Noise level predictions in Human perceptual system

NeuroMatch Academy, 2021 (Computational Neuroscience online summer school)

We quantified neural traces that represent noise present in the visual stimulus using a Shannon entropy-based metric, on the ECoG data, which contained neural activity while subjects viewed face images with different amounts of noise. The quantification was evaluated with a random forest classifier.

Information maximing GAN regualrized ADMM for CS-MRI reconstruction

Advisor: Dr. Deepak Mishra & Dr. J Sheeba Rani, Master’s thesis project, IIST, Trivandrum.

We trained an Info-GAN on MRI brain images which learns the statistics of the T1 brain images, and we futher used this trained model in an Alternating direction method of multipliers optimization paradigm to reconstruct the Compressively sensed MRI Images. Which could potentially reduced the time of MRI acquisition.

Cognitive Load classification using EEG images

Master’s Course Project (course: Deep Learning for Computer Vision)

EEG time series is converted to EEG spectral topography map (EEG Images) which has combined spatial, temporal and frequency domain information. These maps are used for cognitive load classification, using CNNs. This project reproduces the results from ”Bashivan, et al, 2015, Learning Representations from EEG with Deep Recurrent-Convolutional Neural Networks”.

Comparative analysis of receiver position computation algorithms

Advisor: Prof. K Sudarshan Reddy, Bachelor’s thesis project, CBIT, Hyderabad.

A comparative analysis was performed on different GNSS positioning algorithms: Bancroft algorithm, Least mean squares algorithm and Recursive least squares, using IRNSS (Indian Regional Navigation sattelite system) constellation data encoded in RINEX format.