Investigation of motor unit (MU) activity provides a fundamental information of neuromuscular control. High density surface electromyogram (HDsEMG) decomposition based on the blind source separation techniques (BSS) has been performed to estimate discharge patterns of MUs. The decomposition techniques for HDsEMG signals were initially proposed for isometric contractions. Recently, the decomposition methods have been utilized for non-isometric muscle contractions. However, what level of dynamic muscle contraction is acceptable for the EMG decomposition techniques is still an open research problem. Thus, in the present study, we investigated the robustness of the EMG decomposition method in dynamic muscle contractions in the tibialis anterior (TA) muscle using a validation method of EMG decomposition by convolving the synthetic MU spike trains with experimentally identified MU action potentials (MUAPs), which change depending on the ankle joint angle. We found that the decomposition accuracy for dynamic contractions with ankle range of motion (ROM) up to 20° was comparable to that during isometric contraction. When ankle ROM was larger than 20°, discontinuously identified (i.e., misidentified) spike timings of MUs due to joint angle changes was significantly larger than those during isometric contraction. However, rate of misidentification due to ankle joint angle change was still lower than 5% up to ankle ROM of 30°. Additionally, we found that increase of ankle ROM decreased total number of identified MUs. Based on the results, the conventional EMG decomposition method is probably applicable for dynamic contractions with ankle ROM of 30° in TA.