Sensor-based online hand gesture recognition on multi-core DSPs

In this paper we propose a novel approach for parallelized computation of sensor-based online hand gesture recognition on multi-core DSPs. To this end, we dispatch different windows of sensor data segments to different groups of cores. Within a group of cores, we distribute the template-based recognition on the basis of reference recordings. In order to model different configurations of our approach, we use extended data flow graphs. Our experiments show real time performance making our approach suitable for online hand gesture recognition.