Dear Omer, 

What about using the Curvelet Transform, it more powerful technique compared to the Wavelet transform, especially, in capturing the edges and curves, which I believe they are important for recognition.

https://www.google.com.tr/url?sa=t&source=web&rct=j&url=http://www.dcs.warwick.ac.uk/~nasir/papers/harem05.pdf&ved=0ahUKEwidi8_KtOjSAhWCDZoKHYfOCG8QFggeMAA&usg=AFQjCNHTj-E3Aedfi_A5xWyezdHOvfExsQ&sig2=hujasyzt_udUYqVSyYYAXA

https://www.google.com.tr/url?sa=t&source=web&rct=j&url=http://ieeexplore.ieee.org/document/4492604/&ved=0ahUKEwidi8_KtOjSAhWCDZoKHYfOCG8QFggoMAI&usg=AFQjCNENmMjv8w6uufLnVCx_IQPYfIKtRg&sig2=8Vj5i8q_-b1WmV2MZyKeog

I think higher order Daubechies wavelets, since as the order of the DB wavelets increases, the capability to detect events and hence discriminant features increases. 

hello sir,

    In  any researach  it is the work of scholar to investigate and prove which method/tool /algorithm is suitable for a particular appication ,Hence it is better to do experimental investigation on all wavelets and infer about suitability. It is theme of yr work itself. all the best and try all wavelets ....

Try using Haar wavelet.

It achieves better results in face recognition.

thanks