AUTHOR=Xie Liping , Lu Chihua , Liu Zhien , Yan Lirong , Xu Tao 

TITLE=Study of Auditory Brain Cognition Laws-Based Recognition Method of Automobile Sound Quality

JOURNAL=Frontiers in Human Neuroscience

VOLUME=Volume 15 - 2021

YEAR=2021

URL=https://www.frontiersin.org/journals/human-neuroscience/articles/10.3389/fnhum.2021.663049

DOI=10.3389/fnhum.2021.663049

ISSN=1662-5161

ABSTRACT=Those methods that are applied to evaluate car sound quality mainly relies on people's subjective psychological feelings are not objective enough and cannot guarantee the universality of evaluation results. There are studies shown that electroencephalography (EEG) physiological signals can objectively reflect people’s subjective feelings, such as emotions and fatigue. With the purpose of exploring the laws of brain cognition under the stimulation of car sound samples, this paper introduced EEG signal as an evaluation method of car sound quality, and proposed an evaluation method that can effectively identify diversified car sound quality. The brain signal were measured using EEG when subjects hear the cars sound with the sound quality of comfort, powerfulness, acceleration respectively. We summarized the brain cognitive rules by analyzing the EEG power topographic map under the stimulation of three types of car sound samples. This study classified EEG features (namely the differential asymmetry (DASM) and rational asymmetry features (RASM)) of subjects based on machine learning classification algorithm, which realized the recognition of three types of car sound qualities; In addition, this research also used kalman smoothing and MRMR dimensionality reduction algorithm to improve the recognition accuracy of the model. The results proved that the neural characteristics of the three types of car sound samples do exist under the stimulation of different quality sound samples, which is specifically reflected there is a positive correlation between EEG energy and sound intensity , frequency band characteristics can better realize the recognition of car sound patterns; The DASM_DE of γ band is used as the input, and the accuracy of three types of car sound samples identified by SVM is the highest, which is 86.26%; Using kalman smoothing and MRMR algorithm can not only improve the recognition accuracy of the model, but also reduce the amount of model calculation. In this study, EEG signals are used as the evaluation index of car sound quality, which avoids the description of language and improves the accuracy of recognition. It provided new idea and method to explore the cognitive rules of car sound quality from the field of brain-computer interface technology.