Nowadays the development of virtual reality-based application is one of the most dynamically growing areas. These applications have a wide user base, more and more devices which are providing several kinds of user interactions and are available on the market. In the applications where the not-handheld devices are not necessary, the potential is that these can be used in educational, entertainment and rehabilitation applications. The purpose of this paper is to examine the precision and the efficiency of the not-handheld devices with user interaction in the virtual reality-based applications. The first task of the developed application is to support the rehabilitation process of stroke patients in their homes. A newly developed application will be introduced in this paper, which uses the two popular devices, the Shimmer sensor and the Microsoft Kinect sensor. To identify and to validate the actions of the user these sensors are working together in parallel mode. For the problem solving, the application is available to record an educational pattern, and then the software compares this pattern to the action of the user. The goal of the current research is to examine the extent of the difference in the recognition of the gestures, how precisely the two sensors are identifying the predefined actions. This could affect the rehabilitation process of the stroke patients and influence the efficiency of the rehabilitation. This application was developed in C# programming language and uses the original Shimmer connecting application as a base. During the working of this application it is possible to teach five-five different movements with the use of the Shimmer and the Microsoft Kinect sensors. The application can recognize these actions at any later time. This application uses a file-based database and the runtime memory of the application to store the saved data in order to reach the actions easier. The conclusion is that much more precise data were collected from the Microsoft Kinect sensor than the Shimmer sensors.