A. Giri research focuses on the low computational cost of active brain source localization to prevent delay in the diagnosis of epileptic seizure location. The localization performance is limited by the head shape assumption, as the EEG data is spatially sampled over the head for efficient data representation. In literature, the human head is approximated by spherical shape. Hence, spherical harmonics have been the natural choice for EEG source reconstruction and localization. She developed a set of basis functions called Head Harmonics, which improves the quality of non-invasive source localization when compared to spherical or spatial domain processing. Her future research investigates the applicability of developed harmonics under realistic head modeling scenario. She is also interested in developing a head harmonics based adaptive EEG acquisition system prototype.

Anatomical Surface Reconstruction using Hemispherical Harmonics

An accurate representation of a three-dimensional (3D) geometry is crucial for structural analysis in many biomedical applications. In medical shape analysis, there exists a wide range of hemisphere-like anatomical structures such as brain, skull and scalp, which are naturally parameterized using the upper hemisphere only. The representation of such hemispherical objects using basis functions defined over the full spherical domain introduces discontinuities at the boundary of the hemisphere and requires a large number of coefficients. Therefore, the two hemispherical area-preserving parameterization methods for simply-connected open and closed surfaces is developed. The hemispherical harmonics basis functions are therfore utilized to yield an accurate representation of hemisphere-like anatomical surfaces. In near future work, she is interested in doing the adaptive parametrization (not constrainede to be on hemisphere or sphere) of input surfaces and use corresponding harmonics for shape description.

Cortical Source Domain based Brain Computer Interface (BCI)

In BCI, decoding the motor task from non-invasive EEG measurements is a challenging problem. It is due to the fact that encoding is assumed to be deep within the brain and is not easily accessible by the scalp recordings. The ability to know the source generators of the intended motor task from EEG may lead to huge improvements in BCI by providing a continuous task relevant neural signals. To overcome these issues and to study the brain activity on the motor cortex, cortical source domain processing is proposed. Her finding emphasize to use the spatial source distribution knowledge in neuro feedback training of BCI systems. She is also interested on enhancing the user's control abilities, as the performance of any BCI system relies heavily on a user's attention level and ability to modulate sensory motor rhythms. Since, considerable progress has been made in "computer" side, while little work has been done on "brain" perspective. She believes her work opens a wide range of possiblities to patients suffering from neuromuscular disabilities, stress and anxiety.

Kinematic Trajectory Prediction from Brain Signals

The ability to reconstruct the kinematic parameters of hand movement using non-invasive electroencephalography (EEG) is essential for external device control. For system development, the conventional classification based brain computer interface (BCI) controls external devices by providing discrete control signals to the actuator. A continuous kinematic reconstruction from EEG signal is better suited for practical BCI applications. The state-of-the-art multi-variable linear regression (mLR) method provides a continuous estimate of hand kinematics, achieving maximum correlation of upto 0.67 between the measured and the estimated hand trajectory. Three novel source aware deep learning models multi layer perceptron (MLP), convolutional neural network - long short term memory (CNN-LSTM), and wavelet packet decomposition (WPD) CNN-LSTM are proposed for motion trajectory prediction (MTP). Our methods provide statistically significant improvement in kinematic trajectory from EEG signals compared to existing state-of-the-art mLR method. Our work bridges the gap between the control and the actuator block, enabling real time BCI implementation.