Nonlinear Predictive Threshold Model for Real-Time Abnormal Gait Detection.

Area of Science:

• Biomedical Engineering
• Gerontology
• Computer Science

Background:

• Falls pose significant risks to human health, leading to physical and psychological injuries.
• Fall-risk prediction systems aim to prevent falls before they occur.
• Current systems often rely on clinical assessments or real-time gait abnormality recognition.

Purpose of the Study:

• To develop a low-complexity, high-accuracy nonlinear model for recognizing abnormal gait patterns.
• To create a realistic dataset for evaluating fall prediction methods using smartphone sensors.
• To enhance real-time fall risk prediction capabilities.

Main Methods:

• A nonlinear user gait model was developed.
• Threshold-based classification was employed for abnormal gait pattern recognition.
• A dataset was generated using smartphone accelerometer and gyroscope sensors to simulate abnormal walks.

Main Results:

• The proposed approach achieved 93.5% accuracy in predicting abnormal walks.
• The system demonstrated up to 3.5 times greater efficiency compared to state-of-the-art methods.
• The nonlinear gait model effectively identified abnormal gait patterns with high precision.

Conclusions:

• The developed nonlinear gait model offers a promising approach for accurate and efficient fall risk prediction.
• Smartphone-based gait analysis using this model can significantly improve fall prevention strategies.
• This method provides a practical and effective tool for real-time fall detection and risk assessment.