Study On Sound Source Localization For Mobile Robot Based On Time Delay Of Arrival

As an important means of human-computer interaction and environment perception,the robot auditory perception system can be used in conjunction with visual perception to compensate for the limitations of visual perception,such as the limitation of light conditions and the influence of obstacles.Therefore,the robot auditory perception system has a wider application.As the core evaluation index of sound source target positioning,positioning accuracy has become a key problem to be studied in auditory perception system.There are many factors affecting the positioning accuracy,including the influence of environmental noise,the actual installation location of microphone array and the rationality of the positioning calculation model.Aiming at the basic problem of improving positioning accuracy,this paper studies the problem of sound source target localization based on the method of time delay of arrival under the premise of the existence of noise.This paper mainly researches the following three aspects:(1)The traditional time delay estimation method is introduced,and the generalized cross-correlation methods using four different weighting functions are simulated and analyzed.Due to the existence of noise in the actual environment,especially nonlinear noise,in order to improve the signal-to-noise ratio of sound signal acquisition,this paper proposes a nonlinear noise elimination method based LM-BP neural network,taking Volterra filter as a reference model,training it,and finally improving the signal-to-noise ratio by noise filtering.Finally,the nonlinear noise cancellation method proposed in this paper is combined with the generalized cross-correlation method,a generalized cross-correlation time delay estimation method based on nonlinear noise cancellation is proposed,and simulation experiments are carried out on the method and the basic cross-correlation and generalized cross-correlation methods.The simulation results show that the method has better anti-noise performance,and the performance degradation rate is reduced.(2)On the premise that the sound source target is the far-field signal,a geometric positioning calculation model is constructed and derived according to the structure ofthe regular tetrahedron microphone array and the sound propagation model,and the simulation experiment of horizontal Angle and pitch Angle error is carried out.The simulation results show that the error of horizontal Angle and pitch Angle decreases with the increase of sound source distance.(3)Aiming at the difference between the actual microphone array and the ideal microphone array,this paper proposes a calibration method for regular tetrahedral microphone arrays.The sound source with a known distance is moved multiple times to obtain multiple sets of measurement data,and the microphone position is estimated based on the data.This method only needs to measure the distance of the sound source and use LM algorithm to process the data.Simulation results show that the method has good calibration results.