Real-Time Human Fall Detection and Emotion Recognition using Embedded Device and Deep Learning

Abstract

The fall detection is one of the most discussed topic among the researchers because of its usage in home care and medical places for the patients. Because death due to falls is considered as most dangerous and injuries sustained from these incidents are pretty fatal and serious. Sometimes if the young age person falls he might suffer from minor injuries but we cannot say the same to the older aged person. Sometimes even though the person is suffered from the fall we can save the person by providing immediate care. Because there are some cases in which the death occurred by not responding quickly. Raspberry pi is used as a embedded device and Mog2 algorithm is used for video analysis. After detecting the fall the people are notified using internet based REST APIs like TWILIO. And for emotion recognition it is mostly used to detect the emotional state of the person. This system is mostly used in health care and marketing sectors to provide better results for the people. For emotion recognition system we use CNN (Convolutional Neural Network) since it considered as the best classifier for the video based recognition system. A data set with various emotions is trained and introduced to the system. For both these systems by using deep learning algorithm with embedded device and OpenCV we can provide better results and the performance of the systems are improved.