Research on Face Recognition Based on Ensemble Learning

In order to improve the recognition performance of existing face recognition algorithms, the algorithm of face recognition based on ensemble learning was proposed. First of all, the face library image was down sampled and the number of the face library was changed form 1 to 5. Then the global features of PCA, the local features of LBP and the visual features were extracted and identified for each library, and 15 results of recognition were obtained. Nextly the final recognition result was obtained by using the majority voting algorithm in ensemble learning. The hold process could be processed in parallel in addition to the image down sampling and the majority voting integration. Lastly the algorithms before and after integration were compared and verified in ORL face database, and the results shown that the recognition rate based on the ensemble face recognition algorithm is 97.65% when the number of samples for each person is 5, which is 9.46%, 4.73% and 2.96% higher than that of PCA, LBP and GIST respectively. Under the single training set, the recognition rate reaches 81.5%, which is increased by 23.11% in maximum and by 10.88% in minimum comparing with the single algorithm before integration. It can be seen that the proposed face recognition algorithm is effective and feasible, and it has strong classification recognition effect.