Adaptive coding for conversational speech communication
C.R. Raviraj, E.V. Jones · 1989
Three techniques of channel capacity reduction (decreasing the SNR, decreasing the number of bits per sample, and decreasing the sampling rate) have been assessed for their subjective acceptability during double talk. A test-bed system capable of adaptively coding the speech sources during double talk has been constructed, this enabled us to hold conversational tests between two people through conventional looking telephones. The waveform coding/decoding part of the system is handled by a digital signal processing card based on a TMS32010 digital signal processing (DSP) chip. The back end processing system based on a Motorola MC68000 CPU is used for active speech detection and adaptive coding during double talk. This system enables the user to select a variety of adaptive techniques to process the PCM codes during double talk. It can also simulate the effect of end-to-end delay in adaptive coding.