Adaptive Predictive Coding of Speech Signals

The aim of efficient coding methods 1 is to reduce the channel capacity required to transmit a signal with specified fidelity. To achieve this objective, it is often essential to reduce the redundancy of the transmitted signal. One well-known procedure for reducing signal redundancy 1973 1974 THE BELL SYSTEM TECHNICAL JOURNAL, OCTOBER 1970 is predictive coding.* 2-J In predictive coding, redundancy is reduced by subtracting from the signal that part which can be predicted from its past. For many signals, the first-order entropy of the difference signal is much smaller than the first-order entropy of the original signal; thus, the difference signal is better suited to memoryless encoding than the original signal. Predictive coding offers a practical way of coding signals efficiently without requiring large codebook memories. Many previous speech coding methods 0 have employed schemes which attempt to separate the contributions of the vocal excitation from that of the vocal-tract transmission function. The well-known channel vocoder of Dudley 7 was the first attempt in this direction. Although vocoders can reproduce intelligible speech, there is appreciable loss in naturalness and speech quality. This degradation in speech quality arises from various operations in the vocoding process, which are either inaccurately performed or are based on certain idealized approximations of speech production and perception processes. The present paper describes a different approach 8,9 to encoding of speech signals, based on predictive coding, which avoids the difficulties encountered in vocoders and vocoder-like devices.