Learning Marmoset Vocal Patterns with a Masked Autoencoder for Robust Call Segmentation, Classification, and Caller Identification

Fuente: arXiv
Salvato in:
Dettagli Bibliografici
Autori principali: Wu, Bin, Takamichi, Shinnosuke, Sakti, Sakriani, Nakamura, Satoshi
Natura: Preprint
Pubblicazione: 2024
Soggetti:
Accesso online:
Tags: Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
_version_ 1866915440632004608
author Wu, Bin
Takamichi, Shinnosuke
Sakti, Sakriani
Nakamura, Satoshi
author_facet Wu, Bin
Takamichi, Shinnosuke
Sakti, Sakriani
Nakamura, Satoshi
contents The marmoset, a highly vocal primate, is a key model for studying social-communicative behavior. Unlike human speech, marmoset vocalizations are less structured, highly variable, and recorded in noisy, low-resource conditions. Learning marmoset communication requires joint call segmentation, classification, and caller identification -- challenging domain tasks. Previous CNNs handle local patterns but struggle with long-range temporal structure. We applied Transformers using self-attention for global dependencies. However, Transformers show overfitting and instability on small, noisy annotated datasets. To address this, we pretrain Transformers with MAE -- a self-supervised method reconstructing masked segments from hundreds of hours of unannotated marmoset recordings. The pretraining improved stability and generalization. Results show MAE-pretrained Transformers outperform CNNs, demonstrating modern self-supervised architectures effectively model low-resource non-human vocal communication.
format Preprint
id arxiv_https___arxiv_org_abs_2410_23279
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle Learning Marmoset Vocal Patterns with a Masked Autoencoder for Robust Call Segmentation, Classification, and Caller Identification
Wu, Bin
Takamichi, Shinnosuke
Sakti, Sakriani
Nakamura, Satoshi
Sound
Artificial Intelligence
Audio and Speech Processing
The marmoset, a highly vocal primate, is a key model for studying social-communicative behavior. Unlike human speech, marmoset vocalizations are less structured, highly variable, and recorded in noisy, low-resource conditions. Learning marmoset communication requires joint call segmentation, classification, and caller identification -- challenging domain tasks. Previous CNNs handle local patterns but struggle with long-range temporal structure. We applied Transformers using self-attention for global dependencies. However, Transformers show overfitting and instability on small, noisy annotated datasets. To address this, we pretrain Transformers with MAE -- a self-supervised method reconstructing masked segments from hundreds of hours of unannotated marmoset recordings. The pretraining improved stability and generalization. Results show MAE-pretrained Transformers outperform CNNs, demonstrating modern self-supervised architectures effectively model low-resource non-human vocal communication.
title Learning Marmoset Vocal Patterns with a Masked Autoencoder for Robust Call Segmentation, Classification, and Caller Identification
topic Sound
Artificial Intelligence
Audio and Speech Processing
url https://arxiv.org/abs/2410.23279