Application for classifying phonic groups of bats using deep learning
Nowadays, we can observe a constant momentum of technical innovations in artificial intelligence (AI). Computer vision is one of the most studied branches. This dissertation deals with the classification of audio files corresponding to ultrasound signals from chiroptera, commonly known as bats, by means of deep learning models capable of identifying patterns from the spectrograms generated. The identification and classification of species is a fundamental basis of any study on fauna and, therefore, it is essential that this recognition is reliable. To this end, we have developed a specialised tool to determine the phonic group to which a given recording of ultrasounds emitted by local chiropterans belongs. For one of these groups (pipistrelleid species) we have determined their membership of different subgroups.
READ FULL TEXT