Abstract
This article covers the issue of the data preprocessing when highlighting steady-state visual evoked potentials using preliminary band-pass filtering of the EEG signal. In the introduction part the authors illustrate relevance of the system integration such as human-machine interaction and brain-computer interface. The integration of the above-mentioned systems as well as the ways of the signal preprocessing for highlighting of steady-state visual evoked potentials in electroencephalograms were examined. The article contains researches of the electroencephalogram signals with steady-state visual evoked potentials for photostimulation frequencies of 8 and 14 Hz with sampling frequency of 5 kHz based on the multivariate synchronization index method. Influence of preliminary band-pass filtering on recognition accuracy of the signal frequency under study is considered. Ratio of the correctly recognized states is considered in the function of accuracy metric. Butterworth filters, Chebyshev filters of I and II types, elliptic filters as well as Bessel filters of different orders are considered as bend-pass filters. The result of the authors’ investigation is a number of recommendations on parameters used while signal preprocessing for highlighting of steady-state visual evoked potentials in the multivariate synchronization index method. The results obtained are of considerable practical importance as they can be used for brain-computer interface producing on the basis of steady-state visual evoked potentials and later can be taken for building of control theory of robot systems of different application and for implementation of decisions on human-machine interaction within narrow practical tasks.
Access provided by Autonomous University of Puebla. Download conference paper PDF
Similar content being viewed by others
Keywords
- Steady-state visual evoked potentials
- Multivariate synchronization index
- Human-machine interaction
- SSVEP
- Electroencephalography
- EEG
- Signal preprocessing
- Butterworth filter
- Chebyshev filter
- Elliptic filter
- Bessel filter
1 Introduction
Over the last years development of brain-machine interfaces (BCI) of different types has become not only like an experimental trend but also more and more of practical use. The first reference to the brain-machine interface creation is primarily connected to the capability development of the electroencephalography and attempts to interpret isolated impulses of brain signals [1]. The basis of all brain-computer interfaces is a biofeedback training method providing information acquisition about current state of this or that physiological function for human training of a conscious brain control and control of this function [2]. Biofeedback utilization involves the use or either positive or negative reinforcement [3].
Electroencephalogram data as well as electromyography or analysis of a human’s ocular motility can be used for the brain-machine interface implementation. Both real and false responses, used for BCI development while patient’s rehabilitation after nervous system damage can be used as brain responses [4].
One of the ways of realization of such systems is the usage of steady-state visual evoked potentials (SSVEP). Registration of the evoked brain potentials (EP) is a non-invasive method of testing functions of the central nervous system [5]. Evoked potentials are brain electrical responses to visual, auditory or sensory stimuli. The important thing is that the signals of the evoked potentials on any irritant are similar to signals generated in case when a person acts as an irritant. SSVEP are evoked potentials of the brain on visual stimuli, where in most cases the following can be used as an irritant: screen image, image in AR goggles or LED glow of different flicker frequencies. Flicker frequency range of an irritant establishes a command set, which can be used for development of brain-computer interfaces or human-machine interactions.
Electroencephalogram (EEG) is a graphical presentation of an oscillatory electrical process registered when electrodes being placed on a brain or a scalp surface. Usually signals are recorded from electrodes to be placed on a scalp surface of a patient. EEG registration required usage of different sets of electrodes from 8 to 75 and different placement on a scalp [6].
The maim problems of using steady-state visual evoked potentials are spontaneous EEG due to external irritants and low signal-to-noise ratio. While the first problem can be solved only by organizational measures – a patient has to be concentrated on an object during the procedure, increase of signal-to-noise ratio often requires coherent integration and/or averaging on a software or hardware level. In addition, it is relevant to reduce an analysis window while highlighting EEG state as it contributes to development of brain-computer interfaces and human-machine interaction systems on the biofeedback basis, which work in real-time mode.
Analysis of research articles and references at scientific conferences in the area of development [7,8,9], implementation [10,11,12] and practical application [13,14,15,16] of human-machine interaction interfaces and development of brain-computer interfaces shows that relevance of recording of steady-state visual evoked potentials only increases with accuracy increase of their highlighting in the EEG signal.
2 Methods and Results
This article looks at the method which shows the highest results from the perspective of various researchers - multivariate synchronization index method [17,18,19,20]. Synchronization principle of multichannel EEG and reference signals, meant to be in line with visual stimulus frequency (photostimulation frequency) is a core of the multivariate synchronization index method. Selection of reference signals are based on Eq. 1.
where N – the harmonic sequence number, K – the number of samples, Fs – the sampling frequency, Y(t) – the reference signal, X(t) – the original EEG signal. Initially, the correlation matrix between two signals is calculated using Eqs. 2–3.
A linear transformation is calculated to eliminate the effect of autocorrelation on the measure of synchronization (Eq. 5).
After that the correlation matrix is calculated by the Eq. 6.
Then multivariate synchronization index is determined by the Eq. 7.
where \(\lambda_{i}^{^{\prime}}\) - normalized eigenvalues S, P = N + Nh и Nh note the number of lines in the reference signal B(t). The frequency of photostimulation is defined as the frequency corresponding to the maximum value of R.
The use of multivariate synchronization index for highlighting of steady-state visual evoked potentials within brain-computer interfaces makes it possible to use signals from several EEG channels at a time that could in perspective contribute to development of human-machine interaction systems of higher quality (Fig. 1).
Butterworth filter is a filter that is different in that its amplitude-frequency characteristic (AFC) is max smooth on band pass frequencies [21]. Chebyshev filter is a filter that is different in that its AFC possesses steeper roll-off and more significant pulsations on band-pass frequencies (Chebyshev filter of type I) and stopband frequencies (Chebyshev filter of type II) than those of filters of other types. They can be used, where with the help of the filter of a small order it is required to provide suppression of frequencies from the stopband and at the same time smoothness of AFC on band-pass and stopband frequencies is not of a big importance [22]. Elliptic filter is a filter that is different in that its AFC possesses very steep roll-off, that it why with the help of this filter it becomes possible to achieve more efficient frequency separation in comparison with other filters [23]. Bessel filter is a filter, which specific feature is a maximally smooth group delay such that waveform of a filtrated signal in output of this filter within the band pass remains almost the same [24].
EEG signals with steady-state visual evoked potentials were explored basing on the multivariate synchronization index with the use of the MATLAB [25] application program package. 60 signals of 30 adult patients, men and women of various age groups without contraindications to EEG and photostimulation reading were studied. LED glow was used as a source of photostimulation. All signals were preliminarily subjected to coherent integration in order to increase the signal-to-noise ratio because a big number of noises triggered by the brain spontaneous activity is typical for clear EEG signals. Photostimulation frequency for the signals from the first group amounted to 8 Hz and 14 Hz for the second group. Discretization frequency for all signals amounted to 5 kHz. Each signal was divided into windows of 500, 1000 and 1500 samples, which is equivalent of the analysis epoch of 100, 200 and 300 ms respectively. Band-pass filtering with the usage of the Butterworth filter, Chebyshev filter of type I, Chebyshev filter of type II, elliptic filter and Bessel filter was preliminarily used for each analysis window. Filter order used lay in the range from 1 to 4. Band edges of the filter for signals with required frequency of 8 Hz were set as 4 Hz and 12 Hz. Band edges of the filter for signals with required frequency of 14 Hz were set as 7 Hz and 21 Hz. Recognition accuracy was estimated as a window quantity ratio, where required frequency was correctly defined in regard to the whole number of windows within the signal. All signals were comprised of 15000 samples. The following frequencies were used as competing frequencies in both cases: 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20 Hz.
At the first stage of the research signals with 1–4 order filters and without them were processed. As a comparison, the average value of the state recognition accuracy was set as a metrics. It has been noticed that the results for all filters except Bessel filter are getting worse with increase of filter order. In view of this, the use of Butterworth filter, Chebyshev filter and elliptic filter for the fourth order resulted in a situation where almost none of states were correctly recognized. The higher analysis epoch is, the less relevance of the filtering use is, as the recognition accuracy without filtration for the window of 1000 samples and more shows the result of 0.94 and higher. The results of the signal processing within the single-channel mode are presented in the table 1. In the tables below the highest results within one category are bolded.
The results obtained within the dual-channel mode of the multivariate synchronization index correspond to the results for the single-channel mode. With increase of the analysis epoch the highest results were obtained without filtration and recognition accuracy for analysis window of 1000 samples showed the result of at least 0.9. The results of the signal processing within the dual-channel mode are presented in the Table 2.
The results obtained within the four-channel mode of the multivariate synchronization index were mixed. With the photostimulation frequency equal to 8 Hz the results correspond to the single-channel and dual-channel modes of filtration. And for the photostimulation frequency equal to 14 Hz with the window of 1000 samples the highest results are obtained with the use of Chebyshev filter and elliptic filter. In the authors’ opinion, it can be ascribed to non-optimal choice of channels for the analysis. For the rest the results correspond to the preceding frequency highlighting modes with the help of the multivariate synchronization index. The results of the signal processing within the four-channel mode are presented in the Table 3.
3 Conclusion
This research was made to discover influence of prefiltration of signals when using the multivariate synchronization index to highlight SSVEP in EEG within different MSI modes. Single-channel, dual-channel as well as four-channel modes of the multivariate synchronization index were studied. It is shown that for the current research the use of the prefiltration is required with the analysis window of 500 samples, which is equal to 100 ms. When the analysis epoch increases up to 1000 samples (equal to 200 ms) and higher the use of prefiltration becomes no longer of relevance, except Bessel filter, the use of which shows the results similar to recognition without filtration. Besides, it is relevant to use filters of the first order, because the higher order is, the worse results become. On the basis of the results obtained the authors consider the use of Chebyshev filter of type II as the most efficient as it shows the best results with the analysis window of 500 samples in comparison with elliptic filter and Chebyshev filter of type I, but it also shows higher results when increasing analysis window. In the future it makes sense to examine EEG state recognition with sufficient accuracy on a smaller analysis window, as decrease of analysis window makes it possible to count on integration of human-machine interaction systems running in real-time mode on the basis of the biofeedback. In this case sufficient accuracy is considered as 0.5 and higher because in this case the required frequency will be selected among the competing frequencies.
References
Nam, C.S., Nijholt, A., Lotte, F.: Brain–Computer Interfaces Handbook: Technological and Theoretical Advances. CRC Press, Boca Raton, 788 p. (2018)
Pfurtscheller, G., et al.: Current trends in Graz Brain-Computer Interface (BCI) research. IEEE Trans. Rehabil. Eng. 8(2), 216–219 (2000)
Kharchenko, S.S.: Investigation of the influence of the positive reinforcement method using biofeedback. Int. J. Biosen Bioelectron. 2(2), 73–75 (2017)
Gnezditsky, V.V.: Evoked Brain Potentials in Clinical Practice, p. 252. Publishing House of TSURE, Taganrog (1997). (in Russian)
Tyagi, A., Semwal, S., Shah, G.: Article: a review of Eeg sensors used for data acquisition. IJCA Proc. Nat. Conf. Future Aspects Artif. Intell. Ind. Autom. 2012 NCFAAIIA(1), 13–18 (2012). https://www.ijcaonline.org/proceedings/ncfaaiia/number1/6725-1004
Malmivuo, J., Plonsey, R.: Bioelectromagnetism - Principles and Applications of Bioelectric and Biomagnetic Fields. Oxford University Press, Oxford, 641 p. (1995)
Ma, W., Tran, D., Le, T., Lin, H., Zhou, S.: Using EEG artifacts for BCI applications. In: 2014 International Joint Conference on Neural Networks (IJCNN), Beijing, pp. 3628–3635 (2014)
Han, J., Ji, S., Shi, C., Yu, S., Shin, J.: Recent progress of non-invasive optical modality to brain computer interface: a review study. In: The 3rd International Winter Conference on Brain-Computer Interface, Sabuk, pp. 1–2 (2015)
Chueshev, A., Melekhova, O., Meshcheryakov, R.: Cloud robotic platform on basis of fog computing approach. In: Ronzhin, A., Rigoll, G., Meshcheryakov, R. (eds.) ICR 2018. LNCS (LNAI), vol. 11097, pp. 34–43. Springer, Cham (2018). https://doi.org/10.1007/978-3-319-99582-3_4
Wan, X., et al.: A review on electroencephalogram based brain computer interface for elderly disabled. IEEE Access 7, 36380–36387 (2019)
Turovsky, Y.A., Kurgalin, S.D., Vahtin, A.A., Borzunov, S.V., Belobrodsky, V.A.: Event-related brain potential investigation using the adaptive wavelet recovery method. Biophysics (Russ. Fed.) 60(3), 443–448 (2015). article № A018
Lee, M., Fazli, S., Mehnert, J., Lee, S.: Hybrid brain-computer interface based on EEG and NIRS modalities. In: 2014 International Winter Workshop on Brain-Computer Interface (BCI), Jeongsun-kun, pp. 1–2 (2014)
Reda, R., Tantawi, M., shedeed, H., Tolba, M.F.: Analyzing electrooculography (EOG) for eye movement detection. In: Hassanien, A.E., Azar, A.T., Gaber, T., Bhatnagar, R., F. Tolba, M. (eds.) AMLTA 2019. AISC, vol. 921, pp. 179–189. Springer, Cham (2020). https://doi.org/10.1007/978-3-030-14118-9_18
Jo, S., Choi, J.W.: Effective motor imagery training with visual feedback for non-invasive brain computer interface. In: 2018 6th International Conference on Brain-Computer Interface (BCI), GangWon, pp. 1–4 (2018)
Clerc, M., Bougrain, L., Lotte, F.: Brain-Computer Interfaces 2: Technology and Applications. Wiley, Hoboken, 364 p. (2016)
Shepelenko, M., Meshcheryakov, R.: Algorithm for automated calculation of a segmented electromechatronic module of robotic system motion. MATEC Web Conf. 113, article № 02005 (2017)
Zhang, Y., Xu, P., Cheng, K., Yao, D.: Multivariate synchronization index for frequency recognition of SSVEP-based brain-computer interface. J. Neurosci. Methods 221, 32–40 (2014)
Purushothaman, G., Prakash, P.R., Kothari, S.: Investigation of multiple frequency recognition from single-channel steady-state visual evoked potential for efficient brain–computer interfaces application. IET Signal Process. 12(3), 255–259 (2018)
Zhang, Y., Guo, D., Xu, P., Zhang, Y., Yao, D.: Robust frequency recognition for SSVEP-based BCI with temporally local multivariate synchronization index. Cogn. Neurodyn. 10(6), 505–511 (2016). https://doi.org/10.1007/s11571-016-9398-9
Kharchenko, S., Turovsky, Y., Mescheryakov, R., Iskhakova, A.: Restrictions of the measurement system and a patient when using visually evoked potentials. In: 2019 12th International Conference on Developments in eSystems Engineering (DeSE), Kazan, Russia, pp. 15–19 (2019)
Oppenheim, A.V., Schafer, R.W., Yuen, C.K.: Digital signal processing. IEEE Trans. Syst. Man Cybern. 8(2), 146 (1978)
Smith, S.: Digital signal processing. Newnes, Boston (2003)
Daniels, R.: Approximastion Methods for Electronic Filter Design. McGraw-Hill, New York (1974)
Paarmann, L.: Design and Analysis of Analog Filters. Springer, Boston (2003). https://doi.org/10.1007/b100752
Turovsky, Y.A., Kharchenko, S.S., Meshcheryakov, R.V., Iskhakova, A.o., Iskhakov, A.Y.: Algorithmic support of the interface of management of robot-human with the steady state visual evoked potentials based on the multivariate synchronization index. Izvestiya SFedU. Eng. Sci. 1, 66–78 (2020). (in Russian)
Acknowledgment
The reported study was partially funded by RFBR according to the research projects № 19-08-00331 and № 19-29-01156.
Author information
Authors and Affiliations
Editor information
Editors and Affiliations
Rights and permissions
Copyright information
© 2021 Springer Nature Singapore Pte Ltd.
About this paper
Cite this paper
Kharchenko, S., Meshcheryakov, R., Turovsky, Y. (2021). Influence of Signal Preprocessing When Highlighting Steady-State Visual Evoked Potentials Based on a Multivariate Synchronization Index. In: Singh, P.K., Veselov, G., Pljonkin, A., Kumar, Y., Paprzycki, M., Zachinyaev, Y. (eds) Futuristic Trends in Network and Communication Technologies. FTNCT 2020. Communications in Computer and Information Science, vol 1396. Springer, Singapore. https://doi.org/10.1007/978-981-16-1483-5_10
Download citation
DOI: https://doi.org/10.1007/978-981-16-1483-5_10
Published:
Publisher Name: Springer, Singapore
Print ISBN: 978-981-16-1482-8
Online ISBN: 978-981-16-1483-5
eBook Packages: Computer ScienceComputer Science (R0)