Pattern analysis of EEG responses to speech and voice: Influence of feature grouping