摘要 |
PURPOSE: To obtain the voice coding method with less quantization error even when an amplitude of a voice signal is suddenly changed. CONSTITUTION: The ADPCM coding method adopts a quantizer (13) conducting quantization based on a difference dn between a voice input signal xn and a predicted value yn of the input signal xn , a coder (14) coding a quantized value qn by the quantizer (13), and a 1st quantization width update device (15) updating a quantization widthΔn based on the difference dn and the quantization widthΔn and giving the updated quantization widthΔn to the quantizer and also with a 1st prediction device (17) obtaining a succeeding prediction signal yn+1 based on the quantized value qn by the quantizer (13) and the predicted value yn of the input signal. Furthermore, the quantizer (13) conducts quantization of en =dn +Δn /2 in the case of difference dn >=0 and conducts quantization of en =dn -Δn /2 in the case of difference dn <0.
|