본문 바로가기 사이드메뉴 바로가기 주메뉴 바로가기

Media Center

A multimedia mosaic of moments at GIST

GIST Excellence

[Press Release] Development of EEG (electroencephalography) pre-processing and implementation of memristive hardware neural network

  • 엘리스 리
  • REG_DATE : 2015.05.14
  • HIT : 837

[Press Release] Development of EEG (electroencephalography) pre-processing and implementation of memristive hardware neural network

Korean researchers have succeeded in implementing the memristive hardware neural network that recognizes real-time EEG (electroencephalography) by speech imagination. Memristive synapses offers the most promising passive method for synaptic interconnections in artificial neural networks.

*Memristor: A memristor is a two-terminal passive device that modifies or retains its resistance according to the time integral of current flowing through it..

*Hardware neural network (HNN): Describing any very large-scale system of integrated circuits that mimic neuro-biological architectures present in the nervous system.

1

Figure 1: Proposed memristive HNN system for EEG pattern recognition. (a) EEG device were placed on the subject’s head and the subjects imagined speaking the vowels /a/, /i/, /u/ and the EEG signals were captured. (b) Processing to extract the distinct features of the three vowels. (c) Each feature is converted into a series of 32-bit binary code to be used by the 32 pre-neurons to generate a spike signal.

2

Figure 2: The single-layer HNN implemented on printing circuit board is shown. The circuit includes a switch array, switch control logic, 32 × 6 cross-point memristive synapse array, and six neuron circuits.

3

(from left) Prof. Byung Geun Lee, Prof. Bo Reom Lee, Dr. Sang-su Park, Ph.D candidate Myong-lae Chu

The research, supported by the Pioneer Research Center Program through the National Research Foundation of Korea funded by the Ministry of Science, ICT & Future Planning (2012-0009460), was led by Prof. Byung Geun Lee of the School of Mechatronics and Prof. Bo Reom Lee of the School of Medical System Engineering. Dr. Sang-su Park of the Department of Nanobio Materials and Electronics and Ph.D candidate Myong-lae Chu of the School of Mechatronics performed the EEG experiments with the support of the Pioneer Research Center Program through the National Research Foundation of Korea. Also Prof. Byoung Hun Lee of the School of Materials Science and Engineering, Prof. Moongu Jeon of the School of Information and Communications, and Prof. Hyun-sang Hwang of the School of Materials and Engineering from Pohang University of Science and Technology also contributed to this research.

Prof. Byung Geun Lee of the Department of Mechatronics and Prof. Bo Reom Lee of the Department of Medical System Engineering said, "The study was to prove the possibilities of various usage of the hardware neural network system using the memristor device. Hardware neural network system is the future of communicating system from just imagining for people with disabilities or people have difficulties in communication."

The paper  entitled “Electronic system with memristive synapse for pattern recognition” was published in Scientific Reports on May 5, 2015.