| 摘要 |
PURPOSE:To improve reproduced sound quality by multiplexing and outputting a quantized signal which is obtained by predicting linearly an even or odd sample output of <=4kHz linearly, a prediction parameter, and a linear prediction coefficient reproduced by using the prediction parameter. CONSTITUTION:An input speech signal is distributed to an even sample extraction part 1 and an even sample extraction part 2 of <=4kHz. A linear prediction part 3 predicts the output of the sample extraction part 1 linearly to output the quantized signal and prediction parameter to a multiplexing part 5. A reproduction part 4 reproduces a speech from the quantized signal from the prediction part 3 by using the prediction parameter from the prediction part 3. In this case, the linear prediction coefficient is determined so as to minimize the residue of the reproduce speech signal and the sample output of the extraction part 2 and outputted to a multiplexing part 5. Therefore, a sampling frequency can be reduce to half and the number of quantized bits can be increased, so the quality of a reproduced speech signal can be improved. |
