The Research On Sound Source Localization Of Digital Hearing Aids

Digital hearing aids can effectively ameliorate the aural comprehension of the patients. Sound source localization is one of the core technologies of digital hearing aids. This paper focuses on the sound source localization algorithm on digital hearing aids and proposes improved algorithms. The main work includes:Firstly, this paper researches the working principle of digital hearing aids, the key algorithms of digital hearing aids including speech enhancement, loudness compensation and feedback cancellation and so on. By analyzing and comparing with the current sound source localization, this paper finally chooses a sound source localization method based on time delay estimation, which has the advantages of small calculating amount and high real time. And then, this paper analyses the digital model of speech signal and room impulse response model, and the microphone array signal ideal model and the actual model are established respectively.Secondly, this paper places emphasis on the sound source localization method based on time delay estimation. It introduces the widely used generalized cross correlation(GCC) method, and discusses the characteristics of the weighting functions and performs simulation analysis. In order to solve the problem of GCC, in this paper, combined with fine interpolation of correlation peak method(FICP), an improved generalized cross correlation algorithm based on second correlation is proposed. The simulation results show that compared with the traditional GCC method, the proposed method can improve the performance of time delay estimation significantly.Thirdly, to solve the problem that the performance of cross-power spectral phase(CSP) method decreases significantly in the high noise and reverberation environment, this paper proposes an improved CSP method. This method introduces voice activity detection(VAD) to eliminate noise frames to improve the estimation accuracy, and uses a modified weighting function to sharpen the peak of the correlation function, at last accumulates many frames to reduce the computational complexity and improve the estimation accuracy.Finally, on the basis of source localization analysis, this paper introduces geometrical location and the way of the searching target of function space, and design a localization model by four microphones. It has two modules, including improved generalized cross correlation algorithm based on second correlation and geometrical location. The simulation results show that this system is easy to realize and has less computational complexity, so it can be used in practical environment.