Authors: Dr.J.Gajavalli, Dr P.Saranya
Abstract: There is a massive paradigm shift taking place within the field of global health monitoring systems due to the fusion of Internet of Things (IoT) devices and Deep Learning (DL). Conventional health monitoring solutions are episodic, reactive, and clinic-centric, missing out on the essential need to record continuous and real-time data pertaining to vital health parameters. This paper outlines an IoT-assisted real-time health monitoring system that utilizes multi-modal wearable sensor devices (ECG, PPG, accelerometer, temperature) along with a unique hybrid deep neural network. The proposed solution, dubbed HealthNet, incorporates a Temporal Convolutional Network (TCN), Bidirectional Long Short-Term Memory (Bi-LSTM) network, and Multi-Head Attention mechanism to effectively process real-time time-series data. The performance of HealthNet was tested using a massive database containing 1,000 patients' health records consisting of 10 million labeled time-series samples. Results indicate that the model can accurately detect five health emergencies (arrhythmia, hyperthermia, hypothermia, falls, and respiratory distress) with 98.2% accuracy and 97.5% sensitivity, ensuring sub-second latencies of less than 200 milliseconds.
DOI: https://doi.org/10.5281/zenodo.20416044
