Research On Anti-interference Technology For Frogman Formation Speech Communication

Frogman underwater speech communication system is an indispensable equipment for underwater single frogman combat and frogman formation work.Among various methods of modulation,SSB speech communication system holds the advantages of high stability and better retention of speaker tone,however with no capability of suppressing multichannel and being susceptible to noise.In addition,when frogmen are working in formation,their voices alias around the receiver and cause mutual interference,which is another problem seriously to be solved.Under the background of underwater frogman SSB speech communication system,this paper studies a blind multichannel elimination method based on complex cepstrum feature filter with minimum phase decomposition and a kind of blind source separation technology aiming at mixed speech based on Fast ICA,then improves the anti-multichannel capability of frogman SSB speech communication system as well as the user tolerance of frogman formation speech communication system.Firstly,paper studies a multichannel suppression method based on complex cepstrum characteristic peak filter.This method can effectively filter out the amplitude components of multichannel corresponding to different distribution regions of modulated voice and underwater channels in the transform domain.Then,aiming at the difficulty of phase winding and complex phase calculation in complex cepstrum characteristic filter,an improved algorithm based on minimum phase decomposition is put forward to solve this problem.The effectiveness of the improved algorithm is verified by a comparison test.Next,the simulation algorithm is tested under different SNR conditions and speech denoise is proposed to improve the performance of the algorithm in the case of low SNR.Finally,the lake experiment data verifies the performance of this algorithm.Secondly,this paper studies the method of multi-user speech blind separation based on Fast ICA.This method aims at solving the problem of maximizing the objective function of non-Gaussian of negative entropy and accomplishes the separation of aliasing speech.In this part,speech denoise treatment is proposed to improve the performance of the algorithm for seperating three or four users in low SNR.The paper studies influence of simulation near and far effect on the two-user BSS algorithm.The reason of algorithm performance deterioration and the method of critical value calculation under specific scenarios are givenfrom the angle of aliasing matrix.Considering the actual situation of formation speech communication,the effect of BSS algorithm under multichannel interference is simulated.While the algorithm works effectively in two-user system,the effect in three-user system is general and denoise algorithm effect is rather bad.The reason is analyzed and by conducting the simulation with different Numbers of frogmen under different SNR conditions,the performance of this algorithm is verified by the lake experiment data.Eventually,the thesis completes the design and embedded implementation of anti-multichannel frogman speech communication algorithm based on TMS320C6655 platform.This paper briefly introduces the receiving model of anti-multichannel frogman speech communication system and the data receiving process of DSP platform.According to the requirement of real-time receiving,a segmented real-time speech processing model based on overlapping reservation method is proposed.The specific parameters of the model are determined through simulation analysis.Finally,the feasibility of the algorithm is proved by data and the embedded implementation of anti-multichannel frogman speech communication algorithm is realized.