Daniele Sereno

Embedded CELP coding for variable bit-rate between 6.4 and 9.6 kbit/s

The authors consider the design of a variable-bit-rate CELP (code-excited linear prediction) coder which incorporates the facility of producing an embedded bit stream. This characteristic is particularly attractive for packet transmission where some packets can be lost or rejected whenever they are not received within the maximum allowed delay. The basic scheme used for the investigations is a CELP coder in which the innovation signal is split into three separate contributions. The sum of all contributions, together with the side information, determines the operating bit rate of 9.6 kb/s. The reduced bit rates of 8 and 6.4 kb/s can be achieved, dropping respectively the information relevant to one or two contributions to the innovation signal.<<ETX>>

A 5.85 kbits CELP algorithm for cellular applications

Two versions of the RCELP (relaxation-type code excited linear prediction) algorithm are described. They show that the generalized analysis-by-synthesis paradigm provides increased coding efficiency in practical applications, and that it can be implemented in a variety of ways. A novel pitch-period extraction algorithm which improves the effectiveness of RCELP is described. The computational requirements of RCELP are similar to or less that those of an equivalent conventional CELP algorithm. Existing fast procedures for the fixed-codebook contribution to the excitation can be used. At 5.85 kbit/s the two versions of RCELP provide a speech quality which is equal to the current 13 kbit/s GSM speech-coding standard. It is shown that only using the perceptual weighting where needed results in reduction of computational effort and can increase speech quality.<<ETX>>

Some experiments of 7 kHz audio coding at 16 kbit/s

The problem of coding speech at 7 kHz is considered. A possible application of these codecs could be the videophone, especially for hands-free telephone use. The authors propose a split-band coder structure, where both hands are coded with analysis-by-synthesis techniques in order to take advantage of their high coding gain. Two techniques are considered: multipulse coding and codebook-excited linear prediction. The structures of two possible codecs are described, and indications are given of the considerations that led to their design. Their main characteristic is the use, as far as possible, of excitation model parameters optimized within the analysis by synthesis loop, still maintaining a reasonable computational complexity. Subjective and objective results, obtained with high-quality speech, are reported. They show that the obtained speech quality is close to the original.<<ETX>>

A VR-CELP codec implementation for CDMA mobile communications

This paper focuses on the design, implementation and testing of a variable rate (VR) CELP codec aimed to be used in the testbed of one RACE-II project: CoDiT (code division testbed). The project has been conceived to demonstrate the potentiality of CDMA for the UMTS (universal mobile telecommunications system). Because of the flexibility permitted by CDMA to easily convey the information stream over a VR physical channel, the fixed-rate constraint has been removed from the speech coding algorithm design, in order to exploit the time-varying local character of speech. One major feature of the proposed algorithm is the possibility for the average rate to be either source-controlled or network-controlled. This is particularly appealing for cellular communications in order to cope with areas or cells with a high time-varying congestion.<<ETX>>

Method of and device for coding speech signals with analysis-by-synthesis techniques

The set of possible excitation signals is subdivided into a plurality of subsets, the first of which provides the contribution to the coded signal necessary to set up a transmission at a minimum rate guaranteed by the network, while the others supply a contribution which, when added to that of the first subset, causes a rate increase by successive steps. At the receiving side, a decoded signal is generated by using the excitation contribution of the first subset alone if the coded signals are received at the minimum rate, while for rates higher than the minimum rate the contributions of the subsets which have allowed such rate increase are also used.

Comparison of two speech codecs for DMR systems

Abstract Several activities have been undertaken in Italy in the field of Digital Mobile Radio (DMR). In particular, concerning speech coding two codecs were developed to be compared on experirental Single Carrier Per Channel (SCPC) systems. These two codecs are described in the paper and their performances are compared. Both coders are based on a Split-Band (SB) structure that allows an easy exploitation of masking effects: they are a Sub-Band Coder (SBC) and SB-APC (Adaptive Predictive Coder). The main feature of both schemes is the use of Vector Quantization (VQ) to compact the side information. In addition the SB-APC scheme uses the technique of post-filtering to shape the quantizing noise of each sub-band. The main result reported in the paper is that the performances of the two schemes are almost equivalent although their structure is very different.

9.6 kbit/s Piecewise LPC residual excited coder using multiple-stage vector quantization

This paper describes a speech coder which exploits the combination of piecewise LPC analysis, full residual excitation and vector quantization in order to yield very good quality at 9.6 kbit/s. The piecewise approximation of the speech spectrum permits a higher spectral accuracy than standard LPC and reduces the computational load by about 40%. The full residual excitation overcomes the disadvantages of the base-band model, previously used at this bitrate. Finally, the vector quantization approach permits a dramatic bit saving over scalar quantization for LPC parameter compression, and provides a better distortion performance in the residual representation.

A comparison between CELP and MPLPC at 8 kbit/s for mobile communications

Two analysis-by-synthesis schemes, with a different excitation structure, have been optimized for a bit-rate of 8 kb/s. The codebook excited linear predictive (CELP) coder and the multi-pulse linear predictive coder (MPLPC) were considered. Some improvements over conventional coders have been achieved by computing the long-term parameters by a closed-loop procedure and using a split vector quantization of line spectrum pair coefficients for the short-term spectrum representation. The complexity of the two schemes has been considered as a design constraint and the result is the implementation feasibility of both algorithms with standard floating-point digital signal processors. Both coders are shown to provide very high speech quality without transmission errors. The present study is pertinent to the development of the pan-European digital mobile radio system.<<ETX>>

An object oriented generic audio coding architecture

The motivation of this paper is to try to define a framework for object-oriented audio coding, aimed at the development of the challenging idea of a generic coding architecture for audio signals. The inherent flexibility associated with the object-oriented approach allows one to foresee a possible combination of audio and video coding aspects when moving towards multimedia applications. This topic is addressed in the set of requirements of ISO/MPEG4. The MAVT (CSELT-BOSCH-MATRA) audio candidate for ISO/MPEG4 develops this idea mainly by the design of a flexible bit manipulation unit, capable of producing a variety of manipulated bit streams in respect to bit rate, level, and audio object selection. All of these with a unique flexible decoder algorithm.

CELP coding at 6.55 kbit/s for digital mobile radio communications

The work carried out in designing a speech codec suitable for the half-rate channel of the pan-European digital mobile radio (DMR) system is described. A set of CELP (code-excited linear prediction) schemes was selected with bit-rates around 7 kb/s, on the basis of performance without channel errors and complexity. Then the selection of a single scheme was made, taking into account performance with channel errors. The effect of transmission errors was evaluated by measuring the spectral distortion and the segmental signal-to-noise ratio (SNR) for each corrupted bit of the encoded parameters. These data, combined with informal listening tests, resulted in the definition of classes of importance for the bits. The bit error rate (BER) thresholds at which the distortion, due to errors, is perceived were determined for each class. This information was used to optimize the design of a channel coder for the selected speech coder.<<ETX>>