This is the paper they submitted: http://web.cse.ohio-state.edu/~dwang/papers/CWYWH.jasa16.pdf http://web.cse.ohio-state.edu/~dwang/papers/CWYWH.jasa16.pdf I'm

This is the paper they submitted: http://web.cse.ohio-state.edu/~dwang/papers/CWYWH.jasa16.pdf http://web.cse.ohio-state.edu/~dwang/papers/CWYWH.jasa16.pdf I'm not an expert in the hearing field, but makes sense for me: they probably know that some filters work (and are actually used) for hearing, and feeding that preprocessed data would save them training and hyperparameter tuning time.