With the rapid development of computer vision applications, infrared face recognition is receiving much attention because of miniscule sensitivity to the face variations due to illumination changes. The infrared image is merely a gray level image but the recognition is interestingly good enough to look at. In this article we present an infrared face recognition system where we used two methods and two databases to train and test our system. We used the Histogram of Gradient (HOG) algorithm along with the Support Vector Machine (SVM) classifier as a first method. In the second method, we implemented the Backpropagation algorithm. Although the results of the HOG-SVM were very promising and we were able to get a better results compared with a previous work, neural network gave us a perfect results by using the same number of training examples.