Detection of estrus through automated classification approaches using vocalization pattern in Murrah buffaloes

The field that deals with the studying of animal sounds to detect or classify various animal conditions in an open or a closed setup. Bioacoustics has its prospective applications in speaker identification, detection of respiratory distress, estrus phases, stress induced by ecological conditions, separation and weaning and so on. In dairy cattle, especially murrah buffalo breed, inaccurate detection of estrus is a significant constraint that affects the reproductive capacity of small as well as large cattle-herds. Detecting estrus in dairy cattle at correct time assists livestock managers and farmers to augment milk production, profitability and maintaining welfare of cattle.The studies on relationship between vocal behavior and reproductive performance is still at a very primary stage. This is due to inadequate amount of research work that has been done in order to study the association between acoustic features of animal vocal sounds and estrus owing to limited publicly available dataset. Outcomes of animal vocal sounds based studies are primarily dataset-dependent, i.e., the findings are ineffective if recording conditions are unidentified. Modern developments in deep-learning and machine-learning based approaches have led to the development of high-performance non-invasive ‘‘animal sound classification system These systems, however, are built on scarce speech-corpora which is a preferred requisite for machine learning model training, which requires hours of good-quality recordings of sound of a single speaker in a noiseless environment. Developing such a good-quality sound corpus manually requires labour and time which makes the process expensive. In the present paper, the authors have executed machine-learning models like Random Forest, Decision Tree and SVM as well as a custom-built CNN architecture on a dataset and have carried out a comparative analysis of the above-mentioned techniques.