论文部分内容阅读
研究了利用盲源分离理论解决语音识别之前的语音净化问题。基于MSICA算法良好的性能、低复杂度和实时性,设计了以TI公司的TMS320C6416数字信号处理器为内核的语音净化系统。该系统可以使语音识别系统获得纯净语音,从而有效提高语音识别系统的识别率和鲁棒性。
The problem of speech purification before speech recognition is solved by using blind source separation theory. Based on the good performance, low complexity and real-time performance of MSICA algorithm, a speech purification system based on TI’s TMS320C6416 digital signal processor is designed. The system can make the speech recognition system obtain pure speech, so as to effectively improve the speech recognition system recognition rate and robustness.