Brain Computer Interface for Cursor Movement Control by Fuzzy Logic

4
33 International Conference on Issues and Challenges in Networking, Intelligence and Computing Technologies 2-3 September 2011 KIET, Ghaziabad BRAIN COMPUTER INTERFACE FOR CURSOR MOVEMENT CONTROL BY FUZZY LOGIC Irshad Ahmad Ansari #1 , Sachin Sharma #2 , Gaurav Kumar #3 # Department of Instrumentation and Control Engineering, Dr. B. R. Ambedkar National Institute of Technology, Jalandhar, India 1 [email protected] 2 [email protected] 3 [email protected] AbstractBrain-Computer Interfaces (BCI) involves the communication between the user and the system via brain signals. The electroencephalography (EEG) have developed into one of the most important and widely used quantitative diagnostic tools in analysis of brain signals and patterns. The main goal of our study is to help people with sever motor disabilities (i.e., Spinal cord injuries) and provide them a new way of communication and control options by which they can move the cursor in two dimensions. In this paper, we proposed a novel architecture that can acquire bio signals in real-time and process it for controlling of cursor movement. it will detect noisy signal in the analysis and it remove completely. The aim of this paper is to give an idea of developing an application based Brain Computer Interface (BCI) using Fuzzy Logic. Keywords: Brain Computer interface, fuzzy logic, EEG, brain signal I. INTRODUCTION An electroencephalogram (EEG) based BCI provide a feasible and non-invasive way for the communication between the human brain and the computer [1][7].People with severe injuries (i.e., spinal cord injuries) and motor disabilities face daily challenges in living their lives as normal people. One of these difficulties facing disabled people is using the computer. Such people can control the devices by controlling certain waves from their EEG-Electroencephalogram E.g. movement of a mechanical arm, Control of Computer Cursor or control a wheel chair. Previous researches done on BCI used many different methods to search the possibility of connecting a patient's brain with a computer system successfully and so many different methods were studied and tested. With the advancement of sensor technology and information technology it become easy to design low power required sensors and make the cost of production cheaper. This help in physiological signal monitoring area very effectively. A physiological signal monitoring system will be extremely useful in many areas if they are portable and monitoring target physiological signals and analysing them in real time. Lot of algorithms was proposed for the development of brain-computer interface for improved performance. Here to improve such a interface noise removal system or suitable processing technique is needed. Some of these researches used linear methods, non-linear methods in order to classify the brain signals extracted and others used genetic algorithm or classical classification algorithms to classify the signals coming out of the brain. There are some studies regarding the portable BCI devices [8][10]. Our previous studies discovered that some features in human EEG signals are highly related to drowsiness level [11], [12], and they can be used for estimating driver drowsiness. In this research, the (fuzzy inference system) algorithm was used in order to classify the signals where it showed a better performance. Real time analysis of EEG was used in this study for the disable people in order to control the Cursor movement by extracting the feature from the data and classify them in proper way. The better the classification is, the better the application of any BCI system will be. II. METHODOLOGY This dataset was recorded from a normal subject. The subject sat in a normal chair, relaxed arms resting on the table. The experiment consisted of 4 sessions of 10 minutes each. All sessions were conducted on the same day with some minutes break in between. A. Data Acquisition Data used in this experiment consisted of three channels of EEG placed on Cz, C3 and C4 (international 10-20 electrode system). We developed a BCI which uses μ (8-12Hz) and central β (18-25Hz) EEG rhythms recorded over the motor cortex. The data is recorded from each subject in a timed experimental recording procedure where the subject is instructed to imagine

Transcript of Brain Computer Interface for Cursor Movement Control by Fuzzy Logic

Page 1: Brain Computer Interface for Cursor Movement Control by Fuzzy Logic

33 International Conference on Issues and Challenges in Networking, Intelligence and Computing Technologies 2-3 September 2011

KIET, Ghaziabad

BRAIN COMPUTER INTERFACE FOR CURSOR MOVEMENT CONTROL BY FUZZY

LOGIC Irshad Ahmad Ansari #1, Sachin Sharma #2, Gaurav Kumar#3

# Department of Instrumentation and Control Engineering,

Dr. B. R. Ambedkar National Institute of Technology, Jalandhar, India

1 [email protected]

2 [email protected] [email protected]

Abstract— Brain-Computer Interfaces (BCI) involves the

communication between the user and the system via brain signals. The electroencephalography (EEG) have developed into one of the most important and widely used quantitative diagnostic tools in analysis of brain signals and patterns. The main goal of our study is to help people with sever motor disabilities (i.e., Spinal cord injuries) and provide them a new way of communication and control options by which they can move the cursor in two dimensions. In this paper, we proposed a novel architecture that can acquire bio signals in real-time and process it for controlling of cursor movement. it will detect noisy signal in the analysis and it remove completely. The aim of this paper is to give an idea of developing an application based Brain Computer Interface (BCI) using Fuzzy Logic.

Keywords: Brain Computer interface, fuzzy logic, EEG, brain signal

I. INTRODUCTION

An electroencephalogram (EEG) based BCI provide a feasible and non-invasive way for the communication between the human brain and the computer [1]–[7].People with severe injuries (i.e., spinal cord injuries) and motor disabilities face daily challenges in living their lives as normal people. One of these difficulties facing disabled people is using the computer.

Such people can control the devices by controlling certain waves from their EEG-Electroencephalogram E.g. movement of a mechanical arm, Control of Computer Cursor or control a wheel chair. Previous researches done on BCI used many different methods to search the possibility of connecting a patient's brain with a computer system successfully and so many different methods were studied and tested. With the advancement of sensor technology and information technology it become easy to design low power required sensors and make the cost of production cheaper. This help in physiological signal monitoring area very effectively. A physiological signal monitoring

system will be extremely useful in many areas if they are portable and monitoring target physiological signals and analysing them in real time.

Lot of algorithms was proposed for the development of brain-computer interface for improved performance. Here to improve such a interface noise removal system or suitable processing technique is needed. Some of these researches used linear methods, non-linear methods in order to classify the brain signals extracted and others used genetic algorithm or classical classification algorithms to classify the signals coming out of the brain. There are some studies regarding the portable BCI devices [8]–[10]. Our previous studies discovered that some features in human EEG signals are highly related to drowsiness level [11], [12], and they can be used for estimating driver drowsiness.

In this research, the (fuzzy inference system) algorithm was used in order to classify the signals where it showed a better performance. Real time analysis of EEG was used in this study for the disable people in order to control the Cursor movement by extracting the feature from the data and classify them in proper way. The better the classification is, the better the application of any BCI system will be.

II. METHODOLOGY

This dataset was recorded from a normal subject. The subject sat in a normal chair, relaxed arms resting on the table. The experiment consisted of 4 sessions of 10 minutes each. All sessions were conducted on the same day with some minutes break in between.

A. Data Acquisition

Data used in this experiment consisted of three channels of EEG placed on Cz, C3 and C4 (international 10-20 electrode system). We developed a BCI which uses μ (8-12Hz) and central β (18-25Hz) EEG rhythms recorded over the motor cortex. The data is recorded from each subject in a timed experimental recording procedure where the subject is instructed to imagine

Page 2: Brain Computer Interface for Cursor Movement Control by Fuzzy Logic

34 International Conference on Issues and Challenges in Networking, Intelligence and Computing Technologies 2-3 September 2011

KIET, Ghaziabad

moving the left and right hand and foot in accordance to a directional cue displayed on a computer monitor. In each recording session a number of EEG patterns

Fig. 1. Basic model of BCI system

relating to the imagined right or left arm and foot movement are produced by a subject, over a number of trials. All signals are sampled at 128Hz and filtered between 0.5 and 30Hz. The subject can select any movement which is displayed on the computer in order to control the movement of cursor.

B. Architecture and training Procedure

An experiment paradigm was designed for the study and the protocol was explained to each participant before the experiment. In this, the subject was asked to comfortably lie down in a relaxed position with eyes closed. After assuring the normal relaxed state by checking the status of alpha waves, the EEG was recorded, This was used as the baseline reference for further analysis of mental task. The subject was asked to per-form a mental task on presentation of an display cue. Data collected from five subjects performing four mental tasks were analysed. The following mental tasks were used.

Fig. 2. Architecture of BCI system

Left hand movement imagination: The subject was asked to imagine the movement of left hand with open eye.

Right hand movement imagination: The subject was asked to imagine the movement of left hand with open eye.

Left foot movement imagination: The subject was asked to imagine the movement of Left foot with open eye.

Fig. 1. Example of EEG wave

Right foot movement imagination: The subject was asked to imagine the movement of Right foot with open eye. EEG signals were recorded during this period.

C. Feature Extraction

In this work a time-frequency (t-f) approach to EEG feature extraction has been adopted. It rests on the fact that the spectral content of the EEG recorded from bipolar channels over Cz, C3 and C4 locations when a subject performs imagination of hand and foot movements displays relevant changes around μ (8-12Hz) and β (18-25Hz) ranges. When the sensorimotor area of the brain is activated as a result of information processing (i.e. during imagination of hand movement), the amplitude of μ and central β oscillations decreases. This phenomenon is referred to as event-related desynchronization (ERD). The opposite process of amplitude enhancement of the EEG recorded from cortical areas that are not specifically involved in a given mode of activity is called event related synchronization (ERS) [6]. The frequency bands of ERS and ERD vary from subject to subject. The experiments have shown that for the subjects examined, there was ERD of the μ rhythm on the contra lateral side (opposite side to imagined hand movement) and a slight ERS in the central β rhythm on the ipsilateral hemisphere (the same side as the imagined hand movement). Moreover, the most reactive frequency bands from which to extract features for the given subjects have been found. [13] Further studies have proven that utilization of other frequency components does not enhance discriminative properties of the proposed EEG representation.

The short time Fourier transform (STFT) is applied to obtain the t-f representation of the EEGs analysed in this

Page 3: Brain Computer Interface for Cursor Movement Control by Fuzzy Logic

35 International Conference on Issues and Challenges in Networking, Intelligence and Computing Technologies 2-3 September 2011

KIET, Ghaziabad

work. The segment of each EEG trial between 3s and 8s (5s of an event-related signal which corresponds to 640 samples) is divided into Gaussian windows depending on the settings of two parameters: window length, length and the amount of overlap, lap.

The EEG features are calculated separately for each time segment. The first stage amounts to time averaging of spectral content in the specified frequency bands. Next, the square norm of these mean values is calculated (i.e. the square root of the sum of the components squared), which serves as a sub-element yi l of the feature vector R, where l is an index of a recording channel (i.e. either Cz, C3 or C4). The index l implies that a complete feature vector consists of segments of the normalized spectral EEG content for each signal recorded:

R = (R1,.., R2*Nwin ) (1)

The windowing technique allows for effective control over the architecture of the feature space. Firstly, the number of windows dictates its dimension. Secondly, the window length along with the amount of overlap determines the character of information about the t-f evolution of the relevant EEG components exploited.

D. Classification

Among available FISs [15], [16], we chose the Chiu’s FIS (CFIS) [14]. Indeed, CFIS is robust to noise, which is fundamental when dealing with such noisy data as EEG signals. Moreover, according to Chiu, the CFIS is generally more efficient than neural networks. Finally, it is a clustering-based FIS, making it suitable for dealing with small training sets [15]. With the CFIS, fuzzy “if-then” rules can be automatically extracted from data in three successive steps.

1. Clustering of training data. Aclustering algorithm known as “substractive clustering” [14] is applied to the training data of each class separately. This algorithm enables to find automatically the number of clusters and their positions.

2. Generation of the fuzzy rules. A fuzzy “if-then” rule is generated for each cluster found previously. For a cluster j, belonging to class Cli, the generated fuzzy rule is: if X1 is Aj1 and ; . . . ; and XN is AjN then class is Cli where N is the dimensionality of the data, Xk is the kth element of a feature vector X and Ajk is a Gaussian fuzzy membership functionTo increase accuracy, the membership functions can be “two-sided” Gaussians with a plateau and a different standard deviation on each side [14].

3. Fuzzy rule optimization. Each membership function Ajk is tuned according to gradient descent formulas that use a classification error measure E and a learning rate λ [14]

Once trained, the CFIS can use its set of fuzzy rules to classify any new feature vector X. The class assigned to X corresponds to the class associated with the rule j for which its degree of fulfilment is the highest.

E. Conception of “Hand-Made” Fuzzy Rules

It is possible to add handmade fuzzy rules (HMFR) to a FIS as a priori knowledge. Typical a priori knowledge concerning hand motor imagery EEG data concerns the presence of contralateral ERD in the μ and β bands [17]. A human expert could formalize this knowledge using simple rules.

Using schemes of trials and errors on the training sets, the optimal value for λ was chosen to be 15.32 in the four membership functions. It should be noted that such rules cannot be learnt by the CFIS as they describe relationships between features and not the properties of the features.

III. RESULT

The results obtained with the fuzzy logic in classifying each task are on an average of 78.52% of correct classifications, with a peak of 82% (table I).

TABLE I

FUZZY LOGIC CLASSIFICATION

Subject Percentage of correct

classification

Right hand movement

82

Left hand movement

79.1

Right foot movement 77

Left foot movement

76

The error-reject curve is displayed on Fig. 4 for different subjects. The error-reject curves suggest CFIS is able to identify and reject efficiently the outliers, which makes its error rate drop dramatically.

Fig. 4. The error-reject curve for different subjects

Page 4: Brain Computer Interface for Cursor Movement Control by Fuzzy Logic

36 International Conference on Issues and Challenges in Networking, Intelligence and Computing Technologies 2-3 September 2011

KIET, Ghaziabad

If we compare CFIS by other available methods (like MLP, LC, and SVM) of classification then we can see that other classifiers rejects a lot of feature vectors before reaching a low error rate which means they cannot make a clear distinction between outliers This can be explained by the fact that CFIS is a generative classifier which models explicitly the class boundaries using membership functions. The real time application of cursor movement control using fuzzy as classifier is done.

IV. CONCLUSION

Fuzzy based algorithm showed significant results of the accuracy of the classification (above 80%). This means that using CFIS to classify the data extracted from the brain of the patient is efficient and can improve the quality of any BCI application. Thus, improving the life of disabled people. However, using CFIS algorithm requires top-notch PCs. This research dealt with real-time bio signals processing ability to remove effective noise of the BCI where the analysis required. In addition, CFIS is applied to cursor control by motor imagination of hand and feet and further it can be used for word typing,

and even controlling a wheel chair.

REFERENCES

[1] Shao-Hang, Che-Jui Chang, “Development of Real-time wireless brain computer Interface fordrowsiness detection” IEEE

transaction. on bio-medical engineering, vol. 34, no. 3, pp. 1080–1083, 2010.Data Mining and Knowledge Discovery, vol. 9, pp. 59-87, 2004.

[2] Chin-Teng Lin, Yu-Chieh Chen, Tien-Ting Chiu, “Development of Wireless Brain Computer Interface with embedded multitask scheduling and its application on real-Time driver’s drowsiness

detection and warning” IEEE transaction on Bio-medical engineering, vol. 55, no. 5, pp.1582-1591, 2008.

[3] Kenji Ogawa and Mitsuo Shimotani,”A Drowsyness detection system” Annual. technical reports Rev. Biophys. Bioeng, 1973.

[4] J. R. Wolpaw and D. J. McFarland, “Multichannel EEG-based brain– computer communication,” Electroenceph. Clin. Neurophysiol., vol. 90, pp. 444–449, 1994..

[5] M. Cheng, X. Gao, S. Gao, and D. Xu, “Design and

implementation of a brain–computer interface with high transfer rates,” IEEE Trans. Biomed. Eng., vol. 49, no. 10, pp. 1181–1186, Oct. 2002.

[6] B. Obermaier, “Design and implementation of an EEC based virtual keyboard using hiddenMarkov models” Ph.D. dissertation, Tech. Univ.-Graz, Graz, Austria, 2001

[7] B. Obermaier, C. Neuper, C. Guger, and G. Pfurtscheller, “Information transfer rate in a five-classes brain-computer interface,” IEEE Trans. Neural Syst. Rehabil. Eng., vol. 9, no. 3,

pp. 283–288, Sep. 2001 [8] X. Gao, D. Xu, M. Cheng, and S. Gao, “A BCI-based

environmental controller for the motion-disabled,” IEEE Trans.

Neural Syst. Rehabil. Eng., vol. 11, no. 2, pp. 137–140, Jun. 2003.

[9] G. Edlinger, G. Krausz, F. Laundl, I. Niedermayer, and C. Guger,

“Architectures of laboratory-PC and mobile pocket PC brain-computer interfaces,” in Proc. 2nd Int. IEEE EMBS Conf. Neural Eng., Arlington, VA, Mar. 2005, pp. 120–123.

[10] A. K. Whitchurch, B. H. Ashok, R. V. Kumaar, K. Sarukesi, and V. K. Varadan, “Wireless system for long term EEG monitoring of absence epilepsy,” Biomed. Appl.Micro. Nanoeng., vol. 4937, pp.

343–349, 2002. [11] C. T. Lin, R. C. Wu, T. P. Jung, S. F. Liang, and T. Y. Huang,

“Estimating alertness level based on EEG spectrum analysis,”

EURASIP J. Appl. Signal Process., vol. 19, pp. 3165–3174, 2005.

[12] C. T. Lin, R. C. Wu, S. F. Liang, W. H. Chao, Y. J. Chen, and T.

P. Jung, “EEG-based drowsiness estimation for safety driving using independent component analysis,” IEEE Trans. Circuits Syst. I, vol. 52, no. 12, pp. 2726–2738, Dec. 2005

[13] D. Coyle, G. Prasad, T.M. McGinnity. “A Time-Frequency Approach to Feature Extraction for a Brain-Computer Interface with a Comparative Analysis of Performance Measures,”

EURASIP Journal on Applied Signal Processing, 2005 [14] S. L. Chiu, “An efficient method for extracting fuzzy classification

rules from high dimensional data,” J. Adv. Computational

Intelligence, vol. 1, pp. 31–36, 1997 [15] S. Guillaume, “Designing fuzzy inference systems from data:

Aninterpretability- oriented review,” IEEE Trans. Fuzzy Syst., vol.

9, no. 3, pp. 426–443, Jun. 2001 [16] D. Nauck, “Neuro-fuzzy systems: Review and prospects,” in

Proc. 5th Eur. Congress Intelligent Techniques Soft Computing

(EUFIT’97), 1997, pp. 1044–1053 [17] G. Pfurtscheller and C. Neuper, “Motor imagery and direct brain-

computer communication,” Proc. IEEE, vol. 89, no. 7, pp. 1123–

1134, Jul. 2001