噪声背景下语音识别中的端点检测

整个实验系统主要研究的是语音识别中在噪声背景下的端点检测.在实际环境中,加性背景噪声对语音识别的影响非常大,在信噪比小的情况下,无法进行端点检测.在此研究过程中,首先进行加性噪声模型的建立,而后通过滤波对带噪语音信号进行预处理,这里采用维纳滤波和自适应LMS滤波进行滤波.在进行语音信号的预处理后,使用VAD算法进行端点检测.建立整个实验系统后,通过试验比较维纳滤波器和自适应LMS滤波器的优劣,找出最佳方案.

Vertex Examination in Speech Recognition Under the Noise Background

The entire experimental system discusses vertex examination in speech recognition under the noise background.In the actual environment,the additive background noise influences speech recognition greatly.It is impossible to carry on the vertex examination in less-noisy ratio situation.In this study,additive noise model is established,then pretreatment of the belt chirp pronunciation signal through the filter is conducted.The filter adopts the Wiener filter and the auto-adapted LMS filter.Finally,vertex examination is carried on through the VAD algorithm.After establishing the entire experimental system,the best pattern may be discovered by comparing the Wiener filter and the auto-adapted LMS filter.