Abstract:

Endpoint detection of speech signal processing is a very important step,which directly affects the accuracy of the speed and results in voice signal processing.The research on endpoint detection,especially in noisy environments,has always been the hot spots in voice signal processing.Aiming at the front-end detection,which is made for the voice recognition system,this paper compares the use of short-term energy with the use of short-term average zero-crossing rate for endpoint detection.It achieves a double-threshold method of endpoint detection programming and simulation by Matlab.Simulation results show that an accuracy of as high as 93% for endpoint detection can be achieved.

Key words: endpoint detection;speech signal processing;short-term energy;zero crossing rate;dual-threshold