Adversarial Fine-tuning using Generated Respiratory Sound to Address Class Imbalance
This addresses data scarcity and class imbalance in medical audio analysis, particularly for respiratory sounds, with incremental improvements over existing methods.
The paper tackles class imbalance in respiratory sound classification by using an audio diffusion model to generate synthetic sounds and adversarial fine-tuning to align features, improving the ICBHI Score by 2.24% and minority class accuracy by up to 26.58%.
Deep generative models have emerged as a promising approach in the medical image domain to address data scarcity. However, their use for sequential data like respiratory sounds is less explored. In this work, we propose a straightforward approach to augment imbalanced respiratory sound data using an audio diffusion model as a conditional neural vocoder. We also demonstrate a simple yet effective adversarial fine-tuning method to align features between the synthetic and real respiratory sound samples to improve respiratory sound classification performance. Our experimental results on the ICBHI dataset demonstrate that the proposed adversarial fine-tuning is effective, while only using the conventional augmentation method shows performance degradation. Moreover, our method outperforms the baseline by 2.24% on the ICBHI Score and improves the accuracy of the minority classes up to 26.58%. For the supplementary material, we provide the code at https://github.com/kaen2891/adversarial_fine-tuning_using_generated_respiratory_sound.