Naushad Ansari

Impulse denoising for hyper-spectral images

In this work we propose a technique to remove sparse impulse noise from hyperspectral images. Our algorithm accounts for the spatial redundancy and spectral correlation of such images. The proposed method is based on the recently introduced Blind Compressed Sensing (BCS) framework, i.e. it empirically learns the spatial and spectral sparsifying dictionaries while denoising the images. The BCS framework differs from existing CS techniques that employ fixed sparsifying basis; BCS also differs from prior dictionary learning studies which learn the dictionary in an offline training phase. Our proposed formulation has shown over 5dB improvement in PSNR over other techniques.

Signal-matched wavelet design via lifting using optimization techniques

This paper proposes design of signal-matched wavelets via lifting. The design is modular owing to lifting framework wherein both predict and update stage polynomials are obtained from the given signal. Successive predict stages are designed using the least squares criterion, while the update stages are designed with total variation minimization on the wavelet approximation coefficients. Different design strategies for compression and denoising are presented. The efficacy of matched-wavelets is illustrated on transform coding gain and signal denoising.

Image Reconstruction Using Matched Wavelet Estimated From Data Sensed Compressively Using Partial Canonical Identity Matrix

This paper proposes a joint framework wherein lifting-based, separable, image-matched wavelets are estimated from compressively sensed images and are used for the reconstruction of the same. Matched wavelet can be easily designed if full image is available. Also compared with the standard wavelets as sparsifying bases, matched wavelet may provide better reconstruction results in compressive sensing (CS) application. Since in CS application, we have compressively sensed images instead of full images, existing methods of designing matched wavelets cannot be used. Thus, we propose a joint framework that estimates matched wavelets from compressively sensed images and also reconstructs full images. This paper has three significant contributions. First, a lifting-based, image-matched separable wavelet is designed from compressively sensed images and is also used to reconstruct the same. Second, a simple sensing matrix is employed to sample data at sub-Nyquist rate such that sensing and reconstruction time is reduced considerably. Third, a new multi-level L-Pyramid wavelet decomposition strategy is provided for separable wavelet implementation on images that leads to improved reconstruction performance. Compared with the CS-based reconstruction using standard wavelets with Gaussian sensing matrix and with existing wavelet decomposition strategy, the proposed methodology provides faster and better image reconstruction in CS application.

Statistical analysis and parametric yield estimation of standard 6T SRAM cell for different capacities

In advanced CMOS technologies large-scale integration has enabled larger embedded memory capacity in SoCs and it has also necessitated the Static Random Access Memory (SRAM) bitcell qualification requirement of the order of 0.1ppb. This paper presents a qualitative statistical analysis of a 6T standard SRAM cell in read cycle with respect to Static Noise Margin (SNM) due to process parameter fluctuation. The Yield (Y) of SRAM is predicted for different capacities of SRAM array by modeling success/failure boundary through mathematical modeling for one cell. With this frame work, it is demonstrated that the yield can be accurately predicted by increasing the order of the polynomial. The obtained results show that for the first order approximation, the failure probability of a single cell is 2.36×10-6 whereas the failure probability of an SRAM can be decreased to 8.38×10-13 if the success/failure boundary is modeled with a polynomial of order 4.

Tracking the underwater acoustic channel using two-dimensional frequency sampling

Rapidly fluctuating multipath arrivals along with unpredictable surface wave focusing events render the shallow water acoustic channel difficult to track using sparse or least-squared error (LSE) optimization techniques. This fundamental bottleneck is primarily due to the time-varying nature of the underlying distribution. In this work, we propose a complementary channel tracking technique that exploits the dual representation of the acoustic channel in the Fourier domain and employs two-dimensional frequency sampling using an application-inspired input dictionary. Specifically, we reformulate the time-varying channel tracking problem on a MIMO framework and design training symbols that sample the channel in its dual Fourier domain. Ground truths based on experimental field data are presented.

Physics inspired CS based underwater acoustic channel estimation

We propose real-time channel tracking for underwater acoustic communications under dynamic sea conditions. The key idea is to employ sophisticated sparse sensing techniques that are cognizant of stable or slowly time-varying channel components against a transient background. Shallow water acoustic channel is generally challenging to track under moderate to rough sea conditions. This is primarily due to non-stationary highly transient elements within the channel delay spread resulting from rapidly fluctuating multipath arrivals from unpredictable surface wave reflections. The proposed channel estimation method exploits two channel characteristics: (i) Inherent sparsity of the time-varying channel in the two-dimensional dual (Fourier) domain; and (ii) Relative dominance of the direct arrival and slowly varying multipath arrivals against the otherwise non-stationary channel impulse response. Specifically, we utilize this apriori information to compressed sensing (CS) framework and thus, achieve channel sensing cognizant of time-frequency localization across significant channel taps. Numerical evidence based on data-driven channel ground truths are presented.

Joint Framework for Signal Reconstruction Using Matched Wavelet Estimated from Compressively Sensed Data

In compressive sensing (CS), signal xN×1 is recovered from its fewer measured samples yM×1 in noise free case using optimization framework, where AM×N is the measurement matrix and N > M. Here, the original signal x is known to be sparse in some domain, say the wavelet domain W. Any wavelet can be used as the sparsifying basis, but wavelet matched to a given signal should yield better reconstruction results compared to any existing wavelet because it is supposed to provide best representation of a given signal. So far, no method exists in literature for the estimation of matched wavelet from compressively sensed data that can also be utilized at the same time for the efficient signal reconstruction of a compressively sensed signal. In this work, we address this problem.

Shallow water acoustic channel estimation using two-dimensional frequency characterization.

Shallow water acoustic channel estimation techniques are presented at the intersection of time, frequency, and sparsity. Specifically, a mathematical framework is introduced that translates the problem of channel estimation to non-uniform sparse channel recovery in two-dimensional frequency domain. This representation facilitates disambiguation of slowly varying channel components against high-energy transients, which occupy different frequency ranges and also exhibit significantly different sparsity along their local distribution. This useful feature is exploited to perform non-uniform sampling across different frequency ranges, with compressive sampling across higher Doppler frequencies and close to full-rate sampling at lower Doppler frequencies, to recover both slowly varying and rapidly fluctuating channel components at high precision. Extensive numerical experiments are performed to measure relative performance of the proposed channel estimation technique using non-uniform compressive sampling against traditional compressive sampling techniques as well as sparsity-constrained least squares across a range of observation window lengths, ambient noise levels, and sampling ratios. Numerical experiments are based on channel estimates from the SPACE08 experiment as well as on a recently developed channel simulator tested against several field trials.

Underwater acoustic channel estimation via CS with prior information

This paper proposes estimation of underwater acoustic channel in the delay-Doppler domain under dynamic sea conditions. The sparsity of the channel in the delay-Doppler domain is exploited via advanced compressive sensing (CS) technique to estimate the channel. The proposed use of CS with prior information takes care of relatively dominant but stable slowly time-varying channel component and the rapidly fluctuating high energy transients. Simulation results are presented on prior estimated channel of SPACE08 experiment, considered as ground truth, for the validation of the theory presented.

Lifting-based rational wavelet design from a given signal

This paper proposes design of rational wavelet filterbank via lifting scheme. We introduce the concept of rate converters in lifting filters, while designing predict and update stages. This design approach to rational wavelet has all the advantages of lifting, including perfect reconstruction and the design being modular. In addition, both predict and update stage polynomials are obtained from a given signal making the proposed scheme a signal-matched wavelet design approach.