After a spinal cord injury (SCI), some motor neurons can still be functional below the level of the injury and voluntarily controlled. The changes in motor neuron behavior that allow that are not well understood. In this study, we investigated the residual voluntary activity in chronic complete SCI using high-density surface electromyography (HDsEMG). We recorded HDsEMG signals from two groups, SCI (n=8) with no hand function and control (n=12), non-injured young adults. All the participants signed a written informed consent approved by the ethics committee of the Friedrich-Alexander-Universität Erlangen-Nürnberg, Germany (22-150B and 22-138Bm). All the procedures accorded with the Declaration of Helsinki. We placed grids of electrodes around the forearm and wrist of the participants and recorded HDsEMG signals while they were asked to attempt eight different hand tasks shown by a virtual hand (individual fingers, power grasp, two- and three-finger pinches).  By decomposing these EMG signals, we could compare motor units' neural and peripheral characteristics between SCI and control groups. First, we have used a factorization analysis on the smoothed spike trains of motor units to understand if they shared common synaptic input. We have found that the activity of motor units from SCI subjects is controlled by two main neural modules, corresponding to the flexion and extension of the hand digits as in the control group (SCI= 78.3%, control= 73.7% of variance explained with two modules). The comparable number of modules extracted and variance indicate that the synaptic input to the spinal motor neurons might be preserved. Second, we analyzed the behavior of the individual motor units by looking into the phase difference between their smoothed discharge trains and virtual hand kinematics. Through this analysis, we could classify the motor units according to their variance into modulated or non-modulated motor units. We found a significant difference in their variance distribution (generalized linear mixed model: p = 4.3e-10, t(1386)=6.2891) with a high proportion of non-modulated motor units for the SCI group (variance > 0.7; SCI = 40.8%, control = 16.4%), and low proportion of modulated motor units for this same group (variance < 0.3; SCI = 22.5%, control = 47%). Last, we extracted the area of activity of motor units according to their amplitude spatial map. We found that motor units present a larger area of activity (SCI = 584mm² (320,1032), control = 400mm² (300,600); p=1.2e-09, t(1384) = 6.1236).  Overall, at the neural level, motor neuron changes are characterized by a higher number of non-modulated motor units in SCI in comparison to the control group but with a preservation of common input. At the peripheral level, we found changes in motor unit innervation. These properties are important for the characterization of residual activity and to clarify recovery mechanisms after SCI.  Understanding them is crucial to change rehabilitation strategies and exploit this residual activity.