Birds’ Sound Classification using Acoustic Signals

Abstract

Bird sound classification is an important task in bioacoustics research, wildlife conservation, and automated systems for monitoring bird populations. In this paper, we propose a framework to classify bird species based on their vocalizations using acoustic local ternary patterns (acoustic-LTPs) as an acoustic descriptor. It’s an extension of acoustic local binary pattern (acoustic-LBP). We utilize a publicly available bird sounds dataset consisting of 781 sound samples from 18 distinct bird species. After extracting acoustic features using acoustic-LTPs we concatenated the vector with features sets from Mel-Frequency Cepstral Coefficients (MFCC) and Linear predictive coding (LPC). Once the signal representation is done, we classified the signals through different classifiers and achieved a higher accuracy rate of 97.4%. Results show that the proposed method is more powerful and reliable in terms of identifying bird species based on their sounds.  By accurately classifying bird species through their vocalizations, this research not only contributes to the field of bioacoustics research but also offers valuable tools for wildlife conservation efforts. Understanding bird behavior through vocalizations can also provide insights into their ecological roles, habitat preferences, and responses to environmental changes.