In order to improve the recognition rate, efficiency and robustness of face images, a robust sparse linear discriminant analysis method based on Principal Component Analysis (PCA) and support vector machine (SVM) is proposed. The proposed method is compared with linear discriminant analysis, robust linear discriminant analysis, robust linear discriminant analysis based on the Bhattacharyya error bound and -norm, robust adaptive linear discriminant analysis and robust sparse linear discriminant analysis through the ORL and YaleB database, COIL20 object database and some datasets in UCI machine learning database. The experimental results show that under the original image condition, the average recognition rate of this method is 92.80%, 97.76% and 89.61% on the ORL database, COIL20 object image database and some datasets in UCI machine learning database, which is higher than the other five methods. Under the condition of adding salt and pepper noise to the YaleB database, the average recognition rate of proposed method is 81.35%, which is more than 1.37% higher than the other five methods.