To improve the face recognition rate, this paper proposes an incremental learning support vector machine (SVM) face recognition scheme to update the parameters of SVM efficiently. The proposed scheme adopts the Gaussian probability model to depict the parameters of SVM, and updates the parameters of SVM based on the incremental learning SVM without saving the training data. The proposed scheme also employs the rule that minimizes the error of classifications to maximize the distance of the output distributions of two classes. The detailed experimental results and comparisons with the existing schemes show that the proposed scheme can obtain better recognition performance.