Hybrid learning vector quantization (LVQ) algorithm on face recognition using webcam

2nd Nommensen International Conference on Technology and Engineering

Parini^1 ; Mawengkang, Herman^2 ; Efendi, Syahril^3

Stud. in Faculty of Computer Science and Information Technology, Universitas Sumatera Utara, Medan, Indonesia^1

Faculty of Mathematics and Natural Sciences, Universitas Sumatera Utara, Medan, Indonesia^2

Faculty of Computer Science and Information Technology, Universitas Sumatera Utara, Medan, Indonesia^3

关键词: Hybrid algorithms;    Hybrid learning;    Hybrid techniques;    Kohonen algorithm;    Learning data;    Learning vector quantization algorithms;    Pre-processing step;    Sobel operator;   

Others  :  https://iopscience.iop.org/article/10.1088/1757-899X/420/1/012126/pdf
DOI  :  10.1088/1757-899X/420/1/012126

来源: IOP

【 摘 要 】

In this research, hybridization of Learning Vector Quantization (LVQ) algorithm with Self Organizing Kohonen for face recognition using webcam. Hybrid technique performed is Kohonen algorithm used for initial weighting and the result of weighting inserted into LVQ algorithm to get result of training in the form of final weight used for face recognition. The data used in this study is the face of a digital image of the acquisition with a digital tool make use of camera that will be used for learning (learning data set) and a set of images for testing (testing data set).The first step of the process the face image (preprocessing) used for prepare the input face image to be input into network. The preprocessing step in this research is divided into four step like Image Readings, Grayscaling, Sobel operator edge detection and binaryization. The result of this test is percentage of face recognition success with LVQ algorithm is 57.03%, Kohonen is 52.59% and Hybrid is 68.88%.While the average time of the introduction process is for LVQ of 2.64 seconds, Kohonen of 2.61 seconds and Hybrid 2.59 seconds. From the above results can be concluded that in terms of accuracy Hybrid algorithm slightly superior to the Kohonen is 16.29% and in terms of time Hybrid algorithm faster than LVQ algorithm of 0.05 seconds and Kohonen an average of 0.02 seconds.