TWI279693B - Method and device of audio compression - Google Patents

Abstract

The present invention discloses a method and device of audio compression. The present invention utilizes wavelet transform technology to transform audio signal into two folds of sub-band via wavelet function: the low frequency sub-band which represented the important part of audio signal, and the high frequency sub-band which represented the low sensitivity part of audio signal for human hearing. The present invention creates a code block, which includes several sub-bands of the one dimensional wavelet audio signal, then processes the code block through EBCOT and arithmetic coding. The present invention can provide a real time and high compression rate audio compression.

Description

1279693 IX. Description of the Invention: [Technical Field] The present invention relates to an audio signal compression method, and relates to an audio signal compression method using a one-dimensional discrete wavelet transform technique. [Prior Art] The audio is off, most of it is done after Psychoacoustic analysis on _(F Qing Donmn). By encoding the less important high-scales in the frequency domain with the remainder of the round, and

The important low-frequency part is encoded at a high bit rate, and the ratio #0 of the phase # is obtained, and the most famous one is that the audio signal compression method is MPEG丨L III, that is, MP3. MP3's compression ratio can be up to 1 times, but it still retains the sound quality close to the original sound. Therefore, Mp3 related products are currently popular among consumers. ~ However, the general frequency domain audio signal compression technology must use the Fourier Transform to convert the Time D〇main audio signal to #,, after the frequency signal, and then analyze and subsequent compression. Therefore, a considerable amount of difference is required. The compression engine hardware made by this technology, when T 〇 ^ MP3, takes a period of time domain audio signal, so the compression of Mp3 will generate the Latency Delay according to each system. To achieve near real time compression, you need to use a higher operating frequency to increase power consumption. / In order to achieve high compression ratio and low computational complexity, another technique for frequency domain conversion, Wavelet Transform, is a technique that can be employed. By converting the audio signal (Wavelet function), the audio signal is converted, and the low frequency band signal (L〇w frequency sub-band) representing the important data of the audio signal is obtained, and the high frequency band which is less sensitive to the human ear is obtained. Signal

(High frequency sub-band). The above compression method uses the characteristics of the DWT to perform instant audio signal compression. 6 1279693 Μ热新二ΐ applies the wavelet transform technology to the processing of audio signals. 'There is still a lot of room for development and hard work.' I have not seen any use of wavelet transform technology to process audio signals. 1= So this is a technical field that can be developed. SUMMARY OF THE INVENTION The present invention provides an audio signal compression method: hunting for instant compression and high compression ratio. There are 7 purposes for Ming, which can be applied to the current unstable network bandwidth by providing the audio signal compression side p phase ♦ movable phase _ rate. The purpose of the present invention is to provide an N-order one-dimensional discrete wavelet transform by performing the following steps: performing the following steps: performing the following steps: performing the N-order one-dimensional discrete wavelet transform on the _贞 signal, and performing the following steps: Γί元难编斗············································· Perform arithmetic coding: will. The collapsed encoded data is subjected to arithmetic coding; and the output bit method is equipped with an audio compression method as described, and the present invention further provides an audio decompression library, '+, ti the following steps: input bit stream: The audio signal compressed by the first input/frequency compression method of the patent application scope is decoded by bit stream method; the bit compression decoding is performed. · The unit time (1:) is decoded by the unit. The bit compression decoding is performed in a code block manner; the Ν 化, the inverse one-dimensional discrete wavelet transform is performed; and the signal is restored. The shrink/decompression method is embodied in an audio signal buffer memory for storing one unit time (t) of the input by the outside, a two-frequency nickname, one-dimensional discrete wavelet transform or one-dimensional inverse discrete Wavelet-converted chirped frequency signal; 1279693 one-dimensional discrete wavelet converter, interacting with the buffer memory to complete N-dimensional one-dimensional discrete wavelet transform, or N-order one-dimensional inverse discrete wavelet transform; The unit, after the n-times one-dimensional discrete wavelet transform audio signal is formed into a code block, is coded in a bit-plane manner, or is decoded in an inverse bit-plane manner; a compression unit that performs arithmetic compression on the audio channel encoded by the bit plane method, or arithmetically decompresses the compressed audio signal; and a first in first out unit (FIFO) The compressed audio signal is output as a bit το stream, or a bit stream of the compressed audio signal is streamed to the arithmetic compression unit for decompression. In addition, the present invention further provides an audio signal compression device and an audio signal decompression device separately. The above and other objects, features, and advantages of the present invention will become more apparent from the aspects of the appended claims. Wavelet transform (10)) provides the secret (4) (four)

Bit collapse coding and nasal coding work. So, between. The degree of similarity is conducive to the subsequent work. Therefore, the compression ratio can be increased. In addition, since the audio (Audio) signal is a one-dimensional signal

1279693 'Embedded Block Coding with Optimized Truncation' compression technology' because this compression technology has good compression efficiency and quality for wavelet-transformed data. At the same time, the audio signal can be adjusted (scalability). ), random access and other advantages. In particular, the EBCOT technology employed in the present invention uses only the bit compression coding and arithmetic coding portions of the EBC0T technology, and uses a novel algorithm developed to achieve adaptive effects. 々 First, please refer to FIG. 2, the first embodiment of the audio signal compression method of the present invention includes the following steps: sampling (step 11〇); performing N-order call-dimensional discrete wavelet transform (step 120); Quantization is performed (step 130); bit compression coding is performed (step 140); arithmetic coding is performed (step 15A); bit stream is rotated (step 160). After the audio number is input, the audio signal is sampled at a fixed sampling rate (step 110), for example, 44·1 kHz, and the audio signal data is acquired each time. Then, the audio signal in the unit time (t) after sampling can be reversed into the 'one-dimensional discrete wavelet transform, and N times are performed to obtain N+1 sub-bands, which is to perform N-order one-dimensional discrete wavelet transform. (Step 12〇) steps. Next, since the audio information contained in each sub-band is different in importance, the audio signals of all the sub-frequency are individually quantized (step 130). This quantization is scalar quantization, • The case only takes the 16th high bit, 12 high bits, or 8 high bits of the audio signal. Then, all the quantized sub-band audio signals are bit-compress-encoded (step 140). 'The audio signal of the μ-pen is formed into a code block and then bit-encoded; and also, the Μ pen contains all sub-band data. After the data packet (audio signal in one unit time) is composed of a code block (c〇de-block), the code block is divided into different fL element planes, and the bit plane is used as a compression process. Object. Next, 'Arithmetic Coding (AC) is performed (step 150), for example, 'Using 姒-encoder. Finally, the bit stream is output (step, 160), that is, the compressed data completed by the arithmetic coding is output, that is, / all the compression procedures are completed. 9 1279693 The feature of this is that the multiple sub-bands of the audio signal 1D signal are combined into one code block, which is different from the JPEG 2000 method of finely cutting each DWT sub-band into a compressed code block for image compression. In the decompression of the data, the reverse procedure may be performed, such as "the third figure, which includes the following steps: inputting a bit stream (step 210); and decoding (step 220); Perform bit compression decoding (step 23 〇); enter = ri 匕 i step 24 〇); perform an N-th order inverse one-dimensional discrete wavelet transform (step 250), restore the signal (step 260). The string & generated by the compression method of the present invention is input. In step 22, the compression is performed, and the arithmetic decoding is performed. Then, according to the encoding, the compression decoding operation is performed (step 23). Then, according to the original, the grading level, dequantization is performed (step 240). Finally, the inverse-dimensional discrete wavelet transform is performed (step 25), and after the second = material plus (original) , can be output as the original audio = after the change of the beggar in the different (four) seeking, for example, the system bandwidth size is different or yellow in the sound quality requirements, the invention can be different, not compact, J, . ), the quantization level is chest ts, two sounds 22 ^ (64kb ^ · ^ ° ^ storage audio signal storage When the reading is small or the bandwidth is small, the audio signal is compressed at a high sag rate (for example, 128kpbs or 64kbps). The method of squeezing the sound of the seeding, the compression of the seeding horn, the 眚浐 丨 人 人 人 人 人The audio signal of the moon is the second method and the step 320), the input _ _, the sampling 4 the step wm · the second wide, the Shi Wei off the wavelet transform (step 330); the quantization (step 340), the view The reduction rate Α performs bit dust reduction coding (step 里 = 1279693 operative code (step 360); output bit stream (step 370). First, the user selects compression ratio A (step 310), the compression ratio The option is set in advance, as described above. Then, the audio signal is sampled according to a fixed sampling rate (step 320). After sampling, the sampled audio signal is repeatedly subjected to N-order one-dimensional discrete wavelet transform ( Step 330), to obtain the N+1^ frequency band. Next, quantization is performed (step 340), and the quantization level can be unequal to 16 high bits, 12 咼 bits, 8 high bits, etc. Then, according to the compression ratio Α Meta-compression coding (step 350), this step is based on the EBCOT adjustable characteristics, detailed The method will be described in "Figure 10." Finally, the tube code is executed (step 360), and the bit stream is output (step 37). On, in addition, for audio signals, different sampling rates can be used. The purpose of the data volume of the grandfather is low. That is, the present invention can also reduce the amount of data by selecting the sample rate. For example, 取样 441] Ηζ 22, 22 _ ζ 2 2 11 kHz, 8 kHz ... sampling rate, will be different The data density is the same as that of the sampler. The performance of the sound quality is better. You will have a good afternoon, and the embodiment of the "2" ' will turn it into a dusty one. Nothing. It includes the following steps: input=streaming (step 410); performing arithmetic decoding (step 42〇) performing bit pressing code (step side); performing forging (step _ heart is 1 ^ dimensional dispersion), wave Conversion (step 450); restore signal (step 働f. 朽: L1410' is to input the stream generated by the compression method of the present invention, and the step is to compress the A and the (4) decoding. Next, the compression side is adapted according to the rate A used in the encoding (step 430). Then, the 摅 摅 止 还原 曰 • • • • • • • • • ( ( ( ( ( ( ( ( ( ( ( ( ( ) 'Easy to output ί adjustable compression ratio' must have a corresponding: #料压缩技术. As can be seen from the method of No. 4 1279693, the present invention achieves compression ratio adjustment through bit compression coding techniques. Compression coding is the main technique in EBCOT, which compresses each sub-band with a bit-piane. However, the present invention is slightly different from the EBCOT method for image compression. Become the innovation of the present invention. It is used for image compression processing. EBCOT ' + system separates the sub-bands of each two-dimensional matrix, and the audio-data is one-dimensional (10) data. Therefore, there is no so-called code block when EBC 〇T processes image poor materials ( C〇de-block) data. In order to use the EBCOT compression method to compress the audio channel data of the M_ segment processed by the DWT-converted audio step and divide all the sub-rules into a code block. Then, I will do EBC〇T processing together. The choice of eight depends on the hardware requirements. Therefore, the present invention can convert a data into two-dimensional DWT audio signal data. For example, one sub-band (4L) , 5H among 3H, 3H, 2H, 1H), 2H, 1H), DWT data for M time periods are E_ together

The j's 4 achieve a higher compression ratio, and the part of the ιη can be directly deleted, and the data of the five sub-bands is Γ4丨仙叫1

The data is done together with _τ _ material (L, 4H, 3H, 2H, 1H), DWT shrinkage of M time periods. The reason why the 1H data can be deleted directly is that its direct deletion does not cause too much impact on the sound quality. In the case of marriage and wealth - the limit is limited. Below, we will directly remove the 1H example, and compress/decompress the method. "Ding Shui" brother Ming Ming's invention

Audio signal (4) 16-bit quantized digital asset exhibition sentence is the bit plane of Wei (4) of DWT "Fig. 7" (bi==, which becomes 12 1279693 EBCOT is based on the bit plane as the basis of compression processing, this The invention uses the same processing technique. However, in order to achieve the scalability of the scalability, the present invention adopts a method which is somewhat inferior to that of JPEG2000 for image processing. ', EBC0T encoding Japanese order' is in the mother The most significant bit-plane (MSB) of the bit-plane in the block (c〇de-bl〇ck) is coded. For example, in Figure 7, the 16-bit is used. In the code block data 30 composed of the quantized data, the first bit plane is the symbol bit-plane, followed by the MSB, and the last bit plane is the least significant.

Least significant bit-plane (LSB). The encoding of EBC0T is encoded from the MSB to the LSB's and searches for the first significant ("^gnificant" value with 1 to start encoding 2 to 7th from this significant value." The bit plane starts at 3〇_8, which means that there is the first significant value to live. Since the bit plane before the 丨 is not present, its value is zero. Therefore, y is a dragon whose early - (four) represents its element plane, which can greatly increase the compression ratio. Based on this, Ben Fen further developed the characteristics of the audio signal to dynamically select the compression ratio. The present invention utilizes audio signal = color - day 1 when the weak signal will be very inconspicuous, and the human ear is not easy. 'Audio 峨 will be directed at _ in a certain number of bits m is 'as long as The necessary information gathered by the audio signal is preserved, and the unnecessary information is regarded as a flaw, and the purpose of effective compression can be achieved. Private, ίΐίΐ method: from the first opening of a bit plane with a value of 1 (four), and according to the compression ratio of the location, the audio signal (4L, silver Lin Ziqing takes ηι 灭 留 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 The bit opening, n, n, n4 are the bit compression processing of the four processing bit numbers, and the rest 13 13279693 are processed in the 〇 mode. Thus, a good compression efficiency can be obtained. In case, the data distribution will probably be distributed nearby, in which case the number of processing bits larger than the first significant value with a significant value of 丨 will probably be smaller than the number of processing bits. In this case, the present invention directly The first bit plane having a significant value of 1 performs EBCOT compression. The present invention refers to this as an adaptive bit plane adjustment method. - Therefore, the value of the number of processed bits, although already specified in advance, is in the past After analysis, the first bit with a significant value of 1 is interpreted. In the plane, it can be judged whether the setting of the number of processing bits is used. Therefore, the method has an adaptive effect. The values of 111, 112, 113, 114 can be obtained experimentally. For example, the present invention takes one vertical in each of the four sub-bands 4L, 4H, 3H, and 2H.

The cut surfaces Cl, C2, C3, C4, and take (nl, n2, n3, n4> (5, 4, 3, 3). Since the audio signal of the sub-band of 4L has the most useful data amount, 5 processing bits are taken. Yuan '4 Η second, so take 4 processing bits, 3 3 3, 2 3 3. 1, η2, η3, η4), take (5, 4, 3, 3) to take, then EBCOT bit The meta-encoding coding method can still obtain a relatively high sound quality after being restored. Suddenly, the value of (111, 112, 113, 114) is selected according to the demand of the compression ratio. For example, a lower compression ratio may be (6, 5, 4, 4) or (7, 6, 5, 5). ··· Not waiting. y The compression effect of the present invention can be obtained from the following calculation results: assuming that the total amount of data after sampling is 1, the data amount of the original five sub-bands is 1/16, 1/16, 1/8' 1/4, 1 /2 'Because 1H is directly deleted, the sum of the processed data is 1/2. After the adaptive bit plane adjustment method is processed, taking (5, 4, 3, 3) as an example, the overall data volume can be reduced to 'low' to 1/16*5/15+1/16*4/15. +1/8*3/15+1/4*3/15=9/80, even lower, data 1 is greatly reduced. This amount of data is the result of no bit compression coding and arithmetic coding. That is, after bit compression coding and arithmetic editing, 'higher compression effect. For an explanation of the adaptive bit plane adjustment method, please refer to "8A" and 1279693 "8B". "8A" shows that the number of processing bits in the 4L and 4H sub-bands is (5, 4) ''8B'' shows the 3H and 2H sub-bands, and the number of parity bits is (3, 3). Since the 8th bit plane is the first time that a significant value of 1 occurs, after the significant value appears, the bit plane of each sub-band is subjected to bit compression processing according to (5, 4, 3, 3). In addition, in order to make the bit plane of the processing bit number work more smoothly, the present invention shifts the bit plane of the well-processed bit number as a whole to the LSB, and only records the bit number of the translation. . In this way, the efficiency of compression can be increased. As shown in _ "8C" and "8D", the present invention shifts the processing bit/meta plane originally processed from the 8th bit plane to the nth bit (C1), 12th. The bit plane (C2), the 13th bit plane (C3), and the 13th bit plane (C4) start. After panning, the values of other parts are processed in a complementary manner. The entire circuit only needs to record the value of the translation. The above describes how the present invention adjusts the compression ratio by adjusting the "number of processing bits" to achieve the purpose of dynamically adjusting the compression ratio. For the specific process, please refer to "FIG. 9A", which is a specific process of dynamically adjusting the compression ratio of the present invention, including the following steps: · reading the value of the compression ratio ^ step 610); defining the number of processing bits according to the compression ratio (Step 62); Perform bit processing (step 630). • First, the value of the compression ratio is read (step 610), which is defined in advance by the system and selected by the user. Then, the number of processing bits is defined according to the compression ratio (step 620). The value of the processing bit number is smaller than the lower compression ratio; conversely, the lower compression ratio, the value of the processing bit number is relatively higher. Finally, bit-plane coding is performed (step 630). Step 620, which further includes several sub-steps, such as "the 9th figure", searches for the most significant bit (step 621); the bit is processed according to the compression rate, the number is nL (step = 622); The processed bit is adjusted to the LSB (step 623). First, the first search brother dares to significant bits (step 621), and performs (four) flat Φ processing from the level of the most significant bit. Next, the number of bits nL is processed in accordance with the compression ratio to the 15 1279693 stator band (step 622), where L represents the period of synchronization processing. For example, the above four (5, 4, 3, 3), (6, 5, 4, 4), 7'6'5'5)... various combinations will form different compression ratio options. . Finally, the bit is adjusted to the LSB (step 623), and the reserved processing bit 7L is shifted to the LSB, and the portion that is not processed is added as 〇. For the process of solving ^, please refer to "9C", which consists of two steps: sub-plane decoding (step 71G), and the processed bit is adjusted from LSB to in-situ ^. After the reduction to the code block (4) e bi〇ck) The specific circuit device of the audio compression method of the capital invention, please refer to "1st 」"" Λ is the first part of the invention, the surface of the sound surface, the miscellaneous - the specific implementation = The sampling unit η, the sampling unit 12, the buffer unit 12, the wafer W are 70 、, and the nasal compression is performed therein, and the sampling unit is used to reduce the data in the unit time (t) to reduce the sampling. According to the sample data of p.lkHZ, the sampled data is set to _ζ as its sampling rate, that is, the sampling unit u can be reduced to 2 dimensions, and the sample stored in the buffer memory 13 is added with ί ίίί The θ frequency is increased by iii 2 with a reduction factor of the original fine reduction or the audio combination of the one-dimensional inverse discrete wavelet signal is subjected to N-dimensional one-dimensional discrete wavelet conversion, the 3 bis = scatter wavelet converter 14 and the buffer memory 13 mutual match

i material St i discrete small Lin change. After completing the money, Na will encode all the conversion, position, and flat code units 15. The coding method is coded in the bit-plane manner of Eg0T 1279693, so that the compression rate can be matched with the compression factor of the present invention. 16 perform the arithmetic editing of the mother's movement ^ finished material to the arithmetic _ unit out unit 17, the round out of the 4 Ma Yuancheng's poor material, then sent to the advanced first shrinking unit Π to calculate ^ first, the data bit 4 stream input count Meta coding unit. Xie Jingcheng's information is sent to the Shima way, and with the original * selection method, 'the same use of the EBCOT solution, and then through the one-dimensional discrete wavelet ^ cry ^ into the state of decoding. Connected to the body 13 common dimensional discrete wavelet converter 缓冲 and buffer memory sample 12 to increase the sample, the first 雑; Wei by the increase of the hair ϊϊ ΐ ΐ ,:, wavelet conversion 14 n conversion process, dir t = to be This information is added. Therefore, when the sampling is further increased ϊ 12 to do ^^ (4) action, the high frequency (4) is directly added to the Q ^ This ^ ' because the circuit area is directly proportional to the manufacturing cost, therefore, reducing the t area is for the programmer Challenge. In "Pic 10", the memory = 13 system is used to record the unit time to reduce the sampling 1 . Since the memory tends to occupy a considerable area of the circuit, the present invention is to reduce the capacity of the buffer memory 13. Please refer to FIG. 11 , which is a second embodiment of the audio signal compression/decompression of the present invention, including a reduction sampling unit 21, an incrementing unit 22, a buffer memory 23, and a first-dimensional dimension. The discrete wavelet transformer 24, the bit weight unit 25, the arithmetic compression unit 26, the first in first out unit (fif) 27, and the first to first dimensional discrete wavelet converter 28. Compared with "Fig. 10", it can be found that the present invention adds a second one dimension 17 1279693. The volume i of the buffer memory 23 is smaller than the buffer memory 13; iS: L = ri converter 28, And with M nri ^ ^ both 4 minus, the overall circuit area is substantially reduced, after the discarding, only the lower sub-frequency 7 'f, the sub-band of the sub-band 28 increased the circuit area '社于苐二One-Dimensional Discrete Wavelet Converter In addition, it is necessary to save memory in another memory space of the memory space as a buffer memory. That is to say, the buffer memory 23' of the "second picture" is in fact "the entire operation flow" "memory of the first and middle memories. The flow of the description, the picture, and the figure η" is the same, no longer fixed The discussion is not intended to limit the scope of the invention, and the relevant artisan, without departing from the spirit and scope of the present invention, shall be regarded as the finest decoration of the present invention. Prevail. ϊ ί = ί ΐ 五个 五个 经过 五个 五个 五个 五个 五个 五个 五个 五个 五个 五个 五个 五个 ; ; ; ; ; ; ; ; ; ; ; ; ; ; ; ; ; ; ; 五个 五个 五个 五个 五个 五个 五个 五个 五个 五个 五个 五个 五个 五个 五个 五个 五个 五个 五个 五个 五个The second embodiment of the audio service method of the present invention is a second embodiment of the audio signal decompression method of the present invention; = is a DWT audio signal for synchronously processing 4 time periods (M=4) Schematic diagram, the diagram is the bit plane of the code block (bi t-plane); 8A is the diagram of the 8th diagram, the number of processing bits of the 4H two subbands; 18 1279693 ί ί ϊ Λ 观 2H Schematic diagram of the number of processing bits in the two sub-bands; ϋ ® (4) The processing bits of the two sub-bands of the U after the translation are stunned, · Figure 9BB: The processed bits of the 3H, 2H and the two sub-fourths after the translation Fig. 9A is a specific flow of the dynamic adjustment compression ratio of the present invention. Fig. 9B is a process for defining the number of processing bits according to the compression ratio [sub-two-way: ^ 9C diagram is the decompression of the dynamic adjustment compression ratio of the present invention. Fig. 10 is the first embodiment of the audio compression/decompression apparatus of the present invention. Detailed Description

Figure 11 is a second embodiment of the audio signal compression/decompression apparatus of the present invention. 〆, ' [Key element symbol description] 11 Reducer 12 Add sampler 13 Buffer memory 14 One-dimensional discrete wavelet converter 15 Bit compression coding unit 16 Arithmetic compression unit 17 First in first out unit 21 Reducer 22 Add sampler 23 Buffer memory 24 First dimension discrete wavelet converter 25 Bit compression coding unit 26 Arithmetic compression unit 27 First in first out Unit 28 second one-dimensional discrete wavelet converter 30 code block 1279693 30-8 eighth bit plane

Claims (1)

1279693 45 years 丨 月 巧 巧 修 ) ) ) ) ) 、 、 、 、 、 、 、 、 、 、 、 、 、 、 、 、 、 、 、 、 、 、 、 、 、 、 、 、 、 、 、 、 、 、 、 、 、 、 、 、 、 Processing, including performing one-dimensional one-dimensional discrete wavelet transform; performing quantization; private t 兀 compression coding: encoding the OV-order one-dimensional political wavelet converted audio signal of the unit time to form a code block · t, 彳Γ二彳f encoding · The secret element Wei _ the audio ship line arithmetic coding, and the output bit stream. 2) The audio signal compression method described in item i of the patent scope, wherein the step of selecting a compression ratio before the one-dimensional discrete wavelet transform is achieved by an adaptively adjusted compression method. 3. The audio signal compression method according to item 2 of the patent scope, wherein the method is adapted to perform bit compression adjustment, and the method comprises the following sub-steps: defining a compression ratio according to the selected Each sub-band has a processing bit number; a bit-plane analysis is performed on the code block to confirm the first significant position of the ten-curve of i, The bit-level _---the first one has the number of bits in the bit-plane, and the number of non-Willie-bits is set to 0', and the bit plane is performed. coding. 4. The audio signal compression method according to claim 3, wherein the plane adjustment further comprises the step of flattening the bit plane of the number of processing bits to LSB. 5. The audio signal compression method of claim 3, wherein the value of the number of parity bits is obtained experimentally. The sound compression method of the μ method includes the following steps: · 1 ' input bit stream: the audio signal compressed by the audio compression method 21 1279693 method described in the third paragraph of the patent application is in place. Meta-streaming mode input; performing arithmetic decoding; performing bit compression decoding: bit-compressing and decoding the audio signal of the arithmetically decoded audio signal of unit time (1) in a code block manner; performing dequantization; Dimensional discrete wavelet transform ·, and restore signals. 7. The method for decompressing the audio signal according to item 6 of the patent application scope, wherein when the remote user selects the compressed material, the step of decoding the bit surface, that is, the processing bit set according to the adaptive adjustment compression method The number is inversely processed by the bit plane. 8. The audio signal decompression method described in the third paragraph of the patent scope, wherein the second quantization step is scalar quantization, and may take 16 high bits, 12 high bits, 8 high bits, etc. 9. An audio compression/decompression circuit device comprising: a buffered § memory element for storing one unit time (1) input by an external, an audio signal, a one-dimensional discrete wavelet transform or a one-dimensional inverse discrete wavelet Converted audio signal; a one-dimensional discrete wavelet converter, cooperate with the buffer memory to complete N-dimensional one-dimensional discrete wavelet transform, or N-order one-dimensional inverse discrete wavelet transform; vertical two-bit coding unit, The Μ pen is subjected to N-dimensional one-dimensional discrete wavelet transform 曰 汛 形成 形成 形成 形成 形成 形成 形成 形成 形成 形成 形成 形成 形成 形成 形成 形成 形成 形成 形成 形成 形成 形成 形成 形成 形成 形成 形成 形成 形成 形成 形成 形成 形成 形成 形成 形成 形成 形成 形成 形成The audio signal encoded by the method is subjected to arithmetic compression, or the compressed audio signal is subjected to arithmetic decompression; and the compressed audio signal is rounded/translated as a bit stream or a bit string of the compressed audio signal is compressed. The stream is sent to the arithmetic compression unit for decompression. The audio compression/decompression circuit device of claim 9, further comprising: a reduced sampling unit for sampling the audio signal in the unit time (t) Data, in order to reduce the sampling method, reduce the sampling rate of the sampled data; and increase the sampling unit for the N-dimensional one-dimensional inverse discrete wavelet converted audio signal stored in the buffer memory Add a sample to restore to the audio signal. The audio compression/decompression circuit device of claim 9, wherein the buffer memory system is included in the bit coding unit. The audio compression/decompression circuit device of claim 9, wherein the bit coding unit further comprises a memory for storing the audio signal of the one-dimensional discrete wavelet transform. 13. = The audio surface/decompression circuit basin human wave converter described in the scope of claim 9 is a one-dimensional inverse discrete wavelet transform of the audio signal input, whereby 丄.ί 14. ίΓί special; item f1G The audio compression/decompression electric reduction sampling dimension discrete wavelet converter described in the item, which is used to perform the "== s mouth water meter 70' and the buffer memory between the system and the discrete iiT'f Discrete wavelet transform or the last one-dimensional inverse & one-time electrical memory capacity requirements. The audio input Si machine input - unit time _ _ - 3 from today, U Zheng Xiaobo defending the audio signal; 兀 flat horse single 70 'the M pen after N times one-dimensional discrete wavelet transform 23 1279693 audio After the signal is formed into a code block, it is coded in a bit-plane manner. - The compression unit is used to perform arithmetic compression on the audio signal, and the first-in-first-out unit (FIFO) is used to compress the compressed audio. The signal is output as a bit stream. The sound surface reduction circuit device of claim 15, further comprising: a reduction sampling unit for reducing the sampling by the sampled data of the audio signal in the unit time (1), The sampled data is reduced for sampling according to the set sampling rate. 17* The audio compression circuit device of claim 15, wherein the buffering system is included in the bit coding unit. 18. The audio compression circuit device of claim 15, wherein the code encoding unit further comprises a memory for storing the M signal through N times of one-dimensional discrete wavelet transform. The audio compression circuit device of claim 15, further comprising a second one-dimensional discrete wavelet converter interposed between the audio signal input and the buffer memory. Perform the first one-dimensional discrete wavelet transform 'to reduce the capacity requirement of the buffer memory. 20. The audio compression circuit device as described in claim 15 or claim 19, further comprising a second one-dimensional discrete wavelet converter interposed between the reduction sampling unit and the buffer memory Used to perform the first one-dimensional discrete wavelet transform, thereby reducing the capacity requirement of the buffer memory. An audio decompression circuit device for an audio compression circuit device of the first application of the patent application scope includes: a first in first out unit (FIFO), which receives a bit formed by the compressed audio signal; The compressed audio signal is subjected to arithmetic decompression; ~ a one-bit decoding unit that decodes the arithmetically decompressed audio signal to solve the 24 1279693 frequency coded in a bit-plane manner by N-dimensional one-dimensional discrete wavelet transform Audio signal; a one-dimensional inverse discrete wavelet converter for performing one-dimensional inverse discrete wavelet transform of the N-dimensional one-dimensional discrete wavelet converted audio signal; and a monthly buffer memory for storing ^^ One-dimensional inverse discrete wavelet transforms the audio signal and turns it out. The audio decompression circuit device of claim 21, further comprising an additional sampling unit for converting the audio signal stored in the buffer memory by one-dimensional inverse discrete wavelet transform Increase the sampling to output the audio signal. Μ The audio decompression circuit device of claim 21, wherein the buffer memory system is included in the bit decoding each, 24 · as claimed in the patent range 91 + the bit solution, the road device, wherein the sub-dimensional inverse discrete The sound of the second wave of the wavelet is used for the age of the pen. 25. For example, the 21st item of the patent application includes a second-dimensional inverse decompression circuit device, in which the input/output and the (n) n 'between The audio signal wheel discrete small turn _ one-dimensional inverse 26. As claimed in the scope of the 21st item. The direct needs of the body. Set, the = more include - the second decompression circuit is equipped with a small sample unit and the buffer memory / conversion fast, which is between the reduced dimension inverse discrete wavelet transform, thereby reducing the second to perform the last one The capacity requirements of the memory. 25