Taguchi-TOPSIS based HOG parameter selection for complex background sign language recognition

Abstract

This paper presents an approach to design Indian Sign Language (ISL) recognition system for complex background. In many applications, Histogram of Oriented Gradients (HOG) have been proved to be effective. However, it is observed that the choice of HOG parameters affects the feature vector size and its classification capability. The objective is to select the parameter values in order to have maximal accuracy at a minimal computational time and reduced feature vector size. A combined Taguchi and Technique for Order of Preference by Similarity to Ideal Solution (TOPSIS) based decision-making technique is applied to determine the values of these parameters. Results show that the combined TOPSIS-Taguchi based technique is effective in selecting the parameter combination to get high overall performance. For the acquired ISL complex background dataset, the selected values of parameters are further used to obtain multi-level HOG resulting in the overall accuracy of 92% for 280 features.