Deep Learning-Driven Health Monitoring System Integrated with Wi-Fi Technology

Aim: A Wi-Fi-based system utilizing deep learning is designed to classify real-time health conditions from sensor data. The effectiveness of CNN, LSTM, and the hybrid CNN+LSTM model is compared to identify the best method. Materials and Methods: A microcontroller, ESP32, would process and transmit data wirelessly based on data collected through temperature, heart rate, gas, and gyro sensors. In case necessary feature engineering on raw sensor data is required, this can also be done. Two groups are considered to develop and test the model. Group 1 comprises individual CNN and LSTM models that were designed, trained, and separately evaluated. The CNN model gained an accuracy of 85% with 5 sample predictions and the LSTM model gained an accuracy of 88% with 5 sample predictions. Group 2 is the combined CNN+LSTM model, which also was designed, trained, and evaluated to a higher accuracy at 92% with 5 sample predictions. This shows the fusion of CNN and LSTM produces better accuracy in comparison to isolated models. Result: Both the CNN and LSTM models recorded classification accuracy at 85% and 88%, respectively, while their combination in hybrid CNN+LSTM outpaced them both by attaining a record of 92%. The confusion matrices provided visualization of the performance of classifications. Conclusion: The new system effectively fuses deep learning and IoT for health monitoring. The CNN+LSTM model exhibited higher accuracy, which reflected the advantages of spatial and temporal feature extraction together. System optimization and extending its applications to the broader healthcare arena will be emphasized in future research.