Machine-learning-based children's pathological gait classification with low-cost gait-recognition system.

Area of Science:

• Biomedical Engineering
• Rehabilitation Technology
• Pediatric Health

Background:

• Pathological gaits in children can lead to severe musculoskeletal conditions such as osteoarthritis and scoliosis.
• Early detection and therapeutic intervention are crucial for mitigating long-term health consequences.
• Existing automated, low-cost gait recognition systems for children are limited.

Purpose of the Study:

• To design and validate a cost-effective pathological gait recognition system (PGRS) for children.
• To develop an intelligent gait recognition method (IGRM) utilizing only plantar pressure information.
• To achieve high accuracy and real-time performance in identifying pathological gaits in pediatric populations.

Main Methods:

• A PGRS was developed using an 8x8 pressure-sensor array.
• An IGRM based on machine learning and plantar pressure data was implemented for static and dynamic gait analysis.
• Experiments involved 17 children recognizing normal, toe-in, toe-out, and flat gaits, with performance evaluated using cross-validation, recall, precision, and time cost.

Main Results:

• The IGRM demonstrated practical applicability with high average accuracy in both static and dynamic sections.
• Intra-subject recognition accuracy reached 92.41% (static) and 97.79% (dynamic).
• Inter-subject recognition accuracy was 85.78% (static) and 78.81% (dynamic), with static accuracy slightly lower due to less natural postures.

Conclusions:

• A low-cost PGRS is feasible, offering high precision and real-time gait recognition capabilities for children.
• The system shows potential for computer-aided supervision of both normal and pathological gaits using plantar pressure patterns.
• This technology can aid in gait abnormality rectification through timely feedback and intervention.