Research Area

Compressed sensing MRI (CS-MRI) is successful in reducing the MRI scan time by two to five times. It takes only a few measurements in the frequency domain and then applies highly nonlinear recovery algorithms to reconstruct high-resolution MR images from the partial data. L1-norm-based optimization algorithms in convex optimization are popular for the reconstruction of MR images from partial Fourier data as they are stable, guarantee convergence at large scale, and are efficient, and standard solvers are readily available. Recently, fast convex optimization-based reconstruction algorithms have been developed which are quite competent to achieve the standards for the use of CS-MRI in clinical practice as well.


Underwater signal processing for target tracking involves using sensors like sonar systems and hydrophones to detect and localize objects underwater. Key techniques include beamforming, matched filtering, and Fourier and wavelet transforms, while tracking algorithms like Kalman and particle filters, and multi-hypothesis tracking, are used to estimate and follow target movements. The field faces challenges such as signal attenuation, multipath propagation, and environmental dynamics but has seen advancements through machine learning, improved sensors, and networked systems. Applications span military, commercial, and environmental sectors, from submarine and mine detection to marine wildlife tracking and underwater infrastructure monitoring.


Wireless body area networks (WBANs) are increasingly used for remote healthcare surveillance in recent times, where electrocardiogram (ECG) signals are continuously acquired and transmitted to a base station or remote hospital for their storage and subsequent analysis. Multichannel ECG (MECG) is preferred over single-channel ECG as it provides more information from diagnostic point of view. One of the biggest challenges is to minimize the energy required for the WBAN network for continuous transmission of MECG data, which in turn demands for efficient data compression. Compressive sensing is an efficient signal processing tool for simultaneous compression and reconstruction of MECG data without visibly no or minimum loss of diagnostic information. 

The emerging technology and advancement to record the ECG signal without the help of the medical experts in home care or ambulatory conditions with minimal complexity have become more common in recent times. To reduce the intricacy and enhance patient comfort, it would be better to have minimal lead sets to derive the standard 12-lead ECG.


In MRI, data acquisition is carried out in k-space, which is equivalent to center-shifted Fourier transform of the corresponding MR image. During CS measurements in the k-space domain, center region of k-space is densely acquired, which represents gross structure of the corresponding anatomical structure. On the other hand, peripheral region of the k-space is sparsely acquired, which represents the high-frequency information i.e. fine details, edges, and boundaries. As a result, the diagnostic quality of the CSMRI reconstructed images are degraded. Another vital issue associated with most of the MR images are due to low resolution of the images. These limitations led to serious difficulties in clinical investigation and inaccurate diagnosis.


Satellite image super-resolution is the process of increasing the resolution or sharpness of satellite images. Joint dictionary learning is a technique used in super-resolution of satellite images that involves learning a common dictionary of image patches from both low-resolution and high-resolution images. This approach aims to capture the underlying structure and features of the images in a way that is shared between both resolutions, enabling the creation of a high-resolution image from a low-resolution input. Joint dictionary learning is a powerful technique for super-resolution of satellite images because it can leverage the shared information between both the low-resolution and high-resolution images, resulting in higher-quality super-resolved images. However, it is computationally expensive and requires a large amount of training data to learn the dictionary.