TW382094B - Base tone synchronous differential coding method and device thereof - Google Patents

Abstract

A kind of base tone synchronous differential coding method and device thereof. The method is to firstly input a signal representing a Chinese syllable and abstract the base tone message from input signal; then, according to this base tone message, proceed the format conversion on the input signal; after applying first order differentiation on the signal after format conversion to output the data. The base tone synchronous differential coding device includes a base tone abstracting apparatus, a format converter and a differentiating device. The base tone abstracting apparatus is used to receive an input signal representing a Chinese syllable and abstract the base tone message from the signal. The format converter is used for format conversion on the input signal based on the base tone message. The differentiating device will proceed first order differentiation on the signal after format conversion and produce an output signal for output.

Description

No. 86118724 revised page V. Description of the invention (4)

Format converter; 14 ~ differential n; 15 ~ data compressor; i6 ~ psDc code. 20 ~ input voice data module; 21 ~ band pass chirp processing module; Pitch extraction candidate group; 24 ~ best, logical judgement M group; 25 ~ output of the value of the voiced pitch of the pitch; 3Q ~ psDc code; 3 data decompressor; 32 ~ inverse difference; 33 ~ pitch signal; μ ~ Format Inverter; 35 ~ PCM Code; Grab Chinese Articles; 41 ~ Grammar Processing Device; 42 ~ Speech Data Synthesis Device 43 ~ Chinese Speech Library; 44 ~ Chinese Speech Output; 50 ~ Text Preprocessor; 51 ~ Word Segmentation Processor; 52 ~ Chinese word database; 53 ~ Chinese word frequency database; 54 ~ gram processor; M ~ 汉 i # f ± .-u: # Printed part-of-speech table for employees' cooperatives of the Central Standards Bureau of the Ministry of Economic Affairs; 56 ~ Chinese grammatical rules 以及; and 57 ~ syllable converter. Example: The pitch-synchronous differential coding method according to the present invention is a combination of two types of coding methods: parametric coding and waveform coding. Therefore, compared with parametric coding or waveform coding , The processed voice data can be maintained The real human voice and 6-voice data are easy to control. In addition, the amount of calculation required for the voice data: to reduce the effect of 'real-time synthesis can also be achieved. Please refer to FIG. 1'. Pitch synchronization differential encoding device, a block diagram of a preferred embodiment. The pitch synchronization differential encoding device according to the present invention is an input signal with a pulse code modulation (T abbreviated as PCM) code 10, which passes through the pitch. The same ^ difference encoding method is converted into a pitch synchronous differential code (hereinafter abbreviated as pSDC exhaustion) after 16 'output for storage. The input signal 10 of this pcM code represents a Chinese syllable. The pitch synchronous differential encoding device according to the present invention Including: a pitch extractor U, a format converter 13, a differentiator 14,

I ~ ^ ----ί 1 (Please read the notes on the back before filling in this page) Threading Service Department of the Central Standards Bureau of the Ministry of Economic Affairs, Consumer Cooperatives, Yin Li A 7- ______B7__ 5. Description of the invention (1) This invention relates to The speech processing technology, in particular, relates to a pitch synchronous differential coding method and a device thereof. It can cooperate with the grammar processing device to synthesize the pronunciation of arbitrary Chinese characters, words, and sentences, and the synthesized speech can approximate the effect of real human pronunciation. In the technical field of speech synthesis, the speech data is first converted into easy-to-handle speech data by a specific encoding method and stored; when the post-achievement »wants to synthesize speech, the corresponding decoding method is used to capture the data to be synthesized. Wakazen voice data is synthesized. The closer the synthesized speech is to the effect of real-life pronunciation ’, this is what people in this industry want. At present, the coding methods for speech synthesis can be divided into two types: parameter coding and waveform coding. Parametric coding is to analyze the speech waveform after sampling, to retrieve the speech parameters and store them. When synthesizing speech, the speech data is restored based on these parameters. Because the parameter coding is based on the speech generation model, some characteristic parameters related to the sound source and channel can be obtained through analysis, and the sampled speech data is converted into a set of parameters for storage. These references are used when synthesizing speech: Regroup voice data. However, this method must first establish a set of correct pronunciation models, and then describe these models with relevant parameters. Therefore, the key to the technique of this method lies in the establishment of a successful pronunciation model and the need to describe this model with a set of accurate parameters. Since the current understanding of the human pronunciation process is still in the exploratory stage, a survey model has not yet been established, so the parameters The encoding method is easy to introduce into the voice, and the human voice will be lost after speech synthesis. Furthermore, the amount of calculation required is quite large. 'It is difficult to achieve an instant speech synthesis effect. Linear Predictive (Linear Predictive This paper size applies to Chinese National Standards (CNS > Specifications (210X297 mm) ^ —ί—------: _ {^ 他 ^ ~ ml-I n · '(Please first Pavilion On the back of the order, Jiang Jiang will save the item and fill in this purchase] II. V. Invention Description (5

Trouble II and a data compressor 15. … Chinese syllables are different from unvoiced and voiced. Please refer to Figure 7 for waveforms of unvoiced and voiced. It can be seen that the unvoiced sound has a small amplitude, a relatively irregular waveform, and no periodicity; a voiced sound has a large vibration, and the waveform has a regular and obvious period. Printed by the Consumers' Cooperative of the Central Bureau of Standards of the Ministry of Economic Affairs As shown in Figure 1, the PCM code 10 is simultaneously applied to the pitch extractor U and the format converter 13. Because there are differences between unvoiced and unvoiced Chinese syllables, the pitch extractor 11 performs unvoiced / voiced judgment and calculation of the voiced period on the input ship code 10, and outputs it as a pitched message 12. Judgment of unvoiced / voiced voice can be achieved by calculating short term average energy or average cross zero, for example, and the calculation method is as follows: Short-term average energy: Εη = Σ [x (m) w (nm)] 2; where 'x (m) represents speech data as a PCM code, and w (nm) represents a window function. Short-term average zero-crossing rate: Zn = | sgn [x (m) -sgn [x (ml)] | w (nm); where 'sgn [x (n)] = l when x [n] g 0; sgn [x (n)] = − 1 when χ [η] < 〇. As for the voiced period calculation, the Gold-Rabiner method can be used. The method flow chart is shown in Figure 2. First, input voice data in step 20, and the voice data is the PCM code 10 shown in Fig. 1. The input voice data is processed by band-pass filtering in step 21, and the frequency range of this band-pass filtering is as possible as _ 7 This paper standard is applicable to Zhongguanjia Standard (CNS) 8 4 Lai _ (21GX 297 public meals)-1. -——----I.il li II ...... I n---I n ~-. ~ (Please read the notes on the back before filling this page),? Τ A7 B7 V. Invention Instructions (2)

Coding (referred to simply as LPC) is a type of parametric coding. The synthesized speech is quite stiff, and it introduces obvious noise, which cannot meet the requirements of close to real human voice. Waveform coding is a method of sampling and storing voice waveform data. The signal-to-noise ratio is relatively high, which can better maintain the timbre of the original voice data. It is equipped with Adaptive Differential Pulse Code Modulation (referred to as ADPCM) coding, which is a type of waveform coding. 'This method can maintain the original timbre of voice data. Therefore, it is used in speech synthesis. The obtained voice sound is better than the parametric encoder. However, its original sound database has a fixed source and cannot be adjusted for various situations in the sentence. Therefore, it is still stiff when synthesizing Hanyu sentences based on it, and it cannot achieve the effect of close to real pronunciation. Therefore, it is an object of the present invention to provide a pitch synchronous differential encoding and decoding method and a device thereof, which can synthesize a Chinese voice with a near-real-life pronunciation effect, which is coherent and natural. Another object of the present invention is to provide a pitch synchronization differential encoding and decoding method and a device thereof, which can synthesize speech in real time. To achieve the above object, the present invention can be accomplished by providing a pitch synchronization differential coding method. First, after inputting an input signal representing one of the Chinese syllables, the pitch information is extracted from the input signal. Then, the input signal is format-converted based on this pitch information '. Then, a first-order difference is performed on the signal in the converted format, and the output data is output. Furthermore, the present invention can be accomplished by providing a pitch synchronous differential encoding device. This pitch-synchronous differential encoding device includes: a pitch extraction. The paper size is applicable to the Chinese National Standard (CNS) A4 specification (210X297 mm). Please read the notes on the back before filling in the book. Printed by the consumer cooperative! 1 >. I-— ~ II Order — —-I — --- n Ί— n —Ί n --- nn __ A? &Quot; B7 V. Description of the invention (3) Device, one format A converter, and a differentiator. The pitch extractor is used to receive an input signal representing one of the Chinese syllables and extract pitch information from the input signal. The format converter performs format conversion on the input signal based on the pitch information. However, the differentiator performs a first-order difference on the signal in the converted format, and outputs it as an output data. In order to make the above and other objects, features, and advantages of the present invention more comprehensible, a preferred embodiment is given below in conjunction with the accompanying drawings to make a detailed description as follows: Brief description of the drawings: Section 1 阖FIG. 2 is a block diagram showing a preferred embodiment of a pitch synchronization differential encoding device according to the present invention; FIG. 2 is a flowchart showing a method for calculating a voiced period according to the present invention; FIG. 3 is a view illustrating a pitch synchronization according to the present invention Differential decoding device, a block diagram of a preferred embodiment; FIG. 4 is a block diagram showing a speech synthesis system; FIG. 5 is a block diagram showing a grammatical processing device of FIG. 4; Printed by the cooperative (please read the notes on the back before filling out this page) Figures 6A, 6B, and 6C show the coding method of the present invention, coupled with differential pulse code modulation (ADPCM) coding, and linear prediction coding, respectively. (LPC) comparison of the synthesized speech spectrum with the original speech spectrum envelope; and Figure 7 shows the unvoiced and voiced waveforms. Explanation of symbols: 10 ~ input signal; 11 ~ pitch extractor; 12 ~ pitch message; 13 ~ This paper size is applicable to China National Standard (CNS) A4 specification (210X297 mm) No. 86118724 amendment page 5. Description of the invention (4 )

Format converter; 14 ~ differential n; 15 ~ data compressor; i6 ~ psDc code. 20 ~ input voice data module; 21 ~ band pass chirp processing module; Pitch extraction candidate group; 24 ~ best, logical judgement M group; 25 ~ output of the value of the voiced pitch of the pitch; 3Q ~ psDc code; 3 data decompressor; 32 ~ inverse difference; 33 ~ pitch signal; μ ~ Format Inverter; 35 ~ PCM Code; Grab Chinese Articles; 41 ~ Grammar Processing Device; 42 ~ Speech Data Synthesis Device 43 ~ Chinese Speech Library; 44 ~ Chinese Speech Output; 50 ~ Text Preprocessor; 51 ~ Word Segmentation Processor; 52 ~ Chinese word database; 53 ~ Chinese word frequency database; 54 ~ gram processor; M ~ 汉 i # f ± .-u: # Printed part-of-speech table for employees' cooperatives of the Central Standards Bureau of the Ministry of Economic Affairs; 56 ~ Chinese grammatical rules 以及; and 57 ~ syllable converter. Example: The pitch-synchronous differential coding method according to the present invention is a combination of two types of coding methods: parametric coding and waveform coding. Therefore, compared with parametric coding or waveform coding , The processed voice data can be maintained The real human voice and 6-voice data are easy to control. In addition, the amount of calculation required for the voice data: to reduce the effect of 'real-time synthesis can also be achieved. Please refer to FIG. 1'. Pitch synchronization differential encoding device, a block diagram of a preferred embodiment. The pitch synchronization differential encoding device according to the present invention is an input signal with a pulse code modulation (T abbreviated as PCM) code 10, which passes through the pitch. The same ^ difference encoding method is converted into a pitch synchronous differential code (hereinafter abbreviated as pSDC exhaustion) after 16 'output for storage. The input signal 10 of this pcM code represents a Chinese syllable. The pitch synchronous differential encoding device according to the present invention Including: a pitch extractor U, a format converter 13, a differentiator 14,

I ~ ^ ----ί 1 (Please read the notes on the back before filling this page)

Trouble II and a data compressor 15. … Chinese syllables are different from unvoiced and voiced. Please refer to Figure 7 for waveforms of unvoiced and voiced. It can be seen that the unvoiced sound has a small amplitude, a relatively irregular waveform, and no periodicity; a voiced sound has a large vibration, and the waveform has a regular and obvious period. Printed by the Consumers' Cooperative of the Central Bureau of Standards of the Ministry of Economic Affairs As shown in Figure 1, the PCM code 10 is simultaneously applied to the pitch extractor U and the format converter 13. Because there are differences between unvoiced and unvoiced Chinese syllables, the pitch extractor 11 performs unvoiced / voiced judgment and calculation of the voiced period on the input ship code 10, and outputs it as a pitched message 12. Judgment of unvoiced / voiced voice can be achieved by calculating short term average energy or average cross zero, for example, and the calculation method is as follows: Short-term average energy: Εη = Σ [x (m) w (nm)] 2; where 'x (m) represents speech data as a PCM code, and w (nm) represents a window function. Short-term average zero-crossing rate: Zn = | sgn [x (m) -sgn [x (ml)] | w (nm); where 'sgn [x (n)] = l when x [n] g 0; sgn [x (n)] = − 1 when χ [η] < 〇. As for the voiced period calculation, the Gold-Rabiner method can be used. The method flow chart is shown in Figure 2. First, input voice data in step 20, and the voice data is the PCM code 10 shown in Fig. 1. The input voice data is processed by band-pass filtering in step 21, and the frequency range of this band-pass filtering is as possible as _ 7 This paper standard is applicable to Zhongguanjia Standard (CNS) 8 4 Lai _ (21GX 297 public meals)-1. -——----I.il li II ...... I n---I n ~-. ~ (Please read the notes on the back before filling this page),? Τ Central Bureau of Standards, Ministry of Economic Affairs A7 B7 printed by employee consumer cooperative V. Description of invention (6) It is between 100 ~ 900Hz. The speech data processed by the band-pass filtering goes to step 22, and the peak and valley detection is performed to obtain N pitch extraction candidates, starting from blocks 23A ... and to block 23N. Generally, the value of N is 6. Then, in step 24, the best logic judgment is performed on the N pitch extraction candidates, and then in step 25, the periodic value of the voiced pitch is output. Based on this, the pitch message 12 includes the unvoiced / voiced judgment result and the voiced period value. In a word, after the PCM code 10 is subjected to the pitch extraction processing, the pitch extractor 11 outputs a pitch information 12 to the format converter 13, and the pitch information 12 includes a voiced / voiced judgment result and a period value of the voiced pitch. Therefore, the format converter 13 performs format conversion processing on the input PCM code 10 based on the pitch information 12. For example, if the inputted PCM code 10 is unvoiced, its format is converted to: 丨 unvoiced flags 丨 unvoiced length 丨 voice data If the inputted PCM code 10 is voiced, its format is converted to: voiced flag cycle periods The length speech data is converted into different formats by the format converter 13 respectively, and then converted to the differentiator 14 to perform differential processing on the voiced portion. For example, this difference processing can be a first order difference for adjacent voiced periods. If it is simply expressed in a calculation formula, it is as follows: Sd (n) = S (n) -S (nT); this paper Standards are applicable to China National Standard (CNS) A4 specifications (210 X 29 " 7 mm) (Please read the precautions on the back before filling out this page) --Τ-Γ--1-.! ·-^-R -----------, '~, w clothing ------ Order ------ A7- B7 Printed by the Consumers' Cooperative of the Central Standards Bureau of the Ministry of Economic Affairs V. Invention Description (7) Among them, Sd (n) represents the current sampling point after the difference; S (n) represents the current sampling point before the difference; S (nT) represents the current sampling point of the previous cycle before the difference; T represents the current cycle length. Due to the high frequency of unvoiced speech, the effect of -order difference is not obvious, and the pronunciation of Chinese characters is mainly composed of voiced sounds. Therefore, the differentiator 14 mainly performs differential processing on the voiced sounds. This difference processing ′ may be a second-order difference or a higher-order difference processing of a second order or higher. If the amount of data to be stored is reduced for consideration, the data processed by the differential processor 14 can be subjected to data compression at the data compressor 15 'into a PSDC code 16 and output. The data compressor 15 can perform data compression, such as the conventional Huffman method, and reduces the amount of data to reduce the demand for the capacity of the storage device. The Huffman method includes the statistics of the symbol frequency of the sound bank, and the specific steps of establishing the Huffman tree. However, the Huffman method is already a well-known technology and is not the focus of the present invention, so it will not be repeated here. It is worth mentioning that the data compression by the data compressor 15 is an optional step, so the data processed by the differentiator 14 can also be output to 16 in 00 code. The PSDC code 16 produced by this ^ already contains the corresponding pitch information, then it is sent to a storage device (not shown) for data storage. This storage device can be a disk drive, memory, etc. This storage device stores the PSDC codes of voice data such as voiced and unvoiced sounds of various Chinese syllables, which is a _Chinese phonetic database. When you want to pronounce each character of the -text string, you can get the voice data of voiced or unvoiced sound from the Chinese voice library through decoding, and then perform decoding and synthesis. Please refer to the third chart, where: The basic paper size is applicable to the Chinese National Standard (CNS) A4 specification (21 〇χ 297 mm). (Please read the precautions of the West before filling out this page} k.- Order-ΦΙ. Printed by A7 B7, Consumer Cooperative of the Central Standards Bureau of the Ministry of Economic Affairs 5. Description of the invention (8) A block diagram of a preferred embodiment of the synchronous pitch decoding device according to the present invention. The device is an anti-processing device corresponding to the above-mentioned pitch-synchronous differential encoding device, and converts the PSDC code 30 stored in the Chinese speech database into a PCM code 35 by the pitch-synchronous differential decoding method. The pitch-synchronous differential decoding according to the present invention The device includes: a data decompressor 31, an inverse differentiator 32, and a format inverse converter 34. As shown in FIG. 3, if the voice data is compressed by the data compressor 15, the PSDC code 30 is retrieved. The steps must first be decompressed by the data decompressor 31. In short, it is a kind of data reduction processing. If the original compressed data was using the Huffman method, this decompression process must also be The Huffman method is used for reverse processing. If it was not processed by the data compressor 15, this step of the data decompressor 31 can also be omitted. Then, the decompressed PSDC code 30 is corresponding to it by the inverse differentiator 32 The pitch message 33 is subjected to inverse difference calculation. This inverse difference calculation is an inverse operation relative to the difference calculation. At this time, the voiced S sound and the unvoiced sound are respectively in the above-mentioned format, and then converted into the PCM code 35 by the format inverse converter 34 and output. This format conversion step is relative to the inverse conversion processing of the above-mentioned format converter 13. The above-mentioned encoding and decoding methods of the present invention can be applied to a speech synthesis system, and the block diagram of this speech synthesis system is shown in FIG. 4 After receiving a Chinese article 40 from the input terminal, this system can process the speech synthesis into Chinese speech 44 and output it. The speech synthesis system shown in Figure 4 includes a grammar processing device 41, a speech data synthesis device 42, and a Chinese language. Speech library 43 etc. 10 This paper size it uses Chinese national standard (CNS) A4 specifications (210X297 mm) 1 II III ^ ---- I 11 «I ^ I --- II n., (Please read first (Notes on the back please fill out this page) A?

Printed by the Consumer Standards Cooperative of the Central Standards Bureau of the Ministry of Economic Affairs A7 B7 V. Invention Description (ίο) Library 52 performs all possible segmentation paths; After that, two kinds of split paths can be obtained: "weather // forecast // say // tomorrow // weather // clear //" and "weather // forecast // description // day // air // clear // ". Then, through the step of selecting the shortest path, the shortest path is selected. If the shortest segmentation path is not unique, then it will cooperate with the Chinese word frequency database 53 to calculate the word frequency equivalent to determine the first choice as the segmentation result. Then, the grammar processor 54 performs processing. This grammar processor 54 uses the Chinese part-of-speech table 55 as the basis to analyze the constituents of the words in the segmented sentence to mark the parts of speech of each word. Wait. For example, pronouns are used as subjects before verbs. Then, according to the Chinese grammar rule base 56, the volume, pause, speed, frequency, etc. of the corresponding words are adjusted to obtain relevant speech synthesis parameters. For example, the sentence "He // Speak // Go // Park /// 's" purpose is segmented, "He" is the main subject, "Speak" is the predicate, "Go to the park" is the attributive, " The "object" is an object, so the structure of this sentence is "subject + predicate + attributive + object". Therefore, according to the Chinese grammar rule base 56, the object of this structure should be re-read, so the volume parameter of "purpose" should be adjusted. A pause of 200ms, a pause of 150ms between predicates, and so on. As for the syllable converter 57, the internal code of Chinese characters is converted into Zhuyin first, and then from Zhuyin to Pinyin. This conversion process also judges the broken sound words, for example, the word "中" is in the word "China", and its pronunciation is one sound, not four sounds. By contrast, the conversion speed can be greatly increased. As shown in Figure 4, after processing by the grammar processor 41, the number of speeches is 12. The paper size is applicable to the Chinese National Standard (CNS) A4 specification (210'〆297 mm). (Please read the precautions on the back before filling in this paper. Page) — ^ 1 · 1ν —B ^ iy —on * flu n ^ i 'ix. · 1 Printed by the Consumer Cooperative of the Central Bureau of Standards of the Ministry of Economic Affairs A7 B7 5. Description of the invention (11) According to the synthesis device 42. The speech data synthesizing device 42 extracts the speech data of the corresponding syllable from the Chinese character speech database 43 according to the content of the Chinese article 40. The operation of extracting the speech data is based on the above-mentioned pitch synchronous differential decoding method. The decoded speech data is synthesized, and at the same time, the phonology, intonation, volume, pitch, and speed of speech are adjusted in conjunction with the speech synthesis parameters sent by the grammar processor 41, and the Chinese speech 44 is output. Please refer to Figures 6A, 6B, and 6C, respectively, for the speech spectrum and original speech spectrum synthesized by the coding method of the present invention, coupled with differential pulse code modulation (ADPCM) coding, and linear prediction coding (LPC). Comparison of envelopes. In the figure, curve E represents the original speech spectrum envelope, curve F represents the speech spectrum synthesized according to the method of the present invention, and curve G represents the speech spectrum synthesized by conventionally coupled differential pulse code modulation (ADPCM) coding. Η represents the speech spectrum synthesized by conventional linear prediction coding (LPC). It can be seen from the frequency spectrum comparison in FIG. 6 that the pitch synchronous differential coding methods of the present invention are closer to the original speech signal than the conventional differential pulse code modulation (ADPCM) coding and linear prediction coding (LPC). Based on this, it is possible to achieve a close-to-natural Chinese speech that is close to real-life pronunciation effects; furthermore, it is possible to synthesize speech instantly. Although the present invention has been disclosed in the preferred embodiment as above, it is not intended to limit the present invention. Any person skilled in the art can make changes and retouches without departing from the spirit and scope of the present invention. Therefore, the protection of the present invention The scope shall be determined by the scope of the attached patent application. 13 This paper size is in accordance with Chinese National Standard (CNS) A4 (210X297 mm) (铕 Please read the precautions on the back before filling out this page)

Claims (1)

A8 B8 C8 D8 Patent application scope 1. A method of pitch synchronization differential encoding, including: inputting an input signal representing one of the Chinese syllables; extracting pitch information from the input signal; (please read the notes on the back before filling out this page) According to the pitch information, The input signal is format-converted; and after the converted format signal is differentiated, it is output as a data. 2. The pitch synchronous differential encoding method as described in item 丨 of the patent application scope, wherein the input signal is a pulse code modulation code. 3. The pitch-synchronous differential coding method as described in item 1 of the Shen Qing patent range, wherein the method of extracting pitch information is realized by calculating the short-term average energy and the short-time average zero-crossing rate. 4 ’The pitch-synchronous differential coding method as described in item 1 of the scope of the patent application. In the eighth, the five-tone pitch is used to indicate that the input signal is one of voiced ^ unvoiced. 5. The pitch-synchronous differential encoding method as described in item 4 of the scope of the patent application, wherein if the input signal is the voiced sound, the pitch information also includes periodic data of the voiced sound. -LQ. Printed by the Central Standards Bureau of the Ministry of Economic Affairs, Shelley Consumer Cooperative, 6. The pitch-synchronous differential encoding method described in item 5 of the scope of patent application, wherein the converted format signal includes a voiced mark, number of cycles, The period length, and voice data. 7. The pitch-synchronous differential coding method as described in item 4 of the scope of patent application, wherein if the input signal is the unvoiced sound, the 仏 sign of the converted format includes unvoiced mark, unvoiced length, and voice data. 8. According to the pitch synchronization differential encoding method described in item 1 of the scope of the patent application, when the data needs to be decoded, it is decomposed by a pitch synchronization difference and printed by the Consumers ’Cooperative of the Central Standards Bureau of the Ministry of Economic Affairs. -The C8 D8 patent application range code method works; the pitch synchronous differential decoding method includes the following steps. Inverse difference the data; and inverse transform the inverse difference processed signal based on the pitch information. 9. The pitch-synchronous differential editing method as described in item 8 of the scope of patent application, wherein ′ is subjected to compression processing after differential processing. 10. The pitch-synchronous differential coding method described in item 9 of the scope of the patent application, wherein ′ is subjected to decompression processing before inverse-differential processing is performed on the data. 11. A pitch synchronous differential encoding device, comprising: a pitch extractor that receives an input signal representing one of the Chinese syllables and extracts pitch information from the input signal; a format converter 'performs the input signal based on the pitch information Format conversion; and a differentiator, which outputs the data after differentiating the converted format signal. 12. The pitch synchronous differential encoding device as described in item u of the patent application scope, wherein the input signal is a pulse code modulation code. 13. The pitch-synchronous differential encoding device according to item u in the scope of the patent application, where 'the pitch extractor extracts the pitch information by calculating the short-term average energy and the short-time average zero-crossing rate. 14. The pitch-synchronous differential encoding device according to item 1] L of the scope of the patent application, wherein the pitch information is used to indicate that the input signal is one of voiced and unvoiced. (Please read the precautions on the back before filling out this page) A8, B8, C8, D8, patent application 15. The pitch synchronous differential encoding device as described in item 14 of the scope of patent application, wherein if the input signal is the voiced sound, the pitch information also includes the periodic data of the voiced sound . 16. The pitch-synchronous differential encoding device according to item 15 of the scope of patent application, wherein the converted format signal includes a voiced flag, the number of periods, the length of the periods, and speech data. 17. The pitch-synchronous differential encoding device according to item 14 of the scope of patent application, wherein if the input signal is the unvoiced, the converted format signal includes unvoiced flags, unvoiced length, and voice data. 18. The pitch synchronous differential encoding device described in item 11 of the scope of patent application, when the data needs to be decoded, it is performed by a pitch synchronous differential decoding device; the pitch synchronous differential decoding device includes: a contrast A demultiplexer performs inverse difference on the data; and a format inverse converter performs inverse conversion on the inverse difference processed signal based on the pitch information. 19. The pitch synchronous differential encoding device as described in item 18 of the scope of patent application, wherein the data processed by the differentiator is still subjected to compression processing. Printed by the Consumer Cooperative of the Central Standards Bureau of the Ministry of Economic Affairs (please read the notes on the back before filling this page) 20. The pitch synchronous differential encoding device as described in item 19 of the scope of patent application, in which the inverse difference is used to process the Before data is decompressed. This paper size applies to Chinese national standards (CNS > Α4 size (210X297 mm)