; finally the music synthesizer 17 will be recorded in digital (four) synthetic real (four) music waveform, and then with the user's voice - from the mix of Jie 18 (mixer ), Lakou eight 19 exporters. [Advantages of the Invention] [The automatic tune adjustment method of the music accompaniment device disclosed in the present invention, the calculation mode of the user's song voice tone is not taken as an average value, but is calculated every time - time (eg 0- In the manner of 1 second), the basic sequence recorder 12 converts the basic frequency of the user's voice into the user scale parameter sequence 121 · female day η 1 1Z1, that is, the present invention compares the main melody The sequence 14 and the user's vocal order sequence m, rather than the matching pot average pitch, the scale matching unit 13 is the matching degree of the aligning-segment scale sequence, which belongs to a dynamic aligning scale pattern 'and outputs it Take the scale difference when the match is good, because the scale matcher dynamically compares the scale sequence curve of the time-segment time, not just the interval value of the gate 2, so the transposition value obtained has higher accuracy, which is enough Get the best tone adjustment effect, which is more in line with the needs. 2, the technical characteristics of the invention..., directly to the main theme of the recorded song, 13 200813977 does not require a complicated process, with In the lower system operation, it occupies lower system resources, so the overall operation efficiency and immediacy of the whole system can greatly improve the problem of the delay of the conventional system, and achieve more practical and useful progress. The present invention has been specifically described by the embodiments, and is described by specific terms in the text, which is not to be construed as limiting the spirit and principles of the present invention. Changes and modifications are made to achieve equivalent purposes, and such changes and modifications shall be covered in the scope as described later.疋 申 申 申 申 申 14 14 14 14 14 14 14 14 14 14 14 14 14 14 14 14 14 14 14 14 14 14 14 14 14 14 14 14 14 14 14 14 14 14 14 14 14 14 14 14 14 14 14 14 14 14 14 14 14 14 14 14 Figure 2: The scale of the present invention [main component symbol description] Tone estimator scale sequence recorder user scale sequence scale scale matcher main melody scale sequence music shift tone determiner music synthesizer mixer mouth punctual eight automatic A block diagram of the system block action flow of the method. 15