Claude Galand

New quadrature mirror filter structures

This paper introduces new quadrature mirror filter (QMF) structures for the frequency domain analysis and synthesis of digital signals. The conventional QMF technique is first extended to cover complex quadrature mirror filters (CQMF) in which a digital signal is split into N adjacent complex subbands where the real and imaginary parts are subsampled by 1/2N with respect to the original signal. The computational complexity of QMF banks is then analyzed and a new scheme which reduces the computational complexity by about a factor of two over conventional QMF implementations is proposed. Finally, the filter design tradeoffs are discussed and the microprogramed implementation of QMF banks is evaluated.

32 Kbps CCITT Compatible split band coding scheme

This paper deals with the application of the SVCS (Split band Voice Coding Scheme) concept to the coding of a PCM channel at half the rate of the presently used 64 kbps CCITT standard. This 32 kbps coder is shown to meet the specifications as recommended by the CCITT for a PCM channel operating at 64 kbps (8kHz sampling, 8 bits/sample A-Law). In addition, channel performances have been evaluated with and without transmission errors for different types of signals ranging from tones and modem signals to voice signals. It has been observed that transparency and apparent channel performances are not affected and that the behavior of the proposed 32 kbps is considerably less affected by transmission errors than the 64 kbps. Examples of taped results assuming different error rates on a voice signal with both the 32 kbps SVCS and the 64 kbps A-Law will be played at the conference.

Pitch detection for use in a predictive speech coder

A pitch detector to adjust long term prediction in a pulse excitation speech coder. A residual signal r(n) is first derived from the speech signal s(n) by short term filtering. Then, r(n) is processed to calculate a prediction error signal e(n) which is subsequently pulse excitation encoded. The processing of e(n) entails prediction of a residual by measuring a pitch related factor M, employing two steps. First calculating a coarse M value through peak clipping and sign transition detection, and then adjusting the M value by autocorrelation--calculations about the roughly spaced peaks.

Adaptive code excited linear predictive coder (ACELPC)

Code-excited linear predictive coding (CELP) is a recent vector waveform coding technique which permits the encoding of telephone speech with high quality at very low bit rates. The authors show that they can further reduce the computational complexity and the storage requirements of the coder, while improving the perceptual quality of the reconstructed speech. These improvements are achieved by two key factors: the implementation of a noise-shaping effect by alternate estimation of the short-term predictor coefficients, and the use of a fixed/adaptive codebook together with a long-term predictor.<<ETX>>

16kbps Real time QMF sub-band coding implementation

The availability of medium performance microprocessors, in conjunction with different concepts to generate real time efficient signal processing software on microprocessors without hardwired multipliers, allows the real time implementation of sub-band voice compression algorithms. This paper deals with the implementation of a 16kbps (8kHz, 2 bits per sample) 8 sub-band coder assuming a bank of 40 tap QMF decimator and interpolator filters, an adaptive allocation of the bits resource based on channels activity and a straigth block PCM coding of the decimated samples. The paper focuses on the implementation and software techniques which were used to achieve a complete sub-band coder with about 1.3 million of instructions per second on a processor with a 16 bits instruction and data flow.

High-frequency regeneration of base-band vocoders by multi-pulse excitation

In this paper, we propose a new method for the regeneration of the high-frequency band in base-band vocoders. This approach takes advantage of the quasi-harmonic structure of voiced speech in the time domain, while existing techniques capitalize on this structure in the frequency domain (squaring, full wave rectifying, interpolation, etc ...). We show that the high frequency part of the excitation signal, which is usually regenerated by a frequency duplication of the harmonic structure of the base-band, always presents time variations which can be modeled by a pulse/noise combination. The position and amplitudes of the pulses are shown to be critical for the good reconstruction of the speech wave-form, in the sense that a slight jitter can produce an audible roughness. We report simulation results on a complete vocoder implemented at 7.2 kbps.

Multirate subband coding applied to digital speech interpolation

It has been shown that the output of a multirate subband coder has an embedded bit stream, i.e., accepts bit deletion and insertion for dynamic rate conversion. This property is of great importance in a digital voice communication network since it allows any overloaded node to simply flag and drop appropriate bits from the bit stream, without tandeming of encode/decode operations. This paper presents a statistical analysis of a digital speech interpolation (DSI) system using multirate subband coding. Our intent is to estimate the level of speech quality and grade of service that can be provided by such a system, given values of various system parameters. The analysis uses birth-death models for the arrival and departure of calls and talk-spurts, and considers a strategy for allocating bandwidth among the ports of the system that takes advantage of the properties of the multirate subband coder. A simulation model is employed to estimate values of certain performance parameters that are difficult to obtain analytically. Our bandwidth allocation strategy ensures that all traffic offered to the DSI system is carried, with neither blocking nor freezeout. We consider, as an example, a DSI system with a maximum concentration ratio of 4 to 1, in which 96 PCM speech channels at 64 kbits/s each are concentrated on a single 1.544 Mbit/s T1 link. The application of multirate subband coders is shown to be extremely efficient, since completely toll quality speech transmission is provided even at the full 4-to-1 compression ratio, without any freezeout. Preliminary results of informal listening tests verify the level of speech quality predicted by our analysis.

Design and evaluation of parallel quadrature mirror filters (PQMF)

Quadrature mirror filters (QMF) have been proposed in /1/, and are now widely used in digital sub-band coding of speech. The standard approach to the QMF splitting/reconstruction of a signal is based on a tree arrangement of half-band QMF filters. In this paper, we show that, assuming a proper design, a parallel implementation of the QMF filter bank (PQMF) is much more efficient than the tree implementation. In the first part of the paper, the basic tree approach is briefly reviewed. Than the parallel filter bank structure is introduced, the equations for alias-free reconstruction are derived and solved. It is shown that the parallel approach can drastically be simplified by truncating the so computed impulse responses. The effects of this truncation have been experimentally evaluated and some rules of thumb have been established for optimum truncation order. An example of design is given to illustrate the proposed method.

MPE/LTP speech coder for moblie radio application

Abstract In this paper, we describe our contribution to the CEPT effort for the definition of a European standard for the cellular mobile radio. We first give an overview of the Multipulse excitation with long term prediction (MPE/LTP) algorithm. Then, we report details on the implementation (constants, quantizing tables) respectively in the analysis part, the error protection/correction, and the synthesis part of the codec. Finally, we report detailed figures about the implementation efficiency (ROM/RAM/Cycles).

Adaptive code excited predictive coding

A novel way to use the code excited linear prediction (CELP) concept that decreases the processing load while keeping the same speech quality is discussed. Rather than performing individual weighting of each candidate sequence, a global implementation of the perceptual weighting function at the codebook level is proposed. As a result, the analysis-by-synthesis procedure does not require the processing of all the candidate sequences through the synthesis and weighting filters; the complexity requirement of the algorithm is therefore much reduced. The concept is carried out with an adaptive codebook. Two fixed-point implementations of the adaptive CELP (ACELP) algorithm are reported: a 7.2 kb/s block coder (7 MIPS), and a 12 kb/s low-delay coder (11 MIPS). Both coders have been rated to provide the same quality as the 13 kb/s block coder adopted by the GSM for the European cellular telephone. >