Supervised Classification of Baboon Vocalizations - Archive ouverte HAL Accéder directement au contenu
Communication Dans Un Congrès Année : 2013

Supervised Classification of Baboon Vocalizations

Résumé

This paper addresses automatic classification of baboon vocalizations. We considered six classes of sounds emitted by "Papio papio" baboons, and report the results of supervised classification carried out with different signal representations (audio features), classifiers, combinations and settings. Results show that up to 94.1\% of correct recognition of pre-segmented elementary segments of vocalizations can be obtained using Mel-Frequency Cepstral Coefficients representation and Support Vector Machines classifiers. Results for other configurations are also presented and discussed, and a possible extension to the "Sound-spotting'' problem, i.e. online joint detection and classification of a vocalization from a continuous audio stream is illustrated and discussed.
Fichier principal
Vignette du fichier
NIPS4B_2013_JanvierMaxime.pdf (334.01 Ko) Télécharger le fichier
Origine : Fichiers produits par l'(les) auteur(s)
Loading...

Dates et versions

hal-00910104 , version 1 (27-11-2013)

Identifiants

  • HAL Id : hal-00910104 , version 1

Citer

Maxime Janvier, Radu Horaud, Laurent Girin, Frédéric Berthommier, Louis-Jean Boë, et al.. Supervised Classification of Baboon Vocalizations. NIPS4B - Workshop: Neural Information Processing Scaled for Bioacoustics : NIPS4B, Dec 2013, Lake Tahoe, Nevada, United States. 10 p. ⟨hal-00910104⟩
759 Consultations
390 Téléchargements

Partager

Gmail Facebook X LinkedIn More