A Deep Learning Approach for Detecting and Classifying Cat Activity to Monitor and Improve Cat's Well-Being Using Accelerometer, Gyroscope, and Magnetometer

Activity detection of pets using sensor technology and artificial intelligence (AI) approaches has gained momentum in recent years. These technologies have revamped up the monitoring systems for the well-being of pets and are emerging as revolutionary technologies. Most of the research works have been conducted on the activity detection of humans, dogs, sheep, and horses; however, cat activity detection has not been explored meticulously using the aforementioned latest technologies. In this article, we proposed an long short-term memory (LSTM)-based deep learning model and investigated the activity detection of cats using three wearable sensors. Our research content can be summarized as follows. First, we collected 1765825 data samples from ten cats of different ages, breeds, and gender using an accelerometer, gyroscope, and magnetometer. Second, the raw data were preprocessed by applying data transformation approaches and feature engineering techniques. Third, we developed three different kinds of deep learning models, namely; artificial neural network (ANN), 1-D convolutional neural network (CNN), and LSTM, to classify ten different activities of cats. Hyperparameter tuning was performed to optimize the performance of the models. LSTM model outperformed other models and showed good accuracy. The model achieved an accuracy of 95.94%. To the best of our knowledge, this work is the first of its kind, which applied a deep learning-based model for the detection of cat activity using three kinds of sensor data, such as accelerometer, gyroscope, and magnetometer. This proposed model will help improve the well-being of cats by detecting the activities of cats with high accuracy using wearable sensor devices and deep learning technology.