Radiotekhnika
Publishing house Radiotekhnika

"Publishing house Radiotekhnika":
scientific and technical literature.
Books and journals of publishing houses: IPRZHR, RS-PRESS, SCIENCE-PRESS


Тел.: +7 (495) 625-9241

 

Method of data dimensionality reduction in brain-computer interface systems

DOI 10.18127/j20700814-201809-02

Keywords:

R.R. Bakunov - Senior Lecturer, Department of Information Technologies and Automated Systems, Perm National Research Polytechnic University
E-mail: bakunov_roman@mail.ru


BCI technology is based on the measurement of user EEG signals and the recognition of conscious brain electrical activity using digital signal processing techniques.
The functioning of the BCI system can be represented as a cycle. In each iteration, the EEG signal is measured and pre-processed, the characteristic features are extracted, the classification is implemented and the control action corresponding to the recognized command of the operator is generated.
In BCI systems, EEG signal analysis is often performed in the frequency domain. In this case, the set of processed signals can be represented as a set of points in N-dimensional space, where N is the number of allocated spectral components. Often this value can be equal to 32, 64, 128, etc. A large amount of coordinates in the processed vectors generates a number of problems. Therefore, the mathematical support of BCI systems should be adapted to operations on high dimensionality data. It should include tools for effective reduction of the number of coordinates in the processed vectors. At the moment, there are a number of methods that solve the problem of processing multidimensional data. One of them is a linear discriminant analysis (LDA). This is one of the fastest approaches used by researchers to reduce the dimensionality of the processed data in BCI systems. However, its use requires a fairly productive computer, because it is associated with complex calculations. It should be noted that very often BCI systems are required to be mobile. In this case, they are based on computing devices with severely limited performance. Therefore, there is a need for new methods of data dimensionality reduction.
One of these approaches is discussed in the article. Its detailed description is given: theoretical foundations, formulas, algorithm scheme. The proposed algorithm of data dimensionality reduction is based on the concepts of the cross-correlation coefficient and the distance between the vectors.
In this article, much attention is paid to the results of experimental testing. The experiments showed that using the pro-posed approach allows to significantly reduce the time required to perform operations of data dimensionality reduction. In ad-dition, its use does not negative affect on the clustering quality of processed sets of signals. It is experimentally confirmed that the developed algorithm effectively works in conjunction with the LDA, acting as a preprocessor for the LDA. At the same time, the speed of such bundle is much higher than speed of LDA without the preprocessor.
It should be noted that the proposed solution can be used in any information-measuring and control systems, especially in cases when measured signals should be classified in real time. In this case, the calculations can be performed by microprocessor devices with limited performance, low power consumption and low memory capacity.

References:
  1. Tu Dzh., Gonsales R. Princzipy’ raspoznavaniya obrazov. M.: Mir. 1978.
  2. Gajdy’shev I. Analiz i obrabotka danny’x: speczial’ny’j spravochnik. SPb.: Piter. 2001. 752 s.
  3. Ajvazyan S.A., Buxshtaber V.M., Enyukov I.S., Meshalkin L.D. Prikladnaya statistika: Klassifikacziya i snizhenie razmernosti: Sprav. izd. / Pod red. S.A. Ajvazyana. M.: Finansy’ i statistika. 1989. 607 s.
  4. Lan T., Black L., Van Santen J., Erdogmus D. A comparison of different dimensionality reduction and feature selection methods for single trial ERP detection // 2010 Annual International conference of the IEEE engineering in medicine and biology society EMBC'10. 2010. P. 6329−6332.
  5. Ajficher E’.S., Dzhervis B.U. Czifrovaya obrabotka signalov: prakticheskij podxod. Izd. 2-e. Per. s angl. M.: Izdatel’skij dom «Vil’yams». 2004. 992 s.
  6. Barsegyan A.A., Kupriyanov M.S., Xolod I.I., Tess M.D., Elizarov S.I. Analiz danny’x i proczessov: Ucheb. posobie. Izd. 3-e, pererab. i dop. SPb.: BXV-Peterburg. 2009. 512 s.
  7. Fajzraxmanov R.A., Bakunov R.R., Kashin O.A. Primenenie kriteriev kachestva klasterizaczii v informaczionno-izmeritel’ny’x sistemax // Nauchnoe obozrenie. 2015. № 8. S. 231−238.
  8. Fajzraxmanov R.A., Bakunov R.R. O povy’shenii kachestva klasterizaczii signalov v texnicheskix sistemax s pomoshh’yu linejnogo diskriminantnogo analiza // E’lektrotexnika. 2016. № 11. S. 55−59.
  9. Tre’vis Dzh., Kring Dzh. LabVIEW dlya vsex. Izd. 4-e, pererab. i dop. M.: DMK Press. 2011. 904 s.
  10. Fayzrakhmanov R.A., Bakunov R.R. The reduction of learning sample in information-measuring and control systems based on brain-computer interface technology // Proc. of 2nd International Conference on Industrial Engineering, Applications and Manufacturing (ICIEAM). Chelyabinsk, Russia. May 19−20, 2016. South Ural State University (national research university). Electronic data. [S. l.]: IEEE. 2016. URL = http://ieeexplore.ieee.org/document/7911544/.
June 24, 2020
May 29, 2020

© Издательство «РАДИОТЕХНИКА», 2004-2017            Тел.: (495) 625-9241                   Designed by [SWAP]Studio