Human gait has been proven to be of importance when trying to recognize people. In addition gait also conveys the emotional state of someone. The present study propose to objectively and systematically analyze gait data to highlight features that can characterize someone and the emotion conveyed. Rather than using gait stance phase, frequency, footstep length... we use the inverse kinematics data computed from the motion-capture data using a 34 degree of freedom human body model. Then we compute a similarity criteria with respect to a reference motion. We first utilize the 6 components of the base-link velocity for the similarity criteria computation. The motion data are collected on 4 candidates (2 males and 2 females professional actors), 4 emotional states are simulated: neutral, happy, angry, sad. Each is repeated 5 times. The experimental results show that using the gait characteristics it is possible to characterize each candidate and to characterize each emotional state with a good accuracy.