2.4 EEG Time-Frequency analysis
EEG epochs underwent spectral decomposition using Morlet wavelets to extract power data for frequencies from 4 to 30 Hz (with approximately 1 Hz intervals). Three cycles were set for the lowest frequency and cycles increased linearly with frequency scaled at 0.5 using EEGLAB newtimef() function (‘cycles’, [3 0.5]; Delorme and Makeig, 2004). Temporal resolution was around 36-37 ms per time frame. Baseline correction within each 1 Hz frequency interval was performed for each single trial by subtracting the average power between -600 and -100 ms pre-stimulus onset from each time point post-stimulus onset to calculate event-related spectral perturbation (ERSP) using the gain model (Grandchamp and Delorme, 2011). The power data were then logarithmically converted to decibel (dB) for further statistical analysis. Increase in power relative to baseline represents event-related synchronization (ERS), while lower power relative to baseline represents event-related desynchronization (ERD).