VOICE SIGNAL DECODING METHOD, VOICE SIGNAL DECODER, DECODING METHOD FOR ENCODED VOICE SIGNAL AND DECODER FOR ENCODED VOICE SIGNAL

摘要

PROBLEM TO BE SOLVED: To improve the quality of reproduced voice with respect to the voice of background noise in a voice signal decoder which generates a voice signal by driving filters composed of filter coefficients by an exciting signal which is obtained by decoding a sound source signal, gain and the filter coefficients and multiplying the sound source signal by the gain. SOLUTION: A sound and silence discrimination circuit 2020 calculates the variation of spectral parameters from LSP and the long-time average of LSP and recognizes sound or silence from the long-time average of the variation, frame power and pitch predicting gain. Further, a noise classifying circuit 2030 averages the long-term average of the variation for a longer time in the silence part and, using the same, the noise is classified. A first switching circuit and a second switching circuit 1120, 2210 switch the filter which is applied into a decoding parameter based on the recognition of sound and silence and the classification of the noise.