HYBRID BCI SYSTEM
Keywords: EEG, EOG signals, Biomedical Signal Processing, Machine Learning, Decision Fusion, Real-Time Progressing
Timeline: june 2019 - Today
The project aims to provide a realistic and economical system for rehabilitation using a hybrid BCI paradigm. This project aims to mix EOG and EEG signals obtained from the user to accurately estimate the movement intension using Decision Fusion methods like Kalman Filter for EOG-EEG based signals by using attention level for activation.
Currently, the project is under development phase
TDOA - HYDROPHONES
Keywords: Underwater acoustics, Pre-Processing Signals, Signal Processing, Embedded systems, AUV, Real-Time Processing
​
Timeline: May 2019 - today
This project aims to find the pinger location underwater which transmits a particularly high-frequency acoustic signal periodically in a short window as specified during AUV competitions.It uses TDOA ( Time Difference of Arrival) as its major unit to calculate the angle. The module provides an OpenSource start for all the enthusiasts working on pinger localization's real-time implementation, Following were tasks implemented in the project
* Worked on the Acquisition Board for preprocessing of hydrophone signals
* Worked on ADC interfacing for through bcm2835 library on Rpi-3b
* Extensive use of MATLAB for Algorithm testing using packages like DSP Toolbox, Phased Array toolbox, Filter Builder and Python for Data Analysis with matplotlib library along with numpy
* Implemented the Algorithm using C++ and RasberryPi- 3b using FFTW and Liquid-DSP library for real-time implementation in the AUV
IRIS FLOWER CLASSIFIER
Topics: DataAnalysis, Machine Learning, Python
A small vectorised neural network classifier with regularisation for classification of the dataset. This project was a Capstone to Machine Learning course by Standford University. In this project,
* data extraction and organisation was required
* A classifier was designed and optimised for great accuracy up to 95%
* multi-class classification using neural network was successfully implemented
* NO INBUILT functions were used to explore the mathematical aspects of a network, just numpy and pandas were used for vectorised implementation and processing
CRUSADE
Topics: Computer Vision, Robotics, Controller, Embedded System, Programming
An Image processing based Lane project for the IIT Kharagpur tech-fest event "Crusade" Under Cyborg robotics and automation society of NIT Rourkela. This project required the following aspects
* Design of hardware for motor control
* Image processing for feature extraction using OpenCV
* Using the features for path planning and manoeuvre control
* PID algorithm for control of the trajectory of bot
* Entire Code was implemented on Rpi-3b for on board processing
got the BEST ALGORITHM AWARD in the competition