The Convergence of Cutting-Edge Technologies: Leveraging AI and Edge Computing to Transform the Internet of Medical Things (IoMT)
Rajasrikar Punugoti, Narayan Vyas, Ahmad Talha Siddiqui, Abdul W. Basit
Abstract
This research used a wearable sensor to gather photoplethysmography (PPG) signals from 15 healthy subjects. The dataset includes 7,308 PPG segments, each containing 8 seconds of PPG data and corresponding labels indicating the type of physical activity the subject performed. The article proposes a convolutional neural network (CNN) model to classify physical activity from the PPG signals. The proposed model includes several layers: batch normalization, convolutional, max-pooling, dropout, and fully connected. The output layer uses the softmax activation function to compute the probabilities of each class. Regarding performance, the suggested CNN model outperforms conventional models like SVM with RBF kernel, Decision Tree, and Random Forest. The article also suggests several techniques to optimize the model further, which can be beneficial for developing IoMT applications such as activity recognition and vital signs monitoring.