RTFS-Net: Recurrent Time-Frequency Modelling for Efficient Audio-Visual Speech Separation

TL;DR

RTFS-Net introduces a novel time-frequency domain approach for audio-visual speech separation, leveraging RNNs and attention-based fusion to outperform existing methods in speed and quality with fewer parameters.

Contribution

The paper presents RTFS-Net, a new time-frequency domain model that uses RNNs and attention mechanisms for efficient and superior audio-visual speech separation.

Findings

Outperforms prior SOTA in separation quality

Reduces model parameters by 90%

Decreases computational complexity by 83%

Abstract

Audio-visual speech separation methods aim to integrate different modalities to generate high-quality separated speech, thereby enhancing the performance of downstream tasks such as speech recognition. Most existing state-of-the-art (SOTA) models operate in the time domain. However, their overly simplistic approach to modeling acoustic features often necessitates larger and more computationally intensive models in order to achieve SOTA performance. In this paper, we present a novel time-frequency domain audio-visual speech separation method: Recurrent Time-Frequency Separation Network (RTFS-Net), which applies its algorithms on the complex time-frequency bins yielded by the Short-Time Fourier Transform. We model and capture the time and frequency dimensions of the audio independently using a multi-layered RNN along each dimension. Furthermore, we introduce a unique attention-based fusion technique for the efficient integration of audio and visual information, and a new mask separation approach that takes advantage of the intrinsic spectral nature of the acoustic features for a clearer separation. RTFS-Net outperforms the prior SOTA method in both inference speed and separation quality while reducing the number of parameters by 90% and MACs by 83%. This is the first time-frequency domain audio-visual speech separation method to outperform all contemporary time-domain counterparts.