if you want a local features, you can divide the face into 6 regions vertically. then you will have a region around eyes and another around mouth. If not, you can use landmarks detection then extract features around certain landmarks. 

You can use facial proportions using horizontal thirds that can easily isolate eye, nose, lip easily. Then use any landmark detection technique and extract expressive facial features

You can use the shape features (ASM)

see :

http://cmp.felk.cvut.cz/~uricamic/clandmark/

Traditionally, Local Binary Patterns (LBP) has been successfully used as features for facial expression recognition. It is very important to extract LBP from regions in a grid over the face image and concatenate them in a single vector.