EMG muscle electricity sensor, 6 ~ 10 electrode arm, if the goal is to identify gestures, identification of finger movement (naruto dozen of fingerprint, for instance, and the deaf sign language what,,)

Alone myoelectricity identification sign language seems to be more difficult, the muscles of the fingers not much,, as far as possible improve the success rate of recognition, plus bending sensor,

What kind of neural network algorithm with high recognition rate? This is a classification problem