PROGRAMMING ASSIGNMENT

Programming assignment

Programming Assignment

Introduction

This paper is about a vocoder operating at 600 bps. Such vocoders are mainly of interest for military use where communications bandwidth is at a premium, such as over tactical HF radio using frequency hopping as an electronic counter-counter measure (ECCM). In such applications, users will accept the disadvantages (against higher rate vocoders) of greater terminal equipment cost and lower speech quality (McLarnon & Holmes, 1974: 89). There are also potential applications for stored message systems, in particular those requiring a large repertoire of messages and low cost. As usually most of the cost of vocoders is in the analyser, for these applications a low-cost synthesiser would be possible and memory costs for storage of speech would be reduced. Work to date has been on the specification, design and implementation of a non-real-time software emulation of the vocoder. Real-time implementation constraints were taken into account in the specification and design. The software emulation will be used for parameter tuning and evaluation purposes. Work on parameter tuning and preliminary evaluation is expected to start immediately after this (Bridle & Sedgwick, 1977: 121).

The vocoder, invented by Homer Dudley in about 1939, was the first system to achieve bandwidth compression of speech. Since then, many other schemes have been proposed and tried but none thus far has been as worthwhile as the vocoder. Group 6Z has been actively engaged in the design and construction of a vocoder. This report traces the history of Group 62's effort, its motivations, and some of the interesting problems encountered. In order for the reader to better grasp the flavor of these problems, it was thought desirable to include some preliminary discussion of the speech process and of vocoders in general.

Previous Work

Seeviour, Holmes and Judd (1976) and Holmes (1978) onspectral analysis in the closed-glottis phase using discrete Fourier transforms (DFTs), and on estimation of formant frequencies and amplitudes by analysis-by-synthesis. Holmes (1974) and McLarnon, Holmes and Judd (1974) on variable frame-rate (WR) coding using segments with constant and linearly interpolated formant parameters, selected according to thresholds on the spectral coding error. Reported data rates were 1200 and 1500 bps (Holmes, 1978: 14).

Dupree (1984) on VFR coding using a dynamic programming (DP) algorithm to select simultaneously the optimal sequence of frames to transmit, whether segment infill should be by replication (constant parameters) or by linear interpolation, and which assignment between formants and spectral peaks should be used (from several hypothesized alternatives) for the transmitted frames. This DP approach is a continuous version of that proposed by Bridle and Sedgwick (1977). Reported data rates were as low as 580bps (Dupree, 1984: 279). The work on parallel formant synthesis also carried out at JSRU (see Rye and Holmes 1982) complements the above approach to formant analysis to give a complete formant vocoder.

Dudley's Channel Vocoder

The first major effort to encode speech electronically was Homer Dudley's channel vocoder (“voice coder”) developed starting in October of 1928 at AT&T Bell Laboratories. An overall schematic of the channel vocoder is ...