| Robot binaural sound source localization technology is one of the important research directions of robot hearing,and has good application prospects.Mobile robots can provide a basis for human-computer interaction by judging the direction of the sound source.In indoor environments with reverberation and noise,the performance of binaural sound source localization based on the interaural time difference(ITD)is severely degraded.The relative motion between the mobile robot and the sound source will cause a certain error between the calculated value of the localization and the actual value of the sound source.Aiming at these problems,this paper studies the binaural continuous sound source localization technology when the robot and the sound source move relative to the indoor environment with reverberation and noise.The main research contents and work of the paper are as follows:Firstly,according to the characteristics of binaural sound source localization of mobile robots,the characteristics of the robot’s binaural localization cues are studied and analyzed;under the constraints of the far-field model and binaural localization resolution,in order to meet the characteristics of the miniaturization of the robot’s binaural array platform,design the appropriate binaural array spacing;then study the relative position of the robot and the binaural array,design the relative position of the binaural array and the robot according to the motion characteristics of the robot and the sound source and the face-to-face interaction between the human and the robot;A localization model of the binaural array for ear time difference.Secondly,for the performance degradation of binaural sound source localization based on cross-correlation in the environment of reverberation and noise,a method based on signal-to-noise ratio(SNR,signal-to-noise ratio)estimation combined with density-based clustering with noise is proposed.A sound source localization algorithm based on density-based spatial clustering of applications with noise(DBSCAN,density-based spatial clustering of applications with noise).Based on the SNR estimation algorithm,the optimal sub-band is selected,and the signal frequency band with the largest signal-to-noise ratio is reserved,thereby eliminating the interference of the noise sub-band and improving the robustness of sound source localization in reverberation and noise environments.To avoid the interference of residual noise,the DBSCAN algorithm is introduced to eliminate the influence of noise points.Simulation experiments show that the algorithm has good robustness under the conditions of noise and reverberation.Then,in order to solve the problem of localization error caused by the inconsistency of space and time during the relative motion of the mobile robot and the sound source,a sound source azimuth angle compensation method based on Kalman filter is proposed.The motion of the sound source azimuth is introduced into Kalman filter to establish the sound source motion model,and the optimal localization value of the current frame is obtained through the optimal value of the previous frame and the angle compensation value of the current frame.Simulation experiments show that this method can effectively reduce the Error in positioning azimuth.Finally,an experimental platform system for binaural sound source localization based on mobile robots is built,and the proposed algorithm is tested and analyzed.The experiments show that the proposed algorithm can improve the robustness of localization and reduce the process of continuous sound source localization.errors in. |
PDF Full Text Request