Personalized PercepNet: Real-time, Low-complexity Target Voice Separation and Enhancement

TL;DR

Personalized PercepNet is a real-time, low-complexity speech enhancement model that effectively isolates a target speaker from multi-talker environments using speaker embeddings, outperforming existing methods.

Contribution

It introduces a novel speaker-dependent enhancement approach by integrating a perceptually motivated speaker embedder into PercepNet, enabling targeted speech separation in real-time.

Findings

Significantly outperforms PercepNet and baselines in objective metrics.

Achieves higher human opinion scores for speech quality.

Operates in real-time with low complexity.

Abstract

The presence of multiple talkers in the surrounding environment poses a difficult challenge for real-time speech communication systems considering the constraints on network size and complexity. In this paper, we present Personalized PercepNet, a real-time speech enhancement model that separates a target speaker from a noisy multi-talker mixture without compromising on complexity of the recently proposed PercepNet. To enable speaker-dependent speech enhancement, we first show how we can train a perceptually motivated speaker embedder network to produce a representative embedding vector for the given speaker. Personalized PercepNet uses the target speaker embedding as additional information to pick out and enhance only the target speaker while suppressing all other competing sounds. Our experiments show that the proposed model significantly outperforms PercepNet and other baselines, both in terms of objective speech enhancement metrics and human opinion scores.