|
Dear list, We have developed several models allowing to predict speech intelligibility in noise and reverberation. Most of them are using as inputs the signals at the listener’s ears (some use binaural room impulse responses) to
predict differences in speech reception threshold between different conditions. There are different model versions that predict different effects: binaural hearing (spatial release from masking), dip listening, the effect of hearing loss (taking the listener’s
audiogram as additional input), multiple effects of reverberation, and harmonic cancellation.
You can find them here: https://mathieulavandier.wordpress.com/home/models/ With each model, there is an example of the predictions corresponding to one dataset from the literature. You have access to the signals used in the speech intelligibility test, along with the data resulting from this
test (listeners’ SRTs), and the code used to run the model on this dataset; so that you can check how to handle a model before using it to predict intelligibility in other conditions. Do not hesitate to play around with these models and give us feedback! We hope they can be useful. All the best, mat ----------------------------------------------------------------------------------------------- Mathieu Lavandier https://mathieulavandier.wordpress.com/ https://celya.universite-lyon.fr ENTPE, Laboratoire de Tribologie et Dynamique des Systèmes (LTDS, UMR 5513) Rue M. Audin, 69518 Vaulx-en-Velin Cedex, FRANCE |