In order to develop truly intelligent systems, it is necessary to improve their ability to understand non-verbal communication. We propose a novel framework to recognize individuals and emotions from gait, in order to improve HRI. We collected the motion data of the torso from 4 professional actors' gait, using motion capture system, and 7 non-actors' using 2 IMU sensors. We developed Feature Value Method which is a PCA based classifier and finally we achieved high recognition rate through cross-validation. © 2012 Springer-Verlag.