Gender classification of running subjects using full-body kinematics

Abstract

This paper proposes novel automated gender classification of subjects while engaged in running activity. The machine learning techniques include preprocessing steps using principal component analysis followed by classification with linear discriminant analysis, and nonlinear support vector machines, and decision-stump with AdaBoost. The dataset consists of 49 subjects (25 males, 24 females, 2 trials each) all equipped with approximately 80 retroreflective markers. The trials are reflective of the subject's entire body moving unrestrained through a capture volume at a self-selected running speed, thus producing highly realistic data. The classification accuracy using leave-one-out cross validation for the 49 subjects is improved from 66.33% using linear discriminant analysis to 86.74% using the nonlinear support vector machine. Results are further improved to 87.76% by means of implementing a nonlinear decision stump with AdaBoost classifier. The experimental findings suggest that the linear classification approaches are inadequate in classifying gender for a large dataset with subjects running in a moderately uninhibited environment.

Transfer learning using infrared and optical full motion video data for gender classification

“…In [33] several handcrafted features are compared for human identification. Gait analysis is performed by other authors using motion capture data for human subjects [1,5,34]. We demonstrate in [2] that using motion capture type skeleton representation for feature extraction combined with occlusion completion improves human identification performance with our small in house LIDAR data, in comparison to existing human identification work.…”

“…The ability to classify subject gender is one of many tasks that can help a real-world security effort to recognize or analyze human identity and behavior. This work continues an effort by our lab [1][2][3][4][5] to analyze human subject data using special sensors including MoCap and LIDAR.…”

“…The ODU Vision Lab has ongoing work in gender classification and person classification using optical and infrared special purpose sensors [1][2][3][4][5]. Infrared sensors compliment the abilities of visual range sensors for human subject analysis including face recognition [6][7][8][9], action recognition [10][11][12][13][14], and gender [15][16][17][18][19][20] and identity [20][21][22][23][24][25][26] recognition.…”

“…Our group began this project using motion capture (MoCap) data collected at the Air Force Research Lab. Our initial work used feature engineering to capture human motion gait signatures in walking data that are used to classify gender in walking and running subjects [1,5]. Using kinematic features, this work was extended for gender classification in asymmetric limp and aperiodic throwing actions [4].…”

Joint Swing Energy for Skeleton-Based Gender Classification

“…In this case, all subjects are both in the training and testing data. If a machine learining algorithm is trained using this training data and tested on the testing data, the algorithm may artificially achieve a high performance [52], [57], [63], [64]. To avoid this situation, in this study, we used LOPO cross validation.…”