Jaringan Saraf Tiruan Backpropagation untuk Pengenalan Wajah Metode Ekstraksi Fitur Berbasis Histogram

Sigit Kusmaryanto


One common weakness in pattern recognition for face recognition is imperative that accurate input pattern to the pattern identified . This causes the input is often not recognized or not identified so as to be in the input repeatedly . The application of face recognition using Artificial Neural Network (ANN) backpropagation with MATLAB 7.0 is expected to overcome the weakness of pattern recognition systems for face recognition . Histogram -based feature extraction methods used in this study to obtain identification characteristics of the face image and a neural network input data . Face image data using pixel size variations . Trained in facial image pixel size variations 640 x 480 pixels and 600 x 800 pixels with two of distance making on face: average( 2-5m ) , close( < 2m ) . The results obtained from ANN test using 18 images with frontal face : resulting number of units in the hidden layer 6 , the number of input unit 255 , the number of output unit 10 , the maximum epoch 2500, 0001 and learning the target error rate = 0.9 with a percentage of 95 % of face recognition

Keyword - Backpropagation , Face Recognition Feature Extraction , Histogram .

Full Text:



  • There are currently no refbacks.