Learning Filterbanks from Raw Speech for Phone Recognition
This work addresses speech recognition challenges by introducing a learnable front-end, though it is incremental as it builds on existing filterbank methods.
The paper tackled the problem of phone recognition by learning time-domain filterbanks from raw speech, which consistently outperformed traditional mel-filterbanks on TIMIT, achieving improved performance across several architectures.
We train a bank of complex filters that operates on the raw waveform and is fed into a convolutional neural network for end-to-end phone recognition. These time-domain filterbanks (TD-filterbanks) are initialized as an approximation of mel-filterbanks, and then fine-tuned jointly with the remaining convolutional architecture. We perform phone recognition experiments on TIMIT and show that for several architectures, models trained on TD-filterbanks consistently outperform their counterparts trained on comparable mel-filterbanks. We get our best performance by learning all front-end steps, from pre-emphasis up to averaging. Finally, we observe that the filters at convergence have an asymmetric impulse response, and that some of them remain almost analytic.