A method for accurately estimating chest velocity in air bag deployment tests is presented. It employs a state model of chest motion and a Kalman Filter to combine experimental acceleration and displacement measurements to estimate the velocity in a maximum likelihood sense. This technique uses statistics of the experimental and modelling noise to effectively combine the results of both types of measurements in making an accurate estimate of velocity. Results from highly demanding chest injury assessment tests, out-of-position air bag deployments, are presented. These results indicate that the problems associated with attempting to obtain velocity from either acceleration or displacement alone can be significantly reduced. Methods for estimating Kalman Filter parameters and assessing accuracy are presented. Accuracy comparisons are made between the Kalman Filter technique and SAE Recommended Practices for velocity estimation.