Search

The convergence of edge computing systems with Field-Programmable Gate Array (FPGA) technology has shown considerable promise in enhancing real-time applications across various domains. This paper presents an innovative edge computing system design specifically tailored for pavement defect detection within the Advanced Driver-Assistance Systems (ADASs) domain. The system seamlessly integrates the AMD Xilinx AI platform into a customized circuit configuration, capitalizing on its capabilities. Utilizing cameras as input sensors to capture road scenes, the system employs a Deep Learning Processing Unit (DPU) to execute the YOLOv3 model, enabling the identification of three distinct types of pavement defects with high accuracy and efficiency. Following defect detection, the system efficiently transmits detailed information about the type and location of detected defects via the Controller Area Network (CAN) interface. This integration of FPGA-based edge computing not only enhances the speed and accuracy of defect detection, but also facilitates real-time communication between the vehicle’s onboard controller and external systems. Moreover, the successful integration of the proposed system transforms ADAS into a sophisticated edge computing device, empowering the vehicle’s onboard controller to make informed decisions in real time. These decisions are aimed at enhancing the overall driving experience by improving safety and performance metrics. The synergy between edge computing and FPGA technology not only advances ADAS capabilities, but also paves the way for future innovations in automotive safety and assistance systems.

The touchless techniques in human-computer interaction (HCI) can effectively expand computer access capabilities for disabled people. This paper presents Touchless Head-Control (THC), an assistive system method for computer cursor control based on head pose captured with an RGB camera. Our work aimed to replace the standard cursor control using a device on the user’s head. The convolutional neural networks with predicted fine-grained feature maps and binned classification were applied to estimate the head pose angles. The mouse pointer or cursor is moved to actual locations on the screen based on head movement (yaw and pitch) and the center position of the face. Head tilt to the right or left (roll) to control the mouse button. In addition, the proposed method can be used to simulate the movement of the robot or joystick using the head to control objects within three degrees of freedom (DOF). Various participants were involved in the interaction design evaluation, in which target selection accuracy, travel time, and path efficiency were measured. This technology allows people with limited motor skills to easily control a PC cursor and 3D object orientation without the use of additional equipment or sensors.

Dynamic point clouds (DPC) are new media storage formats that allow end-users to watch objects/scenes in a three-dimensional (3D) sense. It can be displayed from different angles throughout time. However, the raw size of a point cloud is huge because there can be millions of points (each containing color triplet and location triplet information) in a point cloud, and there can be multiple point clouds in a DPC. Video-based point cloud compression (V-PCC) is developed to project a 3D point cloud to 2D images: attribute, geometry, and occupancy images. After padding, the 2D images are compressed using the well-established high-efficiency video coding (HEVC). In this study, we first employ an occupancy image to propose a blocky occupancy flag (BOF), to denote the occupancy information on “a block basis”. For coding attribute and geometry images, we use a BOF to develop a fast coding unit (CU) algorithm for early termination of the CU search recursion. We also utilize the geometry images to calculate the 2D and 3D information of each pixel, for 2D/3D spatial homogeneity of the pixels to design fast CU decision. In addition, we proposed a modified rate-distortion optimization for different color components considering the picture order count (POC) structure in HEVC/V-PCC. Finally, we propose an HEVC input pixel modification method based on a BOF to reduce the unnecessary information to be coded for attribute images. Compared with the state-of-the-art fast V-PCC encoding method, the proposed work outperforms by up to 2.31% in Bjøntegaard delta bit rates (BDBR) (with very slight loss by only up to 0.38%), and improves the time saving performances by up to 7.84% for two different testing datasets.

Backlight power-saving algorithms can reduce the power consumption of the display by adjusting the frame pixels with optimal clipping points under some tradeoff criteria. However, the computation for the selected clipping points can be complex. In this paper, a novel algorithm is created to reduce the computation time of the state-of-the-art backlight power-saving algorithms. If the current frame is similar to the previous frame, it is unnecessary to execute the backlight power-saving algorithm for the optimal clipping points, and the derived clipping point from the previous frame can be used for the current frame automatically. In this paper, the motion vector information was used as the measurement of the similarity between adjacent frames, where the generation of the motion vector information requires no extra complexity since it is generated to reconstruct the decoded frame pixels before the display. The experiments showed that the proposed work can reduce the running time of the state-of-the-art methods by 25.21% to 64.22%, while the performances are maintained; the differences with the state-of-the-art methods in PSNR are only 0.02~1.91 dB, and those in power are only −0.001~0.008 W.

As compressed videos are transmitted in the communication networks, video packet loss inevitably occurs. This problem can be solved by error concealment method. We used the motion vector of the available neighboring blocks to estimate the lost motion vector for the lost block. These estimates propagate to predict all other missing motion vectors. We further improved the work by using the idea of the motion vector disparities between neighboring available blocks to modify the motion vector weightings. Furthermore, the differences between the compensated pixels and the decoded pixels in the neighboring blocks are computed for another weighting for improvement. These two novelties are combined as a final indicator to prediction weightings. By comparison against the state-of-the-art method, the four proposed algorithms increase the average peak signal-to-noise ratio (PSNR) by up to 1.86, 1.93, 1.94, and 2.04 dB on average, showing the gradual improvement of our design systems. For other video quality measurements, the average gains of the proposed work against the state-of-the-art work can be up to 0.0575 in structural similarity index metric (SSIM), -0.0278 in video quality metric (VQM) (the lower the better), -0.0008 in motion-based video integrity evaluation (MOVIE) (the lower the better), and 2.77 in subjective evaluation. The proposed work performs slightly worse than a pixel-based state-of-the-art method in PSNR and SSIM but performs better in VQM and MOVIE (both correlate better with human perception) and subjective experiments, with much lower computational complexity.

This paper presents a very large-scale integration (VLSI) circuit design of a micro control unit (MCU) for wireless body sensor networks (WBSNs) in cost-intention. The proposed MCU design consists of an asynchronous interface, a multisensor controller, a register bank, a hardware-shared filter, a lossless compressor, an encryption encoder, an error correct coding (ECC) circuit, a universal asynchronous receiver/transmitter interface, a power management, and a QRS complex detector. A hardware-sharing technique was added to reduce the silicon area of a hardware-shared filter and provided functions in terms of high-pass, low-pass, and band-pass filters according to the uses of various body signals. The QRS complex detector was designed for calculating QRS information of the ECG signals. In addition, the QRS information is helpful to obtain the heart beats. The lossless compressor consists of an adaptive trending predictor and an extensible hybrid entropy encoder, which provides various methods to compress the different characteristics of body signals adaptively. Furthermore, an encryption encoder based on an asymmetric cryptography technique was designed to protect the private physical information during wireless transmission. The proposed MCU design in this paper contained 7.61k gate counts and consumed 1.33 mW when operating at 200 MHz by using a 90-nm CMOS process. Compared with previous designs, this paper has the benefits of increasing the average compression rate by over 12% in ECG signal, providing body signals analysis, and enhancing security of the WBSNs.

With the rise of the Internet of things, the smart environmental issue is becoming increasingly important. Sensor web is one of the best solutions to this issue and provides the advantages of sensor networks and web services. Ontology web language for services (OWL-S) is an OWL-based web services ontology, which provides the ability to describe the semantics of web services and their capabilities in a formal and machine-processable manner. Moreover, it aids semantic service matching, selection and composition. However, automatically annotating semantic web services is a highly complicated and tedious task. In this study, we propose a methodology to uncover information in the history data and profiles of web services and then semantically annotate them. With the proposed approach, semantic relationships between web services could be extracted via a combination of association rules and input/output matching. Our results show that this hybrid automated knowledge-discovery approach works better than traditional approaches do. We also provide a scenario to explain how the proposed methodology works.