Descripció del projecte
The goal of this PhD offer is the application of machine learning technologies to capacitive sensing with the intent to use changes in a base capacitance, measured by a voltage reference, to characterize signals related to certain gestures and activations, classified by different use cases which are linked to Human Machine Interface (HMI). We will study the relative perceived reliability and efficiency compared to classical approaches (signal thresholds and derivatives) or even mechanical push buttons where a tactile pressure release confirms a functional switch transition.
Several studies showed the feasibility of an accurate simple gesture recognition system by using machine learning in low cost general purpose embedded devices, the evaluation of interference factors like the use of gloves remains unsettled as well as the possibility of using recent breakthroughs in the area of deep learning for low cost embedded devices.
The aim is to improve those solutions through the implementation of new algorithm tailored to the Kostal technology for capacitive touch sensing together with the different models of mechanical integration in cars, the embedded System-on-Chip (SoC) processor platforms, new developments in general purpose microcontrollers related to neural networks and the High Performance Computing (HPC) resources for the implementation of the learning phases on collected data.
