Sound Source Localization And Tracking Accurately Under Low SNR

Obtaining the position information and motion trajectory of the sound source from the audio signal collected by the microphone array is one of the advanced researches in signal processing field.With decades years of research,many scholars have put forward a lot of new theories and effective methods of positioning and tracking and some of the achievements have been applied to many fields,such as service robots,intelligent conference rooms,etc.These achievements have also improved the environment of people's life and work.However,with the development of artificial intelligence,people hope to achieve accurate sound source location and tracking in a more complex and changeable environment.But,existing technologies still have problems to be solved in theory and application.Therefore,studying the source localization and tracking algorithm with strong adaptability and unconstrained of background has important academic and application value.Based on the time delay estimation localization algorithm and particle filter tracking algorithm,this paper studies from four aspects which are improving the time delay precision between signals,reducing the computational complexity of near-field sound source location estimation,enhancing the anti-noise ability of the tracking algorithm,and preventing the sound source from randomly disappearing or appearing to influence the tracking accuracy.The main research contents of this paper are as follows:1.Aiming at the problem of low precision and high computational complexity for traditional near-field sound source localization methods under unknown noise and low signal-to-noise ratio,a robust localization method based on improved time arrival difference(TDOA)near-field sound source is proposed.Firstly,the original audio signal is preprocessed by traditional FIR Wiener filter,which enhance the SNR,suppress the additive noise and calculate the TDOA value in the signal frequency domain.Secondly,the airspace shrinkage iterative least squares algorithm is proposed,which replaces the traditional least squares(LS)algorithm to fit the near-field sound source location information.The algorithm reduces the computational complexity of spatial search by narrowing the search region by the iterative region.2.In order to solve the problem that the lower tracking precision in standard particle filter(PF)algorithm,due to the lack of current observational values.With the in-depth analysis of the sound source tracking problem,spherical-radial cubature rule is used to solve the integration problem of the algorithm fusion and linearizes the nonlinear problem.The current measurement information is incorporated into the state estimation by cubature Kalman,and the proposed distribution function isupdated to ensure the diversity of sampled particles and the estimation accuracy of the algorithm.3.Aiming at the problem of low accuracy of traditional methods in alternating acoustic tracking,a cubature Kalman particle filter algorithm was used in this paper to improve the tracking effect of dynamic sound source,and construct an alternating acoustic source tracking framework.The mobile decision factor was introduced and avoided the interference of acoustic source tracking unrelated information before and after the quiet period.At the same time,the initialization system restarted the update iteration,which could shield the system from the unrelated prior information.The simulation results show that the method can obtain more accurate alternating acoustic source trajectory in low SNR environment.