After studying how to select and represent Gabor features of human face, how to combine the recognition results of different Gabor channels, a face recognition method based on multichannel Gaborface representation (MGFR) was proposed. Firstly, feature areas and calculated weight in each channel were selected according to class separability of Gabor features. Then, fuzzy weighted rule was used to combine all the recognition results of different channels. This method can reduce the effect of redundancy information. Meanwhile, it takes advantage of the difference between channels and solves the boundary classification problem. Experiments on AR, CAS-PEAL-R1 and YaleB show that the proposed method has a higher recognition rate than the traditional MGFR. Simultaneously, it was compared with ensemble Gaborface representation (EGFR), which shows it has a great advantage in average recognition time.