An efficient architecture for image descriptor matching that uses a partitioned content-addressable memory (CAM)-based approach is proposed. CAM is frequently used in high-speed content-matching applications. However, due to its lack of functionality to support approximate matching, conventional CAM is not directly useful for image descriptor matching. Our modifications improve the CAM architecture to support approximate content matching for selecting image matches with local binary descriptors. Matches are based on Hamming distances computed for all possible pairs of binary descriptors extracted from two images. We demonstrate an FPGA-based implementation of our CAM-based descriptor-matching unit to illustrate the high matching speed of our design. The time complexity of our modified CAM method for binary descriptor matching is O(n). Our method performs binary descriptor matching at a rate of one descriptor per clock cycle at a frequency of 102 MHz. The resource utilization and timing metrics of several experiments are reported to demonstrate the efficacy and scalability of our design.

Image matching is a key component in computer vision and has many applications. In this work, we propose a circular shifting binary descriptor to increase the speed of rotation invariant image matching algorithms. We can compute the descriptors of the rotated image patches without rotating either the sample points or the image patch by circularly shifting the binary descriptor. Thus, operations such as multiplications and divisions from the orientation estimation step are eliminated from the image matching process which significantly reduces the number of operations for computing the descriptor. In addition, our experiments illustrate that the circular shifting binary descriptor shows limited rotation error in comparison with other descriptors such as ORB while attaining comparable mean average precision.

Binary descriptors have been shown to be faster than nonbinary descriptors while producing comparable results in image matching applications. In recent years, there have been many attempts to design hardware accelerators for extraction of binary descriptors to achieve higher processing rates. One of the well-known methods is the binary robust invariant scalable key point (BRISK) algorithm, which has shown outstanding results in various applications. In this work, we propose a multiscale field-programmable gate array (FPGA)-based hardware architecture for the BRISK descriptor. In addition, a new image sampling pattern for the BRISK algorithm is described which is shown to be more efficient than the original sampling pattern for hardware implementation. Our new sampling pattern decreases the size of the patches containing the key point to one-quarter of the size of that used in the original BRISK algorithm, which leads to a reduction in FPGA resource utilization while maintaining the accuracy of the image matching application. Our proposed design is fully pipelined and achieves a frame rate of 78 fps on images with full HD resolution.

An attack on hardware typically results in a severer and difficult-to-recover damage, hence, it is our interest to focus this work on hardware security. A brief survey of security issues found in various application domains is presented, based on a collection of over seventy papers in IEEE Xplore archived since 2011. Challenges and potential solutions to different kinds of hardware attacks and threats are discussed. Finally, specific hardware security challenges are connected and mapped to each application domain.

Computing platforms deployed in many critical infrastructures, such as smart grid, financial systems, governmental organizations etc., are subjected to security attacks and potentially devastating consequences. Computing platforms often get attacked 'physically' by an intruder accessing stored information, studying the internal structure of the hardware or injecting a fault. Even if the attackers fail to gain sensitive information stored in hardware, they may be able to disrupt the hardware or deny service leading to other kinds of security failures in the system. Hardware attacks could be covert or overt, based on the awareness of the intended system. This work classifies existing hardware attacks. Focusing mainly on covert attacks, they are quantified using a proposed schema. Different countermeasure techniques are proposed to prevent such attacks.

A novel motion retrieval scheme is proposed. Based on semantic analysis and graph model, this scheme involves system learning in the first stage. In system learning, a Motion Semantic Dictionary (MSD) is derived by clustering. A Dynamic Bayesian Network (DBN) graph model is constructed based on the MSD and learning parameters. MSD and DBN are combined to derive motion information as features. Motion categories are recognized based on motion feature queries and matching. Experimental results are presented, showing the proposed method is more effective in execution time as compare to some existing representative algorithms.

Many emerging motion-related applications, such as virtual reality, decision making, and health monitoring, demand reliability and quick response upon input changes. Motion capture has been a well-researched topic in the past decades with applications in many industries. The ability to capture motion goes hand in hand with real-time capability in a system. This paper gives an overview on real-time motion capture, including frameworks, recording devices, and algorithms. Using real-time motion capture as a tool to solve application problems is also described.

Sentiment analysis of the text data available on the web either in the form of blogs or at social media sites such as Twitter, Facebook, and LinkedIn, offers information through which to assess people's perspective of products and services that are of interest to them. Consumers routinely scour the Internet to assess other user's reviews for a product or a service before making their own decision. Likewise, these same data can potentially provide businesses a snapshot in time of the users' response to their products/services and even the trends over time. This information gained through sentiment analysis can then be used by businesses to make decisions about improving their products and services, and gain an increased edge over their competitors. In this paper, we illustrate the potential of sentiment analysis of Twitter data to gauge users' response to popular smart phone brands and their underlying operating systems. Specifically, our objective is to investigate whether the tweets available on the web are sufficient to gain useful insight about the performance of popular smart phone brands, their battery life, screen quality, and on the perceived performance of the phones operating systems. Our results show that although the Twitter data does provide some information about users' sentiments to the popular smart phone brands and their underlying operating systems, the amount of data available for different brands varies significantly. This limitation makes the comprehensive analysis of users' response somewhat more challenging for some brands compared to others and consequently makes the comparison between brands almost impossible.

In this paper, we present a mobile learning tool for ecosystem study using an augmented reality user interface. Our system provides a recommender function of learning content according to a learner's interest, which is extracted in the form of a feature term set that is based on the user's browsing behavior of learning content and its related objects, which are displayed through the augmented reality user interface. Using a prototype, we confirmed that our system can properly personalize the recommended results for ecosystem-related learning content, and we have the promising vision that our system can be utilized in a practical way for mobile learning in fieldwork for ecosystem studies.

With the emergence of powerful computation platforms such as grid, multi-core, and cloud, intelligent nodes and agents in a distributed environment are important for the successful deployment of many complex applications. This work examines the inherent design and implementation issues as well as challenges in distributed intelligent systems. Surveys and important works are introduced from a historical perspective. Applications that employ the distributed intelligence approach are presented and discussed.

Engineering schools worldwide have a relatively high attrition rate. Typically, about 35% of the first-year students in various engineering programs do not make it to the second year. Of the remaining students, quite often they drop out or fail in their second or third year of studies. The purpose of this investigation is to identify the factors that serve as good indicators of whether a student will drop out or fail the program. In order to establish early warning indicators, principal component analysis is used to analyze, in the first instance, first-year engineering student academic records. These performance predictors, if identified, can then be used effectively to formulate corrective action plans to improve the attrition rate.

In recent years, advances in motion detection, tracking, and classification have made applications in telerehabilition possible. Monitoring of the elderlies and physically impaired, sports medicine, and physical therapy, are some of the active research areas in telerehabilition. Furthermore, the gaming industry has provided some low-cost devices to track motion with reasonable accuracy, thus, making telerehabilition a reality for patients who need it. This project investigates the feasibility of using the Nintindo Wii Remote Controller as a handheld device that patients can use to perform repetitive motions in physical therapy sessions in a telerehabilitation setting. Patients can use such an instructional (or e-learning) system at home without the cost, effort, and time associated with visiting a medical office. At the same time, the public health care system also benefits from the deployment of these telerehabilitation systems.

Computer-based instructional systems provide an ideal setting for learning certain types of sports. In particular, the sports that require premium space could leverage the widely available computing and Internet facilities to teach individual users anywhere and anytime. An e-learning tennis instruction system is currently being designed and developed. The Nintendo Wii Remote is selected as the input device for its low cost and racket-handle like shape. After the data from motion sensors are captured, they have to be cleansed, normalised clustered and classified. Data of three common swings, backhand, forehand, and overhand, have been recorded from fifty people of various levels of tennis skill. Experiments are carried out to identify the most suitable techniques to classify a tennis swing. The adaptive nature of a prototype system is also introduced.

The Dynamic Time Warping (DTW) algorithm is a commonly used algorithm in matching time sequence data in many applications that require some kind of similarity measure. Though effective, DTW is computationally intensive, and therefore is not suitable for real-time situations. In the past 30 years, there has been some research work on implementing DTW in hardware as a stand-alone processing unit, or as a co-processor within a larger system. This work gives a brief survey on previous work done in DTW hardware design and implementation. For many modern-day Web and intelligent applications, one must consider the real time and hardware footprint aspects of the system. A DTW single-element processing unit is proposed in order to investigate the suitability of using it as a building block for more complex architecture for embedded applications. This simple unit is designed and simulated as a Field Programmable Gate Array (FPGA) implementation using Xilinx tools. The performance results of both area and speed show great potential and ascertain DTW hardware is a worthwhile endeavor to pursue further in a systematic fashion.

Advances in portable devices and location-aware applications have necessitated the research in sophisticated yet small-footprint hardware and software in embedded systems, while the proliferation of the Web and distributed database systems has led to new data mining applications. We are investigating the utilization of reconfigurable hardware, due to its flexibility and performance, for data mining applications in portable and embedded computing. In this work, we introduce a reconfigurable hardware solution using field programmable gate array (FPGA) for similarity matrix computation, a commonly used data structure to represent the computed similarity among a set of feature vectors. Our hardware design can be dynamically reconfigured to accommodate three different similarity measures. A space-time cost analysis of the proposed multiplexer-based approach is presented. Experiments performed on the implemented reconfigurable hardware show encouraging and promising results that warrant further investigation in dynamically reconfigurable FPGA-based hardware for data mining applications.

Abstract In the rollback recovery of large‐scale long‐running applications in a distributed environment, pessimistic message logging protocols enable failed processes to recover independently, though at the expense of logging every message synchronously during fault‐free execution. In contrast, coordinated checkpointing protocols avoid message logging, but they are poor in scalability with a sharply increased coordinating overhead as the system grows. With the aim of achieving efficient rollback recovery by trading off logging overhead and coordinating overhead, this paper suggests a partitioning of the system into clusters, and then presents a scheme to implement the conversion between these overheads. Using the proposed conversion, coordination can be introduced to reduce the unbearable logging overhead found in some systems, whereas proper logging can be employed to alleviate the unacceptable coordinating overhead in others. Furthermore, heuristics are introduced to address the issue of how to partition the system into clusters in order to speed up the recovery process and to improve recovery efficiency. Performance evaluation results indicate that our scheme can lower the overall system overhead effectively. Copyright © 2008 John Wiley & Sons, Ltd.

An enormous amount of data needs to be processed in many data mining applications. In addition to algorithmic development, hardware support is imperative to improve the effectiveness and efficiency of these applications. We are investigating various hardware architectural design techniques and methodologies to support data mining at the chip level. In this work, we focus on the design of an FPGA-based processor array for the computation of similarity matrix, a commonly used data structure to represent the similarity among a set of feature vectors, with each matrix element representing the computed similarity measure between two vectors. An algorithm is developed to assign computation efficiently to the array of processing elements. Theoretical performance metrics are derived and compared to the experimental results. Performance gains using the processor array over software implementations are also presented and discussed.

For many users with a physical or motor disability, using a computer mouse or other pointing device to navigate the web is cumbersome or impossible due to problems with pointing accuracy. At the same time, web accessibility using a keyboard in major browsers is rudimentary, requiring many key presses to select links or other elements. We introduce KeySurf, a character controlled web navigation system which addresses this situation by presenting an interface which allows a user to activate any web page element with only two or three keystrokes. Through an implementation of a user-centric incremental search algorithm, elements are matched according to user expectation as characters are entered into the interface. We show how our interface can be integrated with a speech recognition input, as well as with specialized on-screen keyboards for people with disabilities. Using the user's browsing history, we improve the efficiency of the selection process and find potentially interesting page links for the user within the current web page. We present the results from a pilot study evaluating the performance of various components of our system.

In memory-based collaborative filtering, the existing methods conduct a prediction based on the overall consistency of two users or items. The major problem with these methods is that it is hard to find users/items that are overall consistent with the test user/item in the system. In addition, these methods are sometimes being over optimistic, and disregard some useful information in user profiles in making a prediction. This paper exposes the drawbacks in these methods and proposes an inference-based recommendation scheme to overcome those drawbacks. This model is based on the fact that any two users may have common interest genres as well as different ones, with the capability of making full use of rating information to capture accurately the relevance between item and user. Experimental results from two popular public datasets, EachMovie and Movielens, show that our approach improves significantly the prediction accuracy.

As a common technology for fault tolerance and load balance, rollback-recovery faces the challenges of scalability and inherent variability in those long-running large-scale applications with grids as the computing infrastructure. Among the rollback recovery schemes, pessimistic message logging protocols (PMLPs) and coordinated checkpointing protocols (CCPs) are the most popular in practice. Although PMLPs are good in scalability, their fault-free overhead sometimes is prohibitive. CCPs introduce relatively lower overhead, but they are poor in scalability. This work employs partition strategy and introduces the concept of pessimism grain to rollback recovery, striking a balance between good scalability and acceptable overhead. For a partitioned system, a coarse-grained pessimistic message-logging protocol is proposed to achieve scalability and asynchrony both in fault-free execution and in fault recovery. The impact of pessimism grain on the performance is evaluated theoretically. Experimental results show that the pessimism grain is one of the key configuration parameters to reach a desired performance level.

For many users with a disability it can be difficult or impossible to use a computer mouse to navigate the web. An alternative way to select elements on a web page is the label typing approach, in which users select elements by typing part of the label. In most cases, these labels are specified by the page authors, but some selectable elements do not have an obvious textual description, thus requiring that a label be generated. The set of element labels on a web page must be both efficient to select by text input and meaningful to the user. This paper discusses our approach to this problem, using page structural analysis and user history to determine important elements of a page, and then matching this information with the efficiency model of the input device.