Pitch synchronous speech coding techniques.

Efficient source coding techniques are necessary to make optimal use of the limited bandwidth available in mobile phone networks. Most current mobile telephone communication systems compress the speech waveform by using speech coders based on the Code Excited Linear Prediction (CELP) model. Such coders give high quality speech at bit rates of 8 kbps and above. Below 8 kbps, the quality of the coded speech degrades rapidly. At rates of 6 kbps and below, parametric speech coders offer better speech quality. These coders reduce the required bit rate by transmitting certain characteristics of the speech waveform to the decoder, rather than attempting to code the waveform itself. The disadvantage of parametric coders is that the maximum achievable quality is limited by assumptions made during the coding of the speech signal. The aim of the research presented is to investigate and eliminate the factors that limit the speech quality of parametric coders. A new pitch synchronous coding model is proposed that operates on individual pitch cycle waveforms of speech rather than longer, fixed length frames as used in classic techniques. In order to implement a pitch synchronous coder, new pitch cycle detection algorithms have been proposed. Pitch synchronous parameter analysis was investigated and several new techniques have been developed. A novel pitch synchronous split-band voicing estimator has been proposed that utilises only the phase of the speech harmonics rather than the periodicity used in traditional techniques. Fixed rate quantisation of pitch synchronous speech parameters has been investigated and a joint quantisation/interpolation scheme has been proposed. This scheme has been applied to the quantisation of the pitch synchronous parameters and has been shown to outperform traditional quantisation techniques. A comparison of a reference parametric coder with its pitch synchronous counterpart has shown that the pitch synchronous paradigm eliminates some of the main factors that limit the speech quality in parametric coders. It is expected that this will lead to the development of speech coders that can produce speech of higher quality than current parametric coders operating at the same bit rate. Key words: Speech Coding, Pitch Synchronous, Sinusoidal Coding, Split-Band LPC Coding