ACTIVITY RECOGNITION BASED ON 3D CNN-LSTM-ASSISTED APPROACH
Computer-vision is the advanced and innovative field for medicinal services, and intelligent monitoring systems are becoming part of healthcare system rapidly. The mechanization in the identification of ordinary or unpredictable activities of a patient can improve the wellbeing results and can likewise reduce the endeavors of manual checkups and monitoring. In this examination, the work centers around different health focused exercises of a patient to figure the range up to a variation from the norm scale. The proposed solution is legitimately reliant on the recordings acquired from the surrounding cameras to identify the physical developments. Present day techniques are totally reliant on the amount and nature of the information used to prepare the system to create an intense reaction to the recent development. To deal with these issues, a deep learning based classification strategy is proposed to address different physical variations from the norm which can prompt a reason for health affliction. A subset of NTU RGB+D dataset with health focused models are utilized to train the system. In the proposed study, 3D CNN model is utilized to extricate the highlights from the recordings and LSTM model is utilized to classify the activities. The proposed monitoring system has additionally used the qualities of different state-of-the-art models and the results are processed by computing the parameters of review, exactness, precision, and F-measure. Tests hold the proof to legitimize the utility of 3D CNN model for posture classification and LSTM model for activity prediction in our proposed framework.