WO2009097763A1 - 一种增益量化方法及装置 - Google Patents

一种增益量化方法及装置 Download PDF

Info

Publication number
WO2009097763A1
WO2009097763A1 PCT/CN2009/070119 CN2009070119W WO2009097763A1 WO 2009097763 A1 WO2009097763 A1 WO 2009097763A1 CN 2009070119 W CN2009070119 W CN 2009070119W WO 2009097763 A1 WO2009097763 A1 WO 2009097763A1
Authority
WO
WIPO (PCT)
Prior art keywords
codebook gain
fixed codebook
gain
subframe
quantization
Prior art date
Application number
PCT/CN2009/070119
Other languages
English (en)
French (fr)
Inventor
Tao Zhang
Hai Zhang
Xin Li
Jialin He
Dejun Zhang
Longyin Chen
Original Assignee
Huawei Technologies Co., Ltd.
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 Huawei Technologies Co., Ltd. filed Critical Huawei Technologies Co., Ltd.
Publication of WO2009097763A1 publication Critical patent/WO2009097763A1/zh

Links

Classifications

    • G—PHYSICS
    • G10—MUSICAL INSTRUMENTS; ACOUSTICS
    • G10L—SPEECH ANALYSIS TECHNIQUES OR SPEECH SYNTHESIS; SPEECH RECOGNITION; SPEECH OR VOICE PROCESSING TECHNIQUES; SPEECH OR AUDIO CODING OR DECODING
    • G10L19/00—Speech or audio signals analysis-synthesis techniques for redundancy reduction, e.g. in vocoders; Coding or decoding of speech or audio signals, using source filter models or psychoacoustic analysis
    • G10L19/04—Speech or audio signals analysis-synthesis techniques for redundancy reduction, e.g. in vocoders; Coding or decoding of speech or audio signals, using source filter models or psychoacoustic analysis using predictive techniques
    • G10L19/08—Determination or coding of the excitation function; Determination or coding of the long-term prediction parameters
    • G10L19/083—Determination or coding of the excitation function; Determination or coding of the long-term prediction parameters the excitation function being an excitation gain

Definitions

  • the present invention relates to coding techniques, and in particular, to a gain quantization method and apparatus.
  • CELP Code Excited Linear Prediction
  • This model is derived from its inverse process, that is, F(z) is used to remove the near-point redundancy of the speech signal, and P(z) is used to remove the far-point redundancy of the speech signal, which is obtained through two-stage prediction.
  • the normalized residual signal approximates the standard normal distribution.
  • the excitation signal of this model is usually selected from the codebook composed of Gaussian white noise sequences, so it is called code excitation linear prediction model.
  • the quantization of the adaptive codebook gain (pitch gain) ga and the fixed codebook gain gs has a great influence on the quality of the synthesized speech. If the gain quantization is coarse, the quantization noise of the synthesized speech is synthesized. Larger, the naturalness and sharpness of synthesized speech are poor. High-quality speech coding algorithms require high-quality gain quantization.
  • the principle of gain quantization for adaptive codebook gain and fixed codebook gain is to make original speech and reconstructed speech. The perceptual weighted mean square error is minimized.
  • the gain of one frame signal is quantized with 30 bits:
  • the estimated energy value for fixed codebook gain prediction is quantized by 2 bits per frame, one frame signal is divided into four sub-frames, and the adaptive codebook and fixed code of each sub-frame are used.
  • the energy of each sub-frame is calculated by equation (1):
  • N 64 is the subframe
  • c (i) is the fixed codebook excitation
  • g c is the fixed-codebook gains
  • codebook is the average energy, which may preclude the use of calculation of formula (2):
  • N-1 can obtain the estimated energy value of the fixed codebook according to formula (1) as shown in equation (3):
  • the average residual energy of each frame can be calculated, which can be calculated by the formula (6):
  • ⁇ is the cross-correlated normalized energy average obtained by removing the two open-loop pitch analysis
  • 10R is the contribution estimate of the adaptive codebook
  • the estimated energy value is quantized with 2 bits to obtain a fixed codebook gain, and the quantization level can be 18, 30, 42, 54.
  • the adaptive codebook gain and the correction factor for each subframe are combined with the 7-bit joint vector quantization, specifically two-dimensional vector quantization, to complete the gain quantization of one frame of the signal.
  • the embodiment of the present invention provides a gain quantization method and apparatus.
  • the technical solution provided by the embodiment of the present invention may not quantize the estimated energy value, thereby saving quantization bits.
  • the invention provides a gain quantization method, comprising:
  • the prediction coefficients of the fixed codebook gains of the remaining subframes are quantized.
  • the present invention also provides a computer program product comprising computer program code, wherein when the computer program code is executed by a computer, the computer program code can cause the computer to perform any of the gain quantization methods A step.
  • the present invention also provides a computer readable storage medium, the computer storing computer program code, when the computer program code is executed by a computer, the computer program code can cause the computer to perform any of gain quantization methods A step.
  • the invention also provides a gain quantization device, comprising:
  • a fixed codebook gain calculation unit for calculating a fixed codebook gain of at least one subframe of a frame signal
  • a fixed codebook gain quantization unit configured to quantize the fixed codebook gain calculated by the fixed codebook gain calculation unit to obtain a quantized value of the fixed codebook gain of the at least one subframe
  • a coefficient calculation unit configured to: Calculating a prediction coefficient of a fixed codebook gain of the remaining subframes using a quantized value of the fixed codebook gain and a fixed codebook gain of the remaining subframes of the one frame signal;
  • a coefficient quantization unit configured to: The prediction coefficients of the fixed codebook gains of the remaining sub-frames are quantized.
  • the embodiment of the present invention uses the quantized value of the fixed codebook gain of the first subframe to calculate the prediction coefficient of the fixed codebook gain of the remaining subframes, thereby making the remaining sub
  • the prediction coefficient of the fixed codebook gain of the frame is related to the quantized value of the fixed codebook gain of the first subframe, which fully utilizes the strong correlation of the fixed codebook gain between each subframe, so that the estimated energy value does not need to be calculated. It is also not necessary to quantify the estimated energy value, so that bits that require quantization of the estimated energy value can be saved.
  • Embodiment 1 is a flowchart of Embodiment 1 of a gain quantization method according to an embodiment of the present invention
  • Embodiment 2 is a flowchart of Embodiment 2 of a gain quantization method according to an embodiment of the present invention
  • Embodiment 3 is a structural diagram of Embodiment 1 of a gain quantization apparatus according to an embodiment of the present invention.
  • FIG. 4 is a structural diagram of Embodiment 2 of a gain quantization apparatus according to an embodiment of the present invention.
  • FIG. 1 is a flowchart of a first embodiment of a gain quantization method according to an embodiment of the present invention, including: Step 101: Calculating a fixed codebook gain of a subframe in a frame signal;
  • the above one subframe is any one of the one frame signals, but the embodiment of the present invention preferably uses the first subframe of the one frame signal;
  • the embodiment of the present invention does not limit the calculation of the fixed codebook gain of only one subframe in a frame signal, and may also calculate the fixed codebook gain of more than one subframe. In the following embodiments, the calculation is used.
  • the fixed codebook gain of one subframe is described, but is not intended to limit the embodiments of the present invention.
  • the process of calculating the fixed codebook gain of one or more subframes is substantially the same as the process of calculating the fixed codebook gain of one subframe, which is not described in detail in the embodiments of the present invention.
  • the fixed codebook gain can be calculated by using a conventional fixed codebook gain calculation method, and the fixed codebook gain can be calculated by other methods; how to calculate the fixed codebook gain does not affect the implementation of the embodiment of the present invention. Therefore, the embodiment of the present invention does not limit the specific calculation process of the fixed codebook gain.
  • a method for calculating a fixed codebook gain according to an embodiment of the present invention is performed by using the following formula (9) Calculation:
  • vector z is the convolution of the fixed codebook vector and the pulse of the perceptually weighted synthesis filter
  • vector ⁇ 2 is the target vector of the fixed codebook search.
  • Step 102 Quantify a fixed codebook gain to obtain a quantized value of a fixed codebook gain of the one subframe.
  • the fixed codebook gain When the fixed codebook gain is quantized, scalar quantization or vector quantization can be performed. If scalar quantization is performed, the fixed codebook gain can be directly quantized by the quantized step size of the fixed codebook, or it can be fixed. The codebook gain is mapped to the logarithmic domain and then subjected to non-uniform scalar quantization; the adaptive codebook gain of the first subframe can be further quantized, and when the fixed codebook gain is quantized, scalar quantization can be performed.
  • Vector quantization can be performed; if scalar quantization is performed, uniform scalar quantization can be performed directly on the adaptive codebook gain with a uniform quantization step size, or the adaptive codebook gain can be mapped to a logarithmic domain and then non-homogenous scalars can be performed.
  • Step 103 Calculate a fixed codebook gain of the remaining subframes of the one frame signal.
  • the calculation of the fixed codebook gain of the remaining sub-frames may be calculated by using equation (9), or may be calculated by other methods. How to calculate the fixed codebook gain of the remaining sub-frames does not affect the embodiment of the present invention.
  • the implementation of the present invention does not limit the specific calculation process of the fixed codebook gain of the remaining subframes;
  • Step 104 Calculate a prediction coefficient of a fixed codebook gain of the remaining subframes by using a quantized value of the fixed codebook gain of the one subframe and a fixed codebook gain of the remaining subframes of the one frame signal;
  • the prediction coefficients of the fixed codebook gains of the remaining subframes may be calculated by using the quantized values of the fixed codebook gains of the one subframe and the fixed codebook gains of the remaining subframes respectively;
  • Step 105 Quantify the prediction coefficients of the fixed codebook gains of the remaining subframes.
  • the prediction coefficients of the fixed codebook gains of the remaining subframes may be vector quantized or scalar quantized. However, in the embodiment of the present invention, when scalar quantization is performed on the fixed codebook gain of the first subframe, a fixed code for the remaining subframes is preferred.
  • the prediction coefficient of the book gain is vector-quantized, thereby combining scalar quantization and vector quantization to improve the quantization precision.
  • this embodiment uses the quantized value of the fixed codebook gain of the first subframe to calculate the remaining children.
  • the prediction coefficient of the fixed codebook gain of the frame so that the prediction coefficient of the fixed codebook gain of the remaining subframe is correlated with the quantized value of the fixed codebook gain of the first subframe, and the fixed codebook gain between each subframe is fully utilized.
  • the correlation is strong, so that there is no need to calculate the estimated energy value, nor is it necessary to quantize the estimated energy value, so that the bit that needs to quantize the estimated energy value can be saved; when the number of quantization bits is constant, the The quantized bits of the estimated energy value are allocated to the sub-frames in one frame of the signal, thereby increasing the quantization accuracy.
  • FIG. 2 is a flowchart of a second embodiment of a gain quantization method according to an embodiment of the present invention, including: Step 201: Calculating a fixed codebook gain and an adaptive codebook gain of a subframe in a frame signal; Step 202: The fixed codebook gain of the subframe is quantized to obtain a quantized value of the fixed codebook gain of the one subframe;
  • Step 203 Quantify an adaptive codebook of the foregoing one subframe.
  • Steps 202 and 203 do not have a sequence of time
  • Step 204 Calculate a fixed codebook gain and an adaptive codebook gain of the remaining subframes of the one frame signal.
  • Step 205 Calculate a prediction coefficient of a fixed codebook gain of the remaining subframes by using a quantized value of the fixed codebook gain of the one subframe and a fixed codebook gain of the remaining subframes of the one frame signal.
  • Step 206 Quantify the prediction coefficients of the fixed codebook gain and the adaptive codebook gain of the remaining subframes.
  • the prediction coefficient of the fixed codebook gain and the adaptive codebook gain may be vector-quantified into a two-dimensional vector, or scalar quantization may be performed separately.
  • the present embodiment uses the quantized value of the fixed codebook gain of the first subframe to calculate the prediction coefficient of the fixed codebook gain of the remaining subframes, so that the prediction coefficients of the fixed codebook gain of the remaining subframes are first.
  • the quantized value of the fixed codebook gain of each subframe is correlated, and the feature of strong fixed codebook gain between each subframe is fully utilized, so that it is not necessary to calculate the estimated energy value, and it is not necessary to quantify the estimated energy value, so
  • the bit that needs to quantize the estimated energy value is saved; when the number of quantization bits is constant, the quantization bit that should be given the estimated energy value can be allocated to the subframe in one frame signal, thereby increasing the quantization precision.
  • the gain quantization method provided by the embodiment of the present invention is described by taking a gain quantization of a frame signal having 4 subframes as an example.
  • the adaptive codebook gain and the fixed codebook gain of the first subframe of a frame signal are calculated.
  • the adaptive codebook gain and the fixed codebook gain algorithm can be used to calculate the adaptive codebook gain and
  • the fixed codebook gain can also be used to calculate the adaptive codebook gain and the fixed codebook gain according to the minimum mean square error criterion.
  • the adaptive codebook gain calculated according to the minimum mean square error criterion is the optimal adaptive codebook gain
  • the fixed codebook gain calculated according to the minimum mean square error criterion is the optimal fixed codebook gain, which is an embodiment of the present invention.
  • One way to solve for the best adaptive codebook gain and the best fixed codebook gain is as follows:
  • the adaptive codebook gain is denoted by ga
  • the fixed codebook gain is denoted by g s .
  • ga and gs need to satisfy equation (10):
  • g a (2xc[ ⁇ ]xc[2]-c[3]xc[4])/(c[4]xc[4]-4xc[0]xc[2])
  • the optimal adaptive codebook gain and the best fixed codebook gain are obtained, and the adaptive code is obtained.
  • the book gain and the fixed codebook gain are quantized, and the embodiment of the present invention preferably performs scalar quantization; the scalar quantization may be a scalar quantization or a non-uniform scalar quantization, for example, the scalar quantization may be performed on the gain directly by the quantization step of the uniform hook, Or the gain is first mapped to the logarithmic domain and then scalar quantized; for example, the 5-bit non-uniform scalar quantization is performed on the best fixed codebook gain, first the logarithm of the best fixed codebook gain is multiplied by 10, and its unit is Decibel (dB), and then quantized, the quantization step size can reach 10*lgl0 (3349.654392/15.848932)/31/2, which is 0.375dB, which can guarantee the quantization precision. Since the quantization accuracy can be improved by using the
  • the optimal adaptive codebook is quantized using 4 bits, and the quantization range may be 0.012445-1.296012; 5 The bits are gain quantized for the best fixed codebook gain, and the quantization range can be 15.848932-3349.654392. Since the quantization bit of 9 bits is allocated for the first subframe, the quantization precision of the first subframe can be further improved, and the estimated energy value is not quantized by using 2 bits, thereby improving the utilization of the quantization bits.
  • the prediction coefficient of the codebook gain is quantized by 7 bits.
  • the adaptive codebook gain and the prediction coefficient of the fixed codebook gain are two-dimensionally vector quantized or separately scalar-quantized.
  • vector quantization is preferably performed.
  • the adaptive codebook gain and the prediction coefficient of the fixed codebook gain can be combined into a two-dimensional vector, and the obtained two-dimensional vector is quantized.
  • the above is only a calculation method of the prediction coefficient of the fixed codebook gain provided by the embodiment of the present invention, and the embodiment of the present invention does not limit the prediction coefficient that can only calculate the fixed codebook gain by using the above manner. .
  • the calculation of the fixed codebook gain of the current subframe may be performed by using the method provided by the embodiment of the present invention, or the calculation method of the conventional fixed codebook gain may be used, and the embodiment of the present invention does not limit the fixed code of the current subframe. How to calculate the book gain.
  • the present embodiment uses the quantized value of the fixed codebook gain of the first subframe to calculate the prediction coefficient of the fixed codebook gain of the remaining subframes, so that the prediction coefficients of the fixed codebook gain of the remaining subframes are first.
  • the quantized value of the fixed codebook gain of each subframe is correlated, and the feature of strong fixed codebook gain between each subframe is fully utilized, so that it is not necessary to calculate the estimated energy value, and it is not necessary to quantify the estimated energy value, so Saving bits that need to quantize the estimated energy value; when the number of quantization bits is constant, the quantization bits that should be given the estimated energy value can be allocated to the subframes in one frame signal, thereby increasing the quantization precision; There are no complicated logarithmic and exponential operations, and there is no need to consume a large amount of computation.
  • FIG. 3 is a diagram showing the structure of the first embodiment of the gain quantization apparatus according to the embodiment of the present invention, including:
  • the fixed codebook gain calculating unit 301 is configured to calculate a fixed codebook gain of one subframe in one frame signal; and calculate a fixed codebook gain of the remaining subframes in one frame signal;
  • the fixed codebook gain quantization unit 302 is configured to quantize the fixed codebook gain of one subframe calculated by the fixed codebook gain calculation unit 301 to obtain a quantized value of the fixed codebook gain of the one subframe;
  • the coefficient calculation unit 303 is configured to calculate a prediction coefficient of the fixed codebook gain of the remaining subframes by using the quantized value of the fixed codebook gain of the one subframe and the fixed codebook gain of the remaining subframes;
  • the coefficient quantization unit 304 is configured to quantize the prediction coefficients of the fixed codebook gains of the remaining subframes calculated by the coefficient calculation unit 303.
  • the present embodiment uses the quantized value of the fixed codebook gain of the first subframe to calculate the prediction coefficient of the fixed codebook gain of the remaining subframes, so that the prediction coefficients of the fixed codebook gain of the remaining subframes are first.
  • the quantized value of the fixed codebook gain of the sub-frames is correlated, and the characteristics of the strong codebook correlation between the sub-frames are fully utilized, so that the predicted energy value does not need to be calculated, and the predicted energy value does not need to be quantized.
  • One subframe allocates more quantization bits, increasing the quantization precision of the first subframe.
  • FIG. 4 is a diagram showing a structure of a second embodiment of a gain quantization apparatus according to an embodiment of the present invention, including: a fixed codebook gain calculation unit 401, configured to calculate a fixed codebook gain of one subframe in a frame signal; Fixed codebook gain for the remaining subframes;
  • the adaptive codebook gain calculation unit 402 is configured to calculate an adaptive codebook gain of the one subframe.
  • the fixed codebook gain quantization unit 403 is configured to quantize the fixed codebook gain of the one subframe calculated by the fixed codebook gain calculation unit 401, to obtain a quantized value of the fixed codebook gain of the one subframe;
  • the adaptive codebook gain quantization unit 404 is configured to quantize the adaptive codebook gain of the one subframe obtained by the adaptive codebook gain calculation unit 402.
  • the coefficient calculation unit 405 is configured to calculate a prediction coefficient of the fixed codebook gain of the remaining subframes by using the quantized value of the fixed codebook gain of the one subframe and the fixed codebook gain of the remaining subframes;
  • the coefficient quantization unit 406 is configured to quantize the prediction coefficients of the fixed codebook gains of the remaining subframes calculated by the coefficient calculation unit 405.
  • the present embodiment uses the quantized value of the fixed codebook gain of the first subframe to calculate the prediction coefficient of the fixed codebook gain of the remaining subframes, so that the prediction coefficients of the fixed codebook gain of the remaining subframes are first.
  • the quantized value of the fixed codebook gain of each subframe is correlated, and the feature of strong fixed codebook gain between each subframe is fully utilized, so that it is not necessary to calculate the estimated energy value, and it is not necessary to quantify the estimated energy value, so
  • the bit that needs to quantize the estimated energy value is saved; when the number of quantization bits is constant, the quantization bit that should be given the estimated energy value can be allocated to the subframe in one frame signal, thereby increasing the quantization precision.
  • the prediction coefficients of the fixed codebook gains of the remaining subframes are quantized.
  • the adaptive codebook gain of the one subframe is further quantized.
  • the adaptive codebook gain of the remaining subframes is quantized.
  • the above-mentioned storage medium may be a read only memory, a magnetic disk or an optical disk or the like.

Landscapes

  • Engineering & Computer Science (AREA)
  • Computational Linguistics (AREA)
  • Signal Processing (AREA)
  • Health & Medical Sciences (AREA)
  • Audiology, Speech & Language Pathology (AREA)
  • Human Computer Interaction (AREA)
  • Physics & Mathematics (AREA)
  • Acoustics & Sound (AREA)
  • Multimedia (AREA)
  • Compression, Expansion, Code Conversion, And Decoders (AREA)

Description

一种增益量化方法及装置
本申请要求于 2008 年 1 月 31 日提交中国专利局、 申请号为 200810006804.7、 发明名称为 "一种增益量化方法及装置" 的中国专利申请的 优先权， 其全部内容通过引用结合在本申请中。
技术领域
本发明涉及编码技术， 具体涉及一种增益量化方法及装置。
背景技术
随着近十几年计算机技术的不断发展，各种多媒体应用越来越受到人们的 青睐。作为多媒体应用不可缺少的语音编码技术， 在最近十几年取得了很大的 进步。 码激励线性预测编码模型 （CELP: Code Excited Linear Prediction )是 在语音编码领域中应用的较为广泛的一种编码模型，利用几乎是白的信号激励 两个时变的线性递归滤波器，每个滤波器反馈环路上有一个预测器， 其中一个 是长时预测器（或基音预测器） P(z) , 用来产生浊音语音的音调结构 （谱的细 致结构）， 另一个是短时预测器 F(z) , 用来恢复语音的短时谱包络。 这种模型 来源于它的逆过程， 即 F(z)用于去除语音信号的近样点冗余度， P(z)用于去除 语音信号的远样点冗余度，经过两级预测得到的归一化残差信号近似服从标准 正态分布。 该模型的激励信号通常从高斯白噪声序列构成的码书中选取， 所以 称为码激励线性预测模型。
在码激励线性预测模型中， 对自适应码书增益 (基音增益 ) ga和固定码书 增益 gs的量化对合成语音的质量有很大影响， 若增益量化的比较粗糙， 则合 成语音的量化噪声比较大，合成语音的自然度和清晰度比较差， 高质量的语音 编码算法需要高质量的增益量化，对自适应码书增益和固定码书增益进行增益 量化的原则是使原始语音和重建语音之间的感知加权均方误差最小化。
现有的可变比特率宽带语音编码中增益量化的方法如下所述：
一帧信号的增益用 30比特进行量化： 用于固定码书增益预测的估计能量 值每帧用 2比特进行量化， 一帧信号分为四个子帧，每个子帧的自适应码书和 固定码书增益使用 7比特进行矢量量化， 共 2+7x4=30比特。 其中每个子帧的 能量釆用式（1 )计算：
N-1
Es (n) = 10 log (- gc 2∑c2 (i)) = 20 \og(gc ) + Ex ( 1 ) 其中 N=64 是子帧的长度， c(i)是固定码书激励， gc是固定码书增益， 是码书的平均能量， 其中 可以釆用式（2)进行计算：
N-1 才艮据式（1 )可以得到固定码书的估计能量值可以如式（3 )所示：
^ =201og(g'c) + ^ (3) 因此根据式（3)得到每个子帧的预测固定码书增益如式（4)所示： g'c=10 。5( ) (4) 所以为了得到预测固定码书增益 , 需要计算出 , 可以首先计算每个 子帧的线性预测残差的能量， 具体可以釆用式（5)计算每个子帧的线性预测 残差的能量：
Figure imgf000004_0001
其中 r( )是线性预测残差。
通过式（5)计算出每个子帧的线性预测残差的能量后， 可以计算每帧的 平均残差能量， 具体可以通过式（6)进行计算：
Figure imgf000004_0002
将得到的平均残差能量去除自适应码书的贡献估计值后可以得到估计能 量值 因此得到的 如式（7)所示：
Figure imgf000004_0003
其中: ^是去除两个开环基音分析得到的互相关归一化能量平均值， 10R是 自适应码书的贡献估计值。
对估计能量值用 2比特进行量化得到固定码书增益 ,量化级可以为 18, 30, 42, 54,估计能量值可以更进一步被强制成大于 £max -37或:^ =54,其中 £max 是 4个子帧中 最大值；
由此可以根据计算得到的: ^和式（4)计算得到预测固定码书增益 g'， 计 算固定码书增益 和预测固定码书增益 之间的校正因子， 计算式如式（8) 所示：
Figure imgf000004_0004
(8) 再对各个子帧的自适应码书增益和校正因子 故 7比特联合矢量量化， 具 体是二维矢量量化， 完成一帧信号的增益量化。
在对现有技术的研究中， 发明人发现： 使用现有的增益量化方法需要用 2 比特对估计能量值进行量化。
发明内容
本发明实施例提供了一种增益量化方法及装置，使用本发明实施例提供的 技术方案， 可以不对估计能量值进行量化， 从而节省量化比特。
本发明提供一种增益量化方法， 包括：
计算一帧信号中至少一个子帧的固定码书增益；
对所述固定码书增益进行量化，获得所述至少一个子帧的固定码书增益的 量化值；
使用所述固定码书增益的量化值和所述一帧信号的其余子帧的固定码书 增益， 计算所述其余子帧的固定码书增益的预测系数；
对所述其余子帧的固定码书增益的预测系数进行量化。
本发明还提供一种计算机程序产品，所述计算机程序产品包括计算机程序 代码， 当所述计算机程序代码被一个计算机执行的时候， 所述计算机程序代码 可以使得所述计算机执行增益量化方法中的任意一项步骤。
本发明还提供一种计算机可读存储介质， 所述计算机存储计算机程序代 码， 当所述计算机程序代码被一个计算机执行的时候， 所述计算机程序代码可 以使得所述计算机执行增益量化方法中的任意一项步骤。
本发明还提供一种增益量化装置， 包括：
固定码书增益计算单元，用于计算一帧信号中至少一个子帧的固定码书增 益；
固定码书增益量化单元，用于对所述固定码书增益计算单元计算得到的固 定码书增益进行量化， 获得所述至少一个子帧的固定码书增益的量化值； 系数计算单元，用于使用所述固定码书增益的量化值和所述一帧信号的其 余子帧的固定码书增益， 计算所述其余子帧的固定码书增益的预测系数； 系数量化单元， 用于对所述其余子帧的固定码书增益的预测系数进行量 化。 从本发明实施例提供的以上技术方案可以看出，由于本发明实施例使用第 一个子帧的固定码书增益的量化值计算其余子帧的固定码书增益的预测系数， 从而使其余子帧的固定码书增益的预测系数与第一个子帧的固定码书增益的 量化值相关， 充分利用了各个子帧间固定码书增益相关性强的特点，从而不需 要计算估计能量值，也不需要对估计能量值进行量化， 因此可以节省需要对估 计能量值进行量化的比特。
附图说明
图 1为本发明实施例中增益量化方法实施例一的流程图；
图 2为本发明实施例中增益量化方法实施例二的流程图；
图 3为本发明实施例中增益量化装置实施例一的结构图；
图 4为本发明实施例中增益量化装置实施例二的结构图。
具体实施方式
为使本发明的目的、 技术方案、 及优点更加清楚明白， 以下参照附图并举 实施例， 对本发明进一步详细说明。
图 1描述了本发明实施例提供的增益量化方法实施例一的流程， 包括： 步骤 101、 计算一帧信号中一个子帧的固定码书增益；
上述一个子帧是一帧信号中的任意一个子帧，但是本发明实施例优选釆用 一帧信号的第一个子帧；
但是需要说明的是，本发明实施例并不限定仅计算一帧信号中一个子帧的 固定码书增益，还可以计算一个以上子帧的固定码书增益； 在后续的实施例中 釆用计算一个子帧的固定码书增益进行说明 ,但是并不是对本发明实施例的限 定。
计算一个以上子帧的固定码书增益的处理过程与计算一个子帧的固定码 书增益的处理过程基本相同， 本发明实施例不对其赘述。
可以釆用传统的固定码书增益计算方法计算上述固定码书增益，也可以釆 用其他方式计算上述固定码书增益；具体如何计算上述固定码书增益并不会影 响本发明实施例的实现，因此本发明实施例并不限定固定码书增益的具体计算 过程。
本发明实施例提供的一种固定码书增益计算方法釆用如下的式（9 )进行 计算：
Figure imgf000007_0001
其中 gs为固定码书增益，矢量 z是固定码本矢量与感知加权合成滤波器的 脉冲相应的卷积， 矢量 χ 2是固定码本搜索的目标矢量。
步骤 102、 对固定码书增益进行量化， 获得上述一个子帧的固定码书增益 的量化值；
对固定码书增益进行量化时， 可以进行标量量化， 也可以进行矢量量化； 如果进行标量量化，可以对固定码书增益直接以均勾的量化步长进行均勾的标 量量化， 也可以将固定码书增益映射到对数域后进行非均勾的标量量化； 可以进一步对第一个子帧的自适应码书增益进行量化，在对固定码书增益 进行量化时， 可以进行标量量化， 也可以进行矢量量化； 如果进行标量量化， 可以对自适应码书增益直接以均匀的量化步长进行均匀的标量量化，也可以将 自适应码书增益映射到对数域后进行非均勾的标量量化；
步骤 103、 计算上述一帧信号的其余子帧的固定码书增益；
其余子帧的固定码书增益的计算可以釆用式（9 )进行计算， 也可以釆用 其他的方式进行计算，具体如何计算其余子帧的固定码书增益并不会影响本发 明实施例的实现，因此本发明实施例并不限定其余子帧的固定码书增益的具体 计算过程；
步骤 104、使用上述一个子帧的固定码书增益的量化值和上述一帧信号的 其余子帧的固定码书增益， 计算其余子帧的固定码书增益的预测系数；
在上述信号有多个其余子帧时，可以使用上述一个子帧的固定码书增益的 量化值分别与其余子帧的固定码书增益计算其余子帧的固定码书增益的预测 系数；
步骤 105、 对其余子帧的固定码书增益的预测系数进行量化。
可以对其余子帧的固定码书增益的预测系数进行矢量量化或标量量化，但 是本发明实施例在对第一个子帧的固定码书增益进行标量量化时，优选对其余 子帧的固定码书增益的预测系数进行矢量量化，从而进行标量量化和矢量量化 的结合， 提高量化精度。
从上可知，本实施例使用第一个子帧的固定码书增益的量化值计算其余子 帧的固定码书增益的预测系数，从而使其余子帧的固定码书增益的预测系数与 第一个子帧的固定码书增益的量化值相关，充分利用了各个子帧间固定码书增 益相关性强的特点，从而不需要计算估计能量值， 也不需要对估计能量值进行 量化， 因此可以节省需要对估计能量值进行量化的比特； 在量化比特数量不变 时， 可以将本应给估计能量值的量化比特分配给一帧信号中的子帧，从而使量 化精度增加。
图 2描述了本发明实施例提供的增益量化方法实施例二的流程， 包括： 步骤 201、 计算一帧信号中一个子帧的固定码书增益和自适应码书增益； 步骤 202、 对上述一个子帧的固定码书增益进行量化， 获得上述一个子帧 的固定码书增益的量化值；
步骤 203、 对上述一个子帧的自适应码书进行量化；
步骤 202和步骤 203并没有时间上的先后顺序；
步骤 204、 计算上述一帧信号的其余子帧的固定码书增益和自适应码书增 益；
步骤 205、使用上述一个子帧的固定码书增益的量化值和上述一帧信号的 其余子帧的固定码书增益， 计算其余子帧的固定码书增益的预测系数；
步骤 206、对其余子帧的固定码书增益的预测系数和自适应码书增益进行 量化。
可以将固定码书增益的预测系数和自适应码书增益组成二维矢量进行矢 量量化， 也可以分别进行标量量化。
从上可知，本实施例使用第一个子帧的固定码书增益的量化值计算其余子 帧的固定码书增益的预测系数，从而使其余子帧的固定码书增益的预测系数与 第一个子帧的固定码书增益的量化值相关，充分利用了各个子帧间固定码书增 益相关性强的特点，从而不需要计算估计能量值， 也不需要对估计能量值进行 量化， 因此可以节省需要对估计能量值进行量化的比特； 在量化比特数量不变 时， 可以将本应给估计能量值的量化比特分配给一帧信号中的子帧，从而使量 化精度增加。
如下以 30个比特对具有 4个子帧的一帧信号进行增益量化为例， 对本发 明实施例提供的增益量化方法进行描述： 首先，计算一帧信号的第一个子帧的自适应码书增益和固定码书增益， 具 体的，可以釆用传统的自适应码书增益和固定码书增益算法计算自适应码书增 益和固定码书增益，也可以才艮据最小均方误差准则计算自适应码书增益和固定 码书增益。 其中，根据最小均方误差准则计算得到的自适应码书增益是最佳自 适应码书增益，根据最小均方误差准则计算得到的固定码书增益是最佳固定码 书增益，本发明实施例提供的一种求解最佳自适应码书增益和最佳固定码书增 益的方式如下所述：
自适应码书增益用 ga表示， 固定码书增益用 gs表示， 为了使对自适应码 书增益和固定码书增益进行增益量化时，原始语音和重建语音之间的感知加权 均方误差最小， 则 ga和 gs需要满足式（10):
N-1
e = ^[x0 (n) - gaxu (n) - g y (n)f ( 10) 其中， N为激励矢量长度； χ。为感知加权语音减去加权合成滤波器^ 的零输入响应的剩余信号； ,(«)和 »分别为自适应码书输出的最佳矢量和 固定码书输出的最佳矢量通过加权合成滤波器 (z)的零状态响应。
从上可知， 式（ 10 )可以表示成如下所示的式（ 11 ):
e = ¾ + g + g¾ - H - 2gsR02 + 2gagsRu
对于每个子帧，只有其中的自适应码书增益 ga和固定码书增益 gs是变量， 所以求最佳 ga和最佳 gs, 可以求解使二元函数（12)取最小值的 （ga, gs)。
e = ga xga x ci°] + x c[l] + ^ x ^ x c[2] + ^ c[3] + x ^ x c[4] ( 12) 其中二元函数（ 12 )是式（ 11 )去除了 RQ后的一种描述方式。
根据二元函数极值的求解方法， 令 e对 ga和 gs的一阶偏导数为零可得式 ( 13 ):
根据二元函数 ( 12 )和式 ( 13 )可以解得使误差最小的最优解， 解得的最 优解如下所示：
ga =(2xc[\]xc[2]-c[3]xc[4])/(c[4]xc[4]-4xc[0]xc[2])
=(c[3] + g。xc[4])/(- 2xc[2])
由此就得到了最佳自适应码书增益和最佳固定码书增益，进而对自适应码 书增益和固定码书增益进行量化， 本发明实施例优选进行标量量化； 标量量化 可以是均勾标量量化或非均勾标量量化，例如可以对增益直接以均勾的量化步 长进行标量量化、或将增益先影射到对数域再进行标量量化； 例如对最佳固定 码书增益进行 5bit非均勾标量量化， 先对最佳固定码书增益取对数后乘以 10, 使其单位为 分贝 （ dB ) , 再进行量化， 量化步长可以 达到 10*lgl0(3349.654392/15.848932)/31/2 , 即 0.375dB , 可以艮好地保证量化精度。 由于使用最佳固定码书增益和最佳自适应码书增益可以提高量化精度，因此本 发明实施例优选釆用最佳固定码书增益和最佳自适应码书增益进行量化。
由于有 4个子帧， 30个量化比特， 因此可以为第一个子帧分配 9个比特， 剩余的 3个子帧每个子帧分配 7个比特进行增益量化。在解得了第一个子帧的 最佳自适应码书和最佳固定码书增益后，使用 4个比特对最佳自适应码书进行 增益量化，量化范围可以为 0.012445-1.296012;使用 5个比特对最佳固定码书 增益进行增益量化，量化范围可以为 15.848932-3349.654392。 由于为第一个子 帧分配了 9个比特的量化比特， 可以进一步提高第一个子帧的量化精度， 同时 不需要使用 2个比特对估计能量值进行量化， 提高了量化比特的利用率。
用第一个子帧的固定码书增益的量化值对第二， 三， 四个子帧的固定码书 增益进行预测， 然后分别对第二， 三， 四个子帧的自适应码书增益和固定码书 增益的预测系数进行 7比特的量化，具体可以是将自适应码书增益与固定码书 增益的预测系数一起进行二维矢量量化或分别进行标量量化，本发明实施例优 选进行矢量量化，具体可以将自适应码书增益和固定码书增益的预测系数组成 一个二维矢量， 进而对得到的这个二维矢量进行量化。
本发明实施例使用的固定码书增益的预测系数可以是当前子帧的固定码 书增益与所述固定码书增益的量化值的比值，即当前子帧的固定码书增益的预 测系数=当前子帧固定码书增益 /第一个子帧的固定码书增益的量化值。 但是， 需要说明的是，上述仅是本发明实施例提供的固定码书增益的预测系数一种计 算方式，本发明实施例并不限定仅能釆用上述方式计算固定码书增益的预测系 数。其中，当前子帧的固定码书增益的计算可以釆用本发明实施例提供的方法， 也可以釆用传统的固定码书增益的计算方法，本发明实施例并不限定当前子帧 的固定码书增益具体如何计算。 从上可知，本实施例使用第一个子帧的固定码书增益的量化值计算其余子 帧的固定码书增益的预测系数，从而使其余子帧的固定码书增益的预测系数与 第一个子帧的固定码书增益的量化值相关，充分利用了各个子帧间固定码书增 益相关性强的特点，从而不需要计算估计能量值， 也不需要对估计能量值进行 量化， 因此可以节省需要对估计能量值进行量化的比特； 在量化比特数量不变 时， 可以将本应给估计能量值的量化比特分配给一帧信号中的子帧，从而使量 化精度增加； 同时， 本实施例并没有复杂的对数运算和指数运算， 不需要耗费 较大的计算量。
再介绍本发明实施例提供的增益量化装置，图 3描述了本发明实施例提供 的增益量化装置实施例一的结构， 包括：
固定码书增益计算单元 301 , 用于计算一帧信号中一个子帧的固定码书增 益； 计算一帧信号中其余子帧的固定码书增益；
固定码书增益量化单元 302 , 用于对固定码书增益计算单元 301计算得到 的一个子帧的固定码书增益进行量化，获得上述一个子帧的固定码书增益的量 化值；
系数计算单元 303 , 用于使用上述一个子帧的固定码书增益的量化值和其 余子帧的固定码书增益， 计算其余子帧的固定码书增益的预测系数；
系数量化单元 304, 用于对系数计算单元 303计算得到的其余子帧的固定 码书增益的预测系数进行量化。
从上可知，本实施例使用第一个子帧的固定码书增益的量化值计算其余子 帧的固定码书增益的预测系数，从而使其余子帧的固定码书增益的预测系数与 第一个子帧的固定码书增益的量化值相关，充分利用各个子帧间固定码书增益 相关性强的特点，从而不需要计算预测能量值，也不需要对预测能量值进行量 化， 可以为第一个子帧分配更多的量化比特， 使第一个子帧的量化精度增加。
图 4描述了本发明实施例提供的增益量化装置实施例二的结构， 包括： 固定码书增益计算单元 401 , 用于计算一帧信号中一个子帧的固定码书增 益； 计算一帧信号中其余子帧的固定码书增益；
自适应码书增益计算单元 402 , 用于计算上述一个子帧的自适应码书增 益； 固定码书增益量化单元 403 , 用于对固定码书增益计算单元 401计算得到 的上述一个子帧的固定码书增益进行量化，获得上述一个子帧的固定码书增益 的量化值；
自适应码书增益量化单元 404, 用于对自适应码书增益计算单元 402计算 得到的上述一个子帧的自适应码书增益进行量化；
系数计算单元 405， 用于使用上述一个子帧的固定码书增益的量化值和其 余子帧的固定码书增益， 计算其余子帧的固定码书增益的预测系数；
系数量化单元 406, 用于对系数计算单元 405计算得到的其余子帧的固定 码书增益的预测系数进行量化。
从上可知，本实施例使用第一个子帧的固定码书增益的量化值计算其余子 帧的固定码书增益的预测系数，从而使其余子帧的固定码书增益的预测系数与 第一个子帧的固定码书增益的量化值相关，充分利用了各个子帧间固定码书增 益相关性强的特点，从而不需要计算估计能量值， 也不需要对估计能量值进行 量化， 因此可以节省需要对估计能量值进行量化的比特； 在量化比特数量不变 时， 可以将本应给估计能量值的量化比特分配给一帧信号中的子帧，从而使量 化精度增加。
本领域普通技术人员可以理解实现上述实施例方法中的全部或部分步骤 是可以通过程序来指令相关的硬件完成，所述的程序可以存储于一种计算机可 读存储介质中， 该程序在执行时， 包括如下步骤：
计算一帧信号中一个子帧的固定码书增益；
对所述固定码书增益进行量化，获得所述一个子帧的固定码书增益的量化 值；
使用所述一个子帧的固定码书增益的量化值和其余子帧的固定码书增益， 计算所述其余子帧的固定码书增益的预测系数；
对所述其余子帧的固定码书增益的预测系数进行量化。
还可以包括如下步骤：
计算所述第一个子帧的自适应码书增益；
进一步对所述一个子帧的自适应码书增益进行量化。
还可以包括如下步骤： 计算所述其余子帧的自适应码书增益；
对所述其余子帧的自适应码书增益进行量化。
上述提到的存储介质可以是只读存储器， 磁盘或光盘等。
以上对本发明实施例所提供的一种增益量化方法及装置进行了详细介绍， 以上实施例的说明只是用于帮助理解本发明的方法及其思想； 同时，对于本领 域的一般技术人员，依据本发明的思想，在具体实施方式及应用范围上均会有 改变之处， 综上所述， 本说明书内容不应理解为对本发明的限制。

Claims

权 利 要 求
1、 一种增益量化方法， 其特征在于， 包括：
计算一帧信号中至少一个子帧的固定码书增益；
对所述固定码书增益进行量化，获得所述至少一个子帧的固定码书增益的 量化值；
使用所述固定码书增益的量化值和所述一帧信号的其余子帧的固定码书 增益， 计算所述其余子帧的固定码书增益的预测系数；
对所述其余子帧的固定码书增益的预测系数进行量化。
2、 如权利要求 1所述的增益量化方法， 其特征在于， 计算所述其余子帧 的固定码书增益的预测系数包括：
计算所述其余子帧中当前子帧的固定码书增益与所述固定码书增益的量 化值的比值， 所述比值为所述当前子帧的固定码书增益的预测系数。
3、 如权利要求 1所述的增益量化方法， 其特征在于， 所述计算一帧信号 中至少一个子帧的固定码书增益包括：
计算所述一帧信号中第一个子帧的固定码书增益。
4、 如权利要求 3所述的增益量化方法， 其特征在于， 计算所述一帧信号 中第一子帧的固定码书增益包括：
计算所述第一个子帧的最佳固定码书增益。
5、 如权利要求 1至 4任一所述的增益量化方法， 其特征在于， 进一步包 括： 计算所述至少一个子帧的自适应码书增益；
对所述至少一个子帧的自适应码书增益进行量化。
6、 如权利要求 5所述的增益量化方法， 其特征在于， 计算所述至少一个 子帧的自适应码书增益包括： 计算所述至少一个子帧的最佳自适应码书增益。
7、 如权利要求 5所述的增益量化方法， 其特征在于， 对所述自适应码书 增益进行均勾的标量量化或非均勾的标量量化；
对所述固定码书增益进行均勾的标量量化或非均勾的标量量化。
8、 如权利要求 1至 4任一所述的增益量化方法， 其特征在于， 进一步包 括：
计算所述其余子帧的自适应码书增益； 对所述其余子帧的自适应码书增益进行量化。
9、 一种计算机程序产品， 其特征在于， 所述计算机程序产品包括计算机 程序代码， 当所述计算机程序代码被一个计算机执行的时候， 所述计算机程序 代码可以使得所述计算机执行权利要求 1至 8项中任意一项的步骤。
10、 一种计算机可读存储介质， 其特征在于， 所述计算机可读存储介质存 储计算机程序代码， 当所述计算机程序代码被一个计算机执行的时候， 所述计 算机程序代码可以使得所述计算机执行权利要求 1至 8项中任意一项的步骤。
11、 一种增益量化装置， 其特征在于， 包括：
固定码书增益计算单元，用于计算一帧信号中至少一个子帧的固定码书增 益；
固定码书增益量化单元，用于对所述固定码书增益计算单元计算得到的固 定码书增益进行量化， 获得所述至少一个子帧的固定码书增益的量化值； 系数计算单元，用于使用所述固定码书增益的量化值和所述一帧信号的其 余子帧的固定码书增益， 计算所述其余子帧的固定码书增益的预测系数； 系数量化单元， 用于对所述其余子帧的固定码书增益的预测系数进行量 化。
12、 如权利要求 11所述的增益量化装置， 其特征在于， 所述固定码书增 益计算单元用于计算所述一帧信号中第一个子帧的最佳固定码书增益。
13、 如权利要求 11或 12所述的增益量化装置， 其特征在于， 还包括： 自适应码书增益计算单元， 用于计算所述至少一个子帧的自适应码书增 益；
自适应码书增益量化单元，用于对所述至少一个子帧的自适应码书增益进 行量化。
14、 如权利要求 13所述的增益量化装置， 其特征在于， 所述自适应码书 增益计算单元还用于计算所述其余子帧的自适应码书增益；
所述自适应码书增益量化单元还用于对所述自适应码书增益计算单元计 算得到的所述其余子帧的自适应码书增益进行量化。
15、 如权利要求 13所述的增益量化装置， 其特征在于， 所述自适应码书 增益计算单元用于计算所述一个子帧的最佳自适应码书增益。
PCT/CN2009/070119 2008-01-31 2009-01-13 一种增益量化方法及装置 WO2009097763A1 (zh)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN 200810006804 CN101499281B (zh) 2008-01-31 2008-01-31 一种语音编码中的增益量化方法及装置
CN200810006804.7 2008-01-31

Publications (1)

Publication Number Publication Date
WO2009097763A1 true WO2009097763A1 (zh) 2009-08-13

Family

ID=40946327

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2009/070119 WO2009097763A1 (zh) 2008-01-31 2009-01-13 一种增益量化方法及装置

Country Status (2)

Country Link
CN (1) CN101499281B (zh)
WO (1) WO2009097763A1 (zh)

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US6104992A (en) * 1998-08-24 2000-08-15 Conexant Systems, Inc. Adaptive gain reduction to produce fixed codebook target signal
CN1468427A (zh) * 2000-05-19 2004-01-14 �����ɭ��ϵͳ��˾ 一种码激励线性预测语音编码器的增益量化
CN1151492C (zh) * 1995-05-03 2004-05-26 艾利森电话股份有限公司 综合-分析线性预测语音编码中的增益量化方法
CN1820306A (zh) * 2003-05-01 2006-08-16 诺基亚有限公司 可变比特率宽带语音编码中增益量化的方法和装置
CN101030378A (zh) * 2006-03-03 2007-09-05 北京工业大学 一种建立增益码书的方法

Family Cites Families (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
FI113571B (fi) * 1998-03-09 2004-05-14 Nokia Corp Puheenkoodaus

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN1151492C (zh) * 1995-05-03 2004-05-26 艾利森电话股份有限公司 综合-分析线性预测语音编码中的增益量化方法
US6104992A (en) * 1998-08-24 2000-08-15 Conexant Systems, Inc. Adaptive gain reduction to produce fixed codebook target signal
CN1468427A (zh) * 2000-05-19 2004-01-14 �����ɭ��ϵͳ��˾ 一种码激励线性预测语音编码器的增益量化
CN1820306A (zh) * 2003-05-01 2006-08-16 诺基亚有限公司 可变比特率宽带语音编码中增益量化的方法和装置
CN101030378A (zh) * 2006-03-03 2007-09-05 北京工业大学 一种建立增益码书的方法

Also Published As

Publication number Publication date
CN101499281B (zh) 2011-04-27
CN101499281A (zh) 2009-08-05

Similar Documents

Publication Publication Date Title
US10720172B2 (en) Encoder for encoding an audio signal, audio transmission system and method for determining correction values
CN101180676B (zh) 用于谱包络表示的向量量化的方法和设备
US10706865B2 (en) Apparatus and method for selecting one of a first encoding algorithm and a second encoding algorithm using harmonics reduction
CN109712633B (zh) 音频编码器和解码器
CN106463134B (zh) 用于对线性预测系数进行量化的方法和装置及用于反量化的方法和装置
US10607619B2 (en) Concept for encoding an audio signal and decoding an audio signal using deterministic and noise like information
JP6456412B2 (ja) Ｃｅｌｐ符号器および復号器で使用するための柔軟で拡張性のある複合革新コードブック
JP5521097B2 (ja) 信号を符号化する方法および信号を復号する方法
US9552821B2 (en) Encoding method, encoder, program and recording medium
JP6457552B2 (ja) 符号化装置、復号装置、これらの方法及びプログラム
JP6173304B2 (ja) Ｃｅｌｐコーダにおける変換領域コードブック装置
TWI751584B (zh) 具有信號相依數和精度控制之音訊編碼器、音訊解碼器及相關方法與電腦程式
RU2707144C2 (ru) Аудиокодер и способ для кодирования аудиосигнала
JP5663461B2 (ja) 符号化方法、符号化装置、プログラム、記録媒体
US6098037A (en) Formant weighted vector quantization of LPC excitation harmonic spectral amplitudes
WO2009097763A1 (zh) 一种增益量化方法及装置
WO2016121826A1 (ja) 符号化装置、復号装置、これらの方法、プログラム及び記録媒体

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 09709264

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE

122 Ep: pct application non-entry in european phase

Ref document number: 09709264

Country of ref document: EP

Kind code of ref document: A1