Target conversation extraction: Source separation using turn-taking dynamics

TL;DR

This paper introduces a novel method for extracting a target conversation from noisy audio by leveraging turn-taking dynamics and speaker embeddings, demonstrating significant improvements in signal clarity across multilingual datasets.

Contribution

The paper proposes a new task of target conversation extraction using turn-taking patterns and neural networks, advancing speaker separation techniques.

Findings

8.19 dB SNR improvement in two-speaker conversations

7.92 dB SNR improvement in multi-speaker conversations

Feasibility demonstrated on English and Mandarin datasets

Abstract

Extracting the speech of participants in a conversation amidst interfering speakers and noise presents a challenging problem. In this paper, we introduce the novel task of target conversation extraction, where the goal is to extract the audio of a target conversation based on the speaker embedding of one of its participants. To accomplish this, we propose leveraging temporal patterns inherent in human conversations, particularly turn-taking dynamics, which uniquely characterize speakers engaged in conversation and distinguish them from interfering speakers and noise. Using neural networks, we show the feasibility of our approach on English and Mandarin conversation datasets. In the presence of interfering speakers, our results show an 8.19 dB improvement in signal-to-noise ratio for 2-speaker conversations and a 7.92 dB improvement for 2-4-speaker conversations. Code, dataset available at https://github.com/chentuochao/Target-Conversation-Extraction.