In video surveillance applications one important task is the re-identification of people leaving the field-of-view then reappearing. Researches on human gait as a biometric feature showed that it can successfully identify individuals. The benefit of using gait as a feature is that it can be obtained from a distance, so it gives us an effective alternative to other biometric features such as face and iris. The Lidar device compared to the traditional cameras is not confused if there are background movements or people cover each other, so the Lidar might be a good decision for the given task. Gait Energy Image (GEI) seems to be the best choice among the gait features mentioned in the literature, thus my thesis work presents a recognition system based on GEIs. People identification is done by neural networks: an ensemble of a multilayer perceptron and a convolutional neural network, which results the best recognition rates in deep learning applications. We demonstrate that the proposed ensemble of nn perfomrs better thant two relevant reference methods The gait recognition system has been implemented in the Lidar-based framework of the Distributed Event Analysis group of MTA SZTAKI.