Start Date: May 9, 2018
Eirini Mathe
The proposed thesis will investigate novel methods for the implementation of deep neural network architectures. The applied techniques will be based on the measurement of physiological and kinematic parameters, as well as on audiovisual data recorded by sensors placed on the user’s environment or wearables. Emphasis will be given on visual representations of captured data, so that the recognition may be based on convolutional neural networks. Information fusion techniques will also be explored so as to fuse data captured from heterogeneous sensors, such the aforementioned ones