Automatic bird species identification employing an unsupervised discovery of vocalisation units

Zakeri, Masoud (2017). Automatic bird species identification employing an unsupervised discovery of vocalisation units. University of Birmingham. Ph.D.

PDF - Accepted Version

Download (9MB)


An automatic analysis of bird vocalisations for the identification of bird species, the study of their behaviour and their means of communication is important for a better understanding of the environment in which we are living and in the context of environmental protection. The high variability of vocalisations within different individuals makes species’ identification challenging for bird surveyors. Hence, the availability of a reliable automatic bird identification system through their vocalisations, would be of great interest to professionals and amateurs alike.

A part of this thesis provides a biological survey on the scientific theories of the study of bird vocalisation and corresponding singing behaviours. Another section of this thesis aims to discover a set of element patterns produced by each bird species in a large corpus of the natural field recordings. Also this thesis aims to develop an automatic system for the identification of bird species from recordings. Two HMM based recognition systems are presented in this research. Evaluations have been demonstrated where the proposed element based HMM system obtained a recognition accuracy of over 93% by using 3 seconds of detected signal and over 39% recognition error rate reduction, compared to the baseline HMM system of the same complexity.

Type of Work: Thesis (Doctorates > Ph.D.)
Award Type: Doctorates > Ph.D.
College/Faculty: Colleges (2008 onwards) > College of Engineering & Physical Sciences
School or Department: School of Engineering
Funders: None/not applicable
Subjects: T Technology > TK Electrical engineering. Electronics Nuclear engineering


Request a Correction Request a Correction
View Item View Item


Downloads per month over past year