Search

A comparison of the numbers of cloudy days and clear days shows a recent increase of the former because of fine or yellow dust. Although imaging systems are used in various fields, the image-quality degradation that is due to fog is an outstanding problem. In this paper, a hardware method for the removal of the haze from a single input image using the dark channel prior and the FPGA is proposed; furthermore, a high-quality depth map can be used for fog removal. The problem of the existing haze-removal algorithms is the recognition of the high RGB values as haze, resulting in a loss of the high RGB values. The use of the average atmospheric values, however, implements a high-level transmission map that prevents the loss of the high RGB values (recognized as haze) during the haze removal. In addition, an 8 x 5 matrix is created to improve the speed of the dark-channel calculation, resulting in a speed of more than 500 fps.

In the machine learning technology, making computers read letters, characters and digits have been a hot issue in many areas of research. Among them, the recognition of handwritten numbers has still a long way to go, unlike the recognition of printed digits or the recognition of handwritten English sentences. In this paper we will introduce to you a multiple handwritten digit recognition system using deep learning.

Vehicle manufacturers have been conducting continuous research on energy conservation and efficient energy management of vehicles in accordance with the increased power consumption due to increasing fuel consumption and increasing number of high performance vehicles. Therefore, an alternative energy management system that utilizes the Electric/Electronics (E/E) architecture of the vehicle has attracted attention. 'Partial Networking' is known as a typical energy saving method, made possible by the progress of E/E architecture. A partial network enables the individual wake-up of each node by using a partial Controller Area Network (CAN) transceiver. However, the Electronic Control Unit (ECU) is subject to the constraints of a wake-up time, which is needed to reset the sequence of the microprocessor and the initialization of each module when the ECU initiates wake-up. This is because the partial network proceeds under the control of the regulator in the ECU, whereby the partial network cannot use ECU, as it requires prompt operation when the operating message is received. Therefore, in this paper, we will build an embedded system and experiment environment based on a partial CAN transceiver for the wake-up time measurement in actual partial networking. In addition, we develop an algorithm that reduces the wake-up time by using an embedded system. Finally, we will describe the efficiency of the proposed method by comparing with the existing partial network.

This paper presents the hardware architecture and template matching using distance transform. In the proposed method, the extracted Canny edge image is used as a template for matching processing. The distance images are transformed from the Canny edge image to produce a searching template. Then matching processing will start to find the template in distance image. We implement the template matching hardware architecture on a Xilinx Virtex5 XC5LVX330 Field Programmable Gate Array (FPGA) and Verilog Hardware Description Language (Verilog HDL). The maximum operating frequency of this implementation is 110 MHz. It can run more than 240 frames per second with VGA (640x480) resolution image.

This paper presents a hardware architecture developed for use with real-time Chain Code. This system applies open-chain code and closed-chain code to solve some problems related to noise and an inexact edge image. The design of the proposed method uses the Verilog HDL (Hardware Description Language) for the dedicated hardware architecture on a pipeline architecture. The hardware architecture is implemented using Virtex5 (xc5vlx330) on Xilinx Field Programmable Gate Arrays (FPGA). The results of the analysis show that this implementation can achieve a maximum operating frequency of 74.74 MHz and possesses adequate performance to process a 640x480 (VGA) resolution image.

In this paper, a method for parallel reprogramming automotive electronic control units that reduces the total reprogramming time is proposed. The method is implemented in an in-vehicle domain control unit which performs the role of the gateway between the backbone network and its own sub-network. We built a test environment that consists of an in-vehicle network system based on Ethernet backbone to measure, analyze and evaluate the reprogramming performance. Experiment results show that the proposed parallel reprogramming method can reduce the total reprogramming time at least 39% compare to the sequential reprogramming method with the increasing number of reprogramming target ECUs connecting different sub-networks.

The roofline 2-axis method refers to the number of floating-point operations to be executed every second as a graph resulting from the difference between the floating-point operation speed between the CPU and the memory, corresponding to the intensity of operations for access to the main memory shown in the log graph. This tells us that the floating-point computation speeds of up to come out in the kernel that was randomly determined by developers. In addition, by presenting the guidelines on what can be pulled out floating-point performance of the best in the kernel that was given shall correct any part in terms of software, programming can be performed efficiently. In this paper, a method of to be realized in ARM mode of the roofline, to be used in general desktops, was designed by applying the calculation method of the other. Furthermore, by providing this through the Android application, the operating system most dominant in the environment, we used the roofline graph fabricated by applying the built environment performance, which was increased gradually and visually. This had the advantage of confirming the rate evaluation of the device users and many developers via the application easily in a portable manner. Then, using the application, different rooflines according to core type were provided, providing us with an intuitive graphical user interface.

Most mobile computing devices use a battery as their main power source. Efficient power use and management are very important in mobile devices. It is necessary to understand the power consumption of the processor in a mobile device. In this paper, we choose to use a PMU (Performance Monitoring Unit) to understand hardware performance. By using a PMU counter, it is possible to analyze mobile AP (Application Processor) quickly and accurately without additional CPU overhead. We built a power consumption model using the number of memory accesses per cycle and the clock frequency of the core of mobile AP. By using the power consumption model that we built, we can easily and quickly calculate the power consumption of each core in a multi-core mobile AP. We measured the power consumption of mobile AP using the Mibench benchmark suite and then compared actual measured power consumption and predicted power consumption using several benchmark applications of the Parmibench benchmark suite. The average RMSE between actual measured power consumption and predicted power consumption was 5.01%.

In this paper, we present the design and implementation of real-time multiple object centroid tracking for gesture recognition. Our multiple object tracking design consists of four stages: preprocessing, local intensity accumulation, object observation, and particle filter. We implemented the proposed hardware architecture using Verilog Hardware Description Language (HDL) on a Xilinx Virtex-5 LX330 field programmable gate array (FPGA). We focus on two main performances: the trajectory accuracy of moving objects and real-time processing. The performance of the proposed system was evaluated through several experiments. In addition, our processing speed was compared with the same algorithm based on software. Based on the results, we can guarantee that our multiple object tracking design is suitable for gesture recognition in cluttered environments.

A benchmark program provides objective and quantitative experimental data that is used in HW architecture design and as performance criteria. For that reason, a benchmark point must be fairly calculated to make a right decision. It is difficult to benchmark with fairness in embedded systems such as smartphones since there are many HW and SW development environments. For example, iOS adopts ObjectC language and Android supports Java and C/C++ to develop applications. Therefore, it is impossible to apply the same native workload without modification to both iOS and Android platforms. Web-based benchmark programs can solve the problem of fairness. If a platform supports a browsing system, then we can run the same web benchmark program on that platform. In this paper, we construct a web-based benchmark program with JavaScript and HTML5 to evaluate an entire HW platform. The JavaScript workload has three components and the HTML5 workload has two components. Experiment shows that similar runtime characteristics are maintained between native and web workloads. Moreover, we use a compared ratio method to relieve result distortion that comes from unexpected system overheads. Finally, we run our benchmark program on the iOS and many Android platforms.

Robots used in industry rely on complex and time-consuming programming, such as a teaching panel and offline programming. Different intuitive programming methods have been proposed to overcome this drawback. A method using a force/torque sensor is practical and has been used in several organizations. However, problems, such as precision, have to be overcome in this method. This paper presents a method using an exoskeleton device that can overcome the problems based on robot teaching. Its usefulness is demonstrated compared to a method using a force/torque sensor.

Today, multi-core processors are being used widely in mobile environments in addition to the existing PC-based environment. In order to use a multi-core processor efficiently, parallel programming skills are required. However, incorrect parallelization can degrade the performance of the overall system in a mobile multitasking environment. Therefore, a profiling technique that measures the performance of the whole system is needed. Also, it should to be possible to compare the performance of parallelization operations. In this paper, we propose a profiling method that analyzes the utilization of each application's processor when multiple applications are loaded at the same time. We also verified the profiling method by using some benchmark programs running on the Android operating system.

Virtualization technology has recently been widely used due to its various benefits, such as reducing system management efforts and increasing system utilization. Each execution environment can be isolated from other environments as each user can have access to virtual resources provided by the virtualization layer. This achieves benefits such as increasing reliability and security, and increases system mobility because applications and operating systems running on a virtual machine can migrate to other systems using the same virtualization layer. Virtualization technology is being extended to embedded systems, such as mobile phones, as well as to traditional IT systems, such as servers and PCs. Such extended virtualization technology to this domain requires migrating virtual machines between heterogeneous platforms, such as PCs, TVs, and mobile phones, beyond the existing virtual machine migration between homogeneous platforms. Virtual machine migration between heterogeneous platforms has additional benefits, such as increasing computing power of the execution environment, increasing device expandability, and achieving real migration of the user environment.In this paper, we present the real user- environment migration between heterogeneous platforms and design the architecture. We execute and evaluate virtual machine migration between the x86 platforms and the SPARC platform using our modified QEMU to verify the concept. Users can experience real user-environment migration between heterogeneous platforms using this method.

This paper addresses the issue of personal service robots not being widely used. Our approach is to offer a new interface paradigm to develop pervasive robotic applications. Cleaning robots are the most famous of personal service robots. However, people do not know the possibilities of what the robot can do. First, we analyze what task components are important for a cleaning robot. Next, we evaluate the efficiency of cleaning robots using the experience method. Third, we search for issues with cleaning robots, based on evaluations. We build the optimal experience of a practical cleaning robot. Finally, we propose a novel intelligent user interface based on optimal experiences and evaluate its performance.

This research aim is to create sensible interactions between multisensors in mobile devices. The proposal covers sensible interactions for all three kinds of methods: Mobile Application Interaction, Mobile to Mobile Interaction, and Mobile to Home Appliance Interaction. Five components are built in the handheld sensible interface device: distance measurement component, acceleration measure component, main processing component, haptic generation component, and wireless communication component. This research can support emotional user experience better than the button type input method or touch type input method can do. In addition, the proposal developed from the research can support Mobile to Mobile Interaction and Mobile to Home Appliance Interaction. The implement of the proposed sensible interaction method will demonstrate an appliance with a navigation application and digital TV control.

In this paper, we propose a motion-based game control system using the flex sensor and the infrared sensor board. This system operates using a combination of player movement data detected from the flex sensor and the direction or positional data detected from the infrared sensor board. The player's motion is detected from a flex sensor band worn on his/her elbow and knee and the flex sensor belt on his/her waist. The infrared sensor board detects the player's direction of movement through the detected sequence or detected position of the infrared sensors. This system does not recognize all the detailed human motion, since detailed motion detection is not necessary to control the game. The proposed system only detects player motion types through multiple sensors. This feature distinguishes the system from other motion-based controllers such as Nintendo Wii, Sony motion controller and Microsoft Natal. Unlike our system, these existing controllers require a high-technology, high-cost system. This system is implemented on an existing game pad (PC, Xbox360, PS3) based combat game to test the detection of the player motion type.

Skin color is widely used in many applications because of its merit in human-machine interactions. However, detecting skin color requires repetitive operations on all pixels in the image, similar to other vision-based applications. Since the per-pixel processing is difficult to perform efficiently in conventional computers, many real-time image processing applications have problems with performance. In this paper, we propose FPGA implementation of a real-time skin color detection system. Among the various skin color detection methods, we chose a parametric skin distribution modeling method based on a Gaussian mixture, due to its acceptable training amount and skin detection performance. In addition, a mean-based surface flattening method was also proposed and implemented to improve the detection performance. The proposed method flattens the surface of objects in the scene by replacing the pixel value with the mean of its similar neighborhoods to remove the color noise. After this flattening process, the pixel values of the analogous adjacent pixels are located within a narrow range and are easily segmented to a different region. To consider the inherent parallelism of local image processing, all these functions are implemented within the FPGA to meet the demands of real-time performance.

Essential surrounding environmental information needed to control remote mobile robots can be garnered from devices such as cameras, and ultrasonic and infrared sensors. Image delay and sensor noise sensitivity may engender inaccuracies in the information collected. Robot control by the user may become difficult. A force feedback controller transmitting a repulsive force to the user can aid recognition of the surrounding environment. This can operate despite the limited image information and uncertain sensor information. It is possible to transmit the magnitude and direction of the repulsive force by a vibration. The controller manages the vibration frequency and the vibration magnitude. A robot displays the information of the surrounding environment on a grid map to calculate the repulsive force of the obstacle. The cells of the grid map are filled with the certainty value of the obstacle. Then, the robot calculates the sum of the repulsive forces of the obstacle by a repulsive force vector calculation. The user can obtain the distance between the obstacle and robot using tactile perception using this method. The user, recognizing the obstacle's distance by the magnitude and duration of the vibration, can navigate the robot at a remote site.

A variety of devices are used for robot control such as personal computers or other human interface devices, haptic devices, and so on. However, sometimes it is not easy to select a device which fits the specific character of varied kinds of robots while at the same time increasing the user's convenience. Under these circumstances, in this study, we have tried to measure user convenience. We tried to understand the characteristics of several devices used to achieve human robot interaction by using each of these devices that could be used with a personal computer: We used a button type device, a joystick, a driving device which consisted of a handle and pedals, and a motion-based human interface device including an acceleration sensor.