Projects

Adaptive data-compression and noise suppression filtering

Developed an optimized signal processing model to compress the transient electromagnetic system signals and suppress the effect of power-line, very-low-frequency, and vibrational noise for real-time underground water mapping in the areas like Canada, Denmark, India, Africa, etc.

NeuroNet: A DeepLearning Framework

Created an optimum time-frequency analysis toolbox to extract instantaneous information about the time-frequency content of nonstationary biomedical signals. A simple and effective deep-learning model has been designed to detect brain disorders and real-time physiological activities.

AI-based BCI systems

Designed a real-time low complexity brain activity detection (Motor Imagery Tasks, Human Emotions, Sleep Stages, Mental States) strategy using adaptive signal analysis, supervised, and unsupervised techniques. In addition, adaptive wavelets and data compression models have been presented for detecting heart disorders and real-time sound classification.

AI-based Tracking System for GPS-denied Areas

Developing a real-time hardware model for an automated detection of GPS denied areas using time series analysis with supervised machine learning techniques and effective prediction of these locations using recurrent neural network for internet of military battlefield things.

Inference Engines for Edge Gateways

Lightweight, power-efficient, real-time, versatile, flexible ML IP Cores for edge gateways. The deep learning-based inference engine deployed on AMD Xilinx FPGA for generalized applications that demands edge intelligence.

Data-driven real-time drowsiness detection

Developed adaptive and data-driven signal analysis and classification models for accurately detecting drowsiness from a real-time electroencephalogram signal.

ATMEGA16 Micro-controller based Human Controlled Wireless Programmable Robot Arm

ATMEGA16 Micro-controller-based Human Controlled Wireless Programmable Robot Arm. The robotic arm copies the movement of the human hand. We generate binary data according to the human hand movement by using variable resistance. This analog signal will then process by a microcontroller (ATMEGA16) and then gives to a mechanical Robotic Arm, which offers robotic arm movement according to the human hand. In this system, we use gear motors and USART wireless communication system leads to a cost-effective and lightweight robot. (BE, Jan 2012)

Motion Display: Persistence of vision-based LED display on fan

The basis of this project was the course studied during the third and fourth semesters of our B.Tech course. The basic idea behind this was the use of IC555 and LEDs that follow the principle of persistence of vision and conversion of one-dimensional string to two-dimensional data with the help of a rotatory object e.g. Fan. This prototype developed using basic components has won many national-level projects in technical events. Multiple prototypes have been tested that cost less than five use dollars. (BE, Dec 2010).