| 摘要 |
PROBLEM TO BE SOLVED: To reduce the amount of computations and to obtain high quality decoded speech signals in the coded speech decoder. SOLUTION: A separating means 2 divides speech signals into plural frequency bands. Then, the frequency band signales are normalized and quantized for every frequency band and coded speech signals are generated. The signals are then separated into quantized values, quantized widths and normalized coefficients. An inverse quantizing means 3 inverse quantizes the quantized values for every frequency band in accordance with the quantized widths and inversely quantized signals are generated. An inversely normalizing means 4 inversely normalizes the inversely quanitzed signals for every frequency band in accordance with the normalized coefficients and inversely normalized signales are generated. A zero inserting means 7 sets a portion of the signals inputted to an inverse mapping means 5 to zero values in accordance with an output specimen frequency reduction magnification. Then, the means 5 generates voice signals in accordance with the output signals of the means 7. A thinning means 8 only outputs a portion of the speech signals outputted by the means 5 in accordance with the magnification. |
