Automatic recognition of bird species using human factor cepstral coefficients

Abstract

Identification of bird species based on their song is very important task from biodiversity point of view. In order to develop an automatic system of recognition of bird species, a system using signal processing and pattern recognition techniques has gained huge importance. In this paper, we compare the performance of mel frequency cepstral coefficients and human factor cepstral coefficients combined with time- and frequency-based features. Gaussian mixture models have been used for developing feature models, and maximum likelihood estimation is used for classification. Further, selective features have been used in order to increase the performance of the system. With the proposed method, a maximum accuracy of 97.72% has been achieved for a data set of ten bird species.