Preliminary Study Of Binaural Auditory Localization Based On EEG

Binaural hearing refers to the process in which humans use both ears to pick up and process sound stimuli to form auditory perception.In the process of monaural hearing,the auditory center of the contralateral cerebral cortex processes the sound stimuli,mainly forming the perception phenomena such as loudness,timbre,and pitch.In the process of binaural hearing,the auditory centers of the cerebral cortex on both sides are simultaneously analyzing the sound stimuli to form the spatial perception of the sound source,such as spatial localization.The effective estimation of the spatial localization of the sound source plays an important role in human evolution as well as daily life.In this work,the brain reflections of binaural auditory localization are explored,including measuring the brain wave changes caused by different spatial localized sound source and stimulating of binaural auditory localization by algorithm.The main works is as followed:(1)A set of auditory-related brain reflection research platform based on event-related potential was built,and EEGs of seven subjects were measured separately.The research platform consists of computer,BP EEG,professional sound card,professional earphone and Matlab-based control program,whose main functions include synthesis of virtual acoustic signals in different directions,synthesis of induced stimulation sequences,and recording of brainwave signals.In the experiment,the virtual sound technology based on the head-related transfer function was used to synthesize binaural sound stimulation;the spatial distribution of virtual sound included horizontal plane and median plane;virtual sound field included freefield and reflected-field;The mismatched negative wave was extracted from the brain wave signal as a physiological representation of the brain response of binaural auditory localization.The results show that the mismatch negative wave changes with the horizontal direction variations of the sound source in the free and reflected fields,as well as with the vertical direction variations of the sound source in the free field;moreover,the amplitude of the mismatch negative wave for the horizontal plane increases with the deviation angle.(2)Simulation of the process of binaural auditory localization using a convolutional neural network(a typical deep neural network),and a Brain-like algorithm for binaural auditory localization is proposed.The results show that the localization accuracy of the algorithm can reach 84.0% in the 45° angular interval localization experiment,and the localization accuracy can be further improved to 93.1% after the localization confusion calibration.While,in the localization experiment at 15° angular interval,the localization accuracy of the algorithm can reach 93.2%,and the localization accuracy can be further improved to 95.2% after the localization confusion calibration.This work not only helps to further understand the pre-conscious processing mechanism of binaural hearing,but also provides a reference for the design and optimization of binaural localization algorithm.