Edge devices play a crucial role in applications where computing can be done locally. As edge devices have limited computing power and memory due to which there is a necessity to design computationally small models that can be deployed on edge devices like Raspberry Pi, Jetson, Intel Edison, etc. In dairy farming livestock’s affective states can be identified with their vocalizations. Cows have separate vocalizations for close-range and long-distance communication. Additionally, each individual’s vocalizations are unique across circumstances. Distinguishably Understanding dairy cows’ vocalizations under anguish, pain, or fear is crucial. This study is conducted to optimize the known CNN-LSTM deep learning model on cow vocalization dataset with model optimization techniques to deploy it on edge devices for local computing on farms to monitor the emotional affective states of livestock animals.