Algorithms for Walking Speed Estimation Using a Lower-Back-Worn Inertial Sensor: A Cross-Validation on Speed Ranges.

Area of Science:

• Biomechanics
• Wearable Technology
• Gait Analysis

Background:

• Walking/gait speed is crucial for assessing daily mobility.
• Existing algorithms for estimating walking speed using lower back inertial sensors lack comprehensive validation across diverse populations and walking conditions.
• There is a need for robust algorithms that perform well for individuals with varying preferred walking speeds and those using walking aids.

Purpose of the Study:

• To develop, implement, and compare original, improved, and new algorithms for estimating gait parameters: cadence, step length, and walking speed.
• To conduct a comprehensive cross-validation of these algorithms under various conditions, including slow, normal, and fast walking, as well as the use of walking aids.
• To evaluate algorithm performance using two distinct datasets with reference data from instrumented mats and shank-worn sensors.

Main Methods:

• Implementation of multiple algorithms for gait parameter estimation.
• Cross-validation using two datasets: 40 subjects with an instrumented mat and 88 subjects with shank-worn inertial sensors.
• Testing algorithms on data from individuals walking at slow, normal, and fast paces, and with the use of walking aids.

Main Results:

• Significant performance improvements of up to 50% were observed with the developed algorithms.
• Training algorithms on diverse data, including individuals with different preferred speeds, enhanced overall performance.
• For slow, normal, and fast walkers, the root mean square errors (RMSE) for speed estimation were 0.10 m/s, 0.18 m/s, and 0.15 m/s, respectively.
• For individuals using walking aids, the RMSE for speed estimation increased to 0.32 m/s, indicating a greater challenge in estimation for this group.

Conclusions:

• The proposed combined approach for speed estimation demonstrates robustness across different walking speeds.
• The algorithms show promising results for slow, normal, and fast walkers, with acceptable error margins.
• While effective for most, the estimation accuracy decreases for individuals using walking aids, highlighting an area for future refinement.