AV-CrossNet: an Audiovisual Complex Spectral Mapping Network for Speech Separation By Leveraging Narrow- and Cross-Band Modeling

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Kalkhorani, Vahid Ahmadi, Yu, Cheng, Kumar, Anurag, Tan, Ke, Xu, Buye, Wang, DeLiang
Format: Preprint
Published: 2024
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866914836764426240
author Kalkhorani, Vahid Ahmadi
Yu, Cheng
Kumar, Anurag
Tan, Ke
Xu, Buye
Wang, DeLiang
author_facet Kalkhorani, Vahid Ahmadi
Yu, Cheng
Kumar, Anurag
Tan, Ke
Xu, Buye
Wang, DeLiang
contents Adding visual cues to audio-based speech separation can improve separation performance. This paper introduces AV-CrossNet, an audiovisual (AV) system for speech enhancement, target speaker extraction, and multi-talker speaker separation. AV-CrossNet is extended from the CrossNet architecture, which is a recently proposed network that performs complex spectral mapping for speech separation by leveraging global attention and positional encoding. To effectively utilize visual cues, the proposed system incorporates pre-extracted visual embeddings and employs a visual encoder comprising temporal convolutional layers. Audio and visual features are fused in an early fusion layer before feeding to AV-CrossNet blocks. We evaluate AV-CrossNet on multiple datasets, including LRS, VoxCeleb, and COG-MHEAR challenge. Evaluation results demonstrate that AV-CrossNet advances the state-of-the-art performance in all audiovisual tasks, even on untrained and mismatched datasets.
format Preprint
id arxiv_https___arxiv_org_abs_2406_11619
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle AV-CrossNet: an Audiovisual Complex Spectral Mapping Network for Speech Separation By Leveraging Narrow- and Cross-Band Modeling
Kalkhorani, Vahid Ahmadi
Yu, Cheng
Kumar, Anurag
Tan, Ke
Xu, Buye
Wang, DeLiang
Audio and Speech Processing
Machine Learning
Adding visual cues to audio-based speech separation can improve separation performance. This paper introduces AV-CrossNet, an audiovisual (AV) system for speech enhancement, target speaker extraction, and multi-talker speaker separation. AV-CrossNet is extended from the CrossNet architecture, which is a recently proposed network that performs complex spectral mapping for speech separation by leveraging global attention and positional encoding. To effectively utilize visual cues, the proposed system incorporates pre-extracted visual embeddings and employs a visual encoder comprising temporal convolutional layers. Audio and visual features are fused in an early fusion layer before feeding to AV-CrossNet blocks. We evaluate AV-CrossNet on multiple datasets, including LRS, VoxCeleb, and COG-MHEAR challenge. Evaluation results demonstrate that AV-CrossNet advances the state-of-the-art performance in all audiovisual tasks, even on untrained and mismatched datasets.
title AV-CrossNet: an Audiovisual Complex Spectral Mapping Network for Speech Separation By Leveraging Narrow- and Cross-Band Modeling
topic Audio and Speech Processing
Machine Learning
url https://arxiv.org/abs/2406.11619