| 摘要 |
PURPOSE:To execute the application of a voice encoding and decoding system with a high quality by quantizing the mean characteristic of a sound source parameter with a vector quantization efficiently, thereafter, extracting important local components as time series from the sound source parameter and correcting them individually. CONSTITUTION:As the quantizing system of the sound source parameter in the voice encoding and decoding system to express a sound signal with a linear prediction parameter by a linear prediction analyzing part 2 and the sound source parameter by a sound source parameter extracting part 11, a constitution to correct the error of the vector quantization by a vector quantizing part 3 by using a scalar quantization with a scalar quantizing part 5, and thereby, a local fluctuation in a time direction to be difficult only by the vector quantization can be followed. Consequently, a flexible quantizing method can be obtained that the mean characteristic of the sound source parameter is extracted by the vector quantization, thereafter, the important local characteristic is extracted as the time series of the sound source parameter and it is individually corrected. Thus, the application of the voice encoding and decoding system with a high quality can be executed. |
