Predictive entropy coding for speech analysis

Entropy coding using codes with a simple structure is used in connection with predictive coding to reduce the transmission rate. The predictive coder uses a quantiser with a large number of levels to minimise the effects of slope overload. An adaptive quantising scheme is used to maintain system operation at a point that yields a minimum redundancy. Experimental results on speech coding show that the adaptive predictive entropy coding scheme achieves a good signal-to-noise ratio (SNR) versus rate performance. At a 2-bits-per-sample (12.8 kbit/sec)rate, a segmental SNR of 14 dB is achieved; at 3 bits-per-sample (19.2 kbit/sec), the segmental SNR is 20 dB.