Development of a smart necklace for stroke warning based on IoT and Convolutional Neural Network Deep Learning Techiniques

In this study, the authors developed a smart wearable necklace that can predict and warn of strokes based on fall detection by combination of the Internet of Things (IoT) and Deep Learning (DL) technologies. This device is able to measure user's health and movement parameters such as heart rate, translation accelerations, angular velocities, and calculating the Euler rotation angles corresponding to coordinate system attached to the device. Through the user's movement data, a Convolutional Neural Network (CNN) DL model based automatic feature extraction is developed to predict fall which may be caused by a stroke. The results show that CNN Deep Learning model is effective to detect fall with prediction accuracy is over 85% and the sensitivity is over 95% within 0.02 second. Short calculation time for fall detection demonstrated that the devices is capable for real-time prediction of fall which may be caused by a stroke. In addition, an application running on computer is developed to facilitate medical doctors, users, caregivers to monitor the wearer's health condition conveniently. Furthermore, the application is capable of notifying medical doctors and caregivers whenever a fall of an user is detected. Additionally, the user's position is exactly located on the GPS map integrated inside the application for the emergency activities in the case a fall caused by stroke occurs.