Attractor-Based Speech Separation of Multiple Utterances by Unknown Number of Speakers

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Wang, Yuzhu, Politis, Archontis, Drossos, Konstantinos, Virtanen, Tuomas
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866916751792406528
author Wang, Yuzhu
Politis, Archontis
Drossos, Konstantinos
Virtanen, Tuomas
author_facet Wang, Yuzhu
Politis, Archontis
Drossos, Konstantinos
Virtanen, Tuomas
contents This paper addresses the problem of single-channel speech separation, where the number of speakers is unknown, and each speaker may speak multiple utterances. We propose a speech separation model that simultaneously performs separation, dynamically estimates the number of speakers, and detects individual speaker activities by integrating an attractor module. The proposed system outperforms existing methods by introducing an attractor-based architecture that effectively combines local and global temporal modeling for multi-utterance scenarios. To evaluate the method in reverberant and noisy conditions, a multi-speaker multi-utterance dataset was synthesized by combining Librispeech speech signals with WHAM! noise signals. The results demonstrate that the proposed system accurately estimates the number of sources. The system effectively detects source activities and separates the corresponding utterances into correct outputs in both known and unknown source count scenarios.
format Preprint
id arxiv_https___arxiv_org_abs_2505_16607
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Attractor-Based Speech Separation of Multiple Utterances by Unknown Number of Speakers
Wang, Yuzhu
Politis, Archontis
Drossos, Konstantinos
Virtanen, Tuomas
Audio and Speech Processing
Sound
This paper addresses the problem of single-channel speech separation, where the number of speakers is unknown, and each speaker may speak multiple utterances. We propose a speech separation model that simultaneously performs separation, dynamically estimates the number of speakers, and detects individual speaker activities by integrating an attractor module. The proposed system outperforms existing methods by introducing an attractor-based architecture that effectively combines local and global temporal modeling for multi-utterance scenarios. To evaluate the method in reverberant and noisy conditions, a multi-speaker multi-utterance dataset was synthesized by combining Librispeech speech signals with WHAM! noise signals. The results demonstrate that the proposed system accurately estimates the number of sources. The system effectively detects source activities and separates the corresponding utterances into correct outputs in both known and unknown source count scenarios.
title Attractor-Based Speech Separation of Multiple Utterances by Unknown Number of Speakers
topic Audio and Speech Processing
Sound
url https://arxiv.org/abs/2505.16607