Convolutional Recurrent Neural Networks for Polyphonic Sound Event Detection

TLDR

Sound events in unstructured environments vary widely in frequency and temporal structure, and recent work shows that CNNs and RNNs can extract invariant features and capture long‑term context, outperforming traditional methods. We combine CNN and RNN into a CRNN for polyphonic sound event detection. The CRNN integrates convolutional layers for feature extraction with recurrent layers for temporal modeling and is evaluated on polyphonic sound event datasets. Compared to standalone CNN, RNN, and other established methods, the CRNN achieves substantial performance gains across four everyday sound event datasets.

Abstract

Sound events often occur in unstructured environments where they exhibit wide variations in their frequency content and temporal structure. Convolutional neural networks (CNN) are able to extract higher level features that are invariant to local spectral and temporal variations. Recurrent neural networks (RNNs) are powerful in learning the longer term temporal context in the audio signals. CNNs and RNNs as classifiers have recently shown improved performances over established methods in various sound recognition tasks. We combine these two approaches in a Convolutional Recurrent Neural Network (CRNN) and apply it on a polyphonic sound event detection task. We compare the performance of the proposed CRNN method with CNN, RNN, and other established methods, and observe a considerable improvement for four different datasets consisting of everyday sound events.