Interpretable Convolutional Filters with SincNet

TL;DR

SincNet is an interpretable CNN model for speech processing that uses parametrized sinc functions to learn meaningful band-pass filters, leading to faster convergence, better performance, and enhanced interpretability.

Contribution

The paper introduces SincNet, a novel CNN architecture that employs sinc functions for filter parametrization, improving interpretability and efficiency in speech recognition tasks.

Findings

SincNet converges faster than standard CNNs.

SincNet outperforms standard CNNs in speech recognition accuracy.

SincNet's filters are more interpretable due to physical meaning of parameters.

Abstract

Deep learning is currently playing a crucial role toward higher levels of artificial intelligence. This paradigm allows neural networks to learn complex and abstract representations, that are progressively obtained by combining simpler ones. Nevertheless, the internal "black-box" representations automatically discovered by current neural architectures often suffer from a lack of interpretability, making of primary interest the study of explainable machine learning techniques. This paper summarizes our recent efforts to develop a more interpretable neural model for directly processing speech from the raw waveform. In particular, we propose SincNet, a novel Convolutional Neural Network (CNN) that encourages the first layer to discover more meaningful filters by exploiting parametrized sinc functions. In contrast to standard CNNs, which learn all the elements of each filter, only low and high cutoff frequencies of band-pass filters are directly learned from data. This inductive bias offers a very compact way to derive a customized filter-bank front-end, that only depends on some parameters with a clear physical meaning. Our experiments, conducted on both speaker and speech recognition, show that the proposed architecture converges faster, performs better, and is more interpretable than standard CNNs.