Behavior Classification of Egyptian Fruit Bat (Rousettus aegyptiacus) From Calls With Deep Learning

Yılmaz B., Şen M., Maşazade E., Beşkardeş V.

Handbook of Research on New Investigations in Artificial Life, AI, and Machine Learning, Maki K. Habib, Editör, IGI Global, Cairo, ss.60-98, 2022

  • Yayın Türü: Kitapta Bölüm / Araştırma Kitabı
  • Basım Tarihi: 2022
  • Yayınevi: IGI Global
  • Basıldığı Şehir: Cairo
  • Sayfa Sayıları: ss.60-98
  • Editörler: Maki K. Habib, Editör
  • İstanbul Üniversitesi-Cerrahpaşa Adresli: Evet

Özet

Sensing the environment with passive acoustic sensors has been used as a very useful tool to monitor 

and quantify the status and changes on biodiversity. In this chapter, the authors aim to classify the social 

calls (biting, feeding, fighting, isolation, mating protest, and sleeping) of a certain bat species, Egyptian 

fruit bat, which lives in colonies with thousands of others. Therefore, classification of their calls not 

only helps us to understand the population dynamics but also helps us to offer distinct environmental 

management procedures. In this work, the authors use the database previously presented in Prat et al. 

and present the social call classification results under both classical machine learning techniques and 

a convolutional neural network (CNN). The numerical results show that CNN improves the classifica-

tion performance up to 20% as compared to the traditional machine learning approaches when all the 

call classes are considered. It has also been shown that the classes of aggressive calls, which can sound 

quite close to each other, can be distinguished with CNN.