JPH06202697A - Gain quantizing method for excitation signal - Google Patents

Gain quantizing method for excitation signal

Info

Publication number
JPH06202697A
JPH06202697A JP5001110A JP111093A JPH06202697A JP H06202697 A JPH06202697 A JP H06202697A JP 5001110 A JP5001110 A JP 5001110A JP 111093 A JP111093 A JP 111093A JP H06202697 A JPH06202697 A JP H06202697A
Authority
JP
Japan
Prior art keywords
gain
quantization
vector
stage
excitation
Prior art date
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.)
Granted
Application number
JP5001110A
Other languages
Japanese (ja)
Other versions
JP3099852B2 (en
Inventor
Takehiro Moriya
健弘 守谷
Kazunori Mano
一則 間野
Satoshi Miki
聡 三樹
Naka Oomuro
仲 大室
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Nippon Telegraph and Telephone Corp
Original Assignee
Nippon Telegraph and Telephone Corp
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by Nippon Telegraph and Telephone Corp filed Critical Nippon Telegraph and Telephone Corp
Priority to JP05001110A priority Critical patent/JP3099852B2/en
Publication of JPH06202697A publication Critical patent/JPH06202697A/en
Application granted granted Critical
Publication of JP3099852B2 publication Critical patent/JP3099852B2/en
Anticipated expiration legal-status Critical
Expired - Lifetime legal-status Critical Current

Links

Abstract

PURPOSE:To provide the efficient quantizing method for the gain of an excitation source which can reduce waveform distortion due to CELP encoding with a small amount of information of a signal series of a speech. CONSTITUTION:The gain quantizing method for the excitation signal used for a linear predictive encoding method which takes a linear predictive analysis 12 of the input speech at each certain sampling period and determines the shapes and gains of plural excitation signals minimizing the distortion between the composite signal X' after being passed through a linear predictive synthesis part 15 consisting of a prediction coefficient and the input speech X performs gain quantization in two stages; the gains are quantized adaptively to the features of the input speech by the gain quantization of the 1st stage and the gains are quantized together into vectors by the gain quantization of the 2nd stage so as to compensate the result of the gain quantization of the 1st stage.

Description

【発明の詳細な説明】Detailed Description of the Invention

【0001】[0001]

【産業上の利用分野】この発明は、励振信号の利得量子
化方法に関し、特に音声の信号系列を少ない情報量のも
とで符号励振型線形予測符号化(Code Excited Linear
Prediction:CELP)による波形歪を小さくすることがで
きる効率のよい励振信号の利得の量子化方法に関する。
BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a method for gain-quantizing an excitation signal, and more particularly to a code-excited linear predictive coding (Code Excited Linear) for a speech signal sequence with a small amount of information.
Prediction: CELP) The present invention relates to an efficient method for quantizing the gain of an excitation signal that can reduce the waveform distortion caused by CELP.

【0002】[0002]

【従来の技術】ディジタル移動無線通信、音声蓄積サー
ビスその他の情報を伝送し或は蓄積する技術分野におい
ては、電波その他の情報伝送媒体或は記憶媒体の効率的
利用を図るために種々の高能率音声符号化方法が採用さ
れている。サンプリング周波数を8kHz としてサンプリ
ングされた音声を8kbit/s程度で符号化する方法として
はCELP符号化方法が有力な方法である。この方法は、要
約するに、複数の励振信号ベクトルの形状および利得を
選択する線形予測合成器を具備し、合成後の信号と入力
音声信号との間の聴感上の歪が最小となるように励振源
を制御して、その符号を伝送するものである。
2. Description of the Related Art In the technical field of transmitting or storing information such as digital mobile radio communication, voice storage service and the like, various high efficiency is achieved in order to efficiently use radio waves and other information transmission media or storage media. A voice coding method is adopted. The CELP coding method is a promising method for coding speech sampled at a sampling frequency of 8 kHz at about 8 kbit / s. This method, in summary, comprises a linear predictive synthesizer that selects the shape and gain of multiple excitation signal vectors so that the perceptual distortion between the synthesized signal and the input speech signal is minimized. The excitation source is controlled to transmit the code.

【0003】以下、図1を参照してCELP符号化方法につ
いて説明する。音声を高能率に符号化する方法として、
原音声をフレームと呼ばれる5〜50ms程度の一定間隔
の区間に分割し、その1フレームの音声を周波数スペク
トルの包絡線形状についての信号と、その包絡線形状に
対応する線形フィルタを駆動する励振信号とに分離して
それぞれを符号化する方法が提案されている。この場
合、励振信号を符号化する方法として、励振信号を音声
の基本周波数(或はピッチ周期)に対応すると考えられ
る周期成分と、それ以外の成分(換言すれば非周期成
分)とに分離して符号化する方法が知られている。この
励振信号の符号化方法の一種として符号励振型線形予測
符号化方法(CELP)がある。この符号化方法は、図1に
示される如く入力端子11に入力される入力音声Xにつ
いて線形予測分析部12においてその周波数スペクトル
の包絡線形状を表すパラメータが計算される。この分析
には通常、線形予測分析法が使用される。この線形予測
パラメータは線形予測パラメータ符号化部13において
符号化され、この符号化出力Aは線形予測パラメータ復
号化部14において復号化され、この復号化された線形
予測パラメータa’は線形予測合成部15のフィルタ係
数として設定される。線形予測合成部15に後で説明さ
れる励振信号(ベクトル)Eを与えることにより再生合
成音声X’が得られる。
The CELP coding method will be described below with reference to FIG. As a method of encoding voice with high efficiency,
The original speech is divided into intervals of about 5 to 50 ms called a frame, and the speech of one frame is a signal about the envelope shape of the frequency spectrum and an excitation signal for driving a linear filter corresponding to the envelope shape. There has been proposed a method of separating and encoding each. In this case, as a method of coding the excitation signal, the excitation signal is separated into a periodic component considered to correspond to the fundamental frequency (or pitch period) of the voice and another component (in other words, a non-periodic component). A method of encoding is known. The code excitation type linear predictive coding method (CELP) is one of the methods for coding the excitation signal. In this encoding method, as shown in FIG. 1, a parameter representing the envelope shape of the frequency spectrum of the input speech X input to the input terminal 11 is calculated in the linear prediction analysis unit 12. Linear predictive analysis methods are commonly used for this analysis. This linear prediction parameter is coded in the linear prediction parameter coding unit 13, the coded output A is decoded in the linear prediction parameter decoding unit 14, and the decoded linear prediction parameter a ′ is the linear prediction synthesis unit. It is set as a filter coefficient of 15. By giving the excitation signal (vector) E, which will be described later, to the linear prediction synthesis unit 15, the reproduced synthetic speech X ′ is obtained.

【0004】ここで、励振信号(ベクトル)Eについて
説明する。符号帳16は一定の励振ベクトルを多数保持
して切り替え使用する様にするか、或は常に直前のフレ
ームの確定された励振ベクトルが保持される様に構成す
る。この励振ベクトルから或る周期(ピッチ周期)に相
当する長さLのセグメントが切り出され、その切り出さ
れたたベクトルセグメントをフレームの長さTになるま
で繰り返し接続して音声の周期成分と対応する符号ベク
トルが出力される。符号帳16に周期符号(切り出し長
と同じ記号Lで表す)として与える切り出し長Lを変え
ることにより異なる周期成分と対応する符号ベクトルを
出力することができる。以下、符号帳から出力される符
号ベクトルを適応符号ベクトルと称す。
Here, the excitation signal (vector) E will be described. The codebook 16 is configured such that a large number of constant excitation vectors are held and used for switching, or that the fixed excitation vector of the immediately preceding frame is always held. A segment having a length L corresponding to a certain period (pitch period) is cut out from this excitation vector, and the cut out vector segment is repeatedly connected until the length T of the frame is reached to correspond to the periodic component of the voice. The code vector is output. By changing the cutout length L given to the codebook 16 as a periodic code (denoted by the same symbol L as the cutout length), it is possible to output code vectors corresponding to different periodic components. Hereinafter, the code vector output from the code book is referred to as an adaptive code vector.

【0005】符号帳17は乱数符号帳であって、これは
1個或はそれ以上設けられるが、以下の説明は2個の乱
数符号帳171 、172 が設けられる場合について説明
である。各乱数符号帳171 、172 は通常白色ガウス
性ランダム雑音を基調とし、1フレーム分の長さLの各
種の内臓ベクトルが入力音声とは独立にあらかじめ記憶
されており、与えられた乱数符号C(C1 ,C2 )によ
りそれぞれ指定されたベクトルが読みだされ、それぞれ
音声の非周期成分と対応する符号ベクトルとして出力さ
れる。以下、乱数符号帳17から出力される符号ベクト
ルを乱数符号ベクトルと称す。
The codebook 17 is a random number codebook, and one or more codebooks are provided. The following description is for a case where two random number codebooks 17 1 and 17 2 are provided. Each of the random number codebooks 17 1 and 17 2 is usually based on white Gaussian random noise, and various built-in vectors having a length L for one frame are stored in advance independently of the input speech, and the given random number code is given. Vectors designated by C (C 1 , C 2 ) are read out and output as code vectors respectively corresponding to the aperiodic components of speech. Hereinafter, the code vector output from the random codebook 17 is referred to as a random code vector.

【0006】符号帳16および符号帳17から出力され
る各符号ベクトルは利得量子化部20において利得調整
される。即ち各符号ベクトルはそれぞれ利得調整部21
0 ,211 ,212 において符号帳23から出力される
利得g0 ,g1 ,g2 により利得調整され、これらの結
果は加算部22において加算される。符号帳23は与え
られた利得符号Gに従って利得g0 ,g1 ,g2 を切り
替え、或は作成する。加算部22の加算出力Eは励振ベ
クトル候補として線形予測合成部15に供給され、合成
部15から合成再生音声X’が出力される。入力端子1
1から入力される入力音声Xに対するこの合成音声X’
の歪dが歪計算部18において計算される。聴感補正部
19は歪dを最小化する基準に基づいて、先ず、符号帳
16における切り出し長さLを検索し、符号帳16の最
適符号ベクトルを決定する。次いで符号帳17から乱数
符号ベクトルを決定し、更に利得量子化部20の最適利
得g0 ,g1 ,g2 を決定する。以上の手順により歪d
が最小になる様な符号の組み合わせが検索され、その時
の励振ベクトル候補として現フレームの励振ベクトルE
が確定される。歪dが最小となったときの符号帳16の
切り出し長を示す周期符号Lと、符号帳171 ,172
の各符号ベクトルを示す乱数符号C1 ,C2と、利得g
0 ,g1 ,g2 を示す利得符号Gと、線形予測パラメー
タ符号Aとが符号化出力として出力され、伝送または蓄
積される。
The gain quantizer 20 adjusts the gain of each code vector output from the code book 16 and the code book 17. That is, each code vector corresponds to the gain adjusting unit 21.
The gains 0 , 21 1 , 21 2 are adjusted by the gains g 0 , g 1 , g 2 output from the codebook 23, and these results are added by the adder 22. The codebook 23 switches or creates the gains g 0 , g 1 and g 2 according to the given gain code G. The addition output E of the adding unit 22 is supplied to the linear prediction synthesizing unit 15 as an excitation vector candidate, and the synthesizing unit 15 outputs the synthesized reproduced voice X ′. Input terminal 1
This synthesized speech X ′ corresponding to the input speech X input from 1
The strain d of is calculated in the strain calculator 18. The auditory sense correction unit 19 first searches the cutout length L in the codebook 16 based on the criterion for minimizing the distortion d, and determines the optimum code vector of the codebook 16. Next, the random code vector is determined from the codebook 17, and the optimum gains g 0 , g 1 and g 2 of the gain quantizer 20 are further determined. Distortion d
A combination of codes that minimizes is searched, and the excitation vector E of the current frame is set as the excitation vector candidate at that time.
Is confirmed. The periodic code L indicating the cutout length of the codebook 16 when the distortion d becomes the minimum and the codebooks 17 1 and 17 2
Random number codes C 1 and C 2 indicating each code vector of G and the gain g
A gain code G indicating 0 , g 1 and g 2 and a linear prediction parameter code A are output as coded outputs and are transmitted or stored.

【0007】[0007]

【発明が解決しようとする課題】これら励振信号ベクト
ルの形状および利得の量子化の内の利得の量子化につい
ては、励振信号ベクトル毎に対応する利得をスカラ量子
化する方法と、複数の励振信号ベクトルに対応する利得
を一括して量子化するベクトル量子化方法とがある。こ
こで、これら量子化方法の特性についてであるが、スカ
ラ量子化方法は必要とされるメモリ量は僅かであるが波
形歪を小さくするには難点のあるものである一方、ベク
トル量子化方法は波形歪を小さくするには好適であるが
大なるメモリ量の符号帳を必要とするものである。
Among these excitation signal vector shapes and gain quantization, gain quantization is performed by a method of scalar quantization of the gain corresponding to each excitation signal vector and a plurality of excitation signals. There is a vector quantization method for collectively quantizing the gains corresponding to vectors. Regarding the characteristics of these quantizing methods, the scalar quantizing method requires a small amount of memory but has a difficulty in reducing the waveform distortion, while the vector quantizing method Although it is suitable for reducing the waveform distortion, it requires a codebook with a large memory amount.

【0008】そして、利得をベクトル量子化する場合、
図2に示される如く利得を切り替え選択するために使用
する複数の符号帳23を具備し、入力音声を特徴分析部
30により分析した結果である入力音声の性質を使用し
て適応的にこれら符号帳23を切り替え使用する手法も
ある。この様にすることにより波形歪を削減することは
できるが、符号帳23のメモリ量は一般に大きく、これ
を複数具備することにおり全メモリ量は更に増大すると
いう問題があった。
Then, in the case of vector quantizing the gain,
As shown in FIG. 2, a plurality of codebooks 23 used for switching and selecting the gain are provided, and these codes are adaptively used by using the property of the input voice which is the result of analyzing the input voice by the feature analysis unit 30. There is also a method of switching and using the book 23. By doing so, the waveform distortion can be reduced, but the memory capacity of the codebook 23 is generally large, and since a plurality of codebooks 23 are provided, the total memory capacity further increases.

【0009】この発明は、少ない情報量のもとでCELP符
号化による波形歪を小さくすることができる効率のよい
励振源の利得の量子化方法を提供するものである。
The present invention provides an efficient method of quantizing the gain of an excitation source which can reduce the waveform distortion due to CELP coding under a small amount of information.

【0010】[0010]

【課題を解決するための手段】入力音声を一定のサンプ
リング周期毎に線形予測分析12し、予測係数より成る
線形予測合成部15通過後の合成信号X’と入力音声X
との間の歪を最小とする複数の励振信号の形状および利
得を決定する線形予測符号化方法に使用される励振信号
の利得量子化方法において、2段階の利得量子化を実施
し、第1段階の利得量子化においては利得を入力音声の
特徴に合わせて適応的に量子化し、第2段階の利得量子
化においては第1段階における利得量子化の結果を補う
様に利得を一括してベクトル量子化する励振信号の利得
量子化方法を構成した。
The input speech is subjected to a linear prediction analysis 12 at regular sampling intervals, and a synthetic signal X'after passing through a linear prediction synthesis section 15 composed of prediction coefficients and an input speech X.
In the excitation signal gain quantization method used in a linear predictive coding method for determining the shape and gain of a plurality of excitation signals that minimize the distortion between In the step-wise gain quantization, the gain is adaptively quantized according to the characteristics of the input speech, and in the second-step gain quantization, the gains are collectively vectored so as to supplement the result of the first-step gain quantization. The gain quantization method of the excitation signal to be quantized is constructed.

【0011】[0011]

【実施例】この発明の励振信号の利得量子化方法は、要
約すれば、複数の励振信号に対する利得を2段階に分け
て量子化するものであり、第1段階においては利得を入
力音声の特徴に合わせて適応的に量子化し、第2段階に
おいては第1段階における利得量子化の結果を補う様に
利得を一括してベクトル量子化するものである。
BEST MODE FOR CARRYING OUT THE INVENTION In summary, the method of gain quantization of an excitation signal according to the present invention is to quantize the gain for a plurality of excitation signals in two stages. In the second step, the vector quantization is collectively performed so as to compensate the result of the gain quantization in the first step.

【0012】図3を参照してこの発明の第1の実施例を
具体的に説明する。先ず第1段階において符号帳16を
選択して得られる第1の励振ベクトルの利得についての
みスカラ量子化を実施し、第2段階において第1段階に
おける利得量子化の結果を補う様に、第1の励振ベクト
ルの利得と符号帳17を選択して得られる第2の励振信
号ベクトルの利得とを一括してベクトル量子化を実施す
る。
A first embodiment of the present invention will be specifically described with reference to FIG. First, in the first stage, scalar quantization is performed only on the gain of the first excitation vector obtained by selecting the codebook 16, and in the second stage, the result of the gain quantization in the first stage is supplemented. And the gain of the second excitation signal vector obtained by selecting the codebook 17 are collectively subjected to vector quantization.

【0013】符号帳16は、通常は、常に直前のフレー
ムの確定された励振ベクトルを保持する様に構成した適
応符号帳であり、第1の励振ベクトルはこの適応符号帳
16からの出力であり、第2の励振信号ベクトルは乱数
符号帳である符号帳17からの出力である。量子化テー
ブル31は利得gの切り替えのためのものであり、情報
量の極く少ないもので事足りる。特徴分析部30により
分析された音声の性質に合わせて適応的に切り替え使用
される。
The codebook 16 is usually an adaptive codebook configured to always hold the fixed excitation vector of the immediately preceding frame, and the first excitation vector is an output from the adaptive codebook 16. , The second excitation signal vector is an output from the codebook 17, which is a random codebook. The quantization table 31 is for switching the gain g, so that the amount of information is extremely small. It is adaptively switched and used according to the nature of the voice analyzed by the feature analysis unit 30.

【0014】励振信号の決定方法は下記の如きものであ
る。 第1。 最初に、第1の励振ベクトルの形状を決定す
る。この処理は適応符号帳16を使用する。通常の場合
はピッチ周期を求めることとほぼ等価である。 第2。ここで、第1段階の利得量子化を第1段階の利得
量子化部201 において実施する。この処理は第1の励
振ベクトルのみを合成したときの合成信号x'と直前の
フレームからの応答分を差し引いた入力音声yについ
て、y−x' の歪を最小とする利得gを決定する。とこ
ろで、当該フレームの入力音声のレベルが例えば0であ
っても、直前のフレームの入力音声の影響は当該フレー
ムにも及ぶところから当該フレームのレベルは直前のフ
レームからの応答分である僅かのレベルを有するもので
ある。従って、真の当該フレームの入力音声yを求める
ために見かけ上の入力音声から直前のフレームからの応
答分を差し引いく。利得gの量子化テーブル31は情報
量の極く少ないもので事足り、特徴分析部30により分
析された音声の性質に合わせて適応的に切り替え使用す
ることにより歪を小さくすることができる。切り替えの
パラメータとしては有声無声の情報、パワ、ピッチ周期
が考えられる。この第1段階の量子化はスカラ量子化で
あるので、数多くの量子化テーブルを使用しても量子化
テーブルのメモリ量自体極く僅かでもあるところから全
メモリ量の増加は問題とするに値しない。
The method of determining the excitation signal is as follows. First. First, the shape of the first excitation vector is determined. This process uses the adaptive codebook 16. In the normal case, it is almost equivalent to obtaining the pitch period. Second. Here, the first-stage gain quantization is performed in the first-stage gain quantization unit 20 1 . This processing determines the gain g that minimizes the distortion of y-x 'for the input signal y obtained by subtracting the response signal from the immediately preceding frame and the combined signal x'when only the first excitation vector is combined. By the way, even if the level of the input voice of the frame is 0, for example, the influence of the input voice of the previous frame extends to the frame, so the level of the frame is a slight level which is a response from the previous frame. Is to have. Therefore, in order to obtain the true input voice y of the frame, the response amount from the immediately preceding frame is subtracted from the apparent input voice. The quantization table 31 of the gain g need only have a very small amount of information, and the distortion can be reduced by adaptively switching and using it according to the nature of the voice analyzed by the feature analysis unit 30. Voiced and unvoiced information, power, and pitch period can be considered as switching parameters. Since the quantization in the first stage is scalar quantization, even if many quantization tables are used, the amount of memory in the quantization table itself is very small. Therefore, the increase in the total amount of memory is a problem. do not do.

【0015】第3。第2の励振ベクトルの形状vを決定
する。通常は乱数符号帳17から選択する。このときy
−gx' に対する歪を最小化する様に選択する。 第4。最後に、第2段階の利得のベクトル量子化を第2
段階の利得のベクトル量子化部202 において実施す
る。第2段階においては第1段階における利得量子化の
結果を補う様に第1の励振ベクトルの利得と符号帳17
を選択して得られる第2の励振信号ベクトルの利得とを
一括してベクトル量子化を実施する。合成後の波形歪が
最小となる様な利得ベクトルを利得の符号帳16および
17から選択する。
Third. Determine the shape v of the second excitation vector. Normally, the random number codebook 17 is selected. Then y
Choose to minimize distortion for -gx '. Fourth. Finally, the second-stage gain vector quantization
This is carried out in the vector quantizer 20 2 of the step gain. In the second stage, the gain of the first excitation vector and the codebook 17 are added so as to supplement the result of the gain quantization in the first stage.
And the gain of the second excitation signal vector, which is obtained by selecting, are collectively vector-quantized. A gain vector that minimizes waveform distortion after synthesis is selected from the gain codebooks 16 and 17.

【0016】図4を参照してこの発明の第2の実施例を
説明する。この場合、利得のベクトル量子化部202
おいて利得のベクトル量子化を実施するに先だって、時
間領域について各励振ベクトル毎にレベルの三角窓40
を乗算し、励振ベクトル信号を計4個の信号に分離す
る。この様に励振ベクトルにレベルの三角窓40を乗算
してフレームの前半および後半を強調する前操作を施す
ことにより、励振ベクトルを直接ベクトル量子化する第
1の実施例の場合と比較して歪をより小さくすることが
できる。
A second embodiment of the present invention will be described with reference to FIG. In this case, prior to performing the gain vector quantization in the gain vector quantization unit 20 2 , the triangular window 40 of the level for each excitation vector in the time domain is used.
And the excitation vector signal is separated into a total of four signals. In this way, the excitation vector is multiplied by the triangular window 40 of the level, and the pre-operation for emphasizing the first half and the second half of the frame is performed. Can be smaller.

【0017】図5を参照してこの発明の第3の実施例を
説明する。第2の実施例と同様に、各励振ベクトルに時
間領域の三角窓40を乗算してこれらを計4個の信号の
系統に分離している。第1段階における利得の量子化は
音声信号の特徴に合わせて量子化レベルを制御する。或
は、複数の量子化テーブルを具備してこれらを切り替え
て使用する。これに対して、第2段階における利得の量
子化は第1段階にける利得の量子化の結果を補う様に利
得を一括してベクトル量子化する。この場合、音声の特
徴とは無関係に全ての場合に共通に符号帳の中から歪を
最小にする利得ベクトルを選択してその符号を伝送す
る。
A third embodiment of the present invention will be described with reference to FIG. Similar to the second embodiment, each excitation vector is multiplied by the triangular window 40 in the time domain, and these are separated into a total of four signal systems. The gain quantization in the first stage controls the quantization level according to the characteristics of the audio signal. Alternatively, a plurality of quantization tables are provided and these are switched and used. On the other hand, in the quantization of the gain in the second stage, the gain is collectively vector-quantized so as to complement the result of the quantization of the gain in the first stage. In this case, a gain vector that minimizes distortion is selected from the codebook and transmitted in common in all cases regardless of the characteristics of the voice.

【0018】上述された何れの実施例の場合も、入力音
声のパワー、線形予測の予測利得、励振ベクトルのパワ
ーを使用して利得の量子化の前に信号の正規化を行なう
ことができ、スカラ量子化のステップ幅やベクトル量子
化の符号ベクトルの変動幅を小さくすることが可能であ
る。そして、通常第1段階にける利得の量子化は比較的
少ないビット数の量子化でよい。適応的に平均値のみを
変化させる量子化(0ビット量子化)を採用することが
できる。
In any of the embodiments described above, the power of the input speech, the prediction gain of the linear prediction, the power of the excitation vector can be used to perform signal normalization prior to gain quantization. It is possible to reduce the step width of the scalar quantization and the fluctuation width of the code vector of the vector quantization. Then, the quantization of the gain in the first stage is usually a quantization of a relatively small number of bits. Quantization (0-bit quantization) that adaptively changes only the average value can be adopted.

【0019】[0019]

【発明の効果】以上の通りであって、この発明は第1段
階における適応的量子化により、フレーム毎に変化する
音声の特徴に合わせた量子化がなされ、第2段階の量子
化において全ての励振信号を考慮したベクトル量子化を
実施することにより波形歪を小さくすることができる。
As described above, according to the present invention, the adaptive quantization in the first stage performs the quantization in accordance with the characteristics of the voice changing for each frame, and all the quantization in the second stage is performed. Waveform distortion can be reduced by performing vector quantization in consideration of the excitation signal.

【0020】即ち、この発明の利得のベクトル量子化方
法は第1段階における量子化により利得の大きな変動を
吸収するので、第2段階における量子化の符号帳の符号
ベクトルの変動範囲を通常のベクトル量子化の場合の変
動範囲と比較して小さくすることができる。従って、符
号誤りがある場合も、第1段階における符号ビットだけ
を保護すれば符号誤りの影響を軽減することができる。
That is, since the gain vector quantization method of the present invention absorbs a large variation in gain by the quantization in the first stage, the variation range of the code vector of the quantization codebook in the second stage is set to a normal vector. It can be made smaller than the variation range in the case of quantization. Therefore, even if there is a code error, the effect of the code error can be reduced by protecting only the code bit in the first stage.

【0021】そして、図1に示される様な従来の利得の
ベクトル量子化方法と比較して演算量およびメモリ量を
殆ど増加させることなくして歪を小さくすることができ
る。また、図2に示される様な適応的にベクトル量子化
の符号帳を切り替える利得のベクトル量子化方法と比較
して、歪を殆ど増加させることなくして符号帳のメモリ
量を大幅に削減することができる。
As compared with the conventional gain vector quantization method as shown in FIG. 1, the distortion can be reduced without substantially increasing the calculation amount and the memory amount. In addition, as compared to the vector quantization method of gain that adaptively switches the codebook of vector quantization as shown in FIG. 2, it is possible to significantly reduce the memory amount of the codebook without increasing the distortion. You can

【図面の簡単な説明】[Brief description of drawings]

【図1】複数の励振信号をもつCELP符号化方法の基本構
成を示す図。
FIG. 1 is a diagram showing a basic configuration of a CELP encoding method having a plurality of excitation signals.

【図2】適応的に利得を量子化するCELP符号化方法の従
来例を示す図。
FIG. 2 is a diagram showing a conventional example of a CELP encoding method for adaptively quantizing a gain.

【図3】この発明の励振信号の利得量子化方法の第1の
実施例を示す図。
FIG. 3 is a diagram showing a first embodiment of the method of gain quantization of an excitation signal according to the present invention.

【図4】この発明の励振信号の利得量子化方法の第2の
実施例を示す図。
FIG. 4 is a diagram showing a second embodiment of the method for quantizing the gain of an excitation signal according to the present invention.

【図5】この発明の励振信号の利得量子化方法の第3の
実施例を示す図。
FIG. 5 is a diagram showing a third embodiment of the excitation signal gain quantization method according to the present invention.

【符号の説明】[Explanation of symbols]

12 線形予測分析 15 線形予測合成部 16 符号帳 17 符号帳 201 利得の量子化部 202 利得のベクトル量子化部 30 特徴分析部 31 量子化テーブル12 linear prediction analysis 15 linear prediction synthesis unit 16 codebook 17 codebook 20 1 gain quantization unit 20 2 gain vector quantization unit 30 feature analysis unit 31 quantization table

───────────────────────────────────────────────────── フロントページの続き (72)発明者 大室 仲 東京都千代田区内幸町1丁目1番6号 日 本電信電話株式会社内 ─────────────────────────────────────────────────── ─── Continuation of the front page (72) Inventor Omuro Naka 1-16, Uchisaiwaicho, Chiyoda-ku, Tokyo Nihon Telegraph and Telephone Corporation

Claims (1)

【特許請求の範囲】[Claims] 【請求項1】 入力音声を一定のサンプリング周期毎に
線形予測分析し、予測係数より成る線形予測合成部通過
後の合成信号と入力音声との間の歪を最小とする複数の
励振信号の形状および利得を決定する線形予測符号化方
法に使用される励振信号の利得量子化方法において、2
段階の利得量子化を実施し、第1段階の利得量子化にお
いては利得を入力音声の特徴に合わせて適応的に量子化
し、第2段階の利得量子化においては第1段階における
利得量子化の結果を補う様に利得を一括してベクトル量
子化することを特徴とする励振信号の利得量子化方法。
1. Shapes of a plurality of excitation signals that minimize distortion between a synthesized signal after passing through a linear prediction synthesis section composed of prediction coefficients and the input speech by linearly analyzing the input speech at regular sampling intervals. And a gain quantization method for an excitation signal used in a linear predictive coding method for determining the gain and
Stage gain quantization is performed. In the first stage gain quantization, the gain is adaptively quantized according to the characteristics of the input speech, and in the second stage gain quantization, the gain quantization in the first stage is performed. A gain quantization method for an excitation signal, characterized in that the gains are vector-quantized in a batch so as to supplement the result.
JP05001110A 1993-01-07 1993-01-07 Excitation signal gain quantization method Expired - Lifetime JP3099852B2 (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
JP05001110A JP3099852B2 (en) 1993-01-07 1993-01-07 Excitation signal gain quantization method

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
JP05001110A JP3099852B2 (en) 1993-01-07 1993-01-07 Excitation signal gain quantization method

Publications (2)

Publication Number Publication Date
JPH06202697A true JPH06202697A (en) 1994-07-22
JP3099852B2 JP3099852B2 (en) 2000-10-16

Family

ID=11492334

Family Applications (1)

Application Number Title Priority Date Filing Date
JP05001110A Expired - Lifetime JP3099852B2 (en) 1993-01-07 1993-01-07 Excitation signal gain quantization method

Country Status (1)

Country Link
JP (1) JP3099852B2 (en)

Cited By (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JPH08272395A (en) * 1995-03-31 1996-10-18 Nec Corp Voice encoding device
WO1998020483A1 (en) * 1996-11-07 1998-05-14 Matsushita Electric Industrial Co., Ltd. Sound source vector generator, voice encoder, and voice decoder
JP2016533528A (en) * 2013-10-18 2016-10-27 フラウンホーファー−ゲゼルシャフト・ツール・フェルデルング・デル・アンゲヴァンテン・フォルシュング・アインゲトラーゲネル・フェライン Audio signal coding and decoding concept using speech-related spectral shaping information
US10304470B2 (en) 2013-10-18 2019-05-28 Fraunhofer-Gesellschaft Zur Foerderung Der Angewandten Forschung E.V. Concept for encoding an audio signal and decoding an audio signal using deterministic and noise like information

Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JPH02144598A (en) * 1988-11-25 1990-06-04 Nippon Telegr & Teleph Corp <Ntt> Sound coding method
JPH041800A (en) * 1990-04-19 1992-01-07 Nec Corp Voice frequency band signal coding system
JPH04270400A (en) * 1991-02-26 1992-09-25 Nec Corp Voice encoding system

Patent Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JPH02144598A (en) * 1988-11-25 1990-06-04 Nippon Telegr & Teleph Corp <Ntt> Sound coding method
JPH041800A (en) * 1990-04-19 1992-01-07 Nec Corp Voice frequency band signal coding system
JPH04270400A (en) * 1991-02-26 1992-09-25 Nec Corp Voice encoding system

Cited By (26)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JPH08272395A (en) * 1995-03-31 1996-10-18 Nec Corp Voice encoding device
WO1998020483A1 (en) * 1996-11-07 1998-05-14 Matsushita Electric Industrial Co., Ltd. Sound source vector generator, voice encoder, and voice decoder
US6330534B1 (en) 1996-11-07 2001-12-11 Matsushita Electric Industrial Co., Ltd. Excitation vector generator, speech coder and speech decoder
US6330535B1 (en) 1996-11-07 2001-12-11 Matsushita Electric Industrial Co., Ltd. Method for providing excitation vector
US6345247B1 (en) 1996-11-07 2002-02-05 Matsushita Electric Industrial Co., Ltd. Excitation vector generator, speech coder and speech decoder
US6421639B1 (en) 1996-11-07 2002-07-16 Matsushita Electric Industrial Co., Ltd. Apparatus and method for providing an excitation vector
US6453288B1 (en) 1996-11-07 2002-09-17 Matsushita Electric Industrial Co., Ltd. Method and apparatus for producing component of excitation vector
US6757650B2 (en) 1996-11-07 2004-06-29 Matsushita Electric Industrial Co., Ltd. Excitation vector generator, speech coder and speech decoder
US6772115B2 (en) 1996-11-07 2004-08-03 Matsushita Electric Industrial Co., Ltd. LSP quantizer
US6799160B2 (en) 1996-11-07 2004-09-28 Matsushita Electric Industrial Co., Ltd. Noise canceller
US6910008B1 (en) 1996-11-07 2005-06-21 Matsushita Electric Industries Co., Ltd. Excitation vector generator, speech coder and speech decoder
US6947889B2 (en) 1996-11-07 2005-09-20 Matsushita Electric Industrial Co., Ltd. Excitation vector generator and a method for generating an excitation vector including a convolution system
US7289952B2 (en) 1996-11-07 2007-10-30 Matsushita Electric Industrial Co., Ltd. Excitation vector generator, speech coder and speech decoder
US7398205B2 (en) 1996-11-07 2008-07-08 Matsushita Electric Industrial Co., Ltd. Code excited linear prediction speech decoder and method thereof
US7587316B2 (en) 1996-11-07 2009-09-08 Panasonic Corporation Noise canceller
US7809557B2 (en) 1996-11-07 2010-10-05 Panasonic Corporation Vector quantization apparatus and method for updating decoded vector storage
US8036887B2 (en) 1996-11-07 2011-10-11 Panasonic Corporation CELP speech decoder modifying an input vector with a fixed waveform to transform a waveform of the input vector
US8086450B2 (en) 1996-11-07 2011-12-27 Panasonic Corporation Excitation vector generator, speech coder and speech decoder
US8370137B2 (en) 1996-11-07 2013-02-05 Panasonic Corporation Noise estimating apparatus and method
JP2016533528A (en) * 2013-10-18 2016-10-27 フラウンホーファー−ゲゼルシャフト・ツール・フェルデルング・デル・アンゲヴァンテン・フォルシュング・アインゲトラーゲネル・フェライン Audio signal coding and decoding concept using speech-related spectral shaping information
US10304470B2 (en) 2013-10-18 2019-05-28 Fraunhofer-Gesellschaft Zur Foerderung Der Angewandten Forschung E.V. Concept for encoding an audio signal and decoding an audio signal using deterministic and noise like information
US10373625B2 (en) 2013-10-18 2019-08-06 Fraunhofer-Gesellschaft Zur Foerderung Der Angewandten Forschung E.V. Concept for encoding an audio signal and decoding an audio signal using speech related spectral shaping information
US10607619B2 (en) 2013-10-18 2020-03-31 Fraunhofer-Gesellschaft Zur Foerderung Der Angewandten Forschung E.V. Concept for encoding an audio signal and decoding an audio signal using deterministic and noise like information
US10909997B2 (en) 2013-10-18 2021-02-02 Fraunhofer-Gesellschaft Zur Foerderung Der Angewandten Forschung E.V. Concept for encoding an audio signal and decoding an audio signal using speech related spectral shaping information
US11798570B2 (en) 2013-10-18 2023-10-24 Fraunhofer-Gesellschaft zur Förderung der angewandten Forschung e.V. Concept for encoding an audio signal and decoding an audio signal using deterministic and noise like information
US11881228B2 (en) 2013-10-18 2024-01-23 Fraunhofer-Gesellschaft Zur Foerderung Der Angewandten Forschung E. V. Concept for encoding an audio signal and decoding an audio signal using speech related spectral shaping information

Also Published As

Publication number Publication date
JP3099852B2 (en) 2000-10-16

Similar Documents

Publication Publication Date Title
US6952671B1 (en) Vector quantization with a non-structured codebook for audio compression
JP2964344B2 (en) Encoding / decoding device
US7392179B2 (en) LPC vector quantization apparatus
EP0607989B1 (en) Voice coder system
JP3346765B2 (en) Audio decoding method and audio decoding device
EP1224662B1 (en) Variable bit-rate celp coding of speech with phonetic classification
JP3114197B2 (en) Voice parameter coding method
US7065338B2 (en) Method, device and program for coding and decoding acoustic parameter, and method, device and program for coding and decoding sound
US5953698A (en) Speech signal transmission with enhanced background noise sound quality
US6134520A (en) Split vector quantization using unequal subvectors
EP0772186A2 (en) Speech encoding method and apparatus
JPH1091194A (en) Method of voice decoding and device therefor
CA2182790A1 (en) High-frequency signal branching device
US6988067B2 (en) LSF quantizer for wideband speech coder
JP2800618B2 (en) Voice parameter coding method
JPH08272395A (en) Voice encoding device
JPH06202697A (en) Gain quantizing method for excitation signal
JP2538450B2 (en) Speech excitation signal encoding / decoding method
JP2613503B2 (en) Speech excitation signal encoding / decoding method
US5943644A (en) Speech compression coding with discrete cosine transformation of stochastic elements
JP3916934B2 (en) Acoustic parameter encoding, decoding method, apparatus and program, acoustic signal encoding, decoding method, apparatus and program, acoustic signal transmitting apparatus, acoustic signal receiving apparatus
JPH11219196A (en) Speech synthesizing method
JPH05113799A (en) Code driving linear prediction coding system
JP3808270B2 (en) Speech coding apparatus, speech decoding apparatus, and codeword arrangement method
JP2002073097A (en) Celp type voice coding device and celp type voice decoding device as well as voice encoding method and voice decoding method

Legal Events

Date Code Title Description
FPAY Renewal fee payment (event date is renewal date of database)

Free format text: PAYMENT UNTIL: 20070818

Year of fee payment: 7

FPAY Renewal fee payment (event date is renewal date of database)

Free format text: PAYMENT UNTIL: 20080818

Year of fee payment: 8

FPAY Renewal fee payment (event date is renewal date of database)

Free format text: PAYMENT UNTIL: 20080818

Year of fee payment: 8

FPAY Renewal fee payment (event date is renewal date of database)

Free format text: PAYMENT UNTIL: 20090818

Year of fee payment: 9

FPAY Renewal fee payment (event date is renewal date of database)

Free format text: PAYMENT UNTIL: 20090818

Year of fee payment: 9

FPAY Renewal fee payment (event date is renewal date of database)

Free format text: PAYMENT UNTIL: 20100818

Year of fee payment: 10

FPAY Renewal fee payment (event date is renewal date of database)

Free format text: PAYMENT UNTIL: 20100818

Year of fee payment: 10

FPAY Renewal fee payment (event date is renewal date of database)

Free format text: PAYMENT UNTIL: 20110818

Year of fee payment: 11

FPAY Renewal fee payment (event date is renewal date of database)

Free format text: PAYMENT UNTIL: 20120818

Year of fee payment: 12

FPAY Renewal fee payment (event date is renewal date of database)

Free format text: PAYMENT UNTIL: 20130818

Year of fee payment: 13

EXPY Cancellation because of completion of term
FPAY Renewal fee payment (event date is renewal date of database)

Free format text: PAYMENT UNTIL: 20130818

Year of fee payment: 13