2021 Vol. 30, No. 6

REVIEW
A Survey: FPGA-Based Dynamic Scheduling of Hardware Tasks
LI Tianyang, ZHANG Fan, GUO Wei, SUN Mingqian, CHEN Li
2021, 30(6): 991-1007. doi: 10.1049/cje.2021.07.021
Abstract(611) PDF(191)
Abstract:
To meet the increasing computing needs of various application fields, Field programmable gate array (FPGA) has been widely deployed. In FPGA-based processing, hardware tasks can be better accelerated by allocating appropriate computing resources. Therefore, FPGA-based hardware task scheduling has become one of the mainstream research directions in academia and industry. However, the optimization objectives of existing FPGA-based hardware task scheduling methods are relatively scattered. In this regard, this paper summarizes the research status of hardware task dynamic scheduling from the three essential elements of FPGA processing:time, resources, and power consumption. This paper analyzes, sorts out, categorizes the ideas and implementations of various scheduling methods and analyzes and evaluates optimization effects of various scheduling methods from multiple dimensions. Then, the shortcomings of the existing methods are summarized and some practical applications are introduced. Finally, the research direction of task scheduling based on FPGA is prospected and summarized.
COMPUTERS AND MICROELECTRONICS
An Parallel FPGA SAT Solver Based on Multi-Thread and Pipeline
LI Tiejun, MA Kefan, ZHANG Jianmin
2021, 30(6): 1008-1016. doi: 10.1049/cje.2021.08.001
Abstract(239) PDF(50)
Abstract:
The Boolean Satisfiability (SAT) problem is the key problem in computer theory and application. A parallel multi-thread SAT solver named pprobSAT+ on a configurable hardware is proposed. In the algorithm, multithreads are executed simultaneously to hide the circuit stagnate. In order to improve the working frequency and throughput of the SAT solver, the deep pipeline strategy is adopted. When all data stored in block random access memory of the field programmable gate array, the solver can achieve maximum performance. If partial data are stored in the external memory, the size of the problem instances the SAT solver can be greatly improved. The experimental results show that the speedup of three-thread SAT solver is approximately 2.4 times with single thread, and shows that the pprobSAT+ have achieved substantial improvement while a solution is found.
Ultra-thin Body Buried In0.35Ga0.65As Channel MOSFETs with Extremely Low Off-current on Si Substrates
WANG Bo, DING Peng, FENG Ruize, WANG Yanfu, LIU Xiaoyu, SUN Tangyou, CHEN Yonghe, LIU Xingpeng, LI Qi, LI Yue, LIU Yingbo, YIN Yihui, ZHAO Hao, ZHANG Wei, LI Haiou, JIN Zhi
2021, 30(6): 1017-1021. doi: 10.1049/cje.2021.07.024
Abstract(273) PDF(62)
Abstract:
In this paper, we investigated the electrical properties of the Metal-oxide-semiconductor gate stack of Ti/Al2O3/InP under different annealing conditions. A minimum interface trap density of 3×1011cm-2eV-1 is obtained without postmetallization annealing treatment. Additionally, utilizing Ti/Al2O3/InP MOS gate stack, we fabricated ultra-thin body buried In0.35Ga0.65As channel MOSFETs on Si substrates with optimized on/off trade-off. The 200nm gate length device with extremely low off-current of 0.6nA/µm, and on-off ratio of 3.3×105, is demonstrated by employing buried low indium (In0.35Ga0.65As) channel with InP barrier/spacer device structure, giving strong potential for future highperformance and low-power applications.
New Secondary Constructions of Generalized Bent Functions
YANG Zhiyao, KE Pinhui, CHEN Zhixiong
2021, 30(6): 1022-1029. doi: 10.1049/cje.2021.08.003
Abstract(223) PDF(45)
Abstract:
Three new secondary constructions of generalized bent functions are presented. We provide a secondary construction of generalized bent functions from indirect sum methods proposed by Carlet et al. A new secondary construction of generalized bent functions from four initial functions is also investigated. We demonstrate that many known constructions can be derived from our proposed construction as special cases by choosing proper initial functions and parameters. By modifying the new construction, a novel secondary construction of generalized bent functions from two initial generalized bent functions is obtained. For the binary case, the dual functions of the bent functions by our method are presented, which share the same formula as the indirect sum.
Quantum Differential Collision Distinguishing Attacks on Feistel Schemes
ZHANG Zhongya, WU Wenling, WANG Bolin
2021, 30(6): 1030-1037. doi: 10.1049/cje.2021.07.026
Abstract(228) PDF(48)
Abstract:
Feistel schemes are important components of symmetric ciphers, which have been extensively studied in the classical setting. We examine the extension methods of differential distinguishers of Feistel key-function and Feistel function-key schemes. The schemes are subjected to quantum differential collision distinguishing attacks based on the methods. The results show that the complexity is lower than that of differential attacks using only Grover algorithm, and the complexity of differential collision attack based on the Brassard-Høyer-Tapp and Grover algorithms is lower than that of quantization when using only the Grover algorithm. The results also show that different algorithms and methods can be combined to produce a more effective cryptanalysis approach. This provides a research direction for postquantum cryptographic analysis and design.
Video-Driven 2D Character Animation
YIN Qinran, CAO Weiqun
2021, 30(6): 1038-1048. doi: 10.1049/cje.2021.07.016
Abstract(128) PDF(23)
Abstract:
Video-driven animation has always been a hot and challenging topic in the field of computer animation. We propose a method of mapping a sequence of human skeletal keypoints in a video onto a two-dimensional character to generate 2D character animation. For a given two-dimensional character picture, we extract the motion of real human in video data, driving the character deformation. We analyze common two-dimensional human body movements, classify the basic posture of the human body, realize the recognition of skeleton posture based on back propagation network, capture human body motion by automatically tracking the position of the human skeleton keypoints coordinates in the video and redirect the motion data to a 2D character. Compared with the traditional method, our work is less affected by video data illumination and background complexity. We calibrate human body motion in videos to a 2D character according to the skeleton topology to avoid motion distortion caused by the difference in skeleton size and ratio. The experimental results show that the proposed algorithm can generate the motion of two-dimensional characters based on the motion of human characters in video data. The animation is natural and smooth, and the algorithm has strong robustness.
Achieving Safe Deep Reinforcement Learning via Environment Comprehension Mechanism
PENG Pai, ZHU Fei, LIU Quan, ZHAO Peiyao, WU Wen
2021, 30(6): 1049-1058. doi: 10.1049/cje.2021.07.025
Abstract(127) PDF(20)
Abstract:
Deep reinforcement learning (DRL), which combines deep learning with reinforcement learning, has achieved great success recently. In some cases, however, during the learning process agents may reach states that are worthless and dangerous where the task fails. To address the problem, we propose an algorithm, referred as Environment comprehension mechanism (ECM) for deep reinforcement learning to attain safer decisions. ECM perceives hidden dangerous situations by analyzing object and comprehending the environment, such that the agent bypasses inappropriate actions systematically by setting up constraints dynamically according to states. ECM, which calculates the gradient of the states in Markov tuple, sets up boundary conditions and generates a rule to control the direction of the agent to skip unsafe states. ECM is able to be applied to basic deep reinforcement learning algorithms to guide the selection of actions. The experiment results show that the algorithm promoted safety and stability of the control tasks.
TELECOMMUNICATIONS
Secure Scalar Product Protocols
LI Shundong, ZHANG Mengyu and XU Wenting
2021, 30(6): 1059-1068. doi: 10.1049/cje.2021.08.002
Abstract(95) PDF(13)
Abstract:
Secure scalar product computation is a special secure multi-party computation problem. A secure scalar product protocol can be used by two parties to jointly compute the scalar product of their private vectors without revealing any information about the private vector of either party. Secure scalar product protocol is of great significance in privacy-preserving scientific computing, privacy preserving data mining, privacypreserving cooperative statistical analysis and privacypreserving geometry computation, etc. Many privacy preserving computing problems can be transformed to secure scalar product computation. At present, existing scalar product protocols cannot be used to privately compute scalar product of private vectors with both positive and negative components. Based on homomorphic encryption scheme, we design three protocols to compute scalar product of three different kinds of private vectors. The components of the first kind vector are arbitrary integers; those of the second kind are positive rational numbers and those of the third kind are arbitrary rational numbers. We use simulation paradigm proving that the protocols are secure in the semi-honest model. Theoretical analysis and experimental results show that the protocols designed in this paper are efficient.
A Blockchain-Based Flexible Data Auditing Scheme for the Cloud Service
FAN Kefeng, LI Fei, YU Haiyang, YANG Zhen
2021, 30(6): 1159-1166. doi: 10.1049/cje.2021.08.011
Abstract(123) PDF(20)
Abstract:
Nowdays, cloud storage technology has become a hot topic, and an increasing number of users are concerned with the security of their data in the cloud. Many auditing schemes on the cloud are proposed and the introduction of a third-party auditor to assist users in verifying the integrity of cloud data. As a centralized node, the third-party auditor has to communicate with all cloud users and cloud service providers, which becomes the bottleneck of the whole scheme. To solve this problem, we design a blockchain-based flexible cloud data auditing scheme. In our scheme, a decentralized auditing framework is proposed to eliminate the dependency on the thirdparty auditor, which increases the stability, security and performance of the whole scheme. Since the cloud service provider can automatically generates auditing proofs, our scheme can relieve the communication burdens of the cloud service provider. The proposed scheme also adapts the Merkle Hash tree to improve the verification performance. Security analysis and experiments show that the proposed scheme is secure and has better stability and verification efficiency.
Low-Complexity Digital Equalizers for High-Speed Underwater Optical Wireless Communication
LI Chao, HU Zhijia
2021, 30(6): 1167-1172. doi: 10.1049/cje.2021.08.012
Abstract(123) PDF(20)
Abstract:
Hundreds of Megabit per second (Mb/s) data transmission over a moderate range (10m) underwater optical wireless channel is implemented with 4-level pulse amplitude modulation scheme using an off-the-shelf light-emitting diode as the optical source and digital equalizers. In particular, the channel nonlinear distortion from light-emitting diode is considered and a matched model is developed. To mitigate these distortions under various 4-level pulse amplitude modulation signal baud rates, the Bit error rate (BER) and Signal-to-noise ratio (SNR) performances of the pre-, post- and matchedequalizers are evaluated and compared. The digital matched-equalizer bank is shown and verified to offer the best BER/SNR performance for 4-level pulse amplitude modulation signal using more than 200 MBaud rate. With the proposed equalizer, we have recorded the highest data rate of 600Mb/s at 20% FEC limit over 10m free-space underwater link.
Fast Fading Channel Estimation for OFDM Systems with Complexity Reduction
ZHANG Chen, LI Dan
2021, 30(6): 1173-1177. doi: 10.1049/cje.2021.07.023
Abstract(128) PDF(13)
Abstract:
Channel estimation plays a significant role in the Inter-carrier interference (ISI) mitigation and symbol detection in Orthogonal frequency division multiplexing (OFDM) systems. As the speed of mobile terminal increases, the channel changes rapidly, causing fast fading which will degrade the system performance. In this paper, a low complexity fast fading channel estimation method suited for scattered pilot OFDM systems is proposed. This method is characterized by estimating channel matrix directly in the frequency domain, which can significantly reduce the channel estimation complexity with good system performance. Simulation results verify its superiority over the existing estimation methods.
Trusted Network Evaluation Model Based on Comprehensive Trust
YIN Ansheng, HUANG Haiping
2021, 30(6): 1178-1188. doi: 10.1049/cje.2021.07.028
Abstract(101) PDF(14)
Abstract:
Trusted network is currently evaluated based on the trustworthiness of the nodes in the network; thus, trust in the nodes is required to achieve trust in the network. The methods for calculating node trust mainly utilize the following three types of models:node behavior model, multiple-attribute decision-making model and reputation model. User behavior model always determines trust values based on harmful node behaviors. Multi-attribute decision-making model can be regarded as refinements of node behavior model. The trust values of the nodes are determined based on compromised trust attribute data and not just harmful behavior data. Reputation model calculates the trust values of nodes using subjective evaluation data for nodes. This paper presents a comprehensive trust model that combines subjective evaluation data and objective attribute data to calculate node trust. The experimental results show that malicious nodes can be effectively identified, which improves the service success rate of the system.
SIGNAL PROCESSING
A Quaternion Two-Stream R-CNN Network for Pixel-Level Color Image Splicing Localization
CHEN Beijing, JU Xingwang, GAO Ye, WANG Jinwei
2021, 30(6): 1069-1079. doi: 10.1049/cje.2021.08.004
Abstract(149) PDF(23)
Abstract:
Recently, Zhou et al. designed a twostream faster Region-Convolutional neural networks (RCNN) model RGB-N for color image splicing localization in CVPR2018. However, the RGB-N locates spliced regions only at block-level and ignores the entirety and inherent correlation of three channels. Therefore, an improved quaternion two-stream R-CNN model is proposed to solve these drawbacks:a mask branch combining fully convolutional network and condition random field is added for locating spliced regions at pixel-level; quaternion representation of color images is used to process color spliced images in a holistical way. In addition, feature pyramid network based on quaternion residual network is considered to extract multi-scale features for color spliced images; attention region proposal network is combined with attention mechanism and is designed to pay more attention to the spliced regions; a high-pass filter designed for image splicing detection specifically is adopted to replace steganalysis rich model filter in the RGB-N to obtain noise input for the noise stream. Experimental results on a new synthetic dataset and three standard forgery datasets demonstrate that the proposed method is superior to the existing methods in the abilities of localization, generalization, and robustness.
Traffic Sign Recognition Using an Attentive Context Region-Based Detection Framework
LIU Zhigang, DU Juan, TIAN Feng, WEN Jiazheng
2021, 30(6): 1080-1086. doi: 10.1049/cje.2021.08.005
Abstract(180) PDF(31)
Abstract:
Accurate small traffic sign recognition is more important for the safety of intelligent transportation systems. A recognition framework named attentive context region-based detection framework (AC-RDF) is proposed in this paper. We construct the attentive context feature for the recognition of small traffic signs, which combines the target information and the contextual information by the concatenation operation following a pointwise convolutional layer. The proposed attentive context feature exploits the surrounding information for a given object proposal. Next, we propose a novel attentive loss function to replace the original crossentropy function. It distinguishes hard negative samples from easy positive ones in the total loss, allows the proposed framework to obtain enough training, and further improve the recognition accuracy. The proposed method is evaluated on the challenging Tsinghua-Tencent 100K dataset. The experimental results indicate that the attentive context region-based detection framework is superior at detecting small traffic signs and achieves stateof-the-art performance compared with other methods.
Human Following for Outdoor Mobile Robots Based on Point-Cloud's Appearance Model
GONG Linxi, CAI Yunfei
2021, 30(6): 1087-1095. doi: 10.1049/cje.2021.07.017
Abstract(152) PDF(9)
Abstract:
In this paper, we propose a point-cloudbased algorithm for human-following robots to detect and follow the target person in a complex outdoor environment. Specifically, we exploit AdaBoost to train a binary classifier in a designed feature space based on sparse point-cloud to distinguish the target person from other objects. Then a particle filter is applied to continuously track the target's position. Motivated by the interference of obstacles in long-distance human-following scenarios, a motion plan algorithm based on vector field histogram is adopted. Experiments are carried out both on the dataset we collected and in real application scenarios. The results show that our algorithm has the ability of real-time target detection and tracking, and is robust to deal with complex situations in outdoor environments.
Porn Streamer Recognition in Live Video Based on Multimodal Knowledge Distillation
WANG Liyuan, ZHANG Jing, YAO Jiacheng, ZHUO Li
2021, 30(6): 1096-1102. doi: 10.1049/cje.2021.07.027
Abstract(308) PDF(18)
Abstract:
Although deep learning has reached a higher accuracy for video content analysis, it is not satisfied with practical application demands of porn streamer recognition in live video because of multiple parameters, complex structures of deep network model. In order to improve the recognition efficiency of porn streamer in live video, a deep network model compression method based on multimodal knowledge distillation is proposed. First, the teacher model is trained with visual-speech deep network to obtain the corresponding porn video prediction score. Second, a lightweight student model constructed with MobileNetV2 and Xception transfers the knowledge from the teacher model by using multimodal knowledge distillation strategy. Finally, porn streamer in live video is recognized by combining the lightweight student model of visualspeech network with the bullet screen text recognition network. Experimental results demonstrate that the proposed method can effectively drop the computation cost and improve the recognition speed under the proper accuracy.
Video Stabilization via Prediction with Time-Series Network and Image Inpainting with Pyramid Fusion
CHENG Keyang, LI Shichao, RONG Lan, WANG Wenshan, SHI Wenxi, ZHAN Yongzhao
2021, 30(6): 1103-1110. doi: 10.1049/cje.2021.08.006
Abstract(128) PDF(9)
Abstract:
Due to the poor filling effect of the video image defect commonly used in the video stabilization field, the video is seemed still unstable after the image stabilization process, which seriously affects the visual effect. To solve this problem, we improve a video stabilization method based on time-series network prediction and pyramid fusion restoration is proposed to optimize the visual effect after image stabilization. The flow of the proposed method is as follows:First, it is adaptive to determine whether the defect of the corresponding frame at the current time needs padding inpainting. Then, for the frame that needs to be inpainting, the frames generated before the current moment are sent to the model combining the convolutional neural networks and the gate recurrent unit to predict the part to be filled. Next the current defect image and the complete image to be filled are brought into the Laplacian pyramid reconstruction, and the improved weighted optimal suture is introduced for splicing during the fusion. Finally, the video frame is cut after reconstruction. The method is tested on a data set composed of videos commonly used in the field of video stabilization. The experimental results show that the average peak signal to noise ratio of the method is 2 to 5dB higher than that of the comparison algorithm, and the average structural similarity index is improved by about 2% to 7% compared with the contrast algorithm.
Just-in-Time Human Gesture Recognition Using WiFi Signals
MENG Xianjia, FENG Lin, CHEN Hao, CHEN Ting, MA Jianfeng, WANG Anwen, LIU Dongdong, ZHAO Yanfeng
2021, 30(6): 1111-1119. doi: 10.1049/cje.2021.07.022
Abstract(120) PDF(19)
Abstract:
In-air gesture recognition using wireless signals acts as a key enabler for various applications including smart homes, remote healthcare, shared autopilot, etc. Although researchers have conducted extensive research on WiFi-based gesture recognition, it remains an open question of providing accurate, robust, and in-time recognition system with the commodity WiFi infrastructure. We present FaSee, a just-in-time WiFibased gesture recognition system by identifying the fine-grained Channel state information (CSI) features upon off-the-shelf WiFi devices. The core of FaSee is essentially a novel hybrid recognition algorithm, which combines the classical K-Means algorithm with Dynamic time warping (DTW) together, to transform the feature matching in traditional gesture recognition schemes into a hierarchical manner, thereby significantly improving the recognition efficiency. Experimental results show that FaSee recognizes 9 representative gestures with an average accuracy of 94.75% without tedious per-person training, while achieving 30% signal processing delay saving when compared with the state-of-the-arts gesture recognition schemes.
Dilated Convolutional Pixels Affinity Network for Weakly Supervised Semantic Segmentation
ZHANG Zhe, WANG Bilin, YU Zhezhou, LI Zhiyuan
2021, 30(6): 1120-1130. doi: 10.1049/cje.2021.08.007
Abstract(189) PDF(23)
Abstract:
This paper studies semantic segmentation primarily under image-level weak-supervision. Most stateof-the-art technologies have recently used deep classification networks to create small and sparse discriminatory seed regions of each interest target as pseudo-labels for training segmentation networks, which achieve inferior performance compared with the fully supervised setting. We propose a Dilated convolutional pixels affinity network (DCPAN) to localize and expand the seed regions of objects to bridge this gap. Although introduced dilated convolutional units enable capture of additional location information of objects, it falsely highlighted true negative regions as dilated rate enlarge. To address this problem, we properly integrate dilated convolutional units with different dilated rates and self-attention mechanisms to obtain pixel affinity measure matrix for promoting classification network to generate high-quality object seed regions as pseudo-labels; thus, the performance of the segmentation network is boosted. Furthermore, although our approach seems simple, our method obtains a competitive performance, and experiments show that the performance of DCPAN outperforms other state-ofart approaches in weakly-supervised settings, which only use image-level labels on the Pascal VOC 2012 dataset.
Detecting Three-Dimensional Associations in Large Data Set
LIU Chuanlu, WANG Shuliang, YUAN Hanning, GENG Jing
2021, 30(6): 1131-1140. doi: 10.1049/cje.2021.08.008
Abstract(125) PDF(4)
Abstract:
The associations detection among variables in the large dataset is recently important due to the rapid growth rate of data. The interested associations can provide references for solving the problems such as dimension reduction and feature selection. Many methods have done on the associations detection of pairwise variables. The multi-dimensional variables, especially three-dimensional variables, is rarely studied. The relationships among them cannot be revealed by the detection of pairwise variables methods. A new method of Maximal three-dimensional information coefficient (MTDIC) is proposed which is able to indicate the associations of three-dimensional variables. The correlation coefficient is calculated from the three-dimensional mutual information. The World Health Organization (WHO) data and the Tara data are selected to evaluate their associations. The experiment is verified by comparing the coefficient results with the Distance correlation (Dcor). The accurate association strength is obtained by an iterative optimization procedure on sorting descending order of coefficients. The MTDIC performs better than the Dcor in generality and equitability properties.
IMM Robust Cardinality Balance Multi-Bernoulli Filter for Multiple Maneuvering Target Tracking with Interval Measurement
YANG Biao, ZHU Shengqi, HE Xiongpeng
2021, 30(6): 1141-1151. doi: 10.1049/cje.2021.08.009
Abstract(129) PDF(8)
Abstract:
This paper presents a novel Interacting multi-model (IMM) Robust Cardinality balance multitarget multi-Bernoulli (R-CBMeMBer) filter to solve the maneuvering target tracking problem in the case of interval measurement, unknown detection probability and unknown clutter density. In essence, IMM R-CBMeMBer filter is an extended application of R-CBMeMBer filter. In the IMM R-CBMeMBer filter, the target state is first extended to distinguish clutter from the real target. The detection probability and model probability of the target can be adaptively updated. Then, generalized likelihood function and IMM algorithm are introduced to interactively predict and update the state of the target in the IMM R-CBMeMBer filtering process. In addition, a particle application of the IMM R-CBMeMBer filter is given, and a numerical experiment is designed under nonlinear conditions. Meanwhile, Doppler information of the target is employed to estimate the velocity of each maneuvering target. Numerical experiments also verify that the IMM R-CBMeMBer filter can effectively estimate the target position, target velocity, target detection probability and clutter number in the condition of unknown detection probability, unknown clutter rate and interval measurement.
High-Order Extended Strong Tracking Filter
SUN Xiaohui, WEN Tao, WEN Chenglin, CHENG Xingshuo, WU Yunkai
2021, 30(6): 1152-1158. doi: 10.1049/cje.2021.08.010
Abstract(171) PDF(20)
Abstract:
A novel High-order extended Strong tracking filter (H-STF) is proposed for a class of nonlinear systems. All high-order polynomial terms in the state model are regarded as implicit variables; the original state model is equivalently formulated into a pseudo-linear form; the dynamic relationship between each implicit variable and all variables is modeled; original state model is rewritten into an augmented linear model; the nonlinear measurement model can be rewritten into linear form; taking into account the problems of modeling errors and state mutations that may be caused by the introduction of implicit variables, a high-order extended strong tracking filter is designed. Examples are presented to demonstrate the effectiveness of the new algorithm.