Performance Comparison of CNN and AST Models with Stacked Features for Environmental Sound Classification

2026-02-10

AI Summary

This work addresses the challenge of efficient environmental sound classification in resource-constrained settings or when large-scale pretraining data are unavailable. The authors propose a multi-acoustic feature stacking strategy that integrates Log-Mel spectrograms, MFCCs, GTCCs, Spectral Contrast, Chroma, and Tonnetz features, evaluated systematically with both CNN and Audio Spectrogram Transformer (AST) architectures on the ESC-50 and UrbanSound8K datasets. Experimental results demonstrate that, without large-scale pretraining, the feature-enhanced CNN significantly outperforms AST in both data and computational efficiency. This approach offers a lightweight and effective alternative for deploying environmental sound classification on edge devices, effectively mitigating the heavy reliance of transformer-based models on extensive pretraining.

Technology Category

• Machine Learning: Multi-class/Multi-label Learning & Extreme Classification
• Natural Language Processing: Text Classification & Sentiment Analysis
• Intelligent Robots: Multimodal Perception & Sensor Fusion

Application Category

• Web Mining and Content Analysis: Large pretrained models with web data
• Systems and Infrastructure for Web, Mobile and WoT: Applied ML and AI for Web-based mobile applications
• Graph Algorithms and Modeling for the Web: Graph neural networks and deep learning approaches for Web-related graphs

Abstract

Environmental sound classification (ESC) has gained significant attention due to its diverse applications in smart city monitoring, fault detection, acoustic surveillance, and manufacturing quality control. To enhance CNN performance, feature stacking techniques have been explored to aggregate complementary acoustic descriptors into richer input representations. In this paper, we investigate CNN-based models employing various stacked feature combinations, including Log-Mel Spectrogram (LM), Spectral Contrast (SPC), Chroma (CH), Tonnetz (TZ), Mel-Frequency Cepstral Coefficients (MFCCs), and Gammatone Cepstral Coefficients (GTCC). Experiments are conducted on the widely used ESC-50 and UrbanSound8K datasets under different training regimes, including pretraining on ESC-50, fine-tuning on UrbanSound8K, and comparison with Audio Spectrogram Transformer (AST) models pretrained on large-scale corpora such as AudioSet. This experimental design enables an analysis of how feature-stacked CNNs compare with transformer-based models under varying levels of training data and pretraining diversity. The results indicate that feature-stacked CNNs offer a more computationally and data-efficient alternative when large-scale pretraining or extensive training data are unavailable, making them particularly well suited for resource-constrained and edge-level sound classification scenarios.

Problem

• Environmental Sound Classification
• Feature Stacking
• CNN
• Audio Spectrogram Transformer
• Resource-Constrained Scenarios

Innovation

• feature stacking
• environmental sound classification
• CNN
• Audio Spectrogram Transformer
• data efficiency