Restoring upper limb functions in patients after a stroke remains a challenging medical and social task. The introduction of a human—computer interface based on computer vision enables combining motor training with the performance of practical tasks on a computer, laying the foundation for a new approach integrating rehabilitation into the patient’s daily activities, thereby increasing adherence to treatment. The article presents two clinical cases of using the author’s method of rehabilitation in male patients (47 and 73 years) with post-stroke paralysis of the upper limb. The method is based on non-contact computer control via shoulder joint movements, which provides core muscle training. After the training course, patients showed improvements: increased shoulder joint range of motion as measured by goniometry, decreased muscle spasticity, and improved performance on the Fugl-Meyer (upper limb) and Frenchay functional scales. The use of non-contact computer control via shoulder movements is promising for restoring motor function of the paretic limb and contributing to the patient’s social adaptation.