loading page

Non-Linear Features of β Brain Rhythms Predict Listener-Specific Neural Signature in Naturalistic Music Listening
  • Pankaj Pandey ,
  • Krishna Prasad Miyapuram ,
  • Derek Lomas
Pankaj Pandey
Indian Institute of Technology Gandhinagar

Corresponding Author:[email protected]

Author Profile
Krishna Prasad Miyapuram
Author Profile
Derek Lomas
Author Profile

Abstract

Studying brain waves elicited while listening to naturalistic music is a rapidly growing interdisciplinary research area encompassing experts from cognitive science, signal processing, and machine learning. Previous works have documented several perspectives, including correlating brain responses to stimulus features, inter-subject and time-varying correlation, and song identification. However, these studies do not provide approaches for understanding the information pertinent to inter-individual subjective differences. This study aims to identify the listener-specific neural signature. We use six EEG naturalistic Music datasets collected in three nations (USA, Greece, and India) comprising 161 listeners to demonstrate feature representation of EEG signals to identify the listener. We provide an interpretable pipeline comprising of decomposition into five primarily brain waves and extraction of 21 features to predict listener-specific neural signatures. We employ linear and non-linear features while training Random Forest classifiers and show the best feature which predicts maximum discrimination among listeners. Our research demonstrates neural oscillations in the higher frequency band are effective in identifying subjective differences. Beta waves show the maximum contribution in predicting the differences and reaching to average accuracy of 91.03% across all datasets. Non-Linear feature Hjorth Mobility provides the maximum predictive ability. This result has a significant implication in biometrics for naturalistic music listening in specific and for brain-computer interaction in general.