Higher-Order Local Autocorrelation Feature Extraction Methodology for Hand Gestures Recognition

A novel feature extraction method for hand gesture recognition from sequences of image frames is described and tested. The proposed method employs higher order local autocorrelation (HLAC) features for feature extraction. The features are extracted using different masks from Grey-scale images for characterising hands image texture with respect to the possible position, and the product of the pixels marked in white. Then features with the most useful information are selected based on mutual information quotient (MIQ). Multiple linear discriminant analysis (LDA) classifier is adopted to classify different hand gestures. Experiments on the NUS dataset illustrate that the HLAC is efficient for hand gesture recognition compared with other feature extraction methods.

I. Introduction

Hand gesture is a natural way of communication between humans. It is as much about nonverbal cues as it is about talking. In reality, people depend on hand gestures and body language when in situations where they are unable to verbally interact with someone. Therefore, analysis and automatic recognition of hand gesture improves human-computer interaction. Prior to this work, there many studies have been conducted in the field of hand gesture recognition [1]–[5]. Mostly hand gesture recognition systems consist of two important stages; feature extraction and feature selection. The feature extraction is very crucial in hand gesture classification. The process of feature extraction results into a proscribed large number of features and subsequently a smaller sub-set of features need to be selected accordingly to some optimality criteria. Selecting the number of features as low as possible is important for accurate classification. In order to avoid using all available variables features in the data, one selectively chooses a subset of features that will be used in the discriminant system [6]. The main role of feature extraction and feature selection is to reduce dimensionality of data and keeping the number of features as low as possible this will automatically minimize training time and increase classification accuracy.