A nonuniform sampling approach to data compression of speech and voiceband data signals

Abstract

Using voiceband data and speech like signals the performance of the Asynchronous Delta Modulator. (ASDM) has been investigated and its capabilities as a potential medium-band speech coder have been evaluated. The asynchronous delta modulator is a versatile source coder in which the sampling rate is determined directly by the activity of the input signal, hence the samples are nonuniform. When the input signal is idling ASDM samples it infrequently. Active segments of the signal are, on the contrary, encoded with many samples. Since the samples are nonuniform in time, the information in the signal is coded into the output bit polar.ities as well as the time intervals between these bits. Therefore quantization of these inter-bit intervals is necessary as well as a buffer to output the asynchronous input information at a synchronous rate. This study has mainly concentrated on the quantization and encoding of the inter-bit intervals. For both speech and QPSK data, ASDM sampling statistics have been investigated. Logarithmic, optimum and vector quantization .of the inter-bit intervals have been considered. For further compression, entropy coding of the inter-bit interval quantizers have been investigated. Buffer behaviour for different system parameters has been evaluated and the results have been compared with popular speech coders.