• P-ISSN 0974-6846 E-ISSN 0974-5645

Indian Journal of Science and Technology


Indian Journal of Science and Technology

Year: 2017, Volume: 10, Issue: 12, Pages: 1-8

Original Article

Effects of Different Color Models in Hand Gesture Recognition


Objectives: This paper presents a comparative analysis between different color models (YCrCb, YUV, HSV, CIE L*a*b*, and CIE L*u*v*) for hand gesture recognition task by extracting skin-color of index 31-35 on Von Luschan’s chromatic scale. Methods/Statistical Analysis: An eight step algorithm is proposed for comparing different color models, which is simple so as to run in real-time and is kept same for all the color models. Only the threshold for extracting skin-color is different for all the color models to extract the desired region. Morphology operations are used to remove noise and Contour and Convex Hull technique is used to extract desired features. Test data was generated using 16 users and each user performed 2 sequences of 6 different gestures for every color model giving a total of 5760 test data points. Accuracy Percentage, False Positive Rate (FPR) and True Positive Rate (TPR) are used to evaluate color models. Findings: Since the options are several, it becomes crucial which color space to use. In this paper, we have analyzed which color space can most accurately preserve the hand shape and recognize it for skin color of index 31-35 on Von Luschan’s chromatic scale. Experimental results have shown that the CIE L*a*b* and YCrCb color space are the most versatile in different environment settings with 73.9% and 71.5% accuracy, respectively. CIE L*a*b* and CIE L*u*v* are found to have highest TPR and HSV had lowest FPR. Whereas YCrCb and YUV are better than HSV in terms of TPR and not so good compared to CIE L*a*b* and CIE L*u*v*. Also YCrCb and YUV have lesser FPR than CIE L*a*b* and CIE L*u*v* but higher than HSV. These findings will help the researchers of the field of Human-Computer Interaction to select appropriate color-model for specific applications. This paper proposed an algorithm that can be improved to include complex gestures for evaluation. Improvements: Algorithm developed for evaluation can be modified to use machine learning techniques to allow complex gestures and can be accommodated according to different color models. Also other color models can also be included to perform exhaustive evaluation.

Keywords: Color Models, Color Spaces, Hand Gesture Recognition, Human Computer Interaction, Skin-Color


Subscribe now for latest articles and news.