Research project

Recognition of hand gestures plays an important role in different areas, e.g. infotainment systems or in the automotive field. The goal of this project is the classification of predefined static gestures using the so-called PMD camera (Photonic Mixer Device, short: PMD). Modulated infrared light signals are emitted by the sender, illuminate a scene, hit objects and are reflected of those onto the sensor. With this technique it is possible to calculate distances to the observed objects in the scene. The camera also provides a conventional gray scale image. Both images can be used for segmentation of hand gestures. The distance values are used to segment hand and arm. Next, the hand is extracted to calculate the features for the classification. The camera runs with 15 frames per second and therefore the classification should de done in real-time. The system is evaluated with different gestures from different persons and an according training set is embedded into the online system. The software, in which the package should be integrated, was provided by Audi Electronics Venture GmbH. Two modules are designed for this software: the module "Recording" for recording different gestures as well as the module "Classification" which classifies gestures in the online system.