Multiple Choice Learning for Efficient Speech Separation with Many Speakers

Fuente: arXiv
Gespeichert in:
Bibliographische Detailangaben
Hauptverfasser: Perera, David, Derrida, François, Mariotte, Théo, Richard, Gaël, Essid, Slim
Format: Preprint
Veröffentlicht: 2024
Schlagworte:
Online-Zugang:
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
_version_ 1866915037496475648
author Perera, David
Derrida, François
Mariotte, Théo
Richard, Gaël
Essid, Slim
author_facet Perera, David
Derrida, François
Mariotte, Théo
Richard, Gaël
Essid, Slim
contents Training speech separation models in the supervised setting raises a permutation problem: finding the best assignation between the model predictions and the ground truth separated signals. This inherently ambiguous task is customarily solved using Permutation Invariant Training (PIT). In this article, we instead consider using the Multiple Choice Learning (MCL) framework, which was originally introduced to tackle ambiguous tasks. We demonstrate experimentally on the popular WSJ0-mix and LibriMix benchmarks that MCL matches the performances of PIT, while being computationally advantageous. This opens the door to a promising research direction, as MCL can be naturally extended to handle a variable number of speakers, or to tackle speech separation in the unsupervised setting.
format Preprint
id arxiv_https___arxiv_org_abs_2411_18497
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle Multiple Choice Learning for Efficient Speech Separation with Many Speakers
Perera, David
Derrida, François
Mariotte, Théo
Richard, Gaël
Essid, Slim
Sound
Machine Learning
Audio and Speech Processing
Training speech separation models in the supervised setting raises a permutation problem: finding the best assignation between the model predictions and the ground truth separated signals. This inherently ambiguous task is customarily solved using Permutation Invariant Training (PIT). In this article, we instead consider using the Multiple Choice Learning (MCL) framework, which was originally introduced to tackle ambiguous tasks. We demonstrate experimentally on the popular WSJ0-mix and LibriMix benchmarks that MCL matches the performances of PIT, while being computationally advantageous. This opens the door to a promising research direction, as MCL can be naturally extended to handle a variable number of speakers, or to tackle speech separation in the unsupervised setting.
title Multiple Choice Learning for Efficient Speech Separation with Many Speakers
topic Sound
Machine Learning
Audio and Speech Processing
url https://arxiv.org/abs/2411.18497