Posture Guard – A Human Sitting Posture Detection And Alarming System

Poor posture due to prolonged screen time is an increasingly prevalent health issue with long-term consequences such as musculoskeletal discomfort and pain. To address this issue, we introduce the “Posture Guard” system, an innovative solution that leverages computer vision, machine learning, and real-time feedback to monitor and improve posture. Utilizing a webcam, the system analyzes users' posture in real-time, providing immediate visual and audio feedback for corrective action. First, data is collected under five classes: namely, “Normal”, “Neck”, “Shoulders”, “Forward Bend”, “Backward Bend”. Then, multiclass CNN is trained to classify the data into different classes. This paper discusses the development and evaluation of the Posture Guard system, highlighting its key features which include robust posture analysis, data tracking, and user-friendly interface. This study employed a three-phase testing approach to evaluate the robustness of this system in various scenarios like different environment and lighting conditions. The model derived through this research gave around 99% precision on people who were a part of training data and gave an average precision of 64% on unseen faces in a different environment. The system demonstrated the potential to substantially improve posture habits, as evidenced by trials showing enhanced awareness and significant posture correction among users. The implications of this study suggest that such technological interventions could mitigate health risks and improve the quality of life for users engaged in sedentary activities.