Special Issue Article Open Access

Automatic Identification of Bird Species from the Recorded Bird Song Using ART Approach

Abstract

ABSTRACT— our goal is to automatically identify which species of bird is present in an audio recording using spectrogram features. Birds song recognition approach is used to provide automatic investigation and remote monitoring of bird species population, which can provide the relevant agencies with sound information to habitat conservation as well as rare\endangered species survival plans and actions. The set of acoustic features developed for bird’s song recognition was generally inspired by feature representations used in speech/speaker recognition or audio/music classification fields. In general these acoustic features are based on the acoustic model of speech production or the perceptual model of the human auditory system. Each spectrogram can be viewed as an image. Each spectrogram can be viewed as grey-level images. The new MPEG-7 Angular Radial Transform (ART) descriptor can be efficiently describing the greylevel variations within an image region in both angular and radial directions, to extract the shape features from the spectrogram image. Bird’s song having distinct frequency and temporal variations will exhibit different shapes in their spectrogram. It extracts the shape features from the sound spectrograms of fixed duration bird’s song segments. A sector expansion algorithm is proposed to transform its spectrogram image into sector image. It will align with the radial and angular directions of the ART basis function. A classification algorithm then employed to exactly identify the bird species based on the extracted features. GMM (Gaussian mixture models) then used for the classification to determine the bird species associated with the input birds song segment. For the classification of bird species using ART descriptor, is better than the traditional descriptor such as LPCC, MFCC and TDMFCC.

To read the full article Download Full Article