IR-GAN: Room Impulse Response Generator for Far-field Speech Recognition

TL;DR

This paper introduces IR-GAN, a GAN-based method for generating realistic synthetic room impulse responses to enhance far-field speech recognition accuracy in new environments.

Contribution

IR-GAN is a novel GAN-based approach that synthesizes realistic RIRs by extracting acoustic parameters, improving speech recognition performance in unseen environments.

Findings

IR-GAN-generated RIRs outperform GAS in recognition accuracy

Synthetic RIRs reduce word error rate by up to 14.3%

Combining IR-GAN RIRs with GAS further improves results

Abstract

We present a Generative Adversarial Network (GAN) based room impulse response generator (IR-GAN) for generating realistic synthetic room impulse responses (RIRs). IR-GAN extracts acoustic parameters from captured real-world RIRs and uses these parameters to generate new synthetic RIRs. We use these generated synthetic RIRs to improve far-field automatic speech recognition in new environments that are different from the ones used in training datasets. In particular, we augment the far-field speech training set by convolving our synthesized RIRs with a clean LibriSpeech dataset. We evaluate the quality of our synthetic RIRs on the real-world LibriSpeech test set created using real-world RIRs from the BUT ReverbDB and AIR datasets. Our IR-GAN reports up to an 8.95% lower error rate than Geometric Acoustic Simulator (GAS) in far-field speech recognition benchmarks. We further improve the performance when we combine our synthetic RIRs with synthetic impulse responses generated using GAS. This combination can reduce the word error rate by up to 14.3% in far-field speech recognition benchmarks.