Automatic gender and unilateral load state recognition for biometric purposes

Area of Science:

• Biomechanics
• Biometrics
• Machine Learning

Background:

• Human gait analysis is crucial for medical, rehabilitation, sports, and biometric applications.
• Automatic recognition of gender and load state from gait is an ongoing research area.

Purpose of the Study:

• To utilize ground reaction forces (GRF) during gait for recognizing gender and unilateral load state.
• To evaluate the effectiveness of various parameters and classification algorithms for gait analysis.

Main Methods:

• Calculated GRF parameters including mean, variance, standard deviation, peak-to-peak amplitude, skewness, kurtosis, and Hurst exponent.
• Employed classification algorithms: k-Nearest Neighbors (kNN), artificial neural networks, decision trees, and random forests.
• Collected data from 214 individuals using Kistler force plates, recording 7,316 gait cycles.

Main Results:

• The kNN classifier achieved 99.37% accuracy for gender recognition.
• Unilateral load state was recognized with 95.74% accuracy.
• Combined gender and load state recognition reached 95.31% accuracy, outperforming existing methods.

Conclusions:

• The study demonstrates effective automatic recognition of gender and asymmetrical load using GRF parameters and the kNN algorithm.
• The presented method shows promise as an initial stage in biometric identification systems.
• Gait analysis provides valuable data for understanding human movement characteristics.