Sound Event Detection with Adaptive Frequency Selection

TL;DR

HIDACT is a novel neural network architecture that adaptively processes frequency bands for sound event detection, reducing computation while maintaining performance and adjusting to data and resource constraints.

Contribution

The paper introduces HIDACT, a new adaptive network architecture that selectively processes frequency bands for efficient sound event detection, outperforming traditional models in computational efficiency.

Findings

HIDACT achieves comparable accuracy to larger models with fewer parameters.

The model can dynamically adjust its computation based on input complexity.

Experimental results demonstrate reduced computational cost without sacrificing performance.

Abstract

In this work, we present HIDACT, a novel network architecture for adaptive computation for efficiently recognizing acoustic events. We evaluate the model on a sound event detection task where we train it to adaptively process frequency bands. The model learns to adapt to the input without requesting all frequency sub-bands provided. It can make confident predictions within fewer processing steps, hence reducing the amount of computation. Experimental results show that HIDACT has comparable performance to baseline models with more parameters and higher computational complexity. Furthermore, the model can adjust the amount of computation based on the data and computational budget.