Analysis of human auditory system on sound source localization using auditory-evoked potentials

Abstract

In order to improve the human auditory perception, many audio processing algorithms have been implemented on audio devices such as sound bars and high-fidelity headsets. Accordingly, it is necessary to explore the effect of such algorithms on actual human perception. In this paper, we analyze the properties of the human auditory system using two electroencephalography (EEG) experiments. First, the magnitude spectra of the EEG signals in several brain cortexes were investigated depending on the sound rendering method. Then, we observed the effects of amplitude panning applied to virtual sound source localization in the human brain using auditory evoked potentials. In particular, the panned audio sources were generated using the stereophonic law of sines, which is a typical amplitude panning technique for stereo loudspeaker listening environments. Depending on the azimuth direction, the event-related potentials (ERPs) between the actually located sound and the virtually panned ones were compared. It is implied from the experiments that the subjects can distinguish actually located sound sources, depending on the azimuth direction, while it is difficult to distinguish between virtually panned sound sources. Finally, it was also demonstrated that the performance of sound source localizations could be measured using EEG signals.