| 摘要 |
One embodiment of a speech recognition system is organized with speech input signal preprocessing and feature extraction followed by a fuzzy matrix quantizer (FMQ). Frames of the speech input signal are represented in a matrix by a vectorf of line spectral pair frequencies and energy coefficients and are fuzzy matrix quantized to respective vector +E,cir f+EE entries of a matrix codeword in a codebook of the FMQ. The energy coefficients include the original energy and the first and second derivatives of the original energy which increase recognition accuracy by, for example, being generally distinctive speech input signal parameters and providing noise signal suppression especially when the noise signal has a relatively constant energy over at least two time frame intervals. To reduce data while maintaining sufficient resolution, the energy coefficients may be normalized and logarithmically represented. A distance measure between f and +E,cir f+EE , d(f, +E,cir f+EE ), is defined as where the constants alpha 1, alpha 2, beta 1 and beta 2 are set to substantially minimize quantization error, ei is the error power spectrum of the speech input signal and a predicted speech input signal at the ith line spectral pair frequency of the speech input signal, the first G LSP frequencies are most likely to be frequency shifted by noise, and the last P+3 coefficients represent the three energy coefficients. This robust distance measure can be used to enhance speech recognition performance in generally any speech recognition system using line spectral pair based distance measures.
|
