Subspace based object recognition using support vector machines

In this paper, we propose an object recognition technique using higher order statistics without the combinatorial explosion of time and memory complexity. The proposed technique is a fusion of two popular algorithms in the literature, Independent Component Analysis (ICA) and Support Vector Machines (SVM). We propose to use ICA to reduce the redundancy in the images and obtain some feature vectors for every image which has lower dimensions and then make use of SVM to classify these feature vectors coming from the ICA step. Experimental results are shown for Coil-20 and an internally created database of 2D manufacturing objects. Comparative analysis of independent component analysis and principal component analysis (PCA) is also given for each experiment.