Falling-incident detection and throughput enhancement in a multi-camera video-surveillance system

Area of Science:

• Gerontology
• Computer Science
• Biomedical Engineering

Background:

• Elderly individuals are prone to falls due to frailty, often occurring in hazardous areas like stairs or corridors.
• Delayed rescue after a fall can lead to severe injuries, particularly if the individual faints.
• Existing surveillance systems require constant human monitoring or wearable sensors, which are inefficient and inconvenient.

Purpose of the Study:

• To develop an automatic falling detection algorithm for elderly individuals.
• To implement this algorithm within a multi-camera video surveillance system.
• To improve the timeliness and efficiency of fall incident detection and subsequent rescue.

Main Methods:

• An automatic falling detection algorithm was developed and integrated into a multi-camera video surveillance system.
• Each camera captures images from monitored areas, feeding them into the falling-pattern recognition algorithm.
• The system is supported by a Digital Signal Processor (DSP)-based hardware acceleration board for performance.

Main Results:

• The falling detection algorithm achieved an accuracy of at least 90% in simulations.
• The multi-camera surveillance system demonstrated a throughput improvement of approximately 2.1 times.
• Upon detecting a fall, the system automatically sends alerts via short messages.

Conclusions:

• The proposed automatic falling detection algorithm offers a highly accurate and efficient solution for elderly fall monitoring.
• The integrated multi-camera system overcomes the limitations of traditional surveillance methods, reducing human power waste and elder inconvenience.
• This technology has the potential to significantly improve safety and response times for elderly individuals at risk of falling.