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

TL;DR

Lungmix is a novel data augmentation method inspired by Mixup that improves the generalization of respiratory sound classification models across different datasets by blending waveforms and interpolating labels.

Contribution

We propose Lungmix, a new Mixup-based augmentation technique that enhances model robustness and generalization in respiratory sound classification tasks.

Findings

Lungmix significantly improves cross-dataset generalization.

Boosts 4-class classification score by up to 3.55%.

Achieves performance comparable to models trained on target datasets.

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.