Integration of harmonic and analysis by synthesis coders

Hybrid coding of speech has been proposed to overcome the limitations of a single model in representing the wide variety of characteristics of human speech. A new hybrid coding algorithm, which combines harmonic and analysis by synthesis coding techniques, is presented. To integrate the harmonic and analysis by synthesis coders, novel phase synchronisation and speech classification techniques are developed. The perceptual quality of the speech synthesised using the unquantised hybrid model is almost indistinguishable when compared with 128 kbit/s linear PCM. Two variable rate coders are developed based on the designed hybrid model, by quantising the parameters at different bit rates. Subjective listening tests show that the speech quality of the variable rate hybrid coders outperform the quality of 5.3 kbit/s and 6.3 kbit/s ITU G.723.1 coders, at maximum bit rates of 4 kbit/s and 6 kbit/s respectively.     

Vector-quantised transform coder for speech coding at 9-6 kbit/s and below

The transform approach to speech coding has been established for some time, and has been shown to be very efficient in controlling the bit allocation and the shape of the noise spectrum. Various transform coders have been reported which produce high-quality digital speech at around 16 kbit/s. Although these coders can maintain good quality down to about 9.6 kbit/s, they perform poorly at lower bit rates. Here we discuss how vector quantisation (VQ) can be used to improve the quality of transform coders. We describe one specific design of vector-quantised transform coder (VQTC) which follows on from earlier work, and which is capable of producing good-quality speech at as low as 4.8 kbit/s.     

Low delay coder (< 25 ms) of wideband audio (20 Hz-15 kHz) scalable from 64 to 32 kbit/s

A low delay coder for speech and music signals sampled at 32kHz is described. Its algorithmic delay does not exceed 25 ms which enables audioconferencing applications without echo cancellation. Its bit rate is scalable between 64 and 32 kbit/s by steps of 8 kbit/s. The transmitter issues the binary code at 64 kbit/s with lower bit rate codes embedded in it. The receiver may operate at lower bit rates with gradual loss of quality. The proposed coder is based on a mixed scheme : the adopted solution contains elements from the CELP speech coder and frequency domain music coders. The perceptual signal is obtained in the time domain, then transformed to the frequency domain where bit allocation is calculated and transform coefficients are quantized. A first solution based on the dft is discussed, then a second solution based on a mdct with small overlap is applied. The quantization of these coefficients is done in the following way. First, a prediction of the whole spectrum is applied. Then, a mean- removed gain- shape split vq is used for amplitude spectrum quantization and a hierarchical 2- dimensional vq is used for phase spectrum quantization with amplitude correction. At the phase quantization stage, each codeword describing the selected vector index is split into parts corresponding to different bit rates. Due to the hierarchical codebook structure, truncated indices may be used, without much affecting the signal quality. Simulation results are presented and the robustness of the proposed coder is examined.     

Low-delay subband CELP coding for wideband speech

Low-delay techniques are proposed for coding 7 kHz speech using subband code-excited linear predictive coding (CELP). The use of separate and joint index codebooks is compared. Specifically, the joint-index-subband CELP (JISBC) algorithm is found to provide good quality with processing delay in the range 2.375-3.375 ms at corresponding bit rates of 16-8 kbit/s     

一种可变速率的LD—CELP编码的研究

设计了一种可变速率的低时延、码激励线性预测编码（ＬＤ－ＣＥＬＰ）的方案，它是通过修改码本来实现的。该方案工作在１１．２ｋｂｉｔ／ｓ。对其做了计算机仿真，并与１６ｋｂｉｔ／ｓ的ＬＤ－ＣＥＬＰ算法在信经（ＳＮＲ）、波形等方面进行了对比，仿真结果表明效果良好。     

Codage algébrique de la parole : prédiction linéaire à excitation par code ternaire

This article presents new speech coding methods for real time application (telephone, videophone) or offline applications (storage). Speech quality is in the classical telephone range, with a 4 kHz bandwidth and a sampling at 8 kHz. An elementary approach leads to a 16 kbit/s codec and a 24 kbit/s codec, using integer codebooks and fast computations. The speech quality of the two codecs has been measured in comparison with more complex ones and in realistic conditions, with noisy telecommunication channels. The elementary approach is completed by a synthetic model, with a systematic generalization of the algorithms (e.g. for a generalized vselp). Some methods for channel protection, which are already known by the speech coding researchers, are summed up in the Appendix. A change of representation for low density codes (less than 1 bit/sample) is proposed.     

Entropy coding of ITU G.728 speech parameters for low bit rateaudiovisual communications

Entropy coding principles are applied to the 16 kbit/s ITU G.728 speech codec. It is shown that the average bit rate can be reduced to 14.5 kbit/s without a significant increase in the codec complexity. In very low bit rate audiovisual communication applications such as the videophone, the saved bits can be used to improve the output video quality     

Highly Sensitive Speech Detector and High-Speed Voiceband Data Discriminator in DSI-ADPCM Systems

This paper describes a highly sensitive speech detector and a high-speed voiceband data classifier capable of discriminating between speech and voiceband data of a 4.8 kbit/s 8-phase PSK and 4.8 kbit/s 8-point QAM, and a 9.6 kbit/s 16-point QAM as described in a CCITT recommendation. The presence of a speech signal is detected by analyzing short-time energies, zero-crossing rates, and sign bit sequences of the input signal. The proposed speech detector, with a short hangover time of 32 ms, is able to reduce the average talk spurt activity in an international satellite link to 36 percent. This detector can also classify the detected speech into narrow-band or wide-band spectrum sounds or a low power sound for a variable rate ADPCM encoding. Discrimination between speech and high-speed voiceband data is based on short-time energies, a zeros-crossing rate and linear prediction coefficients of an adaptive predictor. Classification among a 4.8 kbit/s 8-phase PSK and 8-point QAM, and a 9.6 kbit/s 16-point QAM can also be performed by an average prediction gain and a coefficient of variation of the short-term amplitude distribution of the input signal. Discrimination of voiceband data was performed successfully, and erroneous discrimination of talk spurt of telephone speech as voiceband data were, respectively, four times for two two-party conversations lasting 5 minutes in an international satellite link. This is equivalent to less than 0.09 percent of the conversation time.     

G.729协议在TMS320C6211上的实现

ITU—T提出基于CS—ACELP算法传输速率为8kbit／s的G．729协议，该协议可以用作如PHS（个人手持电话系统）、比特率低于64kbit／s的可视会议等多媒体服务，也可以作为有效利用传输能力的多路复用器。概要介绍了G．729协议的算法结构，对G．729协议在TMS320C6211DSP上的实现进行了研究。     

Design and performance of robust embedded ADPCM coder

An embedded adaptive DPCM (EADPCM) speech coder is described which allows bit rate reductions to be achieved by progressive deletion of bits from output codewords. Optimised step size multipliers are given for a robust implementation using an improved algorithm for adaptive quantisation. Simulation shows that a graceful reduction in speech quality with bit rate is achieved in the range 16-48 kbit/s.<>     

The big squeeze-a CELP speech codec

The coding algorithm widely recognised as offering the best prospects for delivering toll-quality speech at very low bit rates is called CELP (codebook-excited linear prediction) coding. The CELP codec by Delphi Systems operates in real time, uses a standard digital signal processing chip, and encodes speech at 4.8 and 6.5 kbit/s. The use of this speech compression codec (SCC) is also discussed     

Multi-domain speech compression based on wavelet packet transform

The authors present a multi-domain speech compression method based on a wavelet packet transform. The signals are compressed in domains with different time-frequency resolutions according to their energy distribution in these domains. It is shown that this method is simple to implement and is effective at compressing speech and audio signals, even at bit rates as low as 2 kbit/s     

CELP based mixed-source model for very low bit rate speech coding

A CELP based mixed-source model is described. It uses a mixed excitation which combines a lowpass-filtered adaptive source and a highpass-filtered stochastic source. In addition, one more stochastic source is newly employed for more natural sounding speech. In informal listening tests, the proposed model at 3 kbit/s shows very good performance both in speech quality and intelligibility.<>     

基于小波变换的2.4kbit/s波形内插语音编码算法

基于双正交小波滤波器组对波形内插编码中提取的特征波进行多级分解与重构，提出了一种基于小波变换（WT）的2．4kbit／s特征波形内插（CWI）语音编码算法。编码端去除了特征波对齐运算，并对幅度谱进行多级分解，相位谱不传输，鉴于小波变换对信号的压缩特性，仅传输对人耳感知起主要贡献的最后一级特征波幅度谱；解码端对各尺度空间采用单独重建的方法，相位信息在重构的末级与幅度谱结合，并由浊音度标志选择固定或随机相位。此外，根据语音信号的时变特性，由基于子帧的浊音度标志选择需要传输的幅度谱及量化模式。主观R-A／B测试表明，这种基于小波变换的2．4kbit／s编码算法的合成语音质量明显优于标准的2．4kbit／s的MELP编码器及FS1016的4．8kbit／sCELP编码器，亦优于3．8kbit／s的传统CWI编码框架下的合成语音效果。     

Digital Transmission of Commentary-Grade (7 KHz) Audio at 56 or 64 Kbits/s

A two-band coding system has been constructed for the purpose of providing commentary grade (7 kHz bandwidth) speech or music transmission at 56 or 64 kbits/s. The lower band, 0-to-3650 Hz, is coded with 4 bit ADPCM and the upper band, 3600-to-6800 Hz, is coded with 3 bit or 4 bit/sample APCM. The quality of the coded signal makes the method useful for news and sports broadcasts, and possibly for AM remote music broadcasting. The audio sounds better than that produced by two conventional alternatives: 3200 Hz bandwidth with 8-bit/sample coding and 7000 Hz bandwidth with a single 4-bit/sample coder. The sample may be used in any place with access to a 56 kbit/s Dataphone Digital Service port or to other 56 or 64 kbit/s lines. The power consumption is approximately 12 W in the present form; it could be reduced by a factor of at least two by hardware optimization.     

一种基于低比特率语音算法的芯片设计与分析

介绍了一种低比特率语音算法,它是在MBE编码算法的基础上,利用线性预测谱代替了MBE中的傅里叶变换谱对MBE编码算法进行了改进,从而在保持语音质量的情况下使比特率更低,达到2 kbit/s.同时阐述了基于该算法的系统芯片设计,并对该算法中的基音周期参数、V/U等参数的提取进行详细的推导分析.最后对本算法与常用的码激励线性预测编码(CELP)的算法进行比较,并分析其中的原因.     

基于局部余弦变换的低比特变速率语音编码算法研究

提出将局部余弦变换(LCT)算法应用于语音编码中,系统设计了一个平均比特率近1.6kbit/s的低比特变速率语音编码器。在变比特率编码器设计中采用SVM算法进行VAD检测。激活语音帧的语音模式采用GSM半速率编码中的划分方法,但将其中的强浊音模式和中浊音模式合并为一个中强浊音模式。对各类语音模式和无声帧(背景噪声)的局部余弦变换系数采用分维矢量量化算法进行量化,码书设计采用LGB算法。编码中的码书搜索采用树形快速搜索算法。通过主观非正式听力测试表明设计的变比特率编码器编码的重建语音MOS约为3.15,与比特率为2.4kbit/s美国联邦声码器标准MELP的重建语音相当,具有较强的顽健性,适合于对存在各种环境噪声的语音进行编码。     

Conditional entropy encoding of log-PCM speech

The conditional dependence between the current and previously quantised samples in 64 kbit/s ?-law PCM speech is exploited to noiselessly reduce the average transmitted bit rate. From the conditional probabilities, the conditional entropy and average Huffman code word lengths were computed. The average data rate reduction for the first-order noiseless encoder was found to be 12 kbit/s.     

Subjective Evaluation of a 4.8 kbit/s Residual-Excited Linear Prediction Coder

A 4.8 kbit/s residual-excited linear prediction coder (RELP) with two subband coded basebands was systematically evaluated in terms of intelligibility and overall quality. Intelligibility degradation due to RELP-coding is found to be 6 percent without transmission errors, an additional 6.4 percent with 1 percent bit errors, and 9.8 percent in 10 dB SNR acoustic background noise. Quality of the RELP coded speech is midway between those of 3 and 4 bit log-PCM and is significantly higher than that of the pitch-excited linear prediction coder.