Long-term gait pattern assessment using a tri-axial accelerometer.

Area of Science:

• Biomedical Engineering
• Human Movement Analysis
• Wearable Technology

Background:

• Gait pattern classification is crucial for health monitoring and assistive technologies.
• Traditional methods often require complex features and classifiers, limiting long-term continuous application.
• There is a need for efficient and accurate gait analysis systems adaptable to daily life activities.

Purpose of the Study:

• To develop and validate a pervasive gait pattern classification algorithm using waist-mounted inertial sensor data.
• To enable continuous, long-term monitoring of distinct human locomotion activities.
• To simplify the classification process by utilizing minimal features and a straightforward decision tree classifier.

Main Methods:

• Utilized accelerometer data from a waist-mounted inertial sensor for gait analysis.
• Developed a classification algorithm employing three key features and a decision tree.
• Validated the algorithm's performance on data from nine healthy volunteers across 36 tests, totaling 12.5 hours of recorded acceleration data.

Main Results:

• The algorithm achieved high classification accuracy for continuous gait pattern recognition.
• Specific accuracies included: standing (100%), level walking (approximately 99%), stair ascending (approximately 84%), and stair descending (approximately 85%).
• Overall average classification accuracy for the four gait patterns exceeded 92% in long-lasting, continuous applications.

Conclusions:

• The proposed pervasive solution offers an effective and simplified approach to continuous gait pattern classification.
• The algorithm demonstrates robust performance for distinguishing between standing, walking, and stair ambulation using minimal sensor data.
• This method holds potential for long-term health monitoring, rehabilitation, and activity recognition applications.