Sound Array Optimization Mechanism Based On Copressed Sensing And Its Application In Sound Source Direction Finding

With the development of array signal processing technology,more requirements are put forward for acoustic array testing technology.Traditional beamforming and spectrum analysis techniques are limited by the actual conditions,such as the small numbers of snapshots and array elements,and are gradually replaced by new theoretical methods in acoustic signal processing.Compresses Sensing（CS）,as a new signal processing theory,has attracted more and more attention because of its advantages such as less sampling data and no restriction by the Nyquist sampling law.However,the application of compressed sensing theory in array testing is still in its infancy.The physical realization of the measurement matrix—the acoustic array configuration,especially the correspondence between the array configuration parameters and the array performance,lacks further research,and there is an urgent need for relevant theoretical support and systematic research.This paper studies the optimization mechanism of acoustic array based on compressed sensing.The work of the thesis mainly focuses on four aspects:（1）Research on the performance evaluation parameters and influencing factors of acoustic arrays when using compressed sensing for sound source direction finding;（2）In the process of acoustic arrays sparse optimization,a new crossover and mutation rules is proposed in the discrete coding genetic algorithm;（3）In the process of random acoustic array optimization,the mutation rule of real-valued genetic algorithm is adaptively improved;（4）The application of the optimized array in sound source direction finding is simulated and tested.The research was supported by the National Natural Science Foundation of China（61871447,61671662）.The main research work is as follows:1.Using compressed sensing for sound source direction finding,the problem of sound source direction finding is transformed into the compressed reconstruction parameter estimation problem.It is proposed to use the coherence parametersμ_max and mean-G to evaluate the performance of the acoustic array.The analogy relationship betweenμ_max and Peak Side-lobe Level（PSL）is established.It proves that the mean-G parameter is more suitable for evaluating the performance of the acoustic array in a special angle space.2.Under the premise that the sound source signal has spatial sparseness,based on the reconfigurable condition of compressed sensing,the key technology of acoustic array configuration is analyzed.Through theoretical analysis,simulation and experimentation,it is proved that random acoustic array with continuous and uniform array element positions is the best,the minimum number of array elements is determined by the sparsity and the number of angular space divisions the minimum spacing of the array elements needs to be greater than half a wavelength.3.An Improved Binary Encoded Genetic Algorithm（IBGA）is proposed to thin the acoustic array.The new crossover and mutation rules are proposed for the binary coded genetic algorithm and large acoustic arrays are sparsely optimized for faster convergence.The simulation results show that when PSL is the optimization target,the algorithm can obtain lower array sidelobe values than other algorithms under different conditions.And when the mean-G is taken as the optimization target,the redundancy of the acoustic array and the mean-G parameter decrease simultaneously in the optimization process,and the change trend is consistent.During the optimization process of the new genetic algorithm,the number of array elements remains unchanged,which is suitable for practical engineering applications.4.The probability distribution function of the peak sidelobe value（PSL）of the random linear array is deduced.A Goal-directed Adaptive Real Coded Genetic Algrhthom（GD-ARGA）is proposed.The algorithm is applied to the optimization problem of random array element position.The problem of easily falling into a local minimum is solved.The stability of the acoustic array optimization result is improved.