Gesture recognition for robot teleoperation
Abstract
Interactive robotics is a vast and expanding research
field. Interactions must be sufficiently natural, with robots
having socially acceptable behavior by humans, adaptable to user
expectations, thus allowing easy integration in our daily lives in
various fields (science, industry, domestic, health : : : ). To make
such interaction we choose gestures as a way of communication.
Human gestures are certainly natural and flexible. In this
context we developed a robust upper body gesture recognition
system in order to teleoperate in the future a humanoid robot.
Gestures are performed by Kinect camera for skeleton detection
and tracking. A robust descriptor vector is chosen to describe
gestures, named BSM feature vector which can be represent
three important aspects, The connexion between different Body
parts, the Shape changing during gesture and describe gesture
Motion in the space. Three best-known and successful learning
methods used for training and gesture classification, Random
forest classification, Support vector machine and Multi layer
Perceptron. The proposed method has been evaluated on two
public benchmarks, the Microsoft Research Cambrige (MSRC-
12), and MSR Action3D datasets. The results obtained showed
that the proposed recognition system is more relevant than the
traditional methods.
Origin : Files produced by the author(s)
Loading...