Feature Extraction and Application via Self-Organizing Map Neural Network with Operating Preprocess

• Main Authors: Kao-Chi Lin, 林高吉
• Other Authors: Chih-Peng Huang
• Format: Others
• Language: zh-TW
• Published: 2007
• Online Access: http://ndltd.ncl.edu.tw/handle/61960300601010668325

碩士 === 嶺東科技大學 === 資訊科技應用研究所 === 95 === This dissertation mainly investigates feature extraction and application via self-organizing map (SOM) neural network with operating preprocess. Adopting SOM neural network for given samples, we can obtain a geometric map of the trained neurons with its resemblance. Then, for a group of test samples, the trained neurons are applied for clustering or classifying. The image patterns of number and English characters are used as our experiment’s input samples. First, they are transform to binary bitmap with a given threshold value. From the proposed SOM neural network with operating preprocess, we can obtain the characteristic neurons. For the testing patterns, the trained neurons associated with the learning vector quantization (LVQ) method are used for classifying. Thus, the SOM neural network with distinct operating preprocess are compared and analyzed.