Facial recognition is the process of human identification using a picture of facial expression. With the widespread use of computers, it is expected that facial recognition capabilities can be adopted on such smart devices. The adoption process becomes possible with the discovery of facial recognition methods, one of which is the main component analysis or better known as PCA (Principal Components Analysis). The research started by designing a computer program using the Matlab programming language. The Program was used to test the PCA method using a number of facial imagery. Testing is divided into three categories, which are based on the number of the trainer image, based on the number of key vector features, and the determination of the threshold value. In the end it can be concluded that PCA is quite worthy to be a facial recognition method. The research Data shows a pretty good introduction result with a fairly small error rate on testing using ten training imagery, which is one error introduction of 20 Tests.