摘要 |
<p>A speech encoder employing various encoding schemes based upon parameters including the noise-like spectral content. The spectral content may be further classified as being either substantially stationary or non-stationary noise signal. For stationary noise-like signals, Gaussian excitation may be used. For relatively lower bit rates, the excitation may contain both a pulse-like and a Gaussian like component. In certain embodiments of the invention, a smoothing process is performed. For example, the line spectral frequencies (LSFs) and various gain parameters that are used to perform encoding of the speech signal may be adjusted to perform perceptual smoothing of stationary noise-like signals. These smoothing procedures need not be synchronous. The smoothing of the codebook gains may be performed on one time scale basis and the smoothing of the LSFs or another.</p> |