Motor synergy is considered as a motor control strategy deployed by the central nervous system (CNS), and it can be altered due to ageing, disease and injury. A timely assessment and analysis of altered motor synergy patterns will be helpful for the motor rehabilitation process. However, current research has mainly focused on the implementation of automated assessment scales. While the mechanism of the motor synergy structure alteration is not well understood yet. In this study, we proposed an approach to the analysis of altered human motor synergistic structures. By collecting and preprocessing the 3-dimensional motion data from 30 participants (including 15 stroke patients and 15 healthy individuals), synergistic structure features were extracted. We obtain the spatiotemporal vectors of motion by the nonnegative matrix factorization. These vectors were clustered using K-means and matched with the scalar product. The similarity and specificity clustering pairs were obtained through Kuhn-Munkres Algorithm. The above results revealed that the structure of human motor synergy was greatly altered after stroke, and some new synergistic patterns with commonalities emerged during patients' movements. This study presents a new method to identify specific patterns of motor synergy arising from disease-altered biomechanics and central nervous system, providing new targeted protocols for rehabilitation assessment.