Research On Methods Of Sound Source Localization Based On Differential Microphone Array

The technology of sound source localization with microphone array has been a hot research topic in speech signal processing. However, the performance of classical sound source localization methods with small array size is poor and the accuracy can not meet our demand. Therefore, it is a key problem to be solved in the field of microphone array processing to realize the sound source localization with small size array in reverberation and noisy environment. The acoustic intensity estimation method based on differential microphone array provides an important technical way to realize the sound source localization with small size array. In order to estimate the exact direction of the sound source, we take the microphone array as research object and sound intensity estimation as main basic approach, discuss in two aspects from single and multiple sound sources. Specific work is as follows:The basic theory of differential microphone array localization is introduced firstly. And then, the acoustic source localization methods based on sound intensity estimation, time delay estimation and steered response power are described in detail. At the same time, the performance simulation analysis and comparison are carried on. Results show that the sound source localization method based on the sound intensity estimation is better than the other two methods.Because of the poor performance of the basic intensity estimation method in high reverberation, the improved method based on the local standard deviation and instantaneous direction estimation is proposed. Simulation result shows that although this method has a certain effect, the effect is not obvious. Therefore, a method based on the two thresholds and time-frequency masking is proposed in this paper. The method has been improved from time-frequency domain, and the positioning accuracy is greatly improved. In addition, a method based on beam pointing method is proposed from the perspective of spatial domain. Preliminary result shows that this method improves the sound intensity estimation performance effectively. Finally, we combine the method in time-frequency domain and spatial domain and propose another method. Simulation analysis and actual measurement data show that this method is superior to the method improvevd only in time-frequency domain or spatial domain.We research on the multiple sound sources localization methods with intensity estimation and analyze the advantages and disadvantages of the basic histogram method. Then an improved method based on time frequency masking and fuzzy clustering analysis of multiple sound sources is put forward. From the simulation result and experimental data, we are sure that the proposed method is better than the histogram method.