In order to run a listening experiment given some processing on the speech signals, follow the directions in the README section of the script ./human_perception_experiments/listening_experiments/listening_experiment.m.
To generate an experiment, follow the guides in the README section of /human_perception_experiments/generate_experiment.m. Similarly, to play an experiment, follow the README in /human_perception_experiments/play_experiment.m
Currently we have functions to extract:
- modified group delay function [1]
We ran into memory constraints when trying to train the baseline model in MATLAB. Since the model is a simple GMM, we're implementing it in Python so we can run it on a bigger machine. The features are currently being computed via the baseline script in MATLAB.
[1] Hedge, R. M., Murthy, H. A. "Significance of the Modified Group Delay Feature"