As the global population ages, the death and prevalence of atrial fibrillation (AF) continue to rise, posing significant concerns due to its strong association with stroke-related disabilities. Detecting AF early before a stroke occurs has become paramount. However, existing methods face challenges in achieving quick, easy, and affordable detection in complex environments characterized by motion interference and varying light conditions. To address these challenges, we propose a system that is employable for edge computing devices like smartphones, tablets, or laptops. Meanwhile, to ensure that the dataset reflects real-world scenarios, we collect 7,216 30-second segments from 452 subjects, categorized into Atrial Fibrillation (AF), Normal Sinus Rhythm (NSR), and Other Arrhythmias (Others), with a subject ratio of 105:116:231. Our lightweight non-contact facial rPPG atrial fibrillation detection system utilizes a Convolution Neural Network (CNN) with a large receptive field and a bidirectional spatial mapping augmented attention module (BiSME-ATT) coupled with a bidirectional feature pyramid network layer (BiFPN), optimized for deployment on mobile devices by reducing model parameters and floating-point operations per second (FLOPs). Our approach significantly improves AF detection accuracy, sensitivity, specificity, positive predictive value, and negative predictive value to 94.39%, 91.57%, 95.44%, 88.06%, and 96.93%, respectively, in AF vs. Non-AF scenarios. Furthermore, the results demonstrate notable enhancements in AF detection across various motion and light intensity levels.

In the domain of autonomous driving, the integration of multi-modal perception techniques based on data from diverse sensors has demonstrated substantial progress. Effectively surpassing the capabilities of state-of-the-art single-modality detectors through sensor fusion remains an active challenge. This work leverages the respective advantages of cameras in perspective view and radars in Bird’s Eye View (BEV) to greatly enhance overall detection and tracking performance. Our approach, Camera-Radar Associated Fusion Tracking Booster (CRAFTBooster) represents a pioneering effort to enhance radar-camera fusion in the tracking stage, contributing to improved 3D MOT accuracy. The superior experimental results on K-Radaar dataset, which exhibit 5-6% on IDF1 tracking performance gain, validate the potential of effective sensor fusion in advancing autonomous driving.

Remote photoplethysmography (rPPG) has been used to measure vital signs such as heart rate, heart rate variability, blood pressure (BP), and blood oxygen. Recent studies adopt features developed with photoplethysmography (PPG) to achieve contactless BP measurement via rPPG. These features can be classified into two groups: time or phase differences from multiple signals, or waveform feature analysis from a single signal. Here we devise a solution to extract the time difference information from the rPPG signal captured at 30 FPS. We also propose a deep learning model architecture to estimate BP from the extracted features. To prevent overfitting and compensate for the lack of data, we leverage a multi-model design and generate synthetic data. We also use subject information related to BP to assist in model learning. For real-world usage, the subject information is replaced with values estimated from face images, with performance that is still better than the state-of-the-art. To our best knowledge, the improvements can be achieved because of: 1) the model selection with estimated subject information, 2) replacing the estimated subject information with the real one, 3) the InfoGAN assistance training (synthetic data generation), and 4) the time difference features as model input. To evaluate the performance of the proposed method, we conduct a series of experiments, including dynamic BP measurement for many single subjects and nighttime BP measurement with infrared lighting. Our approach reduces the MAE from 15.49 to 8.78 mmHg for systolic blood pressure (SBP) and 10.56 to 6.16 mmHg for diastolic blood pressure(DBP) on a self-constructed rPPG dataset. On the Taipei Veterans General Hospital(TVGH) dataset for nighttime applications, the MAE is reduced from 21.58 to 11.12 mmHg for SBP and 9.74 to 7.59 mmHg for DBP, with improvement ratios of 48.47% and 22.07% respectively.

Atrial Fibrillation (AF) screening from face videos has become popular with the trend of telemedicine and telehealth in recent years. In this study, the largest facial image database for camera-based AF detection is proposed. There are 657 participants from two clinical sites and each of them is recorded for about 10 minutes of video data, which can be further processed as over 10 000 segments around 30 seconds, where the duration setting is referred to the guideline of AF diagnosis. It is also worth noting that, 2 979 segments are segment-wise labeled, that is, every rhythm is independently labeled with AF or not. Besides, all labels are confirmed by the cardiologist manually. Various environments, talking, facial expressions, and head movements are involved in data collection, which meets the situations in practical usage. Specific to camera-based AF screening, a novel CNN-based architecture equipped with an attention mechanism is proposed. It is capable of fusing heartbeat consistency, heart rate variability derived from remote photoplethysmography, and motion features simultaneously to reliable outputs. With the proposed model, the performance of intra-database evaluation comes up to 96.62% of sensitivity, 90.61% of specificity, and 0.96 of AUC. Furthermore, to check the capability of adaptation of the proposed method thoroughly, the cross-database evaluation is also conducted, and the performance also reaches about 90% on average with the AUCs being over 0.94 in both clinical sites.

Complex emotion is an aggregate of two or more others which has highly variable appearances, inter-dependence, and affective dynamics.These properties make the recognition hard to handle via existing recognition techniques like action units or valence-arousal detection. In this study, we propose a bionic two-system structure for complex emotion recognition. The structure mimics the working theory of the human brain responding to problems decision-making. System I is a fast compound sensing module. System II is a slower cognitive decision module that processes data more integratively. System I contains one branch for facial expression feature representation including basic emotion, action units, and valence arousal detection and one for physiological measurement which is an image-only implementation for practicality. In System II, a decision module with segmentation is employed to ensure the chosen period including the emotion occurrence and iteratively optimize the emotion information in a given segment via reinforcement learning. The proposed method outperforms state-of-the-art on emotion recognition tasks with an accuracy of 94.15% in basic emotion recognition on the BP4D and an accuracy of 68.75% for binary valence arousal classification on the DEAP. For a subset of complex emotions, the recognition accuracy exceeds 70% on both databases, that is a significant improvement.

Atrial fibrillation (AF) has been proven highly correlated to stroke; more than 43 million people suffer from AF worldwide. However, most of these patients are unaware of their disease. There is no convenient tool by which to conduct a comprehensive screening to identify asymptomatic AF patients. Hence, we provide a non-contact AF detection approach based on remote photoplethysmography (rPPG). We address motion disturbance, the most challenging issue in rPPG technology, with the NR-Net, ATT-Net, and SQ-Mask modules. NR-Net is designed to eliminate motion noise with a CNN model, and ATT-Net and SQ-Mask utilize channel-wise and temporal attention to reduce the influence of poor signal segments. Moreover, we present an AF dataset collected from hospital wards which contains 452 subjects (mean age, 69.3 ±13.0 years; women, 46%) and 7,306 30-second segments to verify the proposed algorithm. To our best knowledge, this dataset has the most participants and covers the full age range of possible AF patients. The proposed method yields accuracy, sensitivity, and specificity of 95.69%, 96.76%, and 94.33%, respectively, when discriminating AF from normal sinus rhythm. More than previous studies, other arrhythmias are also taken into consideration, leading to a further investigation of AF vs. Non-AF and AF vs. Other scenarios. For the three scenarios, the proposed approach outperforms the benchmark algorithms. Additionally, the accuracy of the slight motion data improves to 95.82%, 92.39%, and 89.18% for the three scenarios, respectively, while that of full motion data increases by over 3%.

Although deep convolutional networks have achieved great performance in face recognition tasks, the challenge of domain discrepancy still exists in real world applications. Lack of domain coverage of training data (source domain) makes the learned models degenerate in a testing scenario (target domain). In face recognition tasks, classes in two domains are usually different, so classical domain adaptation approaches, assuming there are shared classes in domains, may not be reasonable solutions for this problem. In this paper, self-supervised learning is adopted to learn a better embedding space where the subjects in target domain are more distinguishable. The learning goal is maximizing the similarity between the embeddings of each image and its mirror in both domains. The experiments show its competitive results compared with prior works. To know the reason why it can achieve such performance, we further discuss how this approach affects the learning of embeddings.

Remote photoplethysmography (rPPG) is an unobtrusive solution to heart rate monitoring in drivers. However, disturbances that occur during driving such as driver behavior, motion artifacts, and illuminance variation complicate the monitoring of heart rate. Faced with disturbance, one commonly used assumption is heart rate periodicity (or spectrum sparsity). Several methods improve stability at the expense of tracking sensitivity for heart rate variation. Based on statistical signal processing (SSP) and Monte Carlo simulations, the outlier probability is derived and ADaptive spectral filter banks (AD) is proposed as a new algorithm which provides an explicable tuning option for spectral filter banks to strike a balance between robustness and sensitivity in remote monitoring for driving scenarios. Moreover, we construct a driving database containing over 23 hours of data to verify the proposed algorithm. The influence on rPPG from driver habits (both amateurs and professionals), vehicle types (compact cars and buses), and routes are also evaluated. In comparison to state-of-the-art rPPG for driving scenarios, the mean absolute error in the Passengers, Compact Cars, and Buses scenarios is 3.43, 7.85, and 5.02 beats per minute, respectively. Moreover, AD also won the top third place in the first challenge on remote physiological signal sensing (RePSS) with relative low computational complexity.

This paper presents an effective way to extract low-level features based on sparse coding for facial occlusion detection. Masked Neural Sparse Encoder (MNSE) is proposed to be a sparse coding solver that brings out better feature bases for data representation and improvement in the anomaly detection task. To guarantee the representational capability of features, a set of masks is applied to force each feature basis is heeded on learning a specific stroke within a certain area. The mask set is constructed by clustering primary strokes from training samples, and represents them with corresponding centers of clusters. Hence, these masks stand for main strokes in concerned areas with higher probabilities. Experiments show MNSE contains better representational capability in data from different domains. Compared with the standard sparse coding and the auto-encoder based approaches, MNSE lifts the accuracy up around 20%.

Driver's physiological state is highly correlated to the traffic safety. An affordable and convenient way to monitor driver's physiological state is remote Photoplethysmography (rPPG). Earlier algorithms achieved high accuracy on measuring rPPG signals in stationary case. But in real cases, such as driving, rPPG signals might be corrupted with interference. To obtain higher Signal-to-Noise-Ratio (SNR) rPPG signals, three algorithms are proposed. The PCA spectral subtraction (PCA-SS) considers the spectrum of the environmental noise and utilizes the energy subtraction to reduce the noise. The machine learning methods, convolution autoencoder (CAE) and multi-channel convolution autoencoder (Multi-CAE), are adopted in order to enhance the rPPG signal. The test data we used are 187 videos recorded in stationary case, passenger case, and real driving situation. In driving situation, the Multi-CAE method, in comparison with the original method provided by W. Wang et al. [1] and G. De Haan et al. [2], achieves 33% & 35% reduction in MAE, RMSE respectively, and 11% improvement in success rate [3].

Facial action unit detection (FAUD) is aimed to detect subtle facial motions, known as action units (AUs), induced by the traction or towing of facial muscles. In recent years, FAUD has become an attractive task since these slight changes on face reveal some cues of emotions which can then be utilized to infer people’s affective. Many prior works are developed based on physical features of AUs, such as regional occurrence and temporal continuity, but few of them concern the negative impact of individual differences, causing an unrobust result in variant subjects and environments. To deal with the problem, a contrastive feature learning method is proposed to make a convolutional neural network (CNN) learn to extract the contrastive feature which is the difference between the features of a neutral face image and an AU-occurred face image. In this way, the individual information will be mitigated so that it becomes easier to detect AUs according to the features. A great disparity between the number of positive and negative samples, named as data imbalance, makes a serious interference in most detection problems, including FAUD. A lower occurrence rate of an AU leads a more severe data imbalance problem, which results in a biased result for the AU detection. Therefore, the class-weighted loss is proposed to change the weight-ratio between the positive and negative samples so as to encourage the learning of the positive samples and ease that of negative samples. Two widely used databases, BP4D and DISFA, are adopted to be the benchmark of the performance testing. In the experiment, it shows that the proposed method performs well both in accuracy and speed comparing with the state-of-the-art approaches.

In this paper we investigate the latest vision-based method for systolic blood pressure (SBP) and diastolic blood pressure (DBP) measurement. However, constantly blood pressure supervision needs sufficient medical equipment and may require the potential patients to tie a cuff, which is extremely inconvenient for them. What's more, continuously blood pressure measuring requires the patients to stay in the hospital and professional personnel to stand by. From the research before, we have learned that photoplethysmography (PPG) can be used to measure the blood pressure, which is known as cuffless blood pressure measurement. However, for the neonate and patients with empyrosis, photoplethysmography measuring device is still less practical and restricted in use due to the necessary contact for it to measure the systolic and diastolic blood pressure. Certain level of discomfort is still unavoidable with the use of PPG. We thus focus on remote PPG (rPPG); with green red difference (GRD) and Euler video magnification (EVM) and finite impulse response (FIR) bandpass filters, we are able to recover PPG signals from remote photoplethysmography. We propose a feature extraction measuring methods which yields a root mean square error for SBP as 11.22 mmHg and 7.83 mmHg for pulse pressure (PP) combined with the ANN model. For comparison, we've also used K nearest neighbor (KNN) and deep belief network-deep neural network (DBN-DNN).

To handle the customer distribution in the certain areas, crowd counting is necessary for such applications, which is a labor-intensive work for human. Therefore, an automatic crowd counting system is in great demand, but it is still a challenging problem since the human heads and bodies are usually highly overlapping in crowd images. In this paper, a counting-by-regression framework is employed. The human head is modeled as a Guassian distribution. With a crowd density map estimator, the head count can be obtained by integrating over the density map. Most existing approaches only apply density map regression for training a density map estimator, but it is hard to find a suitable training parameters to train a good one; actually, the head count is overestimated easily. To mitigate this problem, counting regression is combined with density map regression. A deeper and lighter fully convolutional network (FCN) is designed to be a crowd density map estimator. The input and output size of the FCN are the same. After training by the proposed method, our model is more competitive comparing with others. The parameter quantity of the model is the lowest, and it needs the least inference time.

To successfully increase athletes' or exercisers' fitness and endurance, the factors of physiological signal, emotion, or the level of fatigue should be considered during the training program. Many clinical decision support systems can assist to monitor the exercisers by some wearable devices. And, the questionnaire should also be taken into account to produce a report. Such process is cumbersome, and the results are not objective. Furthermore, one may feel uncomfortable when wearing the devices during the training program. In this research, the Rating of Perceived Exertion (RPE) is expected to be estimated automatically without any wearable devices and questionnaires. A camera based heart rate detection algorithm and a fatigue expression feature extractor are fused to estimate the RPE value. The results show that our heart rate detection algorithm can be competitive to the wearable devices, and the trend of the detected heart rate is correlated to RPE. Moreover, the fatigue feature can help reduce the error of the estimation.

A Tent-map chaotic Newton-Raphson optimization based neural network predictive control (TCNR-NPC) is developed to apply to the long-delay permanent magnet synchronous motor (PMSM) system in this paper. Due to a nonlinear model utilized in the predictive controller, nonlinear optimization methods turn into an important issue. To overcome the shortcoming of the conventional nonlinear programming on the initial condition sensitivity and maintain the accuracy of optimal solution, chaos optimization algorithm (COA) and Newton-Raphson (NR) are combined. With the comparison of COA and NR based optimization methods, our approach, the Tent-map chaotic Newton-Raphson (TCNR) optimization, is easier to reach the global optimum, thus, it would be employed in neural network predictive control. It is found that TCNR-NPC has a better performance than those of GPC, modified GPC, adaptive extended PSO based NPC, and PSO based PI controllers in real experiments.

Recently, several robotic wheelchairs have been proposed that employ autonomous functions. In designing wheelchairs, it is important to reduce the accompanist load. To provide such a task, the mobile robot needs to recognize and track people. In this paper, we propose to utilize the multisensory data fusion to track a target accompanist. First, the simultaneous localization and map building is achieved by using the laser range finder (LRF) and inertial sensors with the extended Kalman filter recursively. To track the target person robustly, the accompanist, are tracked by fusing laser and vision data. The human objects are detected by LRF, and the identity of accompanist is recognized using a PTZ camera with a pre-defined signature using the speed-up robust features algorithm. The proposed system can adaptively search visual signature and track the accompanist by dynamically zooming the PTZ camera based on LRF detection results to enlarge the range of human following. The experimental results verified and demonstrated the performance of the proposed system.

With the growing concern for driving safety, many driving-assistance systems have been developed. In this paper, we develop a reasoning-based framework for the monitoring of driving safety. The main objective is to present drivers with an intuitively understood green/yellow/red indicator of their danger level. Because the danger level may change owing to the interaction of the host vehicle and the environment, the proposed framework involves two stages of danger-level alerts. The first stage collects lane bias, the distance to the front car, longitudinal and lateral accelerations, and speed data from sensors installed in a real vehicle. All data were recorded in a normal driving environment for the training of hidden Markov models of driving events, including normal driving, acceleration, deceleration, changing to the left or right lanes, zigzag driving, and approaching the car in front. In addition to recognizing these driving events, the degree of each event is estimated according to its character. In the second stage, the danger-level indicator, which warns the driver of a dangerous situation, is inferred by fuzzy logic rules that address the recognized driving events and their degrees. A hierarchical decision strategy is also designed to reduce the number of rules that are triggered. The proposed framework was successfully implemented on a TI DM3730-based embedded platform and was fully evaluated in a real road environment. The experimental results achieved a detection ratio of 99 % for event recognition, compared with that achieved by four conventional methods.

This study proposes a new approach to video-based traffic surveillance using a fuzzy hybrid information inference mechanism (FHIIM). The three major contributions of the proposed approach are background updating, vehicle detection with block-based segmentation, and vehicle tracking with error compensation. During background updating, small-range updating is adopted to overcome environmental changes under congested conditions. During vehicle detection, the proposed approach detects the vehicle candidates from the foreground image, and it resolves problems such as headlight effects. The tracking technique is employed to track vehicles in consecutive frames. First, the method detects edge features in congested scenes. Next, FHIIM is employed to determine the tracked vehicles. Finally, a method that compensates for error cases under congested conditions is applied to refine the tracking qualities. In our experiments, we tested scenarios both inside and outside the tunnel with three lanes. The results showed that the proposed system exhibits good performance under congested conditions.

In this paper, a vehicle detection approach for complex environments is presented. This paper proposes methods for solving problems of vehicle detection in traffic jams and complex weather conditions such as sunny days, rainy days, cloudy days, sunrise time, sunset time, or nighttime. In recent research, there have been many well-known vehicle detectors that utilize background extraction methods to recognize vehicles. In these studies, the background image needs to continuously be updated; otherwise, the luminance variation will impact the detection quality. The vehicle detection under various environments will have many difficulties such as illumination vibrations, shadow effects, and vehicle overlapping problems that appear in traffic jams. The main contribution of this paper is to propose an adaptive vehicle detection approach in complex environments to directly detect vehicles without extracting and updating a reference background image in complex environments. In the proposed approach, histogram extension addresses the removal of the effects of weather and light impact. The gray-level differential value method is utilized to directly extract moving objects from the images. Finally, tracking and error compensation are applied to refine the target tracking quality. In addition, many useful traffic parameters are evaluated. These useful traffic parameters, including traffic flows, velocity, and vehicle classifications, can help to control traffic and provide drivers with good guidance. Experimental results show that the proposed methods are robust, accurate, and powerful enough to overcome complex weather conditions and traffic jams.

This paper presents a safe and comfortable longitudinal automation system which incorporates human-in-the-loop technology. The proposed system has a hierarchical structure that consists of an adaptive detection area, a supervisory control, and a regulation control. The adaptive detection area routes the information from on-board sensors to ensure the detection of vehicles ahead, particularly when driving on curves. Based on the recognized target distance from the adaptive detection area, the supervisory control determines the desired velocity for the vehicle to maintain safety and smooth operation in different modes. The regulation control utilizes a soft-computing technique and drives the throttle to execute the commanded velocity from the supervisory control. The feasible detection range is within 45 m, and the high velocity for the system operation is up to 100 km/h. The throttle automation under low velocity at 10-30 km/h can also be well managed by the regulation control. Numerous experimental tests in a real traffic environment exhibit the system's validity and achievement in the desired level of comfort through the evaluation of international standard ISO 2631-1.

This study presents an effective system for detecting and tracking moving vehicles in nighttime traffic scene for traffic surveillance. The proposed method identifies vehicles based on detecting and locating vehicle headlights and taillights by using the techniques of image segmentation and pattern analysis. First, to effectively extract bright objects of interest, a fast bright-object segmentation process based on automatic multilevel histogram thresholding is applied on the nighttime road-scene images. This automatic multilevel thresholding approach can provide robustness and adaptability for the detection system to be operated well under various illumination conditions at night. The extracted bright objects are processed by a spatial clustering and tracking procedure by locating and analyzing the spatial and temporal features of vehicle light patterns, and then identifying and classifying the moving cars and motorbikes in the traffic scenes. Experimental results demonstrate that the proposed approach is feasible and effective for vehicle detection and identification in various nighttime environments for traffic surveillance.

This study analyzes the absolute stability in static fuzzy logic control systems with certain and uncertain parameters. For certain static fuzzy control systems, the absolute stability can be analyzed with Popov criterion. The uncertain parameters for absolute stability analysis include the reference input, actuator gain and interval linear plant. The parametric robust Popov criterion based on Lurpsilae systems is applied to stability analysis respect to uncertain parameters. In our work, the parametric robust Popov criterion is applied to analyze absolute stability in static fuzzy logic control systems first time. This study can provide a valuable reference in designing fuzzy control systems. Finally, numerical simulations are provided to verify the analytical results.

This paper presents a real-time vision system for assisting driver during nighttime driving. The proposed system provides the following features: 1) effectively detection and tracking of oncoming and preceding vehicles based on image segmentation and pattern analysis techniques. 2) Robust and adaptive vehicle detection under various illuminated conditions at nighttime urban environments benefited by a novel automatic object segmentation scheme. 3) Providing beneficial information for assisting the driver to perceive surrounding traffic conditions outside the car during nighttime driving. 4) Providing a versatile control strategy for in-vehicle facilities of the autonomous vehicles. 5) Offering real-time traffic event-driven video surveillance machinery for recording evidences of possible traffic accidents. Experimental results demonstrate the feasibility and effectiveness of the proposed system on nighttime driver assistance issues.

Vehicle automation is an important research topic of advanced vehicle systems (AVS). Taiwan iTS-1 is the first smart car with autonomous driving in Taiwan. In this paper, an embedded driving-assistance system is presented. The definition of hierarchical-control structure is necessary in the system to deal with sensorial inputs and environmental and procedural knowledge to manage vehicle actuators in order to accomplish various driving tasks. Upper-level control perceives road environment and determines the proper and safe operation modes including lane-keeping, lane-change, cruise control, adaptive cruise control, and stop-and-go. In each mode, the desired-velocity and reference-trajectory are primarily determined, and then are forwarded to vehicle-body control. To incorporate well driver behavior into our system, vehicle-body control utilizes the fuzzy control technique to manage the fundamental actuators of vehicle, steering wheel, throttle, and brake, to adapt to the desired command (velocity and trajectory). The core controller is built-in on a DSP-based embedded computing platform. The aim of our system is to provide the driving-assistance in the same way human drivers do.

In this paper, a real-time vision-based safety assist system is proposed to provide the driver an assistance system improving the security on driving and after parking. The estimated distances and a friendly intuitive graph are shown on the screen for drivers to examine the distance between other vehicles. The driving status can also be recorded in H.264. Furthermore, the users' mobile phones can monitor the images inside the vehicles to prevent the thieves after parking. The system has been implemented with the embedded systems and tested by the real road environment.

In this paper, an intelligent automated lane-keeping system is proposed and implemented on our vehicle platform, i.e., TAIWAN i TS-1. This system challenges the online integrating heterogeneous systems such as a real-time vision system, a lateral controller, in-vehicle sensors, and a steering wheel actuating motor. The implemented vision system detects the lane markings ahead of the vehicle, regardless of the varieties in road appearance, and determines the desired trajectory based on the relative positions of the vehicle with respect to the center of the road. To achieve more humanlike driving behavior such as smooth turning, particularly at high levels of speed, a fuzzy gain scheduling (FGS) strategy is introduced to compensate for the feedback controller for appropriately adapting to the SW command. Instead of manual tuning by trial and error, the methodology of FGS is designed to ensure that the closed-loop system can satisfy the crossover model principle. The proposed integrated system is examined on the standard testing road at the Automotive Research and Testing Center (ARTC)1and extra-urban highways.

This paper addresses the development and hardware implementation of an efficient hierarchical motion estimation algorithm, HMEA, using multiresolution frames to reduce the computational complexity. Excellent estimation performance is ensured using an averaging filter to downsample the original image. At the smallest resolution, the least two motion vector candidates are selected using a full-search block matching algorithm. At the middle level, these two candidate motion vectors are employed as the center points for small range local searches. Then, at the original resolution, the final motion vector is obtained by performing a local search around the single candidate from the middle level. HMEA exhibits regular data flow and is suitable for hardware implementation. An efficient VLSI architecture that includes an averaging filter to downsample the image and two 2-D semisystolic processing element arrays to determine the sum of absolute difference in pipeline is also presented. Simulation results indicate that HMEA is more area-efficient and faster than many full-search and multiresolution architectures while maintaining high video quality. This architecture with 59K gates and 1393 bytes of RAM is implemented for a search range of [ -16.0, +15.5].

In this paper, an integrated headway adaptation (HA) with collision avoidance (CA) system is proposed to be associated with a human factor aspect for driving assistance. The maneuver in the CA strategy is developed based on the idea that warning or potential crash information generally can be graded such that the driver need not react to a discrete on/off warning. The control law for the HA strategy is designed by using a modified sliding surface to achieve driving comfort requirement and also stability in car-following. To simulate the speed controlling of human drivers, the throttle and brake actuation is directed by the regulation control which is designed based on a human reasoning approach. A description of an experimental vehicle and real traffic testing in highways and urban-like tracks shows the potential and validness of the integrated system.

In this paper, the camera calibration approach based on the perspective model for the tilt angle, pan angle and the height of installation is proposed. Before starting the calibration, the lane information is acquired by applying the Sobel horizontal and vertical edged pixels. The calibration method only utilizes a pair of parallel lane markings without manually selecting special features. The tilt angle, pan angle and camera height are calibrated sequentially in our method. This algorithm can be widely applied in vehicle collision avoidance applications and traffic monitoring systems.

In this study, an indoor localization based on the received signal strength indication (RSSI) in wireless sensor networks (WSN) is proposed. The presented approach proceeds in two phases: the first phase is based on the recorded received signal strength at the certain location. The interpolation, curve fitting and an adaptive neural fuzzy inference system (ANFIS) are used to develop the indoor propagation model, respectively. Thus the strength of the received radio signal can be converted to a physical distance approximately; in the second phase, based on the available distances from the positions localized in the test bed are estimated by using an extended Kalman filter (EKF). In comparison among the propagation models based on the interpolation, ANFIS and curve fitting, the experimental results show that the proposed approach provides a precise performance.

This paper describes a novel approach of detecting lanes. The projective width of lanes can be computed precisely by a method of geometric projection to predict the relative position and features of lane markings. Besides, a camera’s accurate tilt angle and the lane width can be acquired by dynamic calibration. Also, an algorithm named Lane Marking Extraction (LME) Finite State Machine (FSM) is proposed to capture the points with features of lane markings and use fuzzy reasoning to discriminate points of the lane markings. Some points are chosen as knots to supplement all of the points on the lane boundary with cubic B-spline interpolation, which reconstructs the area of lanes with a left and a right curve. The proposed approach uses prediction of a lane’s projective width and the way of tracking to accelerate lane detections and to describe positions of lane markings on both sides effectively when one side is occluded.

The main objective of this paper is to design and implement a real-time autonomous navigation system for intelligent vehicles by integrating a real-time kinematics differential global position system (RTK-DGPS) and a laser range finder. Firstly, a fuzzy logic algorithm for target position tracking is proposed to achieve the human driving concept. In addition, an actuator is equipped in the experimental vehicle TAIWAN iTS-1 so that the steering wheel can be controlled to let the vehicle follows a desired trajectory in digital maps. A laser range finder is used to detect obstacles in front of the vehicle for warning announcement and collision avoidance. A human machine interface is also built up to provide information from sensors. The performance and accuracy of the proposed system is verified by real-time experimental results.

In this paper, the stability in static fuzzy control systems for linear systems with different fixed reference inputs and adjustable parameters are analyzed. Under certain conditions, the unique equilibriums of error in fuzzy control systems can be solved with fixed reference. Furthermore, when the unique equilibrium is stable, the steady state error can be obtained from this stable equilibrium. Under the adjustable parameters and reference inputs, the equilibriums may have stable and unstable transformations each other respect to absolutely stability. Additionally, the insight mechanism for oscillation of equilibrium is also given with a numerical example.

Based on some useful frequency domain methods, this paper proposes a systematic procedure to address the limit cycle prediction of a neural vehicle control system with adjustable parameters. A simple neurocontroller can be linearized by using describing function method firstly. According to the classical method of parameter plane, the stability of linearized system with adjustable parameters is then considered. In addition, gain margin and phase margin for limit cycle generation are also analyzed by adding a gain-phase margin tester into open loop system. Computer simulations show the efficiency of this approach.

This study presents a new method for extracting characters from various real-life complex document images. The proposed method applies a multi-plane segmentation technique to separate homogeneous objects including text blocks, non-text graphical objects, and background textures into individual object planes. It consists of two stages-automatic localized multilevel thresholding, and multi-plane region matching and assembling. Then a text extraction process can be performed on the resultant planes to detect and extract characters with different characteristics in the respective planes. The proposed method processes document images regionally and adaptively according to their respective local features. This allows preservation of detailed characteristics from extracted characters, especially small characters with thin strokes, as well as gradational illuminations of characters. This also permits background objects with uneven, gradational, and sharp variations in contrast, illumination, and texture to be handled easily and well. Experimental results on real-life complex document images demonstrate that the proposed method is effective in extracting characters with various illuminations, sizes, and font styles from various types of complex document images.

This paper presents the integration design and implementation of a longitudinal automation system with the interaction of human-in-the-loop (HITL). The proposed system has a hierarchical structure composed and consists of an adaptive sensory processor, a supervisory control and a regulation control. The adaptive sensory processor routes the information from on-board sensors to avoid missing detection of the vehicle ahead. Based on the recognized measurement from the adaptive sensory processor, the supervisory control determines the desired velocity for the vehicle so as to maintain safety and smooth operation in different modes. The regulation control utilizes soft-computing technique and drives the throttle action to execute the desired velocity commanded from the supervisory control. The feasible sensory distance is within 40 m, and the according driving velocity can achieve 100 km/h upward. The challenge in low velocity operation can also be handled by the regulation control against gear changes and torque converter. Among experimental tests under various kinds of traffic flows, the system validness is exhibited and also the preferable comfort is achieved through the examination of international standard ISO 2631.

This study analyzes front wheels steering (FWS) vehicles using results obtained from two unmanned cars: Unmanned Car I (UCI) and Unmanned Car II (UCII). The vehicle model integrates the lane angle derived from the translational system such that data on the lateral position, the lateral velocity and the lateral acceleration are comprehensively obtained. The main function of each block in the two newly developed structures is described as follows. Two angle controllers were used to eliminate the redundant components of the front-wheel steering angles. A lane scheduled gain (LSG) in each system was used to improve the lane angle deflection in UCI and UCII while the feed-forward controller simulates the behavior of a driver. The use of an empirical pre-filter reduces the lane angle error for UCI and UCII hence enhancing the performance of the system. Finally, the numerical calculation has shown that the two proposed systems are capable of tracking the desired course accurately.

In this paper, the intelligent Stop and Go system (S&G) is designed and implemented for low speed traveling vehicles in urbane area. The human-driving like fuzzy logic control is applied for achieving desired speed and safe inter-vehicle spacing. By proposed intelligent Stop and Go system, the following functions can be obtained including collision avoidance, traffic accidence reduction, driving pressure decrease, and increment of traffic capability. Furthermore, the real car equipped with sensors and a longitudinal controller is provided to demonstrate three scenarios that meet the situations of low speed vehicles in urban area. Three seniors include pedestrian crossing, stop in front of obstacle, and Stop and Go. The application of proposed intelligent Stop and Go system is possible for low speed driving assistance in urban area by our experimental results.

In the paper, perspective transformation is used to project points from the front horizon of the camera to an image plane and thus to measure the distance between the detected object and the camera. With the information about points on the ground, the projective positions of every tip in a rigid object can be figured out through transformation. Features of the object's projection at different distances, such as size and shape, can also be predicted. Besides, the paper has analyzed difference in the result of the measurement and errors caused by the application of different parameters. The information assists engineers of vision-based detection system in determining the parameters of the system and identifying features of an object's projection to accelerate the detection. Also, the camera parameters compensate automatically when being influenced by outer force to promote the effects of detection and make a robust system.

In order to prevent the Advanced Encryption Standard (AES) from suffering from differential fault attacks, the technique of error detection can be adopted to detect the errors during encryption or decryption and then to provide the information for taking further action, such as interrupting the AES process or redoing the process. Because errors occur within a function, it is not easy to predict the output. Therefore, general error control codes are not suited for AES operations. In this work, several error-detection schemes have been proposed. These schemes are based on the (n+1, n) cyclic redundancy check (CRC) over GF(2/sup 8/), where n/spl isin/{4,8,16}. Because of the good algebraic properties of AES, specifically the MixColumns operation, these error detection schemes are suitable for AES and efficient for the hardware implementation; they may be designed using round-level, operation-level, or algorithm-level detection. The proposed schemes have high fault coverage. In addition, the schemes proposed are scalable and symmetrical. The scalability makes these schemes suitable for an AES circuit implemented in 8-bit, 32-bit, or 128-bit architecture. Symmetry also benefits the implementation of the proposed schemes to achieve that the encryption process and the decryption process can share the same error detection hardware. These schemes are also suitable for encryption-only or decryption-only cases. Error detection for the key schedule in AES is also proposed and is based on the derived results in the data procedure of AES.

JPEG2000, which provides a higher compression ratio than the traditional JPEG, is an upcoming compression standard for still images. The experimental results imply that the larger tile size used for JPEG2000 results in better image quality. However, processing the large tile image requires more memory in the hardware implementation. To reduce the hardware resources, a QCB (quad codeblock) based DWT method is proposed to support the processing of large tile images with low memory. Based on the QCB-based DWT engine, three code-blocks belonging to LH/sub 0/, HL/sub 0/ and HH/sub 0/ bands can be generated recursively after each fixed time slice, and the EBC (embedded block coding) processors can directly process the three code-blocks. It can save the size of tile memory by up to 75%. Moreover, the remaining 1/4 size of tile memory can be decreased through the zero holding extension for unavailable data. That is, it only requires 24 Kbytes memory to support the processing of a 512/spl times/512 tile image, with slight image degradation, especially at low bit rates. The low memory requirement makes the hardware implementation practicable.

The robust stability analysis of a neural control based vehicle steering system is considered in this paper. The neural controller can be linearized by the use of describing function first. Since the perturbed parameters including velocity and friction are existed in the vehicle steering system, the stability of equivalent linearized system is then analyzed by using the parameter plane method. According to this procedure, the amplitude and frequency of limit cycles caused by the neural controller can be figured out clearly in the parameter plane. Moreover, the limit cycles may be suppressed by adjusting the control factors carefully. Computer simulation shows the efficiency of this approach.

Text is commonly printed on a complex background. Segmenting text is an important part in document analysis. In the past some methods have been shown for the segmentation of texts with images. However, previous studies have not sufficiently addressed complex compound documents. This investigation presents an algorithm for the segmentation of text in various document images. The proposed segmentation algorithm applies a new multilayer segmentation method to separate the text from various compound document images, independent from the text and background overlapping or not. This method solves various problems associated with the complexity of background images. Experimental results obtained using various document images scanned from book covers, advertisements, brochures and magazines, reveal that the proposed algorithm can successfully segment Chinese and English text strings from various backgrounds, regardless of whether the texts are over a simple, slowly varying or rapidly varying background texture.

In this paper, we propose a high-performance and memory-efficient pipeline architecture which performs the one-level two-dimensional (2-D) discrete wavelet transform (DWT) in the 5/3 and 9/7 filters. In general, the internal memory size of 2-D architecture highly depends on the pipeline registers of one-dimensional (1-D) DWT. Based on the lifting-based DWT algorithm, the primitive data path is modified and an efficient pipeline architecture is derived to shorten the data path. Accordingly, under the same arithmetic resources, the 1-D DWT pipeline architecture can operate at a higher processing speed (up to 200 MHz in 0.25-/spl mu/m technology) than other pipelined architectures with direct implementation. The proposed 2-D DWT architecture is composed of two 1-D processors (column and row processors). Based on the modified algorithm, the row processor can partially execute each row-wise transform with only two column-processed data. Thus, the pipeline registers of 1-D architecture do not fully turn into the internal memory of 2-D DWT. For an N/spl times/M image, only 3.5N internal memory is required for the 5/3 filter, and 5.5N is required for the 9/7 filter to perform the one-level 2-D DWT decomposition with the critical path of one multiplier delay (i.e., N and M indicate the height and width of an image). The pipeline data path is regular and practicable. Finally, the proposed architecture implements the 5/3 and 9/7 filters by cascading the three key components.

In this paper, we apply some effective methods, including the gain-phase margin tester, describing function and parameter plane, to predict the limit cycles of dynamic fuzzy control systems with adjustable parameters. Both continuous-time and sampled-data fuzzy control systems are considered. In general, fuzzy control systems are nonlinear. By use of the classical method of describing functions, the dynamic fuzzy controller may be linearized first. According to the stability equations and parameter plane methods, the stability of the equivalent linearized system with adjustable parameters is then analyzed. In addition, a simple approach is also proposed to determine the gain margin and phase margin which limit cycles can occur for robustness. Two examples of continuous-time fuzzy control systems with and without nonlinearity are presented to demonstrate the design procedure. Finally, this approach is also extended to a sampled-data fuzzy control system.

The main purpose of this paper is to predict the limit cycles of neurocontrol system with perturbed parameters by combining the approaches of stability equation, describing function and parameter plane. The neurocontroller is first linearized by using the describing function method. The stability of equivalent linearized system is then analyzed by using stability equations and the parameter plane method. According to this procedure, the amplitude and frequency of limit cycles can be figured out clearly in the parameter plane. Moreover, the limit cycle may be suppressed by adjusting the control parameters carefully. Finally, a vehicle model is illustrated to demonstrate its validity.

In this paper, we extend some effective methods, including the gain-phase margin tester method, the describing function method and the parameter plane method, to predict the limit cycles of fuzzy control systems. In general, fuzzy control systems are nonlinear. By use of the classical method of describing functions, a fuzzy controller may be linearized first. The stability of the equipment linearization system with adjustable parameters is then analyzed due to the stability equations and parameter plane methods. For the robust design, a novel way is proposed to determine the gain margin and phase margin, which limit cycles, can occur. Two examples are offered to illustrate the design procedure. First, the fuzzy control system with a third-order linear plant is considered. In addition, this approach is also extended to the fuzzy control system with nonlinearity. The results of computer simulation can verify its validity.

The paper presents a VLSI design rule, namely, an embedded instruction code (EIC), for the discrete wavelet transform (DWT). Our approach derives from the essential computations of DWT, and we establish a set of multiplication instructions, MUL, and the addition instruction, ADD. In addition, we propose a parallel arithmetic logic unit (PALU) with two multipliers and four adders, called 2M4A. With these requirements, the DWT computation paths can be calculated more efficiently with limited PALUs. Furthermore, since the EIC is operated under the PALU, the number of needed inner registers depends on the wavelet filters' length. Besides, the boundary problem of DWT has also been resolved by the symmetric extension. Moreover, the two-dimensional inverse DWT (2D IDWT) can be completed using the same PALU as for 2D DWT; the only changes needed to be made are the instruction codes and coefficients. Our chip supports up to six levels of decomposition and versatile image specifications, e.g., VGA, MPEG-1, MPEG-2 and 1024/spl times/1024 image sizes.

The set partitioning in hierarchical trees (SPIHT) algorithm is a computationally simple and efficient zerotree coding technique for image compression. However, the high working memory requirement is its main drawback for hardware realization. We present a low memory zerotree coder (LMZC), which requires much less working memory than SPIHT. The LMZC coding algorithm abandons the use of lists, defines a different tree structure, and merges the sorting pass and the refinement pass together. The main techniques of LMZC are the recursive programming and a top-bit scheme (TBS). In TBS, the top bits of transformed coefficients are used to store the coding status of coefficients instead of the lists used in SPIHT. In order to achieve high coding efficiency, shape-adaptive discrete wavelet transforms are used to transformation arbitrarily shaped objects. A compact emplacement of the transformed coefficients is also proposed to further reduce working memory. The LMZC carefully treats "don't care" nodes in the wavelet tree and does not use bits to code such nodes. Comparison of LMZC with SPIHT shows that for coding a 768 /spl times/ 512 color image, LMZC saves at least 5.3 MBytes of memory but only increases a little execution time and reduces minor peak signal-to noise ratio (PSNR) values, thereby making it highly promising for some memory limited applications.

The emphasis in this article is on the study of nonstationary two-dimensional (2-D) random fields with wide-sense stationary increments, wide-sense stationary jumps, and 2-D fractional Brownian motion (fBm) fields. The effort made in this work is to develop a realizable method of stationarization provided for nonstationary 2-D random fields. We also present the correlation functions of the discrete wavelet transform relating to 2-D fBm fields that will decay hyperbolically fast.