一种改进的基于短时能量的端点检测算法

A Modified Endpoint Detection Algorithm Based on Short-time Energy

为了提高语音识别的实时性和识别率,采用动态窗长短时能量进行语音信号的端点检测方法.比对传统短时能量短时平均过零率端点检测方法,通过实验得到的端点检测波形图像及数据证明,动态窗长短时能量方法可以减少不必要的语音帧的处理,提高识别的速度和识别率.

For the purpose of improving the real-time capabilities and the efficiency of the recognition, a method of dynamic window length short-time energy is used for endpoint detection for speech signals. Comparing with the traditional method of short-time energy and short-time zero-crossing rate, the waveform of endpoint detection and the experiment result prove that the method of dynamic window length short-time energy endpoint detection can de- crease the processing of useless speech frame, increase the speed and the efficiency of the recognition.