|Face recognition is one of challenging research topics in computer science because human face is difficult to be modelled. In this research, SOM and LVQ are proposed for frontal face recognition. The purpose is to compare LVQ and SOM based on identification accuracy. Training uses 400 images from 20 different individuals, and the dimension is 180 x 200 pixels. The data are retrieved from University of Essex, UK. Coefficient approximation at Haar wavelet level 6 is used as feature for classification and clustering process. K-fold cross validation with 10-fold is used to divide training and testing data. The experiment is divided into 3 sets, i.e., the experiment using SOM, LVQ, and LVQ initialized by SOM. The highest accuracy achieved by SOM is 97.8974%, while both LVQ and LVQ initialized by SOM achieve 100% accuracy. Based on the accuracy, LVQ proves to be better than SOM for frontal face recognition. This research needs to be improved in order to recognize various poses and changing expressions.