This article presents a bitrate- and bandwidth-scalable coder submitted as a candidate to the qualification phase of ITU-T embedded G.711 wideband (G.711-WB) standardization. The encoder operates on 5 ms frames and bitrate can be set on a frame basis at either 64, 80 or 96 kbit/s. The input signal, which is by default sampled at 16 kHz, is split in two bands. The low band (0-4000 Hz) is coded by a low-complexity embedded pulse code modulation (PCM) coder at 8-10 bits/sample, including noise feedback at the encoder and PCM-specific post-filtering at the decoder. The high-band (4000- 8000 Hz) is coded by a time-domain aliasing cancellation (TDAC) coder derived from ITU-T G.729.1. Frame erasures are concealed using a low-complexity split-band algorithm derived from ITU-T G.722 Appendix IV. The proposed coder passed all requirements of the G.711WB qualification phase. We summarize and discuss formal ITU-T test results.

This paper describes a combined adaptive transform codec (ATC) and code-excited linear prediction (CELP) algorithm, called ATCELP, for the compression of wideband (7 kHz) signals. The CELP algorithm applies mainly to speech, whereas the ATC mode is selected for music and noise signals. We propose a switching scheme between CELP and ATC mode and describe a frame erasure concealment technique. Subjective listening tests have shown that the ATCELP codec at bit rates of 16, 24 and 32 kbit/s achieved performances close to those of the CCITT G.722 at 48, 56 and 64 kbit/s, respectively, at most operating conditions.

In this paper, an algorithm for coding 20 Hz -15 kHz speech signals at 64 kbit/s with a very low delay (frame of 0.16 ms) is presented.To achieve a quality near to transparency, we propose adapting the Low-Delay CELP coder [1] to the 15 kHz bandwidth and suggest a new noise shaping method based on a psychoacoustic model.In this way we take advantage of linear predictive coding and masking properties of the human perception system.Finally, an algebraic codebook is proposed, allowing an important reduction of coder computational complexity, without decreasing the perceived quality of signals.

A novel approach for minimizing the CELP (code excited linear prediction) error criterion in the Z-transform domain is presented. A covariance method and an autocorrelation method of CELP coding are derived using the Z-transform formulation. The intermediate signals in the excitation selection are interpreted in terms of digital filtering, (inter)correlations, spectra, and inner products. For each method, specific algebraic codebooks permit fast algorithms in the selection of the excitation signal, leading to low complexity CELP coders for high-quality narrowband (0-4 kHz) and wideband (0-7 kHz) speech coding at reduced bit rates.>

International audience

In several speech coding applications the constraint of fixed bit rate and frame size is not imposed. For such a situation, a speech coding system performing speech segmentation and segment coding, both with reasonable complexity, is proposed. Speech is segmented according to its statistical properties. Each class of segment is encoded by a specific algorithm using algebraic excitation codebooks. An average bit rate between 5 and 6 kb/s yields a small degradation over a 9.6-kb/s fixed rate coder.>

International audience

A description is given of an efficient code-excited linear predictive (CELP) coder for bit rates between 6 and 16 kb/s, and novel effective algorithms for the selection of the excitation signal. The authors then propose a class of binary codebooks having interesting properties in terms of performance, robustness against transmission errors and flexibility in the choice of the bit rate. Due to the optimal structure of the coder and to the fast algorithms for selecting the excitation signal, a real-time implementation has been made possible on one 32-bit floating-point digital signal processor.>

This paper describes a robust vector adaptive predictive coder which features a novel arrangement of a decoder to provide a finite memory synthesis and a new scheme for generating the excitation signal using an analysis by synthesis procedure. For bit error rates up to 10-2, our 16 kbit/s coder including error protection is shown to work efficiently on speech signals representative of a two-way mobile-network communication.

This paper presents a new dynamic bit allocation scheme applied to sub-band coding. The technique is based on the use of quadrature mirror filters for separating each sub-band signal. The main feature of the scheme consists of suppressing every side information (SI) associated with each sub-band. This leads to only one kind of transmitted bits, the quantization bits, and consequently the frame synchronization is greatly simplified. An added advantage is that an adaptive quantization algorithm could and has been developed due to this new dynamic bit allocation. Autocorrelation of sub-band signals and linear prediction were also studied. Although the scheme can be applied to a wide range of bit rates, the study was focused on 16 Kbps because of some concerns with satellite and mobile-radio communications. Results of our 16 Kbps coder are presented via an objective measure of speech quality based on a frequency representation of signal and noise instead of the conventional temporal representation.

The synchronous tandem property of nonaccumulation of distortion in tandem-connected ADPCM coders with a 64 kbit/s PCM interface is discussed here. The synchronous tandem algorithm used to provide this property in the steady-state mode is described with a case-by-case analysis, so as to show how the synchronous tandem property is realized in an ADPCM coder. A 32 kbit/s ADPCM coder utilizing this algorithm has been standardized by the CCITT (International Telegraph and Telephone Consultative Committee). The synchronous tandem property of the 32 kbit/s ADPCM coder is of great interest in network applications, because the ADPCM coder appears likely to be introduced into digital networks built partially with existing 64 kbit/s PCM circuits.

This paper concerns a new Quantization Scheme for efficient encoding of waveforms below or about one bit per sample. This technique is then applied to the encoding of baseband residual signals to demonstrate the feasibility of (baseband) Residual Excited LPC at 2400 hit/sec. The technique called "Spherical Vector Quantization" is described in which a block of n consecutive samples is quantized as a vector. The magnitude of the vector is transmitted independently of the vector's orientation. This vector's orientation is vector quantized using a codebook which can be seen as representing a set of N points on a unit hypersphere. The cases for blocks of n = 8 and 24 are discussed which make use of results by Conway and Sloane on regular point lattices. For n = 8, the algorithm is detailed which solves both the problem of finding the closest point on the hypersphere and the problem of determining the index of that point. Experiments involving SVQ in coding the speech baseband residual are described which show in particular that subband coding does not contribute any quality improvement when SVQ is used.

Digital techniques are more and more often used in telecommunications networks. How would these techniques affect the quality of transmitted speech? The use of subjective testing is generally necessary to answer this question. In this paper a degradation category rating test, which is a speech-adapted procedure of the one used for image testing, is proposed to distinguish among good telephone quality codecs. A comparative study of the sensitivity of the method and that of a classic opinion rating test is reported. It shows that the resolution of the degradation category rating test is twice that of a classic opinion rating test within good telephone quality domain. Subjective performances of various 32 kbit/s A.D.P.C.M codecs, including the effects of transmission errors and tandem connections, using degradation category rating tests are also presented.

The application of ADPCM coding technique with adaptive predictors (A.D.P.C.M (V)) and its performances, to transmit 7 kHz bandwith speech signals at 64 kbit/s is presented in this paper. Subjective performances of one ADPCM(V) with two different sets of parameters are evaluated by means of direct comparison tests, with respect to the classic 64 kbit/s P.C.M coder, a 128 kbit/s P.C.M coder and a 64 kbit/s subband coder. The results of the tests show that the ADPCM coding technique can be an efficient mean to transmit wideband speech signals. The quality of the A.D.P.C.M(V) coded speech compares well with that obtained with subbands coding technique. The quality of the transmitted speech, though slightly worse than "high fidelity speech", could be sufficient for communications systems such as audio or video conferencing systems, commentary links for broadcasting.

In the view of transmitting telephone speech signals at the bit rate of 32 kbit/s, the implementation of a predictor with lattice filters has been investigated instead of the classical gradient updated transversal filter commonly used in ADPCM coders. Several simplified methods for updating the lattice filters have been developed. Then those codecs are compared against the standard PCM 64 kbit/s, the PCM 56 kbit/s, the fixed predictor and the gradient updated ADPCM coders at 32 kbit/s on a subjective scale derived from listening tests. The improvement given by the adaptive predictors over fixed ones clearly appears. It is also shown that the simplified lattice ADPCM coders perform significantly better than the non simplified ones for high bit error rates (10-4or 10-3). Another feature of the algorithms developed here is their simplicity that makes them suitable for hardware and VLSI implementation not only for ADPCM coding but also for the speech analysis part in linear predictive vocoders.