Fusion of Face and speech for Multi-modal Person Identification System
Keywords:
Eigen faces, Euclidean distance measure, fusion technique, Principal component analysis, singular spectrum analysisAbstract
A proposed new technique for person identification using fusion of both face and speech which
can essentially improve the recognition rate as compared to the single biometric human identification. The proposed
system uses Principal component analysis technique for face feature extraction. The PCA calculates the eigen vectors
and eigen values which are used in fusion. The Singular spectrum analysis is used to extract speech features and the
values of power spectrum are used in the fusion. The fusion of face and speech is done by simple sum rule fusion
technique and normalization of feature values are done before the fusion. Person identification is depending on the
fused results so that a Euclidean distance measures is used to find the variations in fused results.
