A SVD-based classification of bird singing in different time-frequency domains using multitapers

Summary, in English

In this paper, a novel method for analysing a bird’s song is presented. The song of male great reed warblers is used for developing and testing the methods. A robust method for detecting syllables is proposed and a classification of those syllables as compared to reference syllables is done. The extraction of classification features are based on the use of singular vectors in different time-frequency domains, such as the ambiguity and the doppler domains, in addition to the usual sonogram. The analysis is also made using multitaper analysis where the Welch method and the Thomson multi- tapers are compared to the more recently proposed locally stationary process multitapers.

• Author: Maria Sandsten; Maja Tarka; Jessica Caissy-Martineau; Bengt Hansson; Dennis Hasselquist
• Department/s: eSSENCE: The e-Science Collaboration; Statistical Signal Processing Group; Mathematical Statistics; MEMEG; BECC: Biodiversity and Ecosystem services in a Changing Climate; Molecular Ecology and Evolution Lab
• Publishing year: 2011
• Language: English
• Pages: 966-970
• Publication/Series: European Signal Processing Conference
• Volume: 2011
• Links: http://www.eurasip.org/Proceedings/Eusipco/Eusipco2011/papers/1569422961.pdf
• Document type: Conference paper
• Publisher: European Association for Signal Processing (EURASIP)
• Topic: Probability Theory and Statistics
• Conference name: 19th European Signal Processing Conference, EUSIPCO 2011
• Conference date: 2011-08-29 - 2011-09-02
• Conference place: Barcelona, Spain
• Status: Published
• Research group: Stochastics in Medicine; Statistical Signal Processing; Statistical Signal Processing Group; Molecular Ecology and Evolution Lab
• ISBN/ISSN/Other: ISSN: 2219-5491