Sudden Fall Detection of Human Body Using Transformer Model.

Area of Science:

• Computer Science
• Biomedical Engineering
• Gerontology

Background:

• Accurate fall detection is crucial for elderly care and patient monitoring.
• Timely intervention after a fall can prevent severe health consequences.
• Existing methods often struggle with accuracy and false alarms.

Purpose of the Study:

• To develop a novel fall detection model using a transformer architecture.
• To leverage movement speeds of key body points for enhanced fall recognition.
• To improve accuracy and reduce false alarms in elderly fall detection systems.

Main Methods:

• Utilized the MediaPipe library for tracking key body points in video data.
• Developed a transformer-based model focusing on real-time movement speed analysis.
• Employed the transformer's attention mechanism to detect subtle movement shifts.

Main Results:

• Achieved a fall detection accuracy of 97.6%.
• Significantly reduced false alarm rates compared to traditional methods.
• Demonstrated the model's ability to identify subtle movement changes indicative of falls.

Conclusions:

• The transformer-based model offers a highly accurate and reliable solution for fall detection.
• This approach enhances safety and enables faster intervention for vulnerable populations.
• The model shows practical applicability in elderly care facilities and home monitoring systems.