Unified Microphone Conversion: Many-to-Many Device Mapping via Feature-wise Linear Modulation

TL;DR

This paper introduces a unified framework for microphone conversion that uses feature-wise linear modulation to enable many-to-many device mapping, improving sound event classification robustness across different microphones.

Contribution

The proposed method extends previous device conversion techniques by enabling scalable many-to-many mappings through conditioning on frequency response data with FLM.

Findings

Outperforms state-of-the-art by 2.6% in macro F1 score

Reduces variability by 0.8% in macro F1 score

Effective in real-world sound event classification scenarios

Abstract

We present Unified Microphone Conversion, a unified generative framework designed to bolster sound event classification (SEC) systems against device variability. While our prior CycleGAN-based methods effectively simulate device characteristics, they require separate models for each device pair, limiting scalability. Our approach overcomes this constraint by conditioning the generator on frequency response data, enabling many-to-many device mappings through unpaired training. We integrate frequency-response information via Feature-wise Linear Modulation, further enhancing scalability. Additionally, incorporating synthetic frequency response differences improves the applicability of our framework for real-world application. Experimental results show that our method outperforms the state-of-the-art by 2.6% and reduces variability by 0.8% in macro-average F1 score.