Human Activity Recognition Using Improved Cat Swarm Optimization Algorithm and Convolutional Neural Network

Widespread computing has a lot of ability to offer users customised services and support, especially as more people use smartphones for everyday tasks. In the past ten years, there has been a significant advancement in Sensor-Based Human Activity Recognition (HAR), which detects the movements of the user and emerged as a crucial component of pervasive computing. Thus, accurate HAR plays a significant role in numerous applications. For accurate HAR, the current research proposes an Improved Cat Swarm Optimization Algorithm based Convolutional Neural Network (ICSO-CNN) for tuning the parameters of CNN for optimal solution. Using various performance measures, the present research examines the performance of the proposed ICSO-CNN algorithm. The test results demonstrated that the ICSO-CNN outperforms the compared methods, including CNN with Long Short-Term Memory (CNN-LSTM), Particle Swarm Optimization based CNN (PSO-CNN), and CNN-BiLSTM with an accuracy of 99.79%.