Raúl Finker, a researcher in the Department of Electricity and Electronics of the UPV/EHU's Faculty of Science and Technology and who has recently received his PhD, has managed to implement an artificial neural  network and its learning algorithms in a small integrated circuit designed to be embedded into intelligent environments. Neural networks are one of the most widely used techniques in computational intelligence: they are inspired by the brains of living organisms and have the capacity to learn and adapt to changes in the environment thanks to different learning algorithms.

To achieve this, he used programmable logic devices that enable the whole system that is needed to be implemented on a single chip by making use of a hardware/software architecture: the neural network was implemented in the hardware, and the learning algorithms in the software. What is achieved this way is that the data are processed far more quickly than normally and their incorporation into ambient intelligence environments is fast and does not take up a lot of space. An ambient intelligence environment is a model of interaction in which people are surrounded by an electronic environment which picks up their presence, is context-sensitive and responds in an adaptive and non-intrusive way to the needs and habits of the users.

A fully autonomous system capable of adapting to a user

Two intelligent environment applications have been developed to demonstrate that the hardware/software architecture proposed can be used in the ambient intelligence environments for which it has been designed.

The first is in an inhabited environment known as iDorm. This environment is a study-bedroom developed by the Intelligent Environments Group of the University of Essex, which provided the data they used to conduct the study. The researchers concluded that by using the system developed, it is possible to train a neural network so that it will adapt to the needs of the user of the environment and be capable of controlling the responses based on these needs. "We saw that the system was capable of adapting perfectly to a user's behaviour by controlling the various items present in the study-bedroom and that it adapted to the changes in the behaviour that this user might display during different seasons of the year," explained Finker.

The second application involved the development of a driver identifier in real time for environmental intelligence applied to the car environment. To produce this application, the data used were provided by the Drive-Safe Consortium, specifically the data provided by the Sabanci University of Istanbul using a Uyanik saloon car. By using the data provided by the car itself, a system capable of identifying the driver based on his/her way of driving was designed. The significant difference with respect to other identification systems is that it does not require the use of other elements separate from the car, such as cameras or finger print readers, among other things. "By using the data on the car's accelerometers or through the use of pressures brought to bear on the accelerator or brake pedals, it is possible to identify the driver and obtain very good results," concluded the researcher.

This second application is paving the way for future projects, "on which we are currently working", added the researcher. Firstly, it can be used as a security system for the driver, as "it could even detect whether the driver for some reason is not driving in the way he or she usually does" and, secondly, as a security system for the car as "it could detect that the person who is driving is not one of the people who routinely uses the vehicle in question," he explained. These are two of the potential uses of the system achieved, but "it could be implemented in a whole host of applications depending on the needs and the size of network required," pointed out Finker.

Additional information

Raúl Finker (Barakaldo, 1982) studied Industrial Technical Engineering (Industrial Electronics) and Telecommunications Engineering at the University of Deusto, and did a Master's at the UPV/EHU on Advanced Electronic Systems. In September 2015 he finished writing up his PhD thesis entitled Efficient electronic implementations of adaptive systems for ambient intelligence environments, in the Department of Electricity and Electronics of the UPV/EHU's Faculty of Science and Technology; his supervisors were Javier Echanobe and Inés del Campo. The research was carried out in collaboration with the universities of Essex (United Kingdom) and Sabanci (Instanbul, Turkey).

Recent conferences

Echanobe, J.; Finker, R.; del Campo, I., "A Divide-and-Conquer Strategy for FPGA Implementations of Large Neural Network-based Classifiers" The 2015 International Joint Conference on Neural Networks (IJCNN), Killarney, Ireland, Jul. 2015

Del Campo, I.; Echanobe, J.; Asua, E.; Finker, R.; "Controlled-Accuracy Approximation of Nonlinear Functions for Soft Computing Applications. A high performance co-processor for intelligent embedded systems," 2015 IEEE Symposium Series on Computational Intelligence. Cape Town, South Africa, Dec. 2015 (paper accepted)

Photo caption: The identification system does not require the use of other elements not belonging to the car itself (Syda Productions / Dollar Photo Club).