A New Hardware Friendly Vector Distance Evaluation Function for Vector Classifiers

Abstract

This paper proposes a new vector distance evaluation function for vector classifications. The proposed distance evaluation function is the weighted sum of the differences between vector elements. The weight values are determined according to whether the input vector element is in the neighborhood of the prototype vector element or not. If the element is not within the neighborhood, then the weight is selected so that the distance measure is less significant The proposed distance measure is applied to a hardware vector classifier system and its feasibility is verified by simulations and circuit size evaluation. These simulations and evaluations reveal that the performance of the classifier with the proposed method is better than that of the Manhattan distance classifier and slightly inferior to Gaussian classifier. While providing respectable performance on the classification, the evaluation function can be easily implemented in hardware.