Silhouette Analysis of Hand Gesture Dataset Using Histogram Profile Feature Extraction

Abstract

Hand gesture dataset is a collection of hand gesture images. Several hand gesture datasets are freely available and can be used for various purposes, such as comparison or method testing. Processing the distribution of hand gesture image quality in the dataset has the opportunity to find potential models of hand gesture image quality for further research. This study tries to provide answers by exploring the quality of hand gesture images based on various datasets of public hand gestures. Then perform feature extraction based on the image histogram profile to get an overview of the range of color intensity values from the hand gesture image. The Herarchical Clustering method is used to build clusters based on histogram characteristics. The feasibility of the relationship between clusters was tested based on the silhouette index clustering method. The total number of hand gesture test images is 16 thousand data taken from 6 dataset sources that have been used in hand gesture recognition research. Based on the results of the processing, it is shown that the three clusters have no relationship feasibility or in other words the image clusters are independent.