Imitation of hand gestures classified by Principal Component Analysis with fluid particles

In imitation learning, the correspondence problem, which is caused by the dynamical differences between imitator and the demonstrator, is applied on imitation of human hand gestures with a particle colony which is modeled by using fluid dynamics. In this work, firstly human hand gestures which are collected from different human groups are classified by using Principal Component Analysis (PCA) and then these hand gestures are imitated with fluid particles which are modeled by Smoothing Particle Hydrodynamics (SPH). Since the dimension of the collected hand images are huge and different from each other, PCA is used for dimension reduction and also dimension equalization. For classification of the test images, the distance between weight matrices of the training data set and test data set is measured. After that the fluid parameters which belongs to that class are applied to the particle colony and finally human hand gestures are imitated. Moreover, in this work, the effect of the number of principal components on the classification and average consumed time during training and testing steps are analyzed.