The Computation Speed-up Of Source Localization And Distributive Calibration For Microphone Arrays

Microphone Array enables the acquisition of the space-time structure of an acoustic field.Traditional single-array technology has performance bottlenecks in high-accuracy acoustic analysis.In recent years,the focus of research has gradually shifted to distributed-array system for high-accuracy acoustic analysis and large-space sound field detection.This paper mainly focuses on the computation speed-up methods of single-array and the inter-calibration of distributed-array.Firstly,the basic techniques of microphone array are introduced,including the structural characteristics of single-array and distributed-array,and the sound field delay model.The analysis of sound source localization algorithm based on beamforming are given.And the general method of distributed-array calibration is discussed in combination with acoustic camera mapping model.Secondly,for the computation speed-up problem of single-array system,the prototype of the system based on sound source localization is designed and developed,and the computation hotspots are analyzed.For the heavy load problem with large number of array elements,high-efficiency array element pairing and multi-thread speed-up calculation are proposed.Results show that the computational load of single-array system is reduced by half.Thirdly,for the calibration problem of distributed-array,computer vision processing technology and image mapping model of epipolar geometry and normalized-eight-point algorithm for solving are used.The inter-calibration model is further analyzed and established.The effectiveness of the model and algorithm is verified by simulation.Finally,for the self-calibration deviation problem inside the array,a beamforming algorithm based on cross-correlation function and an angled-based image mapping method are proposed.The tolerance performance of proposed method to microphone position is analyzed by simulation.Results show that,compared with the existing solution,the tolerance capability of the proposed method is doubled and the robustness of array calibration is improved.