Research on speech endpoint detection based on the improved dual-threshold algorithm
Xue Sheng-ya · Electronic Design Engineering · 2015
Speech endpoint detection is an important step during the voice signal processing, which directly affects the speed of the detection accuracy and effectiveness of the voice signal processing. The traditional method based on the double threshold algorithm, in the case of the pure voice, the voice judging is more accurate, but in the case of speech in noise, especially in the case of the low signal to noise ratio, the recognition of the endpoint rate is very low, the error rate is very high. In order to improving the recognition rate, we adjust the threshold number, smooth the signal, introduce the minimum length of the speech ended, through the Matlab simulation experiments, we get a good speech endpoint detection accuracy.