Cough Classification using Few-Shot Learning

2025-09-11

AI Summary

Respiratory disease diagnosis using cough audio is hindered by severe scarcity of labeled data. Method: We propose a few-shot learning framework for multi-class respiratory sound classification—specifically, distinguishing COVID-19, influenza, and healthy states—using Mel-spectrograms as time-frequency representations and Prototypical Networks to construct a lightweight, sample-efficient model. Results: With only 15 support samples per class, the multi-class model achieves 74.87% accuracy; all pairwise binary classifiers exceed 70% accuracy. Paired t-tests and Wilcoxon signed-rank tests confirm no statistically significant performance difference between the multi-class and corresponding binary models (p > 0.05). This work constitutes the first systematic validation of few-shot learning for multi-class acoustic diagnosis of respiratory diseases, establishing a novel paradigm for medical AI modeling under data-constrained conditions.

Technology Category

• Machine Learning: Multi-class/Multi-label Learning & Extreme Classification
• Computer Vision: Multi-modal Vision
• Knowledge Representation and Reasoning: Diagnosis and Abductive Reasoning

Application Category

• Web Mining and Content Analysis: Mining multimedia, multimodal, multilingual, cross-lingual Web data
• Search and Retrieval-Augmented AI: Multilingual and cross-lingual Web search
• Responsible Web: Machine-in-the-loop, human agency and autonomy

Abstract

This paper investigates the effectiveness of few-shot learning for respiratory sound classification, focusing on coughbased detection of COVID-19, Flu, and healthy conditions. We leverage Prototypical Networks with spectrogram representations of cough sounds to address the challenge of limited labeled data. Our study evaluates whether few-shot learning can enable models to achieve performance comparable to traditional deep learning approaches while using significantly fewer training samples. Additionally, we compare multi-class and binary classification models to assess whether multi-class models can perform comparably to their binary counterparts. Experimental findings show that few-shot learning models can achieve competitive accuracy. Our model attains 74.87% accuracy in multi-class classification with only 15 support examples per class, while binary classification achieves over 70% accuracy across all class pairs. Class-wise analysis reveals Flu as the most distinguishable class, and Healthy as the most challenging. Statistical tests (paired t-test p = 0.149, Wilcoxon p = 0.125) indicate no significant performance difference between binary and multiclass models, supporting the viability of multi-class classification in this setting. These results highlight the feasibility of applying few-shot learning in medical diagnostics, particularly when large labeled datasets are unavailable.

Problem

• Classifying respiratory sounds using few-shot learning
• Evaluating few-shot versus traditional deep learning approaches
• Assessing multi-class versus binary classification for cough detection

Innovation

• Few-shot learning with Prototypical Networks
• Spectrogram representations of cough sounds
• Multi-class classification with limited samples