Spectron: Target Speaker Extraction using Conditional Transformer with Adversarial Refinement

TL;DR

Spectron is a transformer-based model for target speaker extraction from mixed audio, employing adversarial refinement and novel training objectives to improve speech quality and outperform existing methods.

Contribution

Introduces a transformer-based target speaker extraction model with adversarial refinement and new training objectives for speaker embedding consistency and waveform invertibility.

Findings

Improves speech extraction quality by 3.12 dB over CNN baseline.

Outperforms recent state-of-the-art methods by 4.1 dB on average.

Leverages multi-scale discriminator for perceptual quality enhancement.

Abstract

Recently, attention-based transformers have become a de facto standard in many deep learning applications including natural language processing, computer vision, signal processing, etc.. In this paper, we propose a transformer-based end-to-end model to extract a target speaker's speech from a monaural multi-speaker mixed audio signal. Unlike existing speaker extraction methods, we introduce two additional objectives to impose speaker embedding consistency and waveform encoder invertibility and jointly train both speaker encoder and speech separator to better capture the speaker conditional embedding. Furthermore, we leverage a multi-scale discriminator to refine the perceptual quality of the extracted speech. Our experiments show that the use of a dual path transformer in the separator backbone along with proposed training paradigm improves the CNN baseline by $3.12$ dB points. Finally, we compare our approach with recent state-of-the-arts and show that our model outperforms existing methods by $4.1$ dB points on an average without creating additional data dependency.