TIMIT module

The timit module is used for calculating phoneme-related responses, phonetic selectivity, and creating data structures used by the pitch temporal receptive field analysis.

intonatang.timit.get_average_response_to_phonemes(out, phoneme_order=['d', 'b', 'g', 'p', 'k', 't', 'jh', 'sh', 'z', 's', 'f', 'th', 'dh', 'v', 'w', 'r', 'l', 'ae', 'aa', 'ay', 'aw', 'ow', 'ax', 'uw', 'eh', 'ey', 'ih', 'ux', 'iy', 'n', 'm', 'ng'])

Returns the average response over all instances of each phoneme in TIMIT

intonatang.timit.save_timit_pitch()

This function saves a pandas dataframe of pitch information for TIMIT sentences.

This script takes the pitch information (fundamental frequency in Hz in 10ms bins) written in *.wav.txt files (which are output from Praat and manually examined for doubling or halving errors), and saves timit_pitch, a dataframe that contains absolute and relative pitch values.

Relative pitch values are computed as a z-score across each sentence’s absolute pitch values. Absolute pitch values are saved with two scalings, one is log Hz and the other is erb-rate. These monotonic, non-linear transformations are based on psychophysical data showing that differences in Hz are perceived differently wtih frequency and lead to pitch values that are more linear with respect to pitch perception.

intonatang.timit.save_timit_pitch_phonetic()

This function combines timit_pitch with timit_phonemes and needs to be run after save_timit_pitch and save_timit_phonemes