Recognition Method Of The Time Varying Azimuth Of Single Sound Source Based On Binaural Hearing Mechanism

Sound source localization is a comprehensive research field involving psychoacoustics,auditory physiology,artificial intelligence and many other disciplines.It is widely used in the fields of robotic hearing,military,intelligent driving,medical equipment,etc.At present,the methods of sound source localization are mainly Microphone array method.Humans and mammals can judge the orientation of the sound source based on the information obtained from binaural.This is because the internal physiological structure can effectively extract the information needed for sound source localization.Therefore,this paper proposes a mobile sound source localization method based on the binaural hearing mechanism.The main contents are as follows:(1)A sound source localization experimental system based on binaural hearing mechanism was designed and fabricated.The system is mainly composed of artificial head,acoustic sensor,data acquisition and stepping motor,which can perform synchronous real-time acquisition of binaural sound signals..(2)Studing the measurement principle,measurement flow and calculation method of the Head-Related Transfer Function(HRTF).The sound source localization method based on the relative head correlation transfer function is proposed and the relationship between the HRTF and the distance or azimuth is obtained.According to the sound source localization mechanism and the bandpass filtering effect of the human ear cochlear basement membrane,a method for extracting the localization time of the Interaural Time Difference(ITD)and Interaural Intensity Difference(ILD)is proposed.The sensitive frequency band of the positioning cue is selected.(3)The recognition algorithm of the instantaneous orientation of the sound source is proposed,which is combined with the positioning cues to judge the incident orientation of the acoustic signal.For voice signals,it needs to be pre-processed before positioning to reduce the impact of noise on positioning performance.In the positioning,in order to eliminate the front and rear azimuth confusion problem of the ITD and improve the positioning accuracy,the ITD and the ILD of the sub-band are combined as the combined characteristics of the sound source localization.Combining the support vector machine algorithm to identify the instantaneous position of the sound source.(4)On the basis of the instantaneous position recognition of the sound source,the orientation of the intermittent moving sound source and the continuous moving sound source are judged.For the continuous moving sound source,which is greatly affected by background noise,an acoustic signal extraction method based on short-time fourier transform is proposed to analyze the time period in which the acoustic signal and steady-state noise appear,segmentally extract the acoustic signal and use high-pass filtering to eliminate the stability.Influence of state noise to improve the accuracy of sound source localization.(5)The acoustic signal extraction method based on auditory neural oscillation network is studied,and the separation of acoustic signal and noise in continuous moving sound source is realized.Without the influence of background noise,judging the moving direction of the pure acoustic signal.This method further improves the accuracy of sound source recognition.