Rapid growth in data collection has led to a need for reconditioning the underlying computation and storage solutions. In data-intensive workloads involving machine learning and large-scale simulations, the computations are shifting from CPUs to GPUs. The overhead of input and output (IO) operations in these workloads during the data transfer between GPU and storage becomes a new performance bottleneck. It has thus become evident that the traditional approach of data transfers between GPU memory and device storage that involves CPU as a buffer, has limited GPUs' ability to utilize their vast resources efficiently. Additionally, the bounce buffer approach wastes CPU cycles spent on transferring data. These paths introduce latency, and reduce overall GPU processing performance, especially if the CPU does not need to utilize the data. To solve the resulting bottlenecks, a new technology, NVIDIA GPUDirect Storage (GDS) supported on NVIDIA GPU, accelerates GPU-storage communication by establishing a direct path between local NVMe or remote storage and GPU memory. In this work, we quantify high throughput gains, low latency achievements, and CPU utilization savings of GDS technology with Weka as remote storage cluster. We experiment with various workloads and identify workloads that benefit the most from these employed technologies. We systematically measure the performance of synthetic representative data read and real machine learning workloads optimized with GDS. Finally, we utilize the findings to discuss correlated implications to systems with GDS. Demonstration results are twofold: a) A testing methodology for the performance evaluation of a GPU client with GDS supported file systems for local NVMe and Weka, b) Establish a baseline level of GDS performance for local and remote storage devices.

In this paper, we introduce two indoor Wireless Local Area Network (WLAN) positioning methods using augmented sparse recovery algorithms. These schemes render a sparse user's position vector, and in parallel, minimize the distance between the online measurement and radio map. The overall localization scheme for both methods consists of three steps: 1) coarse localization, obtained from comparing the online measurements with clustered radio map. A novel graph-based method is proposed to cluster the offline fingerprints. In the online phase, a Region Of Interest (ROI) is selected within which we search for the user's location. 2) Access Point (AP) selection and 3) fine localization through the novel sparse recovery algorithms. Since the online measurements are subject to inordinate measurement readings, called outliers, the sparse recovery methods are modified in order to jointly estimate the outliers and user's position vector. The outlier detection procedure identifies the APs whose readings are either not available or erroneous. The proposed localization methods have been tested with Received Signal Strength (RSS) measurements in a typical office environment and the results show that they can localize the user with significantly high accuracy and resolution which is superior to the results from competing WLAN fingerprinting localization methods.

Thermal image quality assessment is essential in evaluating the performance of thermal imaging systems. The existing no-reference methods can be roughly categorized using quantification measures as follows: a) Human Visual System measures, b) information measures, and c) image distribution measures. All the above classes of measures are not appropriate for thermal imaging because the structure of thermal images is quite different from visible light images. In thermal images, the temperature of the background is high, temperature difference between objects and the background is small, and it is difficult to determine the link between objective and subjective image quality assessment. There is a need in new thermal image quality measures which can connect objective representation performance and thermal image characteristics. This paper presents new no-reference thermal-imaging measures and elaborates on their applications. They integrate advantages of above-mentioned conventional of measures. Simulation results demonstrate effectiveness of the presented measures in evaluating thermal image qualities in comparison with other well-known assessment tools. Finally, conclusions on more general applicability of these measures are presented.

Thermography is an imaging technique which is based on the idea that the temperature rises in regions with increased blood flow and metabolism, which could be a sign of a tumor. Recently, it has found promising applications in cancer diagnosis, particularly, in breast cancer exposure. Even though many thermo-medical applications need a good image assessment tool, there were no reports in literature on using a thermal image as a thermal conditioning of breast cancer progress/treatment. In this paper, a novel thermal images assessment tool for detecting the conditions of breast cancer exposure based on thermography is proposed. The presented innovative measure is based on integration of human visual system, intensity and density attributes of observed images. The proposed scheme as thermal conditions of breast cancer has the advantages of: (i) determining cancer risks for prevention by introducing a measuring system for estrogen imbalance, (ii) detecting the progress and treatment by monitoring the conditions as a new measurements and (iii) breast skin temperature measurement. The experiments show the effectiveness of the proposed measuring system on thermal conditions of breast cancer.

The Global Positioning System (GPS) has been the most popular choice for positioning in open-sky environments, but its difficulty in dealing with signal blockage from buildings, trees and the like has limited its operation in urban environments, underground and especially indoors. Recent advancements in location technology have led to the development of Assisted GPS (A-GPS) technology that can operate satisfactorily in many challenging environments using aiding information. As technology testing is a major development component, this paper describes a low cost approach to augment conventional GPS simulators with A-GPS testing support. The A-GPS support is provided using the concept of Wi-Fi positioning technology and IP address location. The communication between the simulator and the receiver is in accordance with the Secure User Plane Location (SUPL) protocol. Currently only high-end simulators support testing of A-GPS receivers as it involves assistance generation and delivery to the receivers in addition to GPS signal generation.

This paper presents a case study implementation of a fading channel model for a recently introduced Global Positioning System (GPS) simulator from National Instruments. Existing models are discussed and implementation aspects are presented for a model which combines statistical properties of different multipath channels. The NI's GPS simulator is implemented in an open development environment, LabVIEW, which allows an incorporation of user-defined models. Computational optimization issues are also discussed.

Information technology over the last few years has made tremendous leaps, reaching out to every aspect of modern society mainly business, entertainment and education. At the core of this information technology revolution has been the Internet with its vast reach into homes, offices, and classrooms making it a ready, easy and reliable tool for self-learning and technical resources. This has helped students pursue their academic interests, honing their skills from basic to advanced levels in an efficient and concise manner. Traditionally, the standard procedure for electrical engineering students to strengthen their theoretical concepts has been to attend various lab classes to gain valuable experience in design and implementation of classroom theories. However, there may be unavoidable circumstances in any engineering student's life arising from social, professional or medical obligations that may prevent him from attending a theory or practical class. The idea of having a virtual laboratory that can be accessed over the Internet is a practical approach to solving this problem and help students to cope up and sharpen their practical skills. This paper throws light upon an on-going virtual laboratory program at the University of Texas at San Antonio that enables students to access and operate hardware and software based experiments physically connected to a web-server using remote desktop technologies conveniently from their PCs.

This paper studies the validity of horizontal dilution of precision (HDOP) measure to evaluate sensor network geometries when localizing impacts in structural health monitoring (SHM). First HDOP is defined similarly to navigation applications. Even though several low complexity closed-form solutions have been proposed recently, HDOP measure has been theoretically justified only for iterative least squares approach. The localization errors of popular impact localization methods are experimentally collected and compared with HDOP data for validation. The experimental setup is also described. It is shown that HDOP in general correlates with the positioning error and can be used to characterize geometry.

This paper presents a family of architectures for FFT implementation based on the decomposition of the perfect shuffle permutation, which can be designed with variable number of processing elements. This provides designers with a trade-off choice of speed vs. complexity (cost and area.). Hardware comparison to other existing pipeline architecture presented based on implementation of 1024-point FFT with 4 processing elements using 45 nm process technology. The proposed architecture is most suitable for handheld and portable multimedia applications.

The paper presents a family of architectures for FFT implementation based on the decomposition of the perfect shuffle permutation, which can be designed with variable number of processing elements. This provides designers with a trade-off choice of speed vs. complexity (cost and area.). A detailed case study is provided on the implementation of 1024-point FFT with 2 processing elements using 45 nm process technology, including area, timing, power and place-and-route results.

Direct sequence spread spectrum (DSSS) correlators synchronize received signals with locally generated replicas. They can be implemented either in time or frequency domains. Time-domain correlators are more flexible in multimode operations, and they are typically preferred in hardware-based implementations. Frequency-domain approaches are more computationally efficient, which makes them attractive for software-based solutions. As an alternative to frequency-domain methods, this letter derives computationally-fast time-domain algorithm using the pseudorandom and binary structure of widely used DSSS signals. Our approach is multiplication-free and accurate, it is efficient for multimode operations requiring limited number of correlator outputs, and no constraints are imposed on sequence lengths.

This letter presents a new class of polynomial threshold operators for denoising signals using wavelet transforms. The operators are parameterized to include classical soft-and hard-thresholding operators and have many degrees of freedom to optimally suppress undesired noise and preserve signal details. To avoid the complicated process of signal model identification for specific type of signals, a least squares optimization method is proposed for the polynomial coefficients. Our study shows that the proposed term-by-term, fixed-threshold operator can perform as well as adaptively applied, scale-dependent soft and hard thresholding approaches.

Distributed arithmetic (DA)-based implementation of linear filters relies on the linear nature of this operation and has been suggested as a multiplication free solution. In this work, we introduce a nonlinear extension of linear filters optimizing under mean-square error criterion the memory function [(MF) multivariate Boolean function with not only binary output] which is in the core of DA based implementation. Such an extension will improve the filtering of noise which may contain non-Gaussian components without increasing the complexity of implementation. Experiments on real images have shown the superiority of the proposed filters over the optimal linear filters. Different versions of these filters are also considered for an impulsive noise removal, faster processing, and filtering using large input data windows.

We propose new multistage interconnection networks (MIN) for scalable parallel Viterbi decoder architectures. The architecture consists of the desired number of processing elements (PE) connected by the suggested MINs, thus allowing a tradeoff between complexity and speed. The structure of the MIN is derived first by transforming the de Bruijn interconnection-based Viterbi algorithm trellis into the equivalent trellis with a perfect shuffle interconnection, and then applying a new decomposition of the perfect shuffle operator. This results in an efficient modular system and data flow is formed by the shuffling in a local PE memory and data exchange through a fixed interconnection between PEs. We suggest several solutions for 1/n and k/n rate codes, where k denotes the number of input bits shifting into k shift registers of the encoder and, at each cycle, the encoder produces n output bits as linear combinations of certain bits in the shift registers.

In array processors, complex data reordering is often needed to realize the interconnection topologies between the computational nodes in algorithms. Several important algorithms, e.g., discrete trigonometric transforms and Viterbi decoding, can be represented in a radix-R form where the principal topology is stride by R permutation. A general factorialization of stride permutations is derived, which can be mapped onto register-based structures for constructing area-efficient multi-port interconnection networks. The networks can be modified to support several stride permutations and sequence sizes.

An array processor architecture for the 2-D discrete cosine transform (DCT) based on the row-column decomposition of the 2-D DCT. The utilized 1-D DCT architectures are derived by applying the principles used to construct pipelined fast Fourier transform architectures. In general, this approach has not been used due to the irregularities found in the fast DCT algorithms. The basis of our architectural derivation is the constant geometry fast algorithms for DCT described earlier. By rescheduling the operations; an in-place algorithm can be obtained, which can be mapped onto a pipelined structure with the aid of vertical projection. In addition, a sequential matrix transposition network is described, which is based on shift-exchange units.

In parallel architectures for discrete trigonometric transforms, the number of processing elements is typically dependent on the transform size. Scalable architectures can be constructed with a partial column approach where the computation is performed iteratively with less number of processing elements. This approach results in a need for complex data reordering for realizing the interconnections between the processing columns. In this paper, such interconnection networks performing temporal and spatial reordering are proposed. These networks realize the data reordering found in constant geometry radix-2/sup r/ algorithms, which exist, e.g., for discrete Fourier, sine, cosine, and Hartley transforms. A general decomposition of stride by 2/sup r/ permutation is shown with corresponding network implementations. Furthermore, modifications to support mixed-size and 2-D transforms are discussed.

This paper investigates a fast multibase transform coding approach which could be formulated as an optimal multiscanning method.

New architecture oriented regular algorithms are proposed for the discrete cosine transform (DCT), used in DSP problems, with a simple data shuffling pattern similar to the Cooley-Tukey FFT algorithm. One can easily extend many of parallel FFT approaches for these algorithms. The only pipeline structure with logN-1 arithmetic units and transform time of order O(N) is considered. It is simpler than the known pipeline structures.

Many of the signal and image processing problems concern the suppression of the noise which is non-Gaussian and non-additive. Stack filters are nonlinear filters which are often successfully used for this kind of noise cancellation. A fault detection method is proposed for stack filters. The core of the method is the sample selection probability vectors of stack filters. A simple implementation for fault diagnosis is derived based on this notion.