International Journal of Scientific Research and Engineering Development

Developed in this research is a multi-modal deep learning system for understanding bovine emotions, or affective states, by analyzing three dimensions of investigation: sound, face expression, and body movement. The objective is to provide a mechanism that is capable of accurately identifying the primary affective states of bovines (i.e., Distress, Hunger, Imminent Parturition, Contentment). Sensors used include cameras, microphones, and accelerometer collars. Acoustic data is processed through a combination of Mel-Frequency Cepstral Coefficients (MFCC) and facial characteristics, which are evaluated via Iris-Pupillary Ratio (IPR) and Facial Action Unit (FAU) metrics. Temporal structures in data collection are retained via a CNN-LSTM model; thus providing a comprehensive data set from all three dimensions for a final classification prediction from a softmax-conceptualized fusion layer. While the potential for performance improvement from the multi-modal approach over unimodal approaches would be expected, the results from this research indeed substantiate an approximate 60–70% accuracy rate on the multi-class classification tasks (of interest) compared to 87+% accuracy rates for the simple task of detecting high-frequent calls

Shifa Bilal Tamboli, Simeen Phiroj Mulani, Arman Tajuddin Shiakh,"Cow Sentience System" International Journal of Scientific Research and Engineering Development, V9(2): Page(2490-2494) Mar-Apr 2026. ISSN: 2581-7175. www.ijsred.com. Published by Scientific and Academic Research Publishing.