论文部分内容阅读
将人耳听觉感知方面的知识用于孤立数字语音识别，提取出感知线性预测倒谱参数（ＰＬＰＣ）．对ＰＬＰＣ参数和线性预测倒谱参数（ＬＰＣＣ）进行比较，得出使用８阶ＰＬＰＣ参数可以获得优于１２阶ＬＰＣＣ参数的性能．采用基于ＰＬＰＣ参数的多种特征参数进行识别，非特定人识别率可以达到近９７％．
The knowledge of human ear auditory perception is used for isolated digital speech recognition, and the perceived linear predictive cepstrum parameter (PLPC) is extracted. Comparing PLPC parameters with linear predictive cepstrum parameters (LPCC), we conclude that using 8-order PLPC parameters can obtain better performance than 12-order LPCC parameters. Using a variety of characteristic parameters based on the PLPC parameters for recognition, non-specific recognition rate can reach nearly 97%.