RESEARCH AREAS: Brain Computer Interface, EEG Signal Processing, Blind Source Separation, Machine Learning and Deep learning,  Data Science and Analytics

 TEACHING AND RESEARCH INTEREST

Teaching and research is a continuous learning process. If I am saying that my knowledge has completed, there is something wrong interpretations among people because I have learned from education that I have to learn new subjects by hard working, the best teaching, cognitive skills and analytical thinking. Teachers should have tenacity, desire to learn new subjects. I aspire to become a good teacher and accordingly like to convert my efforts in potential energy.

RESEARCH INTEREST

Many patients are afflicted with neurological conditions or neurodegenerative diseases [1] that disrupt the normal information flow from the brain to the spinal cord and eventually to the targets of that information, i.e., the muscles that affect the person’s intent. Amyotrophic lateral sclerosis (ALS, or also called Lou Gehrig’s disease), spinal cord injury, stroke, and many other conditions impair either the neural pathways controlling muscles, or impair the muscles themselves. Those individuals that are most affected may lose all abilities to control muscles. Thus, they lose all options to communicate and become completely locked-in to their bodies. In absence of reversing the effects of the disorders, there are three principal options for restoring function. The first option is to substitute the damaged neural pathways or muscles with pathways or muscles that are still functional. While this substitution is often limited, it can still be useful. For example, patients can use eye movements to communicate or hand movements to produce synthetic speech. The second option is to restore function by detecting nerve or muscle activity above the level of the injury. For example, the Free hand prosthesis is restoring hand function to patients with spinal cord injuries. The third option for restoring function is to provide the brain with a new and non-muscular output channel, a brain–computer interface (BCI) [1], for conveying the user’s intent to the external world.

INTRODUCTION TO BCI:

The brain machine interface (BMI) is a very recent development in the area of the human machine interaction (HCI) and emerged as the sister technology of BCI. However, there is not a clear cut difference between the BCI and the BMI. In fact, a computer is always between the brain and the machine for the interface. Different mental thoughts generate the electrical potentials over the surface of the scalp. It is recorded with the sensors placed on the surface of the head. A physiological signal related to these electrical potentials in response of the mental thoughts is known as Electro-encephalogram (EEG) [2] signals. The BMI based on the EEG signals is called as the EEG-based BMI [2] system. The BMI pertains to manipulation or operation of the external machine as per thought of a user and such machine is called as thought controlled machine. Thus there is a direct communication at the level of thought for the action between both human and machine via a computer. The BMI is most commonly known as the BCI because there is a direct communication between the brain and the external machine via a computer, which analyses and interprets the incoming physiological signals (electroencephalograms), which contain the shadow of the mental activity and the different types of artefacts. A multi-channel recording of the electromagnetic waves emerging from the neural currents in the brain generate a large amounts of the EEG data.  The neural activity of the human brain recorded non-invasively is sufficient to control the external machine, if advanced methods of signal analysis and feature extraction are used in combination with the machine learning techniques either supervised or unsupervised. A suitable feature extraction and classification methods are useful to generate a control command for controlling the external machine.

OBJECTIVE:

The concept of thought generation in the human brain is a very complex phenomenon and has not been accessible yet directly as the biological processing and the format of the signal in the brain is not clearly understood. However, the levels of thoughts and types of thoughts have been broadly investigated as change in the rhythm and the patterns of the EEG signals capture at the surface of the skull. Although, the EEG signals are not the direct access to the levels of thoughts and thought related actions but it represents the shadow of it and hence to some extent thought related patterns can be inferred and used for a controlling action. However 100% thought recognition will remain untouched for either of the statistical classifier using a single feature vector. For the real time control, the information transfer rate is not sufficient to control the external machine such as a robot, an electric wheelchair and a computer cursor. The challenges are usability of the EEG signals for the cognitive task selection. If it can be used what will be the best approach of signal processing, feature extraction which will maximize the information transfer rate.

A large number of the brain signals have been employed in the design of EEG-based BCI, e.g., event-related potentials (ERPs) and event-related EEG changes [such as event-related desynchronization/synchronization (ERD/ERS in alpha and beta rhythm) [2]-[3]. Among the current BCI systems, the system based on visual evoked potential (VEP) has been studied for a long period since the 1970. We have transient state VEP (TSVEP) and steady state VEP (SSVEP) [3]-[4]. At the time of recording the EEG Signals, the SSVEP signals is being recorded in response of stimulus of certain frequency in front of user. SSVEP signal can provide higher information rate 70 bits/min [5] and also require less training time as compare to other EEG signals. Due to this reason SSVEP based BCI have got more attention as compare to other EEG Signals. Still the information transfer rate is not 100%. What will be the optimal frequency band for which the users have a good interaction and what will be the best signal processing approach for the detection of the periodic waveforms of a certain frequency in the recorded EEG Signals in order to maximize the information transfer rate.

In my doctoral thesis, I concentrated on Blind Source Separation based EEG Signal processing for BCI. I have investigated the effect of the asymmetry property of the brain due to handness on the classification accuracy over the movement imagery data particularly left and right hand movement imagery. I have modified the second order statistics based BSS algorithm under the supervised learning approach for the pre-processing of each class of the movement imagery data.  I have done the offline analysis of the EEG Signals and its classification.

Now we are focusing on the challenges of real time control of an electric wheel chair by thought of a user to assist disable persons. For the same we are interested to develop the ubiquitous computing system that will estimate the thought of user and command the external machine to do the specific task.

METHODOLOGY:

The objective of this research work is to examine the areas of the BCI/BMI with respect to the classification accuracy and the information transfer rate for real time control of electric wheel chair for the disabled person. In order to develop the real time control of electric wheel chair we have to buy good quality EEG amplifier along with readymade electric wheelchair, In this research work, we proposed to investigate the approach which will maximize the classification accuracy, the information transfer rate and minimize the processing for the real time control of the external machine. These are the fundamental problems and require some advanced signal analysis and feature extraction methods, which will efficiently solve these fundamental problems.  Many researchers have estimated the periodic wave forms from the recorded sensor signals. Instead of sensor signals, if we can use the uncorrelated sensor signals for the detection of a sinusoidal periodic waveform will give a good estimation of certain frequency signal. For the same, we can apply the blind source separation algorithm for the decorrelation of sensors signals.  We are also interested to investigate the best power spectral density approach in order to get the peak value of signal at desired frequency. Also we are interested to apply adaptive filtering algorithm to estimate the desired frequency component signals. During sanctioned period of project, we are planning to develop and investigate the novel idea of signal processing, which will maximize the information transfer rate for real time control of electric wheelchair for disabled person.

References:

1.Gerwin Schalk, Dennis J. McFarland, and Jonathan R. Wolpaw, Brain-Computer Interfaces for Communication and Control, Clinical Neurophysiology 113 ,2002, pp.767–791

2. Manoj Kumar Mukul and Fumitoshi Matsuno: Feature Extraction from Subband Brain Signals and Its Classification, SICE Journal of Control, Measurement, and System Integration Vol. 4 , 2011, No. 5 , P 332-340 .

3. “Brain–Computer Interfaces Based on the Steady-State Visual-Evoked Response”:      IEEE Transactions on Rehabilitation Eng., Vol. 8, No.2. , 2000,

4. Henrik Gollee∗, Member, IEEE, Ivan Volosyak, Angus J. McLachlan,Kenneth J.  Hunt, and Axel Gr¨aser “An SSVEP-Based Brain–Computer Interface for the Control of Functional Electrical Stimulation” IEEE Transactions on Biomedical Engineering, Vol. 57, No. 8,  2010, pp.1847-1855.

5. Ming Cheng, Xiaorong Gao, Shangkai Gao and Dingfeng Xu “Design and Implementation of a Brain-Computer Interface with High Transfer Rates”:  IEEE Transactions on Biomedical Eng., Vol. 49, No. 10, 2002. Pp.1181-1186