These are scripts related to a speech-in-noise intelligibility experiment conducted at the University of Washington.
The stimuli are recordings of (a subset of) the
IEEE sentences. There are 30
talkers, 180 sentences (out of the original 720), and 6 different presentation
SNRs; each of these parameters is stored in params.yaml
. The noise is white
noise filtered to match the long-term average spectrum of the stimuli. The
particular combination of talker, sentence, and SNR on any given trial are
determined by the script counterbalance-trials.py
, which writes the
combinations to design-matrix.csv
.