Abstract : Human motions are now frequently used in several applications ranging from computer animation to biomechanics analysis. Acquiring such data can be performed in several ways, but none of them is fully accurate and usually a de-noising process is required as post-treatment. This paper presents a new method to process motion data that tends to preserve some characteristic features of human motions. It is based on an adaptation of the well-known bilateral filter to orientation data. We give an algorithm that computes the filter response, and we show practical results obtained on real motion capture data.