Intelligent Embedded Systems Research Group



The research activities of the Intelligent Embedded Systems research group include both fundamental and applied researches. The dominant topics of fundamental research projects are related to learning systems, machine learning. The applied research activities could be grouped around the application possibilities of the hardware implemented ANNs and the development of the ambient intelligent systems.

The scientific objectives are correlated with the development of Reconfigurable intelligent embedded systems having learning capabilities and adaptive behavior. The researches will be further focused on the following topics:

Implementation of artificial neural networks using reconfigurable devices

The aim of the research is to continue the design of hardware implemented artificial neural networks using the latest reconfigurable hardware devices. For implementation we will test and compare the method developed using System Generator with the HDL description of the ANN and the new high level synthesis tools like Vivado Design Suite HLx. We intend to use for implementation the new hardware devices like the Xilinx All Programmable FPGAs and 3D ICs (the 7 Series, Ultrascale and Ultrascale+) or the Xilinx’s All Programmable SoC (Zynq-7000 SoC, Zynq UltraScale+ MPSoC) to integrate the software programmability of a processor with the hardware programmability of an FPGA. In this way the training of the neural networks could be done using the on-chip Dual-core ARM® Cortex™-A9 MPCore™. The ANN could be implemented in the FPGA part to take advantage of parallel implementation of the neurons. The ANN implemented in this way will be tested for example in human activity pattern recognition.

e-Health and Ambient assisted living systems

We intend to continue the research related to the development of devices and systems to support the elderly or disabled persons’ everyday independent life, using the latest assistive technologies. The intelligent assistive environment will include prototypes of multi-sensor networks, multi standard communications, smart digital appliances, or smart monitoring systems. Our research will continue in the following directions:

  • Activity and health status integrated platform development

For human activity monitoring we use multiple wearable sensor tags like the CC2541 power-optimized system-on-chip (SoC) that combines the performance of a Bluetooth low energy (BLE) transceiver with the industry-standard enhanced 8051 MCU. The tag contains a 6 axis motion sensor MPU6050 (acceleration + angular velocity) and BMP180 pressure and temperature sensor. We acquire data using these tags and we intend to make publically available a large dataset with human activities in order to support reproducible research.

  • Activity and health status recognition using neural networks modeled in Matlab

We started to develop our own methods related to Human Activity Recognition for feature extraction and feature selection on publically available datasets, using a combination of feature extraction and selection methods from public repositories, and we plan to continue this research on our dataset too.

  • Activity and health status recognition using hardware implemented ANNs

Following the successful completion of the first two goals we will use the results to implement in hardware a real-time human activity recognition system using neural networks. We aim to extend the recognition system to the health status recognition also.

Intelligent embedded systems design and applications

We will continue to constantly improve the previously designed hardware-software co-design platform based on FPGA adding new functions and modules. For fast prototyping of reconfigurable embedded systems, we plan to design new IP blocks that can be easily connected and can manage basic behaviors of the I/O devices, sensors and actuators. The design of the adaptive interfaces using neural networks will represent also a research interest. The designed architecture allows the insertion of intelligent interfaces based on neural networks. This platform will be based on the low cost general purpose FPGA boards without a need for hardware design of the boards.


Updated: 2019.10.09.

Dear User!


The University of Debrecen considers the protection of personal data and the information provided to be of paramount importance. We hereby inform you that the University of Debrecen has reviewed its processes and incorporated the requirements of GDPR into his own data management and data protection activities, which entered mandatory on may 25,2018. The personal data of the users has been carefully handled by the University of Debrecen, it complies with the applicable data management regulations.Following the requirements of GDPR, we updated our Privacy Statement, which you can access by clicking on the link below: Privacy Policy.

I have read, understand and agree to terms.