A New Design of Experimental Schemes of Complex Matrix LDA

Two-Dimension Linear Discriminant Analysis (2DLDA) and complex-matrix LDA are two noticeable improvements to conventional LDA. They can achieve a good performance respectively. However, the complex-matrix LDA is very suitable for bimodal biometrics. In this paper, we indicate the two available implementation procedures of complex-matrix, i.e. the original implementation procedure and one alternative implementation procedure proposed in this work. The following new implementation scheme for complex-matrix LDA would be designed: Firstly, the two available implementation procedures of complex-matrix LDA will be carried out and then their results to perform face recognition using the matching score fusion algorithm integrated. A higher classification accuracy indicates the proposed implementation scheme more useful than both of the two available implementation procedures of complex-matrix LDA.