Explainable automated anuran sound classification using improved one-dimensional local binary pattern and Tunable Q Wavelet Transform techniques

Classification of animal species using animal sounds is a critical issue for bioacoustics work. Especially the determination of anurans (frogs or toads) species can be used as an indicator of climate change. However, counting and classifying anurans in their natural habitat is challenging. Therefore, computer-assisted intelligent systems must be used to determine anuran types correctly. This work collected a new anuran sound dataset and proposed a hand-modeled sound classification system. The collected dataset contains 1536 anuran sounds belonging to 26 anuran species. Furthermore, an improved one-dimensional local binary pattern (1D-LBP) and Tunable Q Wavelet Transform (TQWT) based feature extraction method has been proposed to generate features at both frequency and space domains. Our proposed hand-modeled anuran sound classification architecture comprises of feature extractor (TQWT + improved 1D-LBP), iterative neighborhood component analysis (INCA) selector and k nearest neighbor (kNN) classifier. Our proposed 1D-LBP and TQWT-based anuran sound classification model has obtained a classification accuracy of 99.35% in classifying 26 anuran species. Moreover, we discussed explainable results. In the future, we plan to validate this work by increasing more species in each group.