待识别语音净化处理系统

Purification system for the speech to be recognised

为提高语音识别系统在噪声环境中的性能,文中提出利用分阶段ICA算法净化待识别语音,并基于此算法设计DSP语音净化系统,有效的提高语音识别系统的识别率和鲁棒性。

In order to improve the capability of speech recognition system in noising environment, a method is proposed which can purify the speech before recognition by using MSICA. A DSP system is also designed based on this arithmetic. The system can improve the recognition accuracy and better robustness.