| 摘要 |
<p>For quantization of subframe feature vectors representative of subframes identified by subframe indexes in each frame of a time sequential signal, such as an audio or a video signal, a first feature vector of a first subframe of a current frame is extracted with additional extraction of a second feature vector of a second subframe of the current frame and is quantized by a frame vector quantizer unit into a quantized frame vector for the current frame. Using a quantized frame vector for a previous frame and referring to interpolation coefficient vectors, an interpolator interpolates an interpolated feature vector for the current frame. An error vector is calculated between the interpolated and the second feature vectors and is quantized by an error vector quantizer unit into a quantized error vector. The subframe feature vectors are quantized by using the subframe indexes, the quantized frame vectors, the interpolated feature vector, and the quantized error vector. Preferably, the interpolation coefficient vectors are adjusted according to a mode decided for the current frame by its subframes.</p> |
