Figure 3

The basic architecture of semi-parameterized SV-based local models. The centroid vectors from voronoi region for each training sample x _i used in the SVM decision function. SVM is considered as a purely non-parametric model, whereas SV-HMM is considered as semi-parametric model as it adopts the method of grouping the associated input vectors in each class i. RBF kernel has been used for the SV-based local models.