Common commercially available hand prostheses are based on reading EMG signal from the stump area. This solution is not suitable for some users as sometimes the muscles and nerves around this area are weak or even degenerate and activating them can cause phantom pain. Furthermore, action is performed as long as the user exercises his muscles, which requires a lot of concentration and effort. Take for example grip: as long as the user thinks of the grip, it is in progress. But inadvertently the grip can be released and the object in his hand will fall. The solution on which this project is based offers an internal control system located on user's leg which when an operation is preformed, the hand remains in a fixed position until a different command is received. The first goal of this project is to create a system for collecting signals from the leg in order not to use the muscles around the stump. A second goal is to implement a classifier that can distinguish between 3 different leg movements under the requirement that the number of false identifications is as small as possible. When implementing the classifier, we used the template matching method. From our results it appears that we have successfully implemented a signal recording system, selected simple movements to be performed by the user and were able to identify them so that the false identifications rate is less than five percent.