Automated Koala call detection using a cochlea model

Abstract

Estimating koala populations is challenging, making surveys costly and often imprecise. This study examines the feasibility of recognising koala calls from audio recordings using a bio-inspired acoustic approach. In the work, we propose to use the Cascade of Asymmetric Resonators with Fast-Acting Compression (CAR-FAC) cochlea model to pre-process koala recordings and a support vector machine (SVM) for classification. The frequency-dependent gain and bandwidth properties of the CAR-FAC model in pre-processing noisy data are discussed and investigated. To evaluate the effectiveness of the proposed method, it is compared with an FFT-based system as a baseline method. The baseline methodology uses a Mel-spectrogram pre-processing and the same linear classifier as the proposed method as the back-end. In summary, we achieved 96.7% accuracy for the proposed system.