TOF sensor-based exercise detection with health coaching for office workers

In today’s office landscape, where employees spend 40% of their day at desks neglecting personal health, Wellness Assistant, named ISA, leverages sensor technology and Artificial Intelligence to create personalized health solutions. ISA focuses on targeted exercises, reducing musculoskeletal problems responsible for 21.2% of sick days in Germany. As remote work surges due to COVID-19, ISA acts as a smart health coach, empowering individuals to enhance well-being while cutting costs from employee absenteeism. Here, machine learning and deep learning approaches are used in a novel way to improve office workers’ wellbeing. For TOF sensor-based exercise detection and individualized health coaching, a combination of the convolutional neural network(CNN)+LSTM/LRCN technique, Mediapipe, and OpenPose is used. Comprehensive exercise analysis is made possible by the depth and infrared images provided by the TOF sensor. Office workers frequently lead sedentary lives, which reduces performance and raises the risk of common pains and discomforts. The TOF sensor is used for precise exercise detection to overcome these difficulties, capturing the user’s postures and motions in real-time. The sequential patterns of workout movements recorded by the TOF sensor’s depth and infrared pictures are analyzed using the CNN+LSTM/LRCN approach. The system can recognize and understand complex exercise routines thanks to this deep learning approach, making activity detection and classification more precise. Additionally, real-time pose estimation and monitoring using the Mediapipe and OpenPose frameworks are used to provide comprehensive data on the user’s body position during activities. The system offers individualized health coaching recommendations based on the analysis of workout data by exploiting the capabilities of machine learning and deep learning algorithms and utilizing depth photos and infrared images from the TOF sensor. This thesis intends to enhance office worker performance through the combination of the CNN+LSTM/LRCN technique, Mediapipe, OpenPose, and the use of depth photos and infrared images from the TOF sensor. The findings of this study make a significant contribution to the expanding field of workplace health technology and illustrate the potential advantages of machine learning and deep learning techniques combined with TOF sensor-based exercise identification for office workers.