JPH05119800A - High-efficiency compressing method for digital speech data - Google Patents
- ️Tue May 18 1993
JPH05119800A - High-efficiency compressing method for digital speech data - Google Patents
High-efficiency compressing method for digital speech dataInfo
-
Publication number
- JPH05119800A JPH05119800A JP3277402A JP27740291A JPH05119800A JP H05119800 A JPH05119800 A JP H05119800A JP 3277402 A JP3277402 A JP 3277402A JP 27740291 A JP27740291 A JP 27740291A JP H05119800 A JPH05119800 A JP H05119800A Authority
- JP
- Japan Prior art keywords
- frame
- parameters
- data
- speech data
- digital speech Prior art date
- 1991-10-24 Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Landscapes
- Compression, Expansion, Code Conversion, And Decoders (AREA)
Abstract
PURPOSE:To compress information to be recorded or sent more by further encoding digital speech data, generated by compressing a digitized speech signal by encoding, with high efficiency. CONSTITUTION:Parameters showing the sound source and modulation characteristics of a speech are extracted, frame by frame, from the digital speech data divided into frames of short sections by a parameter division part 32, and the parameters are sectioned, frame by frame; and a code corresponding to the appearance probability of an information source symbol is assigned to parameters having the statistical partiality in each group of respective parameters to decrease the number of bits of each data.
Description
【0001】[0001]
【産業上の利用分野】本発明は電話通信におけるデジタ
ル音声データの記録又は伝送時の情報量を符号化して削
減したデジタル音声データを更に符号化によって圧縮す
るに適した方法に関する。BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a method suitable for further compressing digital voice data, which has been encoded by reducing the amount of information at the time of recording or transmitting digital voice data in telephone communication.
【0002】[0002]
【従来の技術】デジタル化した音声データを低、中ビッ
トレートで伝送、又は記録する場合、音声はその生成機
構に基いて調音特性、音源情報を短区間のフレーム毎に
符号化して圧縮する高能率音声符号化方式により情報量
を圧縮し、記録、伝送に適したものとする技術が知られ
ている。2. Description of the Related Art In the case of transmitting or recording digitized voice data at low and medium bit rates, voice has a high articulation characteristic based on its generation mechanism, and sound source information is encoded and compressed for each frame of a short section. A technique is known in which the amount of information is compressed by an efficient audio coding method to make it suitable for recording and transmission.
【0003】[0003]
【従来技術の問題点】前記従来の高能率音声符号化方法
を用いて圧縮された音声データを半導体メモリ等に記録
しようとすると、、現在のアナログテープに比して記録
媒体のコストがかかるため、記録又は伝送すべき情報量
をより一層圧縮する必要性に迫られる。Problems to be Solved by the Invention When attempting to record audio data compressed using the conventional high-efficiency audio encoding method in a semiconductor memory or the like, the cost of the recording medium is higher than that of the current analog tape. However, it is necessary to further compress the amount of information to be recorded or transmitted.
【0004】[0004]
【問題点を解決するための手段】かくて本発明は前記従
来の高能率音声符号化方式における調音特性が、音声上
の性質からデータのフレーム間で、それ程急激な変化が
なく前フレームと現フレームの係数上の差が小さい値に
集中することに鑑みて、この値に統計的な偏りを利用し
て符号化するのが有利である点と、音声のピッチ等の音
源情報は各個人によって偏りがある点に着眼し、音声の
音源及び調音特性を表すパラメータを短区間毎に抽出
し、このパラメータをフレーム毎に区分するとともに、
このパラメータのグループ毎の統計的な偏り分布をもつ
パラメータに対して情報源シンボルの出現確率に応じた
符号を割り当てることにより、1データ当たりのビット
数を更に圧縮するようにしたものである。Therefore, according to the present invention, the articulatory characteristics of the conventional high-efficiency speech coding system are not so drastically changed between data frames because of the nature of speech, and they are the same as those of the previous frame. Considering that the difference in the coefficient of the frame concentrates on a small value, it is advantageous to use a statistical bias in this value for encoding, and the sound source information such as the pitch of the voice depends on each individual. Focusing on the point where there is a bias, extracting the parameter representing the sound source and articulatory characteristics of the voice for each short section, and dividing this parameter for each frame,
The number of bits per data is further compressed by assigning a code according to the appearance probability of an information source symbol to a parameter having a statistical bias distribution for each group of parameters.
【0005】[0005]
【実施例】以下図面により本発明の一実施例について説
明する。先ず本発明をメモリへ記録する場合について図
1により説明すると、デジタル化された音声データは、
音声符号化器1によりフレーム毎に符号化され、符号化
データとして出力される。この出力されたデータはn種
類の符号化されたパラメータで、パラメータ分割部3で
分けられ、次に各パラメータ毎の統計的な偏りに分布す
るデータをエントロピー符号化器2によって更に符号化
して圧縮される。DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS An embodiment of the present invention will be described below with reference to the drawings. First, the case of recording the present invention in a memory will be described with reference to FIG.
The audio encoder 1 encodes each frame and outputs the encoded data. The output data is divided into n types of encoded parameters by the parameter division unit 3, and then the data distributed in a statistical bias for each parameter is further encoded by the entropy encoder 2 and compressed. To be done.
【0006】次に図2によりコード励起線型予測符号化
方式(CELP)を例にとって本発明を説明すると線型
予測部21の後の残差信号を音源信号と見做し、長期予
測部24、雑音コードブック27によって符号化する方
式で、符号化操作は入力音声を20乃至40mSecの
区間を1フレームとして分けて、その区間毎に処理され
る。ここで符号化される情報は、音声信号の電力、線型
予測における短期予測係数、長期予測におけるピッチ周
期、予測係数、雑音コードブックにおける各雑音ベクト
ルを表す符号及び雑音ベクトルの利得等である。Next, referring to FIG. 2, the present invention will be described by taking the code excitation linear predictive coding (CELP) as an example. The residual signal after the linear predictor 21 is regarded as a source signal, and the long-term predictor 24 and noise In the encoding method by the codebook 27, the encoding operation is performed by dividing the input voice into 20 to 40 mSec sections as one frame and processing each section. The information encoded here is the power of the speech signal, the short-term prediction coefficient in linear prediction, the pitch period in long-term prediction, the prediction coefficient, the code representing each noise vector in the noise codebook, the gain of the noise vector, and the like.
【0007】通常、符号化された情報は1フレーム単位
でまとめて記録されるが、本発明では図1に示すように
符号化されたデータを各パラメータ毎に記録又は伝送に
用いる。この記録又は伝送時に各パラメータの偏り分布
を利用してエントロピー符号化し、更にデータ量の圧縮
を図る。即ち具体的な記録法としてハフマン符号化法を
用いると、情報源シンボルの出現確率に応じて符号を割
当てるものであるから出現確率の高いものには短かい符
号語を、出現確率の低いものには長い符号語を割当てる
ことにより、データ量の圧縮を図ることができるのであ
る。Usually, the coded information is collectively recorded in units of one frame, but in the present invention, the coded data is used for recording or transmission for each parameter as shown in FIG. At the time of this recording or transmission, entropy coding is performed by utilizing the biased distribution of each parameter, and the data amount is further compressed. That is, when the Huffman coding method is used as a concrete recording method, a code is assigned according to the appearance probability of an information source symbol. By allocating long code words, the data amount can be compressed.
【0008】前記したCELPにおける短期予測係数
は、音声の性質からフレーム間ではそれ程急激に変化す
ることがないので、短期予測係数は前フレームの係数と
の差を取ると、その値は小さい値に集中する。この方式
について図3を用いて説明すると、デジタル化された音
声データはCELP符号器31によってフレーム毎に符
号化された後、このデータはパラメータ分割部32によ
って各パラメータ毎に分けられる。上記の短期予測係数
は、通常10個つまり10次の予測のパラメータで符号
化され、全てのパラメータ個々について同様に処理さ
れ、前フレームの同じパラメータとの差を採って、その
差がハフマン符号化器37によって符号化され伝送に供
し得る処理が施されることになる。又、この時点のパラ
メータは、次のフレームのパラメータとの差を取るため
にディレー部34によって1フレーム分が遅延される。
尚、長期予測については話者に応じたピッチ周期の偏
り、有、無声に応じた情報の偏りを利用し、ハフマン符
号化してデータ量の圧縮を図ることができ、最終的に音
声によっては実に3分の1程度にまで圧縮できるのであ
る。Since the short-term prediction coefficient in CELP described above does not change drastically between frames due to the nature of speech, if the difference between the short-term prediction coefficient and the coefficient of the previous frame is taken, the value becomes a small value. concentrate. This method will be described with reference to FIG. 3. After the digitized voice data is encoded by the CELP encoder 31 for each frame, this data is divided by the parameter dividing unit 32 for each parameter. The above-mentioned short-term prediction coefficient is normally coded with 10 or 10th-order prediction parameters and processed in the same manner for all parameters individually, and the difference from the same parameter in the previous frame is taken, and the difference is Huffman coded. The processing that is encoded by the device 37 and can be used for transmission is performed. The delay unit 34 delays the parameter at this point by one frame in order to take a difference from the parameter of the next frame.
For long-term prediction, it is possible to compress the amount of data by Huffman coding using the bias in pitch period depending on the speaker, and the bias in information depending on the presence or absence of voice. It can be compressed to about one third.
【0009】実用上用いることができるエントロピー符
号としてはハフマン符号、ランレングス符号、算術符号
等、情報の性質に応じて決めることができる。The entropy code which can be practically used can be determined according to the property of information such as Huffman code, run length code, arithmetic code and the like.
【0010】[0010]
【発明の効果】本発明は音源と調音夫々の符号化された
パラメータが、フレーム間又はブロック間で変化の少な
い統計的な偏り分布に応じてデジタルデータを、全体的
に圧縮するためにエントロピー符号化手法を用いて情報
源シンボルの出現確率の高いものには短い符号語を、出
現確率の低いものには長い符号語を割り当てて圧縮して
いるので、音声によっては従来の高能率符号化法に比し
てほぼ3分の1にまでデータ量を圧縮することができ、
半導体メモリに記録する場合にはその記憶量が大幅に減
り、又データ伝送時にはデータ伝送効率の向上に寄与す
ることができる。According to the present invention, the entropy code is used in order to compress the digital data as a whole in accordance with the statistical bias distribution in which the coded parameters of the sound source and the articulator are changed little between frames or blocks. Since a short codeword is assigned to a symbol with a high probability of occurrence of an information source symbol and a long codeword is assigned to a symbol with a low probability of occurrence, the conventional high-efficiency coding method is used for some speech. The data amount can be compressed to about one-third compared to
When the data is recorded in the semiconductor memory, the storage amount is significantly reduced, and it can contribute to the improvement of the data transmission efficiency during the data transmission.
【0011】[0011]
【図面の簡単な説明】[Brief description of drawings]
【図1】本発明デジタル音声データの高能率圧縮方法を
実施するための回路構成図。FIG. 1 is a circuit configuration diagram for implementing a high-efficiency compression method for digital audio data according to the present invention.
【図2】コード励起線型予測符号化方式を説明するため
の回路構成図。FIG. 2 is a circuit configuration diagram for explaining a code excitation linear predictive coding system.
【図3】図2に本発明の方法を適用した回路構成図。FIG. 3 is a circuit configuration diagram in which the method of the present invention is applied to FIG.
【符号の説明】[Explanation of symbols]1 音声符号化器 2 エントロピー符号化器 3、32 パラメータ分割部 4 メモリ 21 線型予測部 23 短期予測部 24 長期予測部 25 誤差評価部 27 雑音コードブック 31 CELP符号化器 35、36 符号化器 37 ハフマン符号化器 1 Speech Encoder 2 Entropy Encoder 3 and 32 Parameter Divider 4 Memory 21 Linear Predictor 23 Short Term Predictor 24 Long Term Predictor 25 Error Evaluator 27 Noise Codebook 31 CELP Encoder 35, 36 Encoder 37 Huffman encoder
Claims (1)
【特許請求の範囲】[Claims]
【請求項1】デジタル化された音声信号を符号化して圧
縮するデジタル音声データの高能率音声符号化方法にお
いて、音声の音源及び調音特性を表すパラメータを、短
区間毎のフレームに分割した上記デジタル音声データか
ら1フレーム毎に抽出して上記パラメータをフレーム毎
に区分し、上記パラメータ毎の各グループの統計的偏り
の分布をもつパラメータに対して、情報源シンボルの出
現確率に応じて符号を割当てることにより1データ当た
りのビット数を更に圧縮するようにしたことを特徴とす
るデジタル音声データの高能率圧縮方法。1. A high-efficiency audio encoding method for digital audio data, which encodes and compresses a digitized audio signal, wherein the parameters representing the sound source and articulation characteristics of the audio are divided into frames for each short section. The above parameters are extracted for each frame from the audio data, and the above parameters are divided into each frame, and codes are assigned to the parameters having the statistical bias distribution of each group for each parameter according to the appearance probability of the information source symbol. Thus, the number of bits per data is further compressed, which is a highly efficient compression method for digital audio data.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
JP3277402A JPH05119800A (en) | 1991-10-24 | 1991-10-24 | High-efficiency compressing method for digital speech data |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
JP3277402A JPH05119800A (en) | 1991-10-24 | 1991-10-24 | High-efficiency compressing method for digital speech data |
Publications (1)
Publication Number | Publication Date |
---|---|
JPH05119800A true JPH05119800A (en) | 1993-05-18 |
Family
ID=17583049
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
JP3277402A Pending JPH05119800A (en) | 1991-10-24 | 1991-10-24 | High-efficiency compressing method for digital speech data |
Country Status (1)
Country | Link |
---|---|
JP (1) | JPH05119800A (en) |
Cited By (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JP2002517019A (en) * | 1998-05-27 | 2002-06-11 | マイクロソフト コーポレイション | System and method for entropy encoding quantized transform coefficients of a signal |
WO2006075605A1 (en) * | 2005-01-12 | 2006-07-20 | Nippon Telegraph And Telephone Corporation | Long-term prediction encoding method, long-term prediction decoding method, devices thereof, program thereof, and recording medium |
JP2009193073A (en) * | 2001-02-13 | 2009-08-27 | Qualcomm Inc | Method and apparatus for reducing undesired packet generation |
-
1991
- 1991-10-24 JP JP3277402A patent/JPH05119800A/en active Pending
Cited By (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JP2002517019A (en) * | 1998-05-27 | 2002-06-11 | マイクロソフト コーポレイション | System and method for entropy encoding quantized transform coefficients of a signal |
JP2009193073A (en) * | 2001-02-13 | 2009-08-27 | Qualcomm Inc | Method and apparatus for reducing undesired packet generation |
WO2006075605A1 (en) * | 2005-01-12 | 2006-07-20 | Nippon Telegraph And Telephone Corporation | Long-term prediction encoding method, long-term prediction decoding method, devices thereof, program thereof, and recording medium |
JP2010136420A (en) * | 2005-01-12 | 2010-06-17 | Nippon Telegr & Teleph Corp <Ntt> | Long-term prediction encoding method, long-term prediction decoding method, device thereof, program thereof |
US7970605B2 (en) | 2005-01-12 | 2011-06-28 | Nippon Telegraph And Telephone Corporation | Method, apparatus, program and recording medium for long-term prediction coding and long-term prediction decoding |
US8160870B2 (en) | 2005-01-12 | 2012-04-17 | Nippon Telegraph And Telephone Corporation | Method, apparatus, program, and recording medium for long-term prediction coding and long-term prediction decoding |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
KR100518640B1 (en) | 2005-12-14 | Data Compression / Restoration Using Rice Encoder / Decoder and Method |
US7133832B2 (en) | 2006-11-07 | Recording and reproducing apparatus for use with optical recording medium having real-time, losslessly encoded data |
US6970479B2 (en) | 2005-11-29 | Encoding and decoding of a digital signal |
US20020049586A1 (en) | 2002-04-25 | Audio encoder, audio decoder, and broadcasting system |
JP4800379B2 (en) | 2011-10-26 | Lossless coding of information to guarantee maximum bit rate |
EP1187338A2 (en) | 2002-03-13 | Method and apparatus for performing variable-size vector entropy coding |
JPH10173541A (en) | 1998-06-26 | Compressed encoding and decoding system |
US20070143118A1 (en) | 2007-06-21 | Apparatus and method for lossless audio signal compression/decompression through entropy coding |
JPH05119800A (en) | 1993-05-18 | High-efficiency compressing method for digital speech data |
Bruekers et al. | 1997 | Improved lossless coding of 1-bit audio signals |
JPS6031326A (en) | 1985-02-18 | Method and apparatus for coding and decoding dpcm |
EP1617411A1 (en) | 2006-01-18 | Code conversion method and device |
JPH11109996A (en) | 1999-04-23 | Voice coding device, voice coding method and optical recording medium recorded with voice coding information and voice decoding device |
KR100686354B1 (en) | 2007-02-22 | Huffman decoding method and apparatus using variable tree |
JPS623440B2 (en) | 1987-01-24 | |
JPH0934493A (en) | 1997-02-07 | Acoustic signal encoding device, decoding device, and acoustic signal processing device |
KR100490638B1 (en) | 2005-09-26 | Encoding and Decoding Systems for Lossless Encoding of Digital Audio Signals |
JPH05150800A (en) | 1993-06-18 | Speech coder |
JPS58131835A (en) | 1983-08-05 | Multichannel signal encoding method |
JPS6146621A (en) | 1986-03-06 | Binary coding method of voice data |
JPH07121198A (en) | 1995-05-12 | Device for storing sound |
JPS63220300A (en) | 1988-09-13 | Voice analyzer/synthesizer |
JPH05175916A (en) | 1993-07-13 | Audio transmission system |