3D Motion Trajectories Recognition using Optimal Set of Geometric Primitives, Angular and Statistical Features
Deval Verma1, Himanshu Agarwal2, A. K. Aggarwal3
1Deval Verma*, Mathematics Department, Jaypee Institute of Information Technology, Noida, India.
2Himanshu Agarwal, Mathematics Department, Jaypee Institute of Information Technology, Noida, India.
3A. K. Aggarwal, Mathematics Department, Jaypee Institute of Information Technology, Noida, India.
Manuscript received on November 15, 2019. | Revised Manuscript received on 26 November, 2019. | Manuscript published on December 10, 2019. | PP: 2972-2979 | Volume-9 Issue-2, December 2019. | Retrieval Number: B7594129219/2019©BEIESP | DOI: 10.35940/ijitee.B7594.129219
Open Access | Ethics and Policies | Cite | Mendeley | Indexing and Abstracting
© The Authors. Blue Eyes Intelligence Engineering and Sciences Publication (BEIESP). This is an open access article under the CC-BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/)
Abstract: This paper presents the 3D motion trajectories (lower case 3D alphabetic characters) recognition using optimal set of geometric primitives, angular and statistical features. It has been observed that the different combinations of these features have not been used in the literature for recognition of 3D characters. The standard dataset named “CHAR3D” has been used for analysis purpose. The dataset consists of 2858 character samples and each character sample is 3 dimensional pen tip velocity trajectory. In this dataset only single pen down segmented characters have been considered. The recognition has been performed using Random Forest (RF) and multiclass support vector machine (SVM) classifier on the optimal subset of extracted features. The best obtained recognition accuracy of 83.4% has been recorded using 3D points, angular and statistical features at 10 fold cross validation using SVM classifier. Moreover, the highest recognition accuracy of 96.88% has been recorded using an optimal subset of 32 dimensional features namely, geometric primitives, angular and statistical features at 10 fold cross validation by RF classifier.
Keywords: Random Forest, CHAR3D, Orthocenter, Curvature, Primitive Features, SVM, Statistical features.
Scope of the Article: Approximation and Randomized Algorithms