Project Proposal

EEG Based action classification

Pratyush Sinha

Mentor: Dr. Amitabha Mukherjee

 

Introduction:

Cognition depends on neural activity. To study cognition, we need a way to study the neural activity. Electroencephalograms or EEGs are electrical signals created due to the activity of neurons in the brain. These can be recorded non-invasively from outside the scalp. EEG is an important tool in the diagnosis of functional brain disorders, and in sleep and epilepsy research.

EEG are complex spatiotemporal signals. Their statistical properties depend on the state of the subject and on external factors. Sensory stimuli, cognitive tasks, motor tasks etc induce changes in the EEG activity. These may either be an increase in the power corresponding to a certain frequency band known as event-related synchronization (ERS) or the decrease in the power corresponding to certain frequency band known as event-related desynchronization (ERD).

 

Capture.PNG

Figure-1 Running power spectra computed for three frequency bands of an EEG recording.

Taken from : http://cognet.mit.edu/library/erefs/arbib/images/figures/A076_fig002.gif

 

Motivation:

EEG signals can be analysed to discriminate between different actions or imaginations. As shown by Pfurtscheller et al. 1997, EEG signals can be used to classify the imagination of motor actions by left and right hands. I plan to do something similar.

Proposed Methodology:

I plan to use a free online database (such as the one available at Physionet). The database contains recordings of EEG signals corresponding to motor actions and their imaginations for both the left and the right hands.

I will do an ERD/ERS analysis of the signals for various frequency bands. I will also try to find if such a characteristic is invariant across subjects using statistical correlation.

Further possible work:

The project can be extended to analyse motor imaginations with more than 2 possible outputs (Ex- motion in 2D). van Gervan et al. 2009 have tried to do something similar.

 

References:

1.       EEG and MEG Analysis - Fernando H. Lopes da Silva and Jan Pieter Pijn

2.       Event-Related Potential – Steven L. Bressler

3.       EEG-based discrimination between imagination of right and left hand movement

4.       Physionet (G. Pfurtscheller, Ch. Neuper, D. Flotzinger, M. Pregenzer)

5.       Attention modulations of posterior alpha as a control signal for two-dimensional brain–computer interfaces (Marcel van Gerven, Ole Jensen)