jdirani / speech-onset-time

Toolkit to detect the speech onset time from audio files.

Speech-Onset Detection Toolkit

This toolkit was developed in order to easily detect the onset of speech in voice recordings for psychology experiments.

Automatically detect utterance time:

auto_utterance_times(): To return the onset time of a single .wav file
auto_utterance_times_batch: Takes list of paths (.wav) as input and returns speech onset time for all audio files. Option to plot the results
plot_utterance_time(): Plot results to monitor accuracy.

Semi-automatic detection of utterance time:

semi_auto_utterance_times(): Automatically detects utterance times and plots them, allowing to manually edit the predictions if needed.
To edit prediction, double click on plot to move the vertical line. Press enter on terminal to go to the next plot. Figures are saved to file.

Example

fs, signal = wav.read('path_in.wav')
idx, rt = auto_utterance_times(signal, fs)
plot_utterance_time(signal, fs, rt)