Vinh Dang

Parallelizing Fast Multipole Method for Large-Scale Electromagnetic Problems Using GPU Clusters

This letter investigates the solution of large-scale electromagnetic problems by using the single-level Fast Multipole Method (FMM). Problems of large scale require high computational capability that cannot be accommodated using conventional computing systems. We investigate a parallel implementation of FMM on a 13-node graphics processing unit (GPU) cluster that populates Nvidia Tesla M2090 GPUs. The implementation details and the performance achievements in terms of accuracy, speedup, and scalability are discussed. The experimental results demonstrate that our FMM implementation on GPUs is much faster than (up to 700 ×) that of the CPU implementation. Moreover, the scalability of the GPU implementation is very close to the theoretical linear expectations.

Analysis of Moving Human Micro-Doppler Signature in Forest Environments

Automatic detection of human motion is important for security and surveillance applications. Compared to other sensors, radar sensors present advantages for human motion detection and identiflcation because of their all-weather and day-and-night capabilities, as well as the fact that they detect targets at a long range. This is particularly advantageous in the case of remote and highly cluttered radar scenes. The objective of this paper is to investigate human motion in highly cluttered forest medium to observe the characteristics of the received Doppler signature from the scene. For this purpose we attempt to develop an accurate model accounting for the key contributions to the Doppler signature for the human motion in a forest environment. Analytical techniques are combined with full wave numerical methods such as Method of Moments (MoM) enhanced with Fast Multipole Method (FMM) to achieve a realistic representation of the signature from the scene. Mutual interactions between the forest and the human as well as the attenuation due to the vegetation are accounted for. Due to the large problem size, parallel programming techniques that utilize a Graphics Processing Unit (GPU) based cluster are used.

Joint DoA-range-Doppler tracking of moving targets based on compressive sensing

In this paper, the tracking of moving targets using a stepped frequency linear antenna array is addressed through the compressive sensing (CS) framework. Multiple targets are resolved in the three-dimensional directions of arrival (DoA)-range-Doppler space for each time instant. The Orthogonal Matching Pursuit (OMP) algorithm is utilized to reconstruct the target space using measurements at a small number of randomly selected frequencies. The performance of the proposed approach is evaluated through simulation result in terms of accuracy and computational complexity.

GPU acceleration of nonlinear diffusion tensor estimation using CUDA and MPI

Diffusion MRI is a non-invasive magnetic resonance technique and has been increasingly used in imaging neuroscience. It is currently the only method capable of depicting the complex structure of white matter of the brain in vivo. One of the most popular techniques is diffusion tensor imaging (DTI) which is commonly used clinically to produce in vivo images of biological tissues with local microstructural characteristics such as water diffusion. Diffusion tensor maps are usually computed on a voxel-by-voxel basis by fitting the signal intensities of diffusion weighted images as a function of their corresponding data acquisition parameters (b-matrices). This processing is computation-intensive and time-consuming which can constraint the clinical practice of DTI. This study presents the application of using high performance GPU clusters in addition to CPUs for diffusion tensor estimation by accelerating the multivariate non-linear regression. The results are tested using both simulated and clinical diffusion datasets and show significant performance gain in tensor fitting. The proposed GPU implementation framework can significantly reduce the processing time of DTI data especially for the datasets with high spatial and temporal resolution, and hence further promote the clinical use of DTI. It also can be used to accelerate statistical analysis of DTI where Monte Carlo simulations are employed, be readily applied to quantitative assessment of DTI using bootstrap analysis, robust diffusion tensor estimation and should be applicable to other MR imaging techniques that use univariate or multivariate regression to fit MRI data to a model.

GPU Cluster Implementation of FMM-FFT for Large-Scale Electromagnetic Problems

The fast multipole method (FMM) combined with fast Fourier transform (FFT) is investigated for the solution of large-scale electromagnetic problems, which require high computational capability that cannot be accommodated using conventional computing systems. The implementation is parallelized on a 13-node graphics processing unit (GPU) cluster that populates Nvidia Tesla M2090 GPUs. The experimental results based on our FMM-FFT implementation on GPUs demonstrate up to 957 times speedup compared to that of the single-core, single-node CPU implementation. The implementation details and the performance achievements in terms of accuracy, speedup, and scalability are discussed.

An Elegant Solution: An Alternative Ultra-Wideband Transceiver Based on Stepped-Frequency Continuous-Wave Operation and Compressive Sensing

For more than 50 years, since early work beginning in 1960, ultra-wideband (UWB) technology has been developed to analytically verify the transient behavior of a class of transverse electromagnetic mode microwave networks. The U.S. Federal Communications Commission has defined UWB as operating with a minimum fractional bandwidth of 20% or with a minimum -10-dB bandwidth of 500 MHz.

Compressive sensing based approach for detection of human respiratory rate

In this paper, the non-invasive detection of human respiratory rate using a stepped-frequency continuous wave (SFCW) radar is addressed through the compressive sensing (CS) framework. Range profiles and the respiratory signatures are resolved using measurements at small numbers of randomly selected frequencies and slow-time samples. The performance of the proposed approach is evaluated through simulation results in terms of accuracy and efficiency.

Detection of moving human micro-Doppler signature in forest environments with swaying tree components by wind

The objective of this paper is to investigate human motion in forest medium with swaying tree components due to time-varying wind effects and to observe the characteristics of the received Doppler signature from the scene. We provide the results of an accurate model accounting for the key contributions to the Doppler signature in this scenario. A realistic walking motion is generated using an analytical model extracted from empirical data. The swaying canopy motion is modeled by employing a spring response mechanism to the wind force. The backscattered field calculations from the scene comprise of contributions from the forest (including trunks, branches, and the ground) and human, and the interactions between them. An analytical forest scattering model, which accounts for the ground effects, is used to calculate the contribution from the forest. The attenuation effects due to the vegetation are accounted for. In order to characterize the effects of human motion accurately, a full wave technique, namely, method of moments (MOM) enhanced with fast multipole method (FMM), is employed for the human scattering calculations. A parallel version of MOM-FMM is implemented on a graphics processing unit based cluster to handle the large problem size. The human walking signatures created by the model are analyzed for different winds.

Graphics processing unit accelerated Fast Multipole Method - Fast Fourier Transform

In this paper, we investigate the parallelization of the single level Fast Multipole Method (FMM) combined with Fast Fourier Transform (FFT) for solving large-scale electromagnetic scattering problems. We evaluate the performance on our 13-node GPU cluster supported by Nvidia Tesla M2090 GPUs. The paper discusses the implementation details and the performance achievements in terms of accuracy, speed up and scalability.

Accelerating nonlinear diffusion tensor estimation for medical image processing using high performance GPU clusters

Diffusion Tensor Imaging (DTI) is a non-invasive magnetic resonance technique that produces in vivo images of biological tissues with local microstructural characteristics such as water diffusion. It can be used, for example, to localize white matter lesions, or in neuro-navigation surgery of brain tumors. Diffusion tensor maps are usually computed on a voxel-by-voxel basis by fitting the signal intensities of diffusion weighted images as a function of their corresponding data acquisition parameters. This processing is highly computation-intensive and can be time-consuming which constraints the clinical use of DTI. This study presents the application of using high performance GPU clusters in diffusion tensor estimation by accelerating the multivariate non-linear regression. The results are tested in simulated DTI brain datasets and show significant performance gain in tensor fitting in addition to favorable scalability characteristics. The proposed GPU implementation framework can further promote the clinical use of DTI, and can be used to accelerate statistical analysis of DTI where Monte Carlo simulations are employed, or readily applied to quantitative assessment of DTI using bootstrap analysis.