PROPOSING A POULTRY SOUND CLASSIFICATION APPROACH BASED ON DCT AND YOLO V8M TO IDENTIFY CHICKEN STRESS TYPE

Most animals make sounds and calls to express their condition or to communicate with other members of the group communicate with other group members. Poultry chickens live in groups and communicate with each other socially through their sounds. The lack of food or stress that chickens are exposed to negatively affects the global production of eggs and meat. Therefore, in this paper, we present a deep learning-based approach to classifying the status of farm chickens by their vocalization, which helps improve production and monitor the animals. First, all the sounds are converted to the time domain, and the Discrete Cosine Transform (DCT) is calculated to produce distinct features. Then, A two-dimensional grayscale image is generated from the coefficient matrix. Secondly, these gray images are utilized as input into the YOLO v8m model to perform classification. The results showed that the proposed model achieved a high classification accuracy of 91.6% without data augmentation and 98.6% with data augmentation, although there is noise interfering with the recordings or the sounds of chickens nearby. This study may play an important role in animal acoustic signal analysis studies