Gait based human identification: a comparative analysis

摘要: Thanks to gait analysis, many examinations such as person identification, disease detection, and evaluation of neuromusculoskeletal system functions can be performed. In the study, the used dataset includes three different gait parameters obtained from 16 different individuals (7 females and 9 males) using wearable gait analysis sensors, and here there are 321 parameters for one gait of each person. In addition, we classify this data using Linear Discriminant, Ensemble Subspace Discriminant, Ensemble Bagged Trees, Optimizable Ensemble-1, and Optimizable Ensemble-2 classifiers. Two different optimization techniques were employed to increase the performance metrics of the classifiers. From the results, it is seen that the Accuracy (%), Error (%), Sensitivity (%), Specificity (%), Precision (%), F1 Score (%), and Matthews Correlation Coefficient (MCC) of Optimizable Ensemble-2 that is the most successful classifier are equal to 97.92, 2.08, 97.92, 99.86, 98.44, 97.86, and 0.9790, respectively.