Effect of musical training on EEG measures of spectro-temporal processing

Abstract

Recent studies show that musical training enhances auditory processing abilities such as sensitivity to temporal fine structure and narrowband frequency resolution. Little is known about the effect of musical training on broadband spectro-temporal processing using rippled noise. This study evaluated whether musician enhancement in frequency resolution can be generalized to broadband spectro-temporal resolution using electrophysiological measures. We tested the hypothesis that musicians have enhanced broadband spectral resolution abilities compared to non-musicians. Spectral processing was measured using P1 N1 P2 responses to spectro-temporally modulated rippled noise. The test stimulus was spectro-temporally modulation wideband (4 octaves) rippled noise from 350 to 5600 Hz. EEG recordings were made using a 32-channel standard 10/20 configuration BioSemi system with a sampling rate of 4 kHz. For each condition, 1000 repetitions of the test stimuli were presented at 100% modulation with 2 Hz temporal modulation via electrically shielded ER-3A earphones, using a passive listening paradigm. The amplitude and latency of the P1-N1-P2 components were measured to determine whether musicians had better sensitivity to spectro-temporal modulation compared to age-matched non-musicians. Results to date show slight musician enhancement in spectral processing. The clinical implications of these findings for auditory-rehabilitation will be discussed.

Spectro-Temporal Modeling for Human Activity Recognition Using a Radar Sensor Network

“…Although both frequency-based modelling and temporal modelling have been investigated for radar-based human activity recognition, it is still waiting for further research to perform spectro-temporal modelling that integrates frequency features on spectrograms and temporal features within sequential signals. Spectro-temporal modelling has been explored in audio analysis [36] and speech recognition [37]. Most current spectro-temporal models [38]- [40] are the combinations of CNNs and RNNs.…”