Research On Speech Enhancement Of Underwater Speech Communication Devices

Frogman underwater communication is one of the hot issues in the current research.In order to improve the speech quality,this thesis studies the speech enhancement methods for the sending and receiving of the frogman underwater speech communication system.The paper first conducted research and simulation tests of the SSB method of modulation at the sending of the frogman underwater speech communication system.The principles and advantages and disadvantages of various SSB method of modulation were introduced and analyzed,including filtering method,phase migration method and weaver method.and finally selected the phase migration method.Then,for the modulation method of phase migration method,the basic characteristics of the water sound channels were fully considered,The results show the stability of the communication in the water acoustic channel and provide theoretical support for the realization of phase migration method at the sending of the frogman underwater speech communication system.At the sending,a targeted speech enhancement solution is designed,including the design of the test scheme and processing algorithm.First of all,according to the sending process at the underwater frogman sending,the thesis point that the algorithm processing should be arranged before SSB modulation,and the design tasks of the mask of test,the pool test and the algorithm processing process are completed.For the collecting test data,The thesis studied the endpoint detection algorithm based on different parameters,and the spectrum reduction method,the wiener filtering method,the MMSE-based spectrum estimation method and the subspace method are also studied.The simulation test shows that the performance of the voice endpoint detection algorithm of combined energy and spectrum entropy has a stable performance;the logmmse algorithm compared to other traditional speech enhancement algorithms,under the bubble noise background,it has better enhancement effect,robustness and performance better.At the end,the receiving speech is mainly disturbed by underwater noise.From the perspective of increasing the signal-to-noise ratio of the receiving speech,the thesis completes the speech enhancement task.First of all,this thesis uses passive time mirror with multipath focusing to process the receiving speech.Based on the LFM signal,the underwater channel is estimated by OMP algorithm.Which has a stable channel estimate ability.At the same time,in order to solve the problem of ship noise.the thesis uses the non-negative matrix algorithm based on the wavelets to enhance speech.The simulation results show that passive time mirror has a certain enhancement effect on receiving speech and can achieve the energy focus of the speech,the DWPT-NMF algorithm has good processing performance with ship noise distributed in the water sound channels.Finally,tests were conducted in the channel pool,Weihai,Shandong,and Sanya,Hainan for the speech enhancement algorithm of this thesis.The results show that the speech enhancement scheme used at the sending can complete the endpoint detection and noise reduction tasks of the speech which has good stability and reliability;the speech enhancement methods used at the receiving perform well with different real tests data,the algorithm gain is high,and the robustness is fine.