Lungmix: A Mixup-Based Strategy for Generalization in Respiratory Sound Classification

2024-12-29

AI Summary

Weak cross-dataset generalization and significant performance degradation on unseen data plague respiratory sound classification models. To address this, we propose Lungmix, a novel mixed-sample augmentation strategy specifically designed for auscultatory audio. Lungmix introduces the first loudness-aware time-domain waveform mixing mechanism, integrating random masking with loudness-weighted interpolation, and jointly enforces semantic-consistent label mapping to achieve synchronized waveform–label augmentation. Extensive multi-source transfer evaluation across three public benchmarks—ICBHI, SPR, and HF—demonstrates that Lungmix improves cross-dataset classification accuracy by up to 3.55% across four respiratory sound classes, approaching the performance of models trained directly on the target domain. This work pioneers the incorporation of loudness modeling into stethoscope-sound augmentation, establishing a new paradigm for generalizable learning in low-resource medical audio.

Technology Category

• Search and Optimization: Mixed Discrete/Continuous Search
• Machine Learning: Mixture of Experts (MoE)
• Planning, Routing, and Scheduling: Mixed Discrete/Continuous Planning

Application Category

• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Search and Retrieval-Augmented AI: Multilingual and cross-lingual Web search
• User Modeling, Personalization and Recommendation: Explainable and interpretable methods for personalization

Abstract

Respiratory sound classification plays a pivotal role in diagnosing respiratory diseases. While deep learning models have shown success with various respiratory sound datasets, our experiments indicate that models trained on one dataset often fail to generalize effectively to others, mainly due to data collection and annotation emph{inconsistencies}. To address this limitation, we introduce emph{Lungmix}, a novel data augmentation technique inspired by Mixup. Lungmix generates augmented data by blending waveforms using loudness and random masks while interpolating labels based on their semantic meaning, helping the model learn more generalized representations. Comprehensive evaluations across three datasets, namely ICBHI, SPR, and HF, demonstrate that Lungmix significantly enhances model generalization to unseen data. In particular, Lungmix boosts the 4-class classification score by up to 3.55%, achieving performance comparable to models trained directly on the target dataset.

Problem

• Breath Sound Classification
• Generalization Performance
• Unseen Data

Innovation

• Lungmix
• Mixup variant
• Breath sound classification