A Blind Positioning Method for Sound Signal Source

Abstract

With the development of silicon microphone technology, high-precision spatial positioning of sound sources becomes possible. When the sound source position of its near-field and far-field models is known, three-dimensional positioning is relatively easy. However, it is difficult to predict the range of the near-field and far-field of the source, which brings difficulties to the subsequent algorithm processing. In addition, the reflection and noise of the sound source signal, reverberation and the phase inconsistency of the silicon microphone also affect. In order to solve these problems, this paper proposes a fusion localization method, and optimized through software tools. The simulation results show that the algorithm is better than the traditional algorithm. The error has been reduced by more than three times. In the far-field model, the influence of the amplitude is eliminated by generalization, thereby optimizing its positioning accuracy. The stability of the positioning is increased by covariance calculation in the near-field model. By comparing the variances of the two models for the calculation of the same sound source, the calculation method with smaller error is selected to determine the model, and the final sound location. Simulations demonstrate the effectiveness of the proposed method. Then the five-sensor cross microphone array is used to verify the experimental results. The experimental results show that the method can effectively improve the positioning accuracy of blind sound sources and has high practical value.