New Abnormal Behavior Detection for Patient Surveillance System

In contemporary healthcare settings, nursing staff encounter significant workload challenges due to high patient-to-nurse ratios, often ranging from 16 to 25 patients per nurse. This issue is particularly acute in psychiatric wards, where patients exhibit unpredictable behaviors. Continuous monitoring is essential for the care of these mental health patients, yet chronic staff shortages hinder the fulfillment of these needs. To address these challenges, this study proposes a new abnormal behavior detection for patient surveillance system, utilizing real-time video surveillance to identify abnormal behaviors in psychiatric wards. The system’s accuracy has been enhanced through well-defined characteristic behaviors and a rule-based detection algorithm. Built on the DeepStream SDK, the system exhibits excellent performance, processing each frame in an average of 0.01 seconds and achieving a recognition accuracy exceeding 90%. This patient surveillance system significantly improves emergency response capabilities, mitigates potential risks, and enhances the work environment for medical personnel by providing accurate and timely patient status information. Ultimately, this approach is anticipated to contribute not only to improved health outcomes for patients but also to enhanced job satisfaction and safety for healthcare professionals.