ConcateNet: Dialogue Separation Using Local And Global Feature Concatenation

TL;DR

ConcateNet introduces a novel architecture for dialogue separation that effectively combines local and global features, demonstrating superior generalization to out-of-domain signals compared to existing noise-reduction methods.

Contribution

The paper presents ConcateNet, a new approach that enhances dialogue separation by processing local and global features for improved out-of-domain generalization.

Findings

Competitive performance on in-domain datasets.

Superior out-of-domain generalization compared to state-of-the-art methods.

Effective processing of local and global features for dialogue separation.

Abstract

Dialogue separation involves isolating a dialogue signal from a mixture, such as a movie or a TV program. This can be a necessary step to enable dialogue enhancement for broadcast-related applications. In this paper, ConcateNet for dialogue separation is proposed, which is based on a novel approach for processing local and global features aimed at better generalization for out-of-domain signals. ConcateNet is trained using a noise reduction-focused, publicly available dataset and evaluated using three datasets: two noise reduction-focused datasets (in-domain), which show competitive performance for ConcateNet, and a broadcast-focused dataset (out-of-domain), which verifies the better generalization performance for the proposed architecture compared to considered state-of-the-art noise-reduction methods.