Kinematic data measured via a motion capture system tends to be contaminated by noise. In order to obtain velocities and accelerations, kinematic data is usually numerically differentiated using finite difference method. The differentiation process could amplify those high frequency components in the data, which are probably from noise, and consequently, filtering out the high frequency components is preferred before differentiation takes place. This paper explores the application of ensemble empirical mode decomposition (EEMD), a newly-emerged, noise-assisted data analysis method, to smoothing kinematic data. The EEMD based smoothing technique is first introduced, and its efficiency and applicability are then evaluated using typical kinematic data from the literature and also data from the experiments conducted by the authors in order to measure walking induced load. Results prove that the EEMD based filtering technique is efficient in smoothing kinematic data and is suitable for automatic batch data processing since the selection of filtering parameters is not difficult. Compared with the commonly used Butterworth filter, the proposed filtering technique avoids choosing a cut-off frequency and filter order, a proper selection of which highly depends on the user's experience. Compared with the wavelet-based filter, the suggested technique is adaptive since it does not require a base function as the mother wavelet in the wavelet-based filter. Therefore, the EEMD-based filtering technique is applicable to kinematic data with various characteristics.