Deep Space Separable Distillation for Lightweight Acoustic Scene Classification

TL;DR

This paper introduces a lightweight deep space separable distillation network for acoustic scene classification that reduces computational complexity and improves performance using novel operators and frequency decomposition.

Contribution

The paper proposes a novel deep space separable distillation network with specialized lightweight operators and frequency decomposition for improved acoustic scene classification.

Findings

Achieves 9.8% performance gain over existing methods

Reduces model parameters and computational complexity

Maintains high accuracy with lightweight design

Abstract

Acoustic scene classification (ASC) is highly important in the real world. Recently, deep learning-based methods have been widely employed for acoustic scene classification. However, these methods are currently not lightweight enough as well as their performance is not satisfactory. To solve these problems, we propose a deep space separable distillation network. Firstly, the network performs high-low frequency decomposition on the log-mel spectrogram, significantly reducing computational complexity while maintaining model performance. Secondly, we specially design three lightweight operators for ASC, including Separable Convolution (SC), Orthonormal Separable Convolution (OSC), and Separable Partial Convolution (SPC). These operators exhibit highly efficient feature extraction capabilities in acoustic scene classification tasks. The experimental results demonstrate that the proposed method achieves a performance gain of 9.8% compared to the currently popular deep learning methods, while also having smaller parameter count and computational complexity.