Classification of Human Postural and Gestural Movements Using Center of Pressure Parameters Derived From Force Platforms

MOspace/Manakin Repository

Breadcrumbs Navigation

Classification of Human Postural and Gestural Movements Using Center of Pressure Parameters Derived From Force Platforms

Please use this identifier to cite or link to this item: http://hdl.handle.net/10355/9613

[+] show full item record


Title: Classification of Human Postural and Gestural Movements Using Center of Pressure Parameters Derived From Force Platforms
Author: Saripalle, Sashi K. (Sashi Kanth)
Keywords: Pressure measurement
Date: 2011-01-20
2010
Publisher: University of Missouri--Kansas City
Abstract: The human body, while standing, can be imagined as a complex feedback system that produces continuous sway patterns. Subtle body movements that can be caused by sensory cues such as visual or auditory, affective, cognitive, pathological or many other factors besides intended movements can be easily captured in the sway patterns derived from ground reaction forces and the body's center of pressure (COP). The purpose of this research is to classify human body movements, even the subtle movements, using a carefully selected feature set. For the first time, we propose a method to classify postural and gestural movements using data from force platforms collected from participants performing 11 choreographed movements. Twenty-three different displacement and frequency based features were initially extracted from COP time series, and ranking and wrapper methods were used for classification-guided feature extraction. Linear classifiers such as Fisher's Linear Discriminant analysis classifier and nonlinear classifiers such as nearest neighbor classifiers, support vector machines (SVM), and neural networks were explored and successfully applied to the aforementioned movement classification. The average classification rates on test sets ranged from approximately 79% to 92%. All the methods proposed in this experiment performed well by themselves over at least one movement type, but none could outperform the others for all movement types and therefore a set of movement-specific features and classifiers is proposed.
URI: http://hdl.handle.net/10355/9613

This item appears in the following Collection(s)

[+] show full item record