Extended Data Fig. 2: Acoustic phonetic feature encoding model.
From: Dissecting neural computations in the human auditory pathway using deep neural networks for speech
Example of feature extraction for a sample sentence, read by a male speaker:’It is well liked by the children and faculty.’ From top to bottom: 1) raw waveform; 2) high-gamma (z-scored) activity at an example electrode; 3) Mel-scaled spectrogram; 4) intensity of voicing; 5) sentence onset; 6) time course of peak rate; 7) absolute pitch (binned into 10 bins); 8) relative pitch (binned into 10 bins); 9) pitch change (binned into 10 bins); 10) phonetic features.
