A Wearable Self-Powered Sensor Tag for Deep Learning-Based Cow Monitoring System

Abstract

Cow’s milk and its products provide many nutrients that contribute to the healthy growth of the body. In addition, milk production brings massive benefits to producers and ensures household livelihoods, food security, and nutrition. However, the ability to exploit milk can be restricted by poor quality feed, diseases, and the capacity of conventional farms. To enhance the efficiency of dairy farming, we proposed a dairy cow monitoring system based on movement recognition and wireless communication. The core of our system is a self-powered sensor tag equipped with an accelerometer sensor and radiofrequency energy harvesting technology, which is mounted on the cow’s neck to measure movements as acceleration data. The energy supplied to this device is harvested from radio frequency waves at the ultra-high frequency band of 915 MHz transmitted by a 1-W reader. As a result, the sensor tag obtains up to 60% power conversion efficiency and consumes only 160- µA of average current. To assess the feasibility of the proposed system, experiments were conducted to recognize three behaviors, often dairy cows, for a week with the assistance of the camera system and observers. These behaviors consist of standing, walking, and grazing. Furthermore, a one-dimensional convolutional neural network model was developed to perform the behavioral classification task. The result showed that the average accuracy achieves 92.52%. Our achievement could make the premise for developing smart and precision farm systems.