RCT: Random Consistency Training for Semi-supervised Sound Event Detection

Fuente: arXiv
Salvato in:
Dettagli Bibliografici
Autori principali: Shao, Nian, Loweimi, Erfan, Li, Xiaofei
Natura: Preprint
Pubblicazione: 2021
Soggetti:
Accesso online:
Tags: Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
_version_ 1866913180817555456
author Shao, Nian
Loweimi, Erfan
Li, Xiaofei
author_facet Shao, Nian
Loweimi, Erfan
Li, Xiaofei
contents Sound event detection (SED), as a core module of acoustic environmental analysis, suffers from the problem of data deficiency. The integration of semi-supervised learning (SSL) largely mitigates such problem while bringing no extra annotation budget. This paper researches on several core modules of SSL, and introduces a random consistency training (RCT) strategy. First, a self-consistency loss is proposed to fuse with the teacher-student model to stabilize the training. Second, a hard mixup data augmentation is proposed to account for the additive property of sounds. Third, a random augmentation scheme is applied to flexibly combine different types of data augmentations. Experiments show that the proposed strategy outperform other widely-used strategies.
format Preprint
id arxiv_https___arxiv_org_abs_2110_11144
institution arXiv
publishDate 2021
record_format arxiv
spellingShingle RCT: Random Consistency Training for Semi-supervised Sound Event Detection
Shao, Nian
Loweimi, Erfan
Li, Xiaofei
Audio and Speech Processing
Machine Learning
Sound
Sound event detection (SED), as a core module of acoustic environmental analysis, suffers from the problem of data deficiency. The integration of semi-supervised learning (SSL) largely mitigates such problem while bringing no extra annotation budget. This paper researches on several core modules of SSL, and introduces a random consistency training (RCT) strategy. First, a self-consistency loss is proposed to fuse with the teacher-student model to stabilize the training. Second, a hard mixup data augmentation is proposed to account for the additive property of sounds. Third, a random augmentation scheme is applied to flexibly combine different types of data augmentations. Experiments show that the proposed strategy outperform other widely-used strategies.
title RCT: Random Consistency Training for Semi-supervised Sound Event Detection
topic Audio and Speech Processing
Machine Learning
Sound
url https://arxiv.org/abs/2110.11144