Voiceprint Identification Endpoint Detection Research Of Intelligent Residential Entrance Guard System

Voiceprint identification endpoint detection accuracy of intelligent residential entrance guard system directly affects the performance of the voiceprint recognition system.Voice signal endpoint detection research is an important issue in voiceprint recognition system。Research shows that more than half of the speech recognition errors are due to the incorrect voice signal endpoint detection. To make the voiceprint recognition technology better serve the intelligent production and life, Voiceprint recognition system with high accuracy voice signal endpoint detection, and short endpoint detection time is necessary.But inevitably there will be a variety of noise and interference in real environment, so it is necessary to consider the impact of noise in the speech signal analysis and processing, especially in this cases that has stronger noise and interference, when the signal in low signal noise ratio conditions signal will be seriously declined. In order to improve the accuracy of voiceprint recognition, endpoint detection of speech signals at low signalnoise ratio is very necessary.In this paper,we discussed the voiceprint recognition system endpoint detection principle,and then make an overview of the relevant theoretical knowledge of the Voice signal processing, reviewed the existing common endpoint detection algorithm and made some simulation to verify its detection performance.Simulation and analysis proves that the detection performance of the common algorithm is relatively well in the high signal noise ratio situation, but the detection performance will be significantly decreased in relatively low signal to noise ratio environments or complex situation.In order to improve the above problem, a endpoint detection algorithm is proposed based on logarithmic energy weighted Band-partitioning Spectral Entropy feature.The algorithm combined the advantages of the logarithmic energy and the Band-partitioning Spectral Entropy in voice signal endpoint detection, improved the miscarriage of justice in speech and noise which is because of the amplitude as well as overcame the impact of a single frequency noise on the voice, and improved the accuracy of endpoint detection with high robustness.Experimental and analytical results show that the proposed algorithm can achieve accurate endpoint detection in low SNR situation, and the detection performance is better than the voice signal endpoint detection algorithm based on logarithmic energy method.