In this paper, a person is identified with face as a biometric feature using modified fisher face and fuzzy fisher face. The premise of this paper is to introduce modified fisher face, fuzzy fisher face and include gradual level of assignment to class being regarded as a membership grade which helps to improve recognition results. Performance of the said system is compared with traditional fisher face methods using Fisher's linear discriminate analysis. For all these methods of face recognition an assumption was made about the same level of relevance of each face to the corresponding category. The comprehensive experiments completed on ORL, and Yale face databases show improved recognition rates and reduced sensitivity to variations between face images caused by changes in illumination and viewing directions.