Normalization Technique

[amit deshmukh]

Hi friends,

I am currently working on Feature Level Fusion of Face and Fingerprint Modalities.

I have used Gabor wavelet (at 8 orientations and 5 scales) to extract features. Later I have used PCA and LDA to facilitate better inter-class separation and dimension reduction.

Which technique should be used to normalize feature vectors (as they are from heterogeneous sources) for feature level fusion?

When should the normalization be applied  before fusion (simple concatenation) of feature vectors or after fusion of feature vectors?

Looking for some guidance in this regard...

Thank You