Research On The Speech Enhancement Technique Of Laser Detector System

Laser detection is a new type of non-contact detection technology,with the advantages of no need placing detection equipment close to the target,easy operation,and good invisibility,which has important value in the military and police field.Since lase r detection is more sensitive to external environmental conditions,and the detected speech signal is susceptible to environmental noise.So in order to meet the demands of actual intelligence collection work,the laser detected signal will be processed by speech enhancement techniques.Due to the properties of laser detection techniques,the noise in the detected signal is quite different from the noise in the traditional speech signal,and the existing algorithms cannot meet the practical needs.This paper aims at the speech enhancement algorithms of laser detection system,and the main work is generalized as follows:1.Based on the analysis of the broadband background noise in the laser detected speech signal and the spectral enhancement algorithms,a modified Wiener filtering speech enhancement algorithm with phase spectrum compensation is proposed.The proposed method updates the noise power spectrum estimation according to the current speech frame state by voice acti vity detection,and compensates the speech spectrum with phase spectrum compensation function,which retains more speech details information.Simulation results show that the proposed algorithm can improve the speech Signal-to-Noise Ratio(SNR)and the speech quality.2.Based on the analysis of the high energy noises properties in the laser detected speech signal,which include truncation noise and impulsive noise.This paper proposes the noise suppression algorithms.Firstly,the time-domain distribution of truncation noise is determined by detecting the difference value of signal,and then the data segments in which the error occurred are corrected accordingly.Then for the impulsive noise caused by environmental causes,its time-domain distribution is determined by detecting the energy of the signal in the high frequency segment of the speech spectrum,and the speech disturbed is estimated by the linear prediction analysis with the AR speech model.The simulation results show that the proposed algorithm has a significant effect on the impulse noise suppression and greatly improves the speech quality.3.The Hiss noise in the laser detected signal is analyzed,and a modified algorithm for suppressing Hiss noise based on modulus maximum value wavelet transfo rm is proposed.The proposed algorithm sets different thresholds at different wavelet scales,and the thresholds are associated with the noise power spectrum.The simulation results show that the algorithm suppresses the Hiss noise well and reduces the auditory fatigue sense.4.The proposed algorithms are tested with the actual laser detected signal,and the influence of the detected target distance,the laser reflector and the surrounding environment condition on the detected signal quality is discussed.