Discrimination of speech and high-speed data using an adaptive predictor for DSI application

A description is presented of a discrimination technique regarding high-speed voiceband data. The discrimination is based on measurements of short-time energy and zero-crossing rate of the input signal, and linear prediction coefficients of an adaptive predictor. Classification concerning 4.8-kbit/s 8-point Quadrature Amplitude Modulation (QAM) and 9.6-kbit/s 16-point QAM voiceband data is performed by making use of a coefficient of variation of the short-term amplitude distribution of the input signal. The coefficient of variation provides a measure of the degree to which the amplitude is dispersed. The considered technique can be used to discriminate between speech and high-speed voiceband data.