Table of Content

31 January 2023, Volume 41 Issue 1

Previous Issue    Next Issue

Special Issue on Computer Applications

Named Entity Recognition Algorithm Enhanced with Entity Category Information

LIU Minghui, TANG Wangjing, XU Bin, TONG Meihan, WANG Liming, ZHONG Qi, XU Jianjun

2023, 41(1):  1-9.  doi:10.3969/j.issn.0255-8297.2023.01.001

Asbtract ( 1760 )   PDF (1423KB) ( 422 )  

References | Related Articles | Metrics

To solve the problem that the character level model of Chinese named entity recognition (NER) may ignore word information in sentences, a Chinese NER method based on entity category information enhancement in knowledge graph was proposed. Firstly, a training set was segmented with word segmentation tool, and all possible words were selected to construct a vocabulary. Secondly, the category information of entities in the vocabulary was retrieved by using generic knowledge graph, to construct a word set related to characters in a simple and effective way, and an entity category information set is generated according to the category information of entities in the word set. Finally, word embedding method was used to convert the set of category information into embeddings and concatenation of character embeddings, so as to enrich features in embedding layer. The proposed method can either be used as a module to expand feature diversity of embedding layer, or jointly applies with a variety of encoder-decoder models. Experiments on the Chinese NER dataset proposed by Microsoft Research Asia (MSRA) show the superiority of the proposed model. Compared with the models of Bi-directional long short-term memory (Bi-LSTM) and Bi-LSTM plus with conditional random field (CRF), the proposed method increases F₁ by 11.00% and 3.09% respectively, verifying that the category information of entities in knowledge graph performs high effectiveness in the enhancement of Chinese NER.

A Spatio-Temporal Similarity Query Algorithm for Trajectory Based on Graph Structure

XIONG Wei, XIONG Shuyi, CAO Jingzhi, CHEN Hao, GAO Jiayuan

2023, 41(1):  10-22.  doi:10.3969/j.issn.0255-8297.2023.01.002

Asbtract ( 1901 )   PDF (643KB) ( 358 )  

References | Related Articles | Metrics

To address the problem of slow similarity query of massive spatio-temporal trajectory data, a similarity query algorithm based on graph structure is proposed. Firstly, the trajectory is modeled as a path with spatial and temporal dimensions in a graph, and a trajectory similarity metric function is designed to match spatial and temporal distances simultaneously. Secondly, based on the similarity metric function, an edge-based inverted index structure combined with time filtering is designed, which supports spatio-temporal similarity query of trajectories while improving query performance using a pruning strategy with distance upper bound. The query algorithm performs distance calculation for each trajectory in the returned set of similar trajectories and sorts to obtain the top k results with the highest similarity. Finally, synthetic dataset and real dataset are used to compare the proposed algorithm with NTrajI algorithm, SHQ algorithm and SHQT algorithm. Experimental results show that the proposed method outperforms the comparison methods in index building and query processing, and obtains higher quality of query results. Therefore, the proposed algorithm is feasible and effective.

Link Prediction in Multiplex Networks Based on Naïve Bayes Model

ZHANG Yakun, LI Longjie, CHEN Xiaoyun

2023, 41(1):  23-40.  doi:10.3969/j.issn.0255-8297.2023.01.003

Asbtract ( 1531 )   PDF (3957KB) ( 246 )  

References | Related Articles | Metrics

To solve the problem of information fusion between layers in link predictions of multiplex networks, this paper proposes a new link prediction method based on the naïve Bayes model. The proposed method predicts links by combining the neighborhood information of target layers with the global information of distinct auxiliary layers relevant to the target layers. In a target layer, according to the neighborhood information of a node pair, the connection probability of the node pair is computed using the naïve Bayes model. In an auxiliary layer, based on whether there is a link between the node pair, the probability that the node pair has a link in the target layer is calculated. Experimental results on real and synthetic networks show that the proposed method achieves superior performance in both positively and negatively correlated multiplex networks.

Medical Electronic Data Feature Learning Method Based on Deep Learning

WANG Ting, WANG Na, CUI Yunpeng, LIU Juan

2023, 41(1):  41-54.  doi:10.3969/j.issn.0255-8297.2023.01.004

Asbtract ( 1809 )   PDF (746KB) ( 287 )  

References | Related Articles | Metrics

How can we effectively carry out the feature learning of high-dimensional and heterogeneous medical electronic data to optimize the risk prediction of concurrent medical use in patients? To address the problem, this paper proposed a method of multi-stage deep feature learning. Firstly, we performed the feature learning of medical use data with temporal properties by combining deep learning models of long short-term memory (LSTM) and auto-encoder (AE), and generated the synthetic factor of concurrent medical use with bisecting k-means clustering method. Secondly, we constructed two types of feature vectors for patients to predict adverse event risk, and analyzed the associated factors of high risk. Finally, we compared the performance of the proposed method with existing methods on real-word dataset, and the results show that the proposed method increases the accuracy by 5%~10%, and reduces the false rate by 3%~5% in the risk prediction of concurrent medical use.

Multi-modal Emotion Recognition Using Speech, Text and Motion

JIA Ning, ZHENG Chunjun

2023, 41(1):  55-70.  doi:10.3969/j.issn.0255-8297.2023.01.005

Asbtract ( 2384 )   PDF (1376KB) ( 527 )  

References | Related Articles | Metrics

For the problems of low accuracy and weak generalization ability in the process of human emotion recognition, a fusion method of multi-modal emotion recognition based on speech, text and motion is proposed. In the speech mode, a depth wavefield extrapolation-improved wave physics model (DWE-WPM) is designed to simulate the sequence information mining process of long short-term memory (LSTM) network; In the text mode, a transformer model with multi-attention mechanism is used to capture the potential semantic expression of emotion; In the motion mode, sequential features of facial expression and hand action are combined by using two-way three-layer LSTM model with attention mechanism. Accordingly, a multi-modal fusion scheme is designed to achieve high-precision and strong generalization ability of emotion recognition. In the general emotion corpus IEMOCAP, the method proposed in this paper is compared with existing emotion recognition algorithms. Experimental results show that the proposed method has higher recognition accuracy both in single modality and multi-modals, with average accuracy improved by 16.4% and 10.5% respectively, effectively improving the ability of human emotion recognition in human-computer interaction.

Force Haptic-Enhanced Virtual Reality Factory System

XU Wenbiao, XU Chi, SHI Hongyan, LI Lin

2023, 41(1):  71-79.  doi:10.3969/j.issn.0255-8297.2023.01.006

Asbtract ( 1594 )   PDF (7922KB) ( 102 )  

References | Related Articles | Metrics

Existing virtual reality factories that depend on helmets and handles can only provide one-way visual and auditory sensing for virtual operation, and cannot support immersive two-way tactile interaction. To deal with the problem, a haptics-enhanced virtual reality factory system is developed using force feedback controller. With Unity3D, 3DMax is used to model multiple kinds of tools and parts with 3D vision in the factory, and enhances their stereo vision by components such as reflection probes and light probes. Furthermore, by integrating components such as rigid bodies, collision bodies, joints to the virtual model, a tactile feedback model for the virtual space is established, which supports the real-time interaction between the haptic controller and virtual bodies. Experiments show that the system can provide a variety of force tactile perception capabilities such as mass, friction, and constant force. It supports tactile operations by touching and grasping more than 20 tools and parts, where the model, contour, and other physical attributes such as mass are provided. In this way, the system realizes the immersive operation experience of multi-dimensional audio-visual-haptic operation in the virtual factory.

Video Anomaly Detection Method Based on Secondary Prediction of Multi-layer Memory Enhancement Generative Adversarial Network

ZENG Jing, LI Ying, QI Xiaosha, JI Genlin

2023, 41(1):  80-94.  doi:10.3969/j.issn.0255-8297.2023.01.007

Asbtract ( 1637 )   PDF (7047KB) ( 159 )  

References | Related Articles | Metrics

In order to improve the accuracy of video anomaly detection, we propose a video anomaly detection method based on secondary prediction of multi-layer memory enhancement generative adversarial networks. Firstly, a spatiotemporal cube is extracted from target detection, and sent into encoder to obtain a prediction frame. Secondly, the apparent feature of the prediction frame and the optical flow feature of corresponding real frames are fused to form fusion features. Finally, a secondary prediction future frame is generated by using multi-layer memory enhancement generative adversarial networks, for learning normal feature patterns of different levels and capturing the semantic information of context. Experimental results on UCSD Ped2 and CUHK Avenue datasets show that the proposed method can effectively improve the performance of video anomaly detection compared with other video anomaly detection methods, and its frame level AUC reaches 99.57% and 91.59%, respectively.

Chinese Event Trigger Extraction Based on Span Regression

ZHAO Yuhao, CHEN Yanping, HUANG Ruizhang, QING Yongbin

2023, 41(1):  95-106.  doi:10.3969/j.issn.0255-8297.2023.01.008

Asbtract ( 1512 )   PDF (640KB) ( 235 )  

References | Related Articles | Metrics

In Chinese event trigger word extraction tasks, word-based models suffer from errors caused by word separation, while character-based models have difficulty in capturing the structural and contextual semantic information of trigger words. In view of the problem, a spanwise regression-based trigger word extraction method is proposed. Considering that a specific length of character subsequence (span) in a sentence may constitute an event trigger word, the method obtains the feature representation of the sentence with a pre-trained model of bidirectional encoder representation from Transformer (BERT), and generates the candidate span of the trigger word on the sentence feature representation. Then the model filters the candidate span with low confidence using a classifier, and adjusts the boundaries of the candidate span by regression to accurately locate the trigger word. Finally, the adjusted candidate spans are classified, and extraction results are obtained. Experimental results on the ACE2005 Chinese dataset show that the F₁ value of the span-based regression method is 73.20% for trigger word recognition task and 71.60% for trigger word classification task, better than existing models. Also, experimental comparison with span-based method without regression verifies that the regression adjustment of span boundaries can improve the accuracy of event trigger word detection.

Person Re-identification Algorithm Based on Channel Feature Aggregation

XU Zengmin, LU Guangjian, CHEN Junyan, CHEN Jinlong, DING Yong

2023, 41(1):  107-120.  doi:10.3969/j.issn.0255-8297.2023.01.009

Asbtract ( 1535 )   PDF (1940KB) ( 286 )  

References | Related Articles | Metrics

In deep-learning person re-identification algorithms, channel characteristics may be neglected, leading to a degraded model-expression ability. Address to the problem, we choose the ResNeSt50 as backbone network, and add an SE block to the end of residual blocks by using characteristics of SENet channel attention for enhancing features extraction of channels in networks. In addition, due to lack of control factors, ReLU function may reduce the correct responses of different feature graphs to activation values. Thus, we present two new activation functions. One is named as Weighted ReLU (WReLU) by combining ReLU with weight bias term, which can effectively improve feature selection ability in neural networks, and the other is Leaky Weighted ReLU (LWReLU), which is applied in Split-Attention and SE block, and enables Split-Attention to promote the weight learning ability from feature maps. Moreover, a new loss function with circle loss is also proposed for optimizing the convergence of objective function. Experimental results show that the proposed algorithm outperforms original backbone by 19.08%, 0.98%, and 2.02% in Rank-1, and 17.13%, 2.11%, and 2.56% in mAP respectively on CUHK03-NP, Market1501, and DukeMTMC-ReID datasets.

Proactive Self-Adaptive Approach Driven by LSTM Prediction for Software System

XIE Shenglong, WANG Lu, LIU Ruijia, PU Ying, LIU Xiao

2023, 41(1):  121-140.  doi:10.3969/j.issn.0255-8297.2023.01.010

Asbtract ( 1641 )   PDF (1389KB) ( 274 )  

References | Related Articles | Metrics

Aiming at the adjustment lag problem of reactive self-adaptive software systems, a proactive self-adaptive approach based on long short-term memory (LSTM) prediction driven is proposed. In this approach, LSTM neural network prediction technology is embedded in the analysis phase of monitor-analyze-plan -execute-knowledge (MAPE-K) control model; Operating data relating to self-adaptive environments, self-adaptive qualities, and self-adaptive goals, and historical data are used for classification prediction to form a self-adaptive early warning mechanism, which can effectively improve the proactive selfadaptive ability of software systems and reduce the lag influence of reactive self-adaptive decision-making at the same time. In order to illustrate the initiative, robustness and effectiveness of this approach, evaluation on the classic distribution tele-assistance system (dTAS) platform is carried out. Experimental results show that the proposed approach can provide early warning to self-adaptive demand, and enable software systems to complete proactive self-adaptive adjustment when necessary.

Design and Implementation of Decentralized Trusted Crowdsourcing Platform Based on Commitment Scheme

WANG Huajian, LI Renwei, ZHOU Huan, YANG Guogui

2023, 41(1):  141-152.  doi:10.3969/j.issn.0255-8297.2023.01.011

Asbtract ( 1648 )  

References | Related Articles | Metrics

In order to remove the dependence of traditional crowdsourcing on third-party central institutions, and at the same time ensure the fair distribution of crowdsourcing tasks and credible submission of results, a design scheme of trusted crowdsourcing platform based on blockchain smart contracts is proposed. First, a commitment-based two-stage submission mechanism is proposed and applied to the data submission process of recipients to solve the data transparency problem on the blockchain, so that the recipients cannot steal data from each other. Second, an unbiased random selection algorithm is designed to select relatively independent receivers from scattered receiver pools, preventing possible collusion among receivers. Finally, the randomness of the selection algorithm and the feasibility of the overall scheme are verified through selection simulation experiment and the implementation of the prototype system on Ethereum.

Optimization Algorithm for Dark Edge Detection of Deep-Sea Image Based on Particle Swarm Optimization

ZOU Qianying, CHEN Huiyang, LI Yongsheng, HU Liwen, WANG Xiaofang

2023, 41(1):  153-169.  doi:10.3969/j.issn.0255-8297.2023.01.012

Asbtract ( 1603 )   PDF (6687KB) ( 426 )  

References | Related Articles | Metrics

In order to solve the problem of image recognition for deep-sea resource detection, an optimization algorithm of image dark edge detection based on particle swarm optimization is proposed. The algorithm improves activation functions by using exponential linear unit and Gaussian error linear unit, constructs a dark edge detection algorithm in combination with the improved activation function according to the detection results of Marr-Hildreth operator, and uses particle swarm to train and optimize the improved dark edge detection algorithm. Finally, the proposed and several existing algorithms are applied and compared on 11 underwater data sets. Experimental results show that the proposed algorithm has the highest peak signal-to-noise ratio, structural similarity and edge retention index, reaching 18.769 6 dB, 0.660 7 and 0.834 5, respectively, and has the lowest mean square error of image of 3 750.225 3. Its average detection time is 0.667 4 s, about 14% shorter than that of the second best performance algorithm in the experiment.

Quantum Key Lifecycle Management Based on Blockchain

LIN Yusheng, CHANG Yan, CHEN Tiansu, YU Shipeng, ZHANG Shibin

2023, 41(1):  170-182.  doi:10.3969/j.issn.0255-8297.2023.01.013

Asbtract ( 1736 )   PDF (706KB) ( 300 )  

References | Related Articles | Metrics

In order to ensure a higher security of quantum key from generation, distribution, storage, use, update and destruction, this paper proposes a quantum key lifecycle management scheme based on blockchain. The two-party which has the requirement of confidential communication generates a truly random symmetric quantum negotiation key pool through quantum key distribution devices, and stores it in the quantum device administrator of each party. Then the quantum device administrators of two parties generate quantum key files according to negotiated numbering rules of quantum keys. Users of the two parties respectively apply for quantum keys from their quantum device administrators for communication. In the process of communication, the log information related to the generation, distribution, use, update and destruction of quantum keys is uploaded to a blockchain, and the quantum device administrators and communication users cooperate with the blockchain administrator to complete the management and traceability of the full lifecycle of quantum keys. Theoretical analysis shows that this scheme can solve the problem that quantum key cannot be effectively traced and managed in communication system, and realize the transparency and reliability of management and traceability of quantum key in whole lifecycle.