Biofeedback (BFB) is a rehabilitation method, which, among other things, is used for the restitution of motor and gait function. As of now, it has become technically feasible to use BFB training based on target gait parameters to improve the gait function in stroke patients. The walking patterns of stroke patients are generally characterized by significant gait phase asymmetries, mostly of the stance phase and the single stance phase. The aim of the study was to investigate the restoration of gait function using BFB training with gait phases as feedback targets. The study included two patient groups, each of 20 hemiparetic patients in the subacute stage of stroke and a control group of 20 healthy subjects. Each patient group received BFB training with either stance phase or single stance phase as the feedback target, respectively. The patients received a total of 8 to 11 training sessions. Assessments based on clinical scales and gait analysis data (spatiotemporal, kinematic, and EMG parameters) were performed before and after the training course. The score-based clinical assessments showed a significant improvement in both patient groups. According to the assessments of gait biomechanics, the subjects in the Single Stance Phase group had significantly more severe dysfunctions. In both patient groups, the unaffected limb responded to the BFB training, while the stance phase significantly changed after training in the unaffected limb only. The other patient group, trained using the single stance phase as the feedback target, showed no changes in the target parameter either in the affected or in the contralateral limb. The clinical and instrumental assessments showed different, non-equivalent sensitivity. The results of the study demonstrated the possibility to use targeted BFB training to improve walking function. However, a significant effect of such training was only observed with stance phase as the target parameter. A response to training was observed predominantly in the unaffected limb and facilitated the desired increase in the functional ability of the paretic limb. Training based on stance phase as the target parameter is probably preferable for the patient population under study.