- Feng Wang;Huaping Liu;Fuchun Sun;Haihong Pan;
In this work, we use a deep learning method to tackle the Zero-Shot Learning(ZSL) problem in tactile material recognition by incorporating the advanced semantic information into a training model. Our main technical contribution is our proposal of an end-to-end deep learning framework for solving the tactile ZSL problem. In this framework, we use a Convolutional Neural Network(CNN) to extract the spatial features and Long Short-Term Memory(LSTM) to extract the temporal features in dynamic tactile sequences, and develop a loss function suitable for the ZSL setting. We present the results of experimental evaluations on publicly available datasets, which show the effectiveness of the proposed method.
2019年06期 v.24 645-653页 [查看摘要][在线阅读][下载 551K] [下载次数：64 ] |[网刊下载次数：0 ] |[引用频次：5 ] |[阅读次数：1 ] - Feng Wang;Huaping Liu;Fuchun Sun;Haihong Pan;
In this work, we use a deep learning method to tackle the Zero-Shot Learning(ZSL) problem in tactile material recognition by incorporating the advanced semantic information into a training model. Our main technical contribution is our proposal of an end-to-end deep learning framework for solving the tactile ZSL problem. In this framework, we use a Convolutional Neural Network(CNN) to extract the spatial features and Long Short-Term Memory(LSTM) to extract the temporal features in dynamic tactile sequences, and develop a loss function suitable for the ZSL setting. We present the results of experimental evaluations on publicly available datasets, which show the effectiveness of the proposed method.
2019年06期 v.24 645-653页 [查看摘要][在线阅读][下载 551K] [下载次数：64 ] |[网刊下载次数：0 ] |[引用频次：5 ] |[阅读次数：1 ] - Bin Fang;Xiang Wei;Fuchun Sun;Haiming Huang;Yuanlong Yu;Huaping Liu;
With the accelerated aging of the global population and escalating labor costs, more service robots are needed to help people perform complex tasks. As such, human-robot interaction is a particularly important research topic. To effectively transfer human behavior skills to a robot, in this study, we conveyed skill-learning functions via our proposed wearable device. The robotic teleoperation system utilizes interactive demonstration via the wearable device by directly controlling the speed of the motors. We present a rotation-invariant dynamicalmovement-primitive method for learning interaction skills. We also conducted robotic teleoperation demonstrations and designed imitation learning experiments. The experimental human-robot interaction results confirm the effectiveness of the proposed method.
2019年06期 v.24 654-662页 [查看摘要][在线阅读][下载 786K] [下载次数：82 ] |[网刊下载次数：0 ] |[引用频次：14 ] |[阅读次数：1 ] - Bin Fang;Xiang Wei;Fuchun Sun;Haiming Huang;Yuanlong Yu;Huaping Liu;
With the accelerated aging of the global population and escalating labor costs, more service robots are needed to help people perform complex tasks. As such, human-robot interaction is a particularly important research topic. To effectively transfer human behavior skills to a robot, in this study, we conveyed skill-learning functions via our proposed wearable device. The robotic teleoperation system utilizes interactive demonstration via the wearable device by directly controlling the speed of the motors. We present a rotation-invariant dynamicalmovement-primitive method for learning interaction skills. We also conducted robotic teleoperation demonstrations and designed imitation learning experiments. The experimental human-robot interaction results confirm the effectiveness of the proposed method.
2019年06期 v.24 654-662页 [查看摘要][在线阅读][下载 786K] [下载次数：82 ] |[网刊下载次数：0 ] |[引用频次：14 ] |[阅读次数：1 ] - Qi Dang;Jianqin Yin;Bin Wang;Wenqing Zheng;
Human pose estimation has received significant attention recently due to its various applications in the real world. As the performance of the state-of-the-art human pose estimation methods can be improved by deep learning, this paper presents a comprehensive survey of deep learning based human pose estimation methods and analyzes the methodologies employed. We summarize and discuss recent works with a methodologybased taxonomy. Single-person and multi-person pipelines are first reviewed separately. Then, the deep learning techniques applied in these pipelines are compared and analyzed. The datasets and metrics used in this task are also discussed and compared. The aim of this survey is to make every step in the estimation pipelines interpretable and to provide readers a readily comprehensible explanation. Moreover, the unsolved problems and challenges for future research are discussed.
2019年06期 v.24 663-676页 [查看摘要][在线阅读][下载 1890K] [下载次数：271 ] |[网刊下载次数：0 ] |[引用频次：108 ] |[阅读次数：1 ] - Qi Dang;Jianqin Yin;Bin Wang;Wenqing Zheng;
Human pose estimation has received significant attention recently due to its various applications in the real world. As the performance of the state-of-the-art human pose estimation methods can be improved by deep learning, this paper presents a comprehensive survey of deep learning based human pose estimation methods and analyzes the methodologies employed. We summarize and discuss recent works with a methodologybased taxonomy. Single-person and multi-person pipelines are first reviewed separately. Then, the deep learning techniques applied in these pipelines are compared and analyzed. The datasets and metrics used in this task are also discussed and compared. The aim of this survey is to make every step in the estimation pipelines interpretable and to provide readers a readily comprehensible explanation. Moreover, the unsolved problems and challenges for future research are discussed.
2019年06期 v.24 663-676页 [查看摘要][在线阅读][下载 1890K] [下载次数：271 ] |[网刊下载次数：0 ] |[引用频次：108 ] |[阅读次数：1 ]
- Kaiming Nan;Sicong Liu;Junzhao Du;Hui Liu;
Despite the rapid development of mobile and embedded hardware, directly executing computationexpensive and storage-intensive deep learning algorithms on these devices' local side remains constrained for sensory data analysis. In this paper, we first summarize the layer compression techniques for the state-of-theart deep learning model from three categories: weight factorization and pruning, convolution decomposition, and special layer architecture designing. For each category of layer compression techniques, we quantify their storage and computation tunable by layer compression techniques and discuss their practical challenges and possible improvements. Then, we implement Android projects using TensorFlow Mobile to test these 10 compression methods and compare their practical performances in terms of accuracy, parameter size, intermediate feature size,computation, processing latency, and energy consumption. To further discuss their advantages and bottlenecks,we test their performance over four standard recognition tasks on six resource-constrained Android smartphones.Finally, we survey two types of run-time Neural Network(NN) compression techniques which are orthogonal with the layer compression techniques, run-time resource management and cost optimization with special NN architecture,which are orthogonal with the layer compression techniques.
2019年06期 v.24 677-693页 [查看摘要][在线阅读][下载 794K] [下载次数：61 ] |[网刊下载次数：0 ] |[引用频次：25 ] |[阅读次数：1 ] - Kaiming Nan;Sicong Liu;Junzhao Du;Hui Liu;
Despite the rapid development of mobile and embedded hardware, directly executing computationexpensive and storage-intensive deep learning algorithms on these devices' local side remains constrained for sensory data analysis. In this paper, we first summarize the layer compression techniques for the state-of-theart deep learning model from three categories: weight factorization and pruning, convolution decomposition, and special layer architecture designing. For each category of layer compression techniques, we quantify their storage and computation tunable by layer compression techniques and discuss their practical challenges and possible improvements. Then, we implement Android projects using TensorFlow Mobile to test these 10 compression methods and compare their practical performances in terms of accuracy, parameter size, intermediate feature size,computation, processing latency, and energy consumption. To further discuss their advantages and bottlenecks,we test their performance over four standard recognition tasks on six resource-constrained Android smartphones.Finally, we survey two types of run-time Neural Network(NN) compression techniques which are orthogonal with the layer compression techniques, run-time resource management and cost optimization with special NN architecture,which are orthogonal with the layer compression techniques.
2019年06期 v.24 677-693页 [查看摘要][在线阅读][下载 794K] [下载次数：61 ] |[网刊下载次数：0 ] |[引用频次：25 ] |[阅读次数：1 ] - Jiawei Tao;Tao Zhang;Yongfang Nie;
In this paper, a flexible spacecraft attitude control scheme that guarantees vibration suppression and prescribed performance on transient-state behavior is proposed. Here, parametric uncertainty, external disturbance, unmeasured elastic vibration, actuator saturation, and even configuration misalignment are considered. To guarantee prescribed performance bounds on the transient-and steady-state control errors, a performance constrained control law is formulated with an error transformed function. An elastic modal observer is employed to estimate the unmeasured flexible modal variables, and a command filter is adopted to avoid the tedious analytical computations of time derivatives of virtual control inherent in the control design. Subsequently, a novel auxiliary system is developed to compensate the adverse effects of the actuator saturation constraints, and a compensation term is integrated into the control law to tackle the configuration misalignment. A comparative simulation study is carried out to illustrate the effectiveness and advantages of the proposed approach.
2019年06期 v.24 694-705页 [查看摘要][在线阅读][下载 1056K] [下载次数：46 ] |[网刊下载次数：0 ] |[引用频次：7 ] |[阅读次数：2 ] - Jiawei Tao;Tao Zhang;Yongfang Nie;
In this paper, a flexible spacecraft attitude control scheme that guarantees vibration suppression and prescribed performance on transient-state behavior is proposed. Here, parametric uncertainty, external disturbance, unmeasured elastic vibration, actuator saturation, and even configuration misalignment are considered. To guarantee prescribed performance bounds on the transient-and steady-state control errors, a performance constrained control law is formulated with an error transformed function. An elastic modal observer is employed to estimate the unmeasured flexible modal variables, and a command filter is adopted to avoid the tedious analytical computations of time derivatives of virtual control inherent in the control design. Subsequently, a novel auxiliary system is developed to compensate the adverse effects of the actuator saturation constraints, and a compensation term is integrated into the control law to tackle the configuration misalignment. A comparative simulation study is carried out to illustrate the effectiveness and advantages of the proposed approach.
2019年06期 v.24 694-705页 [查看摘要][在线阅读][下载 1056K] [下载次数：46 ] |[网刊下载次数：0 ] |[引用频次：7 ] |[阅读次数：1 ] - Yanhua Yu;Jie Li;
In this paper, we propose a novel Residuals-Based Deep Least Squares Support Vector Machine(RBDLSSVM). In the RBD-LSSVM, multiple LSSVMs are sequentially connected. The second LSSVM uses the fitting residuals of the first LSSVM as input time series, and the third LSSVM trains the residuals of the second, and so on. The original time series is the input of the first LSSVM. Additionally, to obtain the best hyper-parameters for the RBD-LSSVM, we propose a model validation method based on redundancy test using Omni-Directional Correlation Function(ODCF). This method is based on the fact when a model is appropriate for a given time series, there should be no information or correlation in the residuals. We propose the use of ODCF as a statistic to detect nonlinear correlation between two random variables. Thus, we can select hyper-parameters without encountering overfitting,which cannot be avoided by only cross validation using the validation set. We conducted experiments on two time series: annual sunspot number series and monthly Total Column Ozone(TCO) series in New Delhi. Analysis of the prediction results and comparisons with recent and past studies demonstrate the promising performance of the proposed RBD-LSSVM approach with redundancy test based model selection method for modeling and predicting nonlinear time series.
2019年06期 v.24 706-715页 [查看摘要][在线阅读][下载 481K] [下载次数：52 ] |[网刊下载次数：0 ] |[引用频次：15 ] |[阅读次数：1 ] - Yanhua Yu;Jie Li;
In this paper, we propose a novel Residuals-Based Deep Least Squares Support Vector Machine(RBDLSSVM). In the RBD-LSSVM, multiple LSSVMs are sequentially connected. The second LSSVM uses the fitting residuals of the first LSSVM as input time series, and the third LSSVM trains the residuals of the second, and so on. The original time series is the input of the first LSSVM. Additionally, to obtain the best hyper-parameters for the RBD-LSSVM, we propose a model validation method based on redundancy test using Omni-Directional Correlation Function(ODCF). This method is based on the fact when a model is appropriate for a given time series, there should be no information or correlation in the residuals. We propose the use of ODCF as a statistic to detect nonlinear correlation between two random variables. Thus, we can select hyper-parameters without encountering overfitting,which cannot be avoided by only cross validation using the validation set. We conducted experiments on two time series: annual sunspot number series and monthly Total Column Ozone(TCO) series in New Delhi. Analysis of the prediction results and comparisons with recent and past studies demonstrate the promising performance of the proposed RBD-LSSVM approach with redundancy test based model selection method for modeling and predicting nonlinear time series.
2019年06期 v.24 706-715页 [查看摘要][在线阅读][下载 481K] [下载次数：52 ] |[网刊下载次数：0 ] |[引用频次：15 ] |[阅读次数：1 ] - Yunlei Zhang;Bin Wu;Yu Liu;Jinna Lv;
Local community detection aims to find a cluster of nodes by exploring a small region of the network.Local community detection methods are faster than traditional global community detection methods because their runtime does not depend on the size of the entire network. However, most existing methods do not take the higher-order connectivity patterns crucial to the network into consideration. In this paper, we develop a new Local Community Detection method based on network Motif(LCD-Motif) which incorporates the higher-order network information. LCD-Motif adopts the local expansion of a seed set to identify the local community with minimal motif conductance, representing a generalization of the conductance metric for network motifs. In contrast to PageRanklike diffusion methods, LCD-Motif finds the community by seeking a sparse vector in the span of the local spectra,such that the seeds are in its support vector. We evaluate our approach using real-world datasets across various domains and synthetic networks. The experimental results show that LCD-Motif can achieve a higher performance than state-of-the-art methods.
2019年06期 v.24 716-727页 [查看摘要][在线阅读][下载 853K] [下载次数：69 ] |[网刊下载次数：0 ] |[引用频次：16 ] |[阅读次数：1 ] - Yunlei Zhang;Bin Wu;Yu Liu;Jinna Lv;
Local community detection aims to find a cluster of nodes by exploring a small region of the network.Local community detection methods are faster than traditional global community detection methods because their runtime does not depend on the size of the entire network. However, most existing methods do not take the higher-order connectivity patterns crucial to the network into consideration. In this paper, we develop a new Local Community Detection method based on network Motif(LCD-Motif) which incorporates the higher-order network information. LCD-Motif adopts the local expansion of a seed set to identify the local community with minimal motif conductance, representing a generalization of the conductance metric for network motifs. In contrast to PageRanklike diffusion methods, LCD-Motif finds the community by seeking a sparse vector in the span of the local spectra,such that the seeds are in its support vector. We evaluate our approach using real-world datasets across various domains and synthetic networks. The experimental results show that LCD-Motif can achieve a higher performance than state-of-the-art methods.
2019年06期 v.24 716-727页 [查看摘要][在线阅读][下载 853K] [下载次数：69 ] |[网刊下载次数：0 ] |[引用频次：16 ] |[阅读次数：1 ] - Qin Shi;Sihao Zhao;Xiaowei Cui;Mingquan Lu;Mengdi Jia;
Localization systems utilizing Ultra-WideBand(UWB) have been widely used in dense urban and indoor environments. A moving UWB tag can be located by ranging to fixed UWB anchors whose positions are surveyed in advance. However, manually surveying the anchors is typically a dull and time-consuming process and prone to artificial errors. In this paper, we present an accurate and easy-to-use method for UWB anchor self-localization,using the UWB ranging measurements and readings from a low-cost Inertial Measurement Unit(IMU). The locations of the anchors are automatically estimated by freely moving the tag in the environment. The method is inspired by the Simultaneous Localization And Mapping(SLAM) technique used by the robotics community. A tightly-coupled Error-State Kalman Filter(ESKF) is utilized to fuse UWB and inertial measurements, producing UWB anchor position estimates and six Degrees of Freedom(6 DoF) tag pose estimates. Simulated experiments demonstrate that our proposed method enables accurate self-localization for UWB anchors and smooth tracking of the tag.
2019年06期 v.24 728-737页 [查看摘要][在线阅读][下载 622K] [下载次数：197 ] |[网刊下载次数：0 ] |[引用频次：59 ] |[阅读次数：1 ] - Qin Shi;Sihao Zhao;Xiaowei Cui;Mingquan Lu;Mengdi Jia;
Localization systems utilizing Ultra-WideBand(UWB) have been widely used in dense urban and indoor environments. A moving UWB tag can be located by ranging to fixed UWB anchors whose positions are surveyed in advance. However, manually surveying the anchors is typically a dull and time-consuming process and prone to artificial errors. In this paper, we present an accurate and easy-to-use method for UWB anchor self-localization,using the UWB ranging measurements and readings from a low-cost Inertial Measurement Unit(IMU). The locations of the anchors are automatically estimated by freely moving the tag in the environment. The method is inspired by the Simultaneous Localization And Mapping(SLAM) technique used by the robotics community. A tightly-coupled Error-State Kalman Filter(ESKF) is utilized to fuse UWB and inertial measurements, producing UWB anchor position estimates and six Degrees of Freedom(6 DoF) tag pose estimates. Simulated experiments demonstrate that our proposed method enables accurate self-localization for UWB anchors and smooth tracking of the tag.
2019年06期 v.24 728-737页 [查看摘要][在线阅读][下载 622K] [下载次数：197 ] |[网刊下载次数：0 ] |[引用频次：59 ] |[阅读次数：1 ] - Yuzhao Wu;Yongqiang Lyu;Yuanchun Shi;
With ever greater amounts of data stored in cloud servers, data security and privacy issues have become increasingly important. Public cloud storage providers are semi-trustworthy because they may not have adequate security mechanisms to protect user data from being stolen or misused. Therefore, it is crucial for cloud users to evaluate the security of cloud storage providers. However, existing security assessment methods mainly focus on external security risks without considering the trustworthiness of cloud providers. In addition, the widely used thirdparty mediators are assumed to be trusted and we are not aware of any work that considers the security of these mediators. This study fills these gaps by assessing the security of public cloud storage providers and third-party mediators through equilibrium analysis. More specifically, we conduct evaluations on a series of game models between public cloud storage providers and users to thoroughly analyze the security of different service scenarios.Using our proposed security assessment, users can determine the risk of whether their privacy data is likely to be hacked by the cloud service providers; the cloud service providers can also decide on strategies to make their services more trustworthy. An experimental study of 32 users verified our method and indicated its potential for real service improvement.
2019年06期 v.24 738-749页 [查看摘要][在线阅读][下载 427K] [下载次数：23 ] |[网刊下载次数：0 ] |[引用频次：14 ] |[阅读次数：1 ] - Yuzhao Wu;Yongqiang Lyu;Yuanchun Shi;
With ever greater amounts of data stored in cloud servers, data security and privacy issues have become increasingly important. Public cloud storage providers are semi-trustworthy because they may not have adequate security mechanisms to protect user data from being stolen or misused. Therefore, it is crucial for cloud users to evaluate the security of cloud storage providers. However, existing security assessment methods mainly focus on external security risks without considering the trustworthiness of cloud providers. In addition, the widely used thirdparty mediators are assumed to be trusted and we are not aware of any work that considers the security of these mediators. This study fills these gaps by assessing the security of public cloud storage providers and third-party mediators through equilibrium analysis. More specifically, we conduct evaluations on a series of game models between public cloud storage providers and users to thoroughly analyze the security of different service scenarios.Using our proposed security assessment, users can determine the risk of whether their privacy data is likely to be hacked by the cloud service providers; the cloud service providers can also decide on strategies to make their services more trustworthy. An experimental study of 32 users verified our method and indicated its potential for real service improvement.
2019年06期 v.24 738-749页 [查看摘要][在线阅读][下载 427K] [下载次数：23 ] |[网刊下载次数：0 ] |[引用频次：14 ] |[阅读次数：1 ] - Yuling Tian;Xiangyu Liu;
The extraction of rolling bearing fault features using traditional diagnostic methods is not sufficiently comprehensive and the features are often chosen subjectively and depend on human experience. In this paper, an improved deep convolutional process is used to extract a set of features adaptively. The hidden multi-layer feature of deep convolutional neural networks is also exploited to improve the extraction features. A deterministic detection of low-confidence samples is performed to ensure the reliability of the recognition results and to decrease the rate of false positives by evaluating the diagnosis of the deep convolutional neural network. To improve the efficiency of the continuous learning elements of the rolling bearing fault diagnosis, a clone learning strategy based on cloning and mutation operations is proposed. The experimental results show that the proposed deep convolutional neural network model can extract multiple rolling bearing fault features, improve classification and detection accuracy by reducing the false positive rate when diagnosing rolling bearing faults, and accelerate learning efficiency when using low-confidence rolling bearing fault samples.
2019年06期 v.24 750-762页 [查看摘要][在线阅读][下载 676K] [下载次数：145 ] |[网刊下载次数：0 ] |[引用频次：66 ] |[阅读次数：1 ] - Yuling Tian;Xiangyu Liu;
The extraction of rolling bearing fault features using traditional diagnostic methods is not sufficiently comprehensive and the features are often chosen subjectively and depend on human experience. In this paper, an improved deep convolutional process is used to extract a set of features adaptively. The hidden multi-layer feature of deep convolutional neural networks is also exploited to improve the extraction features. A deterministic detection of low-confidence samples is performed to ensure the reliability of the recognition results and to decrease the rate of false positives by evaluating the diagnosis of the deep convolutional neural network. To improve the efficiency of the continuous learning elements of the rolling bearing fault diagnosis, a clone learning strategy based on cloning and mutation operations is proposed. The experimental results show that the proposed deep convolutional neural network model can extract multiple rolling bearing fault features, improve classification and detection accuracy by reducing the false positive rate when diagnosing rolling bearing faults, and accelerate learning efficiency when using low-confidence rolling bearing fault samples.
2019年06期 v.24 750-762页 [查看摘要][在线阅读][下载 676K] [下载次数：145 ] |[网刊下载次数：0 ] |[引用频次：66 ] |[阅读次数：1 ]