TRF models predict EEG from speech representations via cross-validated regression

Evidence

For each representation of the speech, the result is a set of regression weights referred to as multivariate temporal response functions (TRFs). k -fold cross-validation (k = 10) was employed to build predictions of the EEG signal from the TRFs of each distinct speech representation model. The prediction accuracies were obtained as an average of the correlation values over a set of 12 best predictive electrodes (6 on the left side of the scalp and the symmetrical counterparts on the right), on a time-lag window from 0 to 250 ms (Lalor et al. 2006 ; Di Liberto et al. 2015 ). — Dijk, Pim van, Physiology, Psychoacoustics and…, 4386

Categorization

• XREF: Temporal response function approach is the standard method from Di Liberto’s speech-tracking work.