Computer Sciences and Information Technology
Title : | Development of robust multi-sensor non-stationary data analysis methods using array signal processing and deep learning techniques for electroencephalogram (EEG) signal analysis. |
Area of research : | Computer Sciences and Information Technology |
Principal Investigator : | Dr. Abhijit Bhattacharyya, National Institute Of Technology (NIT) Hamirpur, Himachal Pradesh |
Timeline Start Year : | 2022 |
Timeline End Year : | 2025 |
Contact info : | abhijit@nith.ac.in |
Equipments : | Laptop
Computer Desktop |
Details
Executive Summary : | Multiple sensors have become increasingly used in various fields to measure real-world physical phenomena, such as sonar, telecommunications, and biomedical instruments. These sensors are used to exploit spatial diversity and hypothesize spatial characteristics in recorded signals. However, these signals are often recorded under low signal to noise ratios, which can be contaminated with noise. The primary objective is to extract underlying information from these signals for various purposes, such as change detection and signal classification. A multi-stage approach for processing multidimensional signals includes preprocessing, transformation, information measurement, and discrimination. In many real scenarios, recorded multi-sensor signals show non-stationary nature, as their spectral characteristics vary over time. This can lead to better estimation of parameters related to time-varying characteristics of the applied signals. This project aims to propose, combine, and extend signal processing and deep learning techniques for analyzing multi-sensor recorded signals. The focus will be on considering synergically both the spatial information provided by the array of sensor signals and the time-varying nature of individual sensor/channel data. Existing signal processing methods for nonstationary signal analysis are univariate and not efficient for utilizing spatial information in multi-sensor recordings. The project will adapt non-stationary array signal processing for multi-sensor time-frequency representations and combine it with deep neural networks to build robust methodologies suitable for various applications, including analyzing multiple-sensor EEG signals recordings. |
Total Budget (INR): | 19,22,456 |
Organizations involved