Improving Acoustic Scene Classification in Low-Resource Conditions

TL;DR

This paper introduces DS-FlexiNet, a novel efficient model for acoustic scene classification that combines advanced convolutional techniques, model compression, data augmentation, and knowledge distillation to improve performance in low-resource and heterogeneous device environments.

Contribution

The paper presents DS-FlexiNet, a new model architecture that integrates depthwise separable convolutions, residual connections, and domain-specific normalization for low-resource acoustic scene classification.

Findings

DS-FlexiNet outperforms existing models in low-resource settings.

Quantization Aware Training reduces model size with minimal accuracy loss.

Knowledge Distillation improves cross-device generalization.

Abstract

Acoustic Scene Classification (ASC) identifies an environment based on an audio signal. This paper explores ASC in low-resource conditions and proposes a novel model, DS-FlexiNet, which combines depthwise separable convolutions from MobileNetV2 with ResNet-inspired residual connections for a balance of efficiency and accuracy. To address hardware limitations and device heterogeneity, DS-FlexiNet employs Quantization Aware Training (QAT) for model compression and data augmentation methods like Auto Device Impulse Response (ADIR) and Freq-MixStyle (FMS) to improve cross-device generalization. Knowledge Distillation (KD) from twelve teacher models further enhances performance on unseen devices. The architecture includes a custom Residual Normalization layer to handle domain differences across devices, and depthwise separable convolutions reduce computational overhead without sacrificing feature representation. Experimental results show that DS-FlexiNet excels in both adaptability and performance under resource-constrained conditions.