Table of Content

30 May 2021, Volume 39 Issue 3

Previous Issue    Next Issue

Column of CCF NCCA 2020

Emotion Classification Based on EEG Deep Learning

HAO Yan, SHI Huiyu, HUO Shoujun, HAN Dan, CAO Rui

2021, 39(3):  347-346.  doi:10.3969/j.issn.0255-8297.2021.03.001

Asbtract ( 2076 )   PDF (2087KB) ( 365 )  

References | Related Articles | Metrics

Electroencephalograph (EEG) research of emotion, as an important task in the advanced stage of artificial intelligence, has received more and more attention in recent years. Emotional EEG classification is widely used in human-computer interaction, medical research and other fields. This study presents the design of an EEG classification system on a lightweight convolutional neural network (CNN). DEAP (dataset for emotion analysis using physiological signals) provides EEG data of two kinds of emotion: arousal and valence. In order to obtain frequency domain information, the power spectral density features of theta, alpha, beta and gamma bands are extracted for evaluation, and each power spectral density matrix is expressed as a two-dimensional gray-scale image. The images were input into the convolutional neural network to train the classification model and complete the task of two classification. Experimental results show that compared with traditional machine learning, CNN has better classification effect. The accuracy of the two classification is 82.33% (Arousal) and 75.46% (Valence) respectively.

Fully Expression Frame Localization and Recognition Based on Dynamic Face Image Sequences

SIMA Yi, YI Jizheng, CHEN Aibin, ZHOU Mengna

2021, 39(3):  357-356.  doi:10.3969/j.issn.0255-8297.2021.03.002

Asbtract ( 1797 )   PDF (1528KB) ( 89 )  

References | Related Articles | Metrics

Considering that the evolution of facial expressions is a continuous process, compared to static images, dynamic image sequences are more suitable as the research objects for facial expression recognition. This paper proposes a sequence frame positioning model based on embedding network. The pre-trained Inception ResNet v1 network extracts the feature vectors of each frame, and then calculates the Euclidean distance between the feature vectors to position the complete frame with the maximum expression intensity, so a standardized facial expression sequences are obtained. In order to further verify the accuracy of the positioning model, we adopt VGG16 network and ResNet50 network to perform facial expression recognition on the positioned complete frame, respectively. Experiments were conducted on the CK+ and MMI facial expression databases. The average accuracy of the sequential frame positioning model proposed in this paper reached 98.31% and 98.08%, respectively. As using the VGG16 network and ResNet50 network to perform expression recognition on the positioned complete frame, the recognition accuracies on the two databases reached 96.32% and 96.5%, 87.23% and 87.88%, respectively. These experimental results show that the proposed model can pick up the complete frame from the facial expression sequence accurately and achieve better performance on facial expression recognition as well.

Seismic Fault Identification Method Based on ResUNet and Dense CRF Model

DU Chengze, DUAN Youxiang, SUN Qifeng

2021, 39(3):  367-366.  doi:10.3969/j.issn.0255-8297.2021.03.003

Asbtract ( 1758 )   PDF (11549KB) ( 116 )  

References | Related Articles | Metrics

Aiming at the problems of time-consuming, low efficiency, and high subjective influence in artificial interpretation of seismic data, a crack identification method based on ResUNet and dense conditional random field (Dense CRF) model is proposed. First, the method uses the ResUNet model to extract the features of different resolution levels from the cracks in the seismic amplitude data volume to achieve seismic crack identification, then it uses the Dense CRF model to further optimize the recognition results, so as to achieve accurate recognition of seismic cracks. The proposed method is compared with the traditional UNet and ResUNet methods based on the synthetic seismic amplitude data volume and the seismic amplitude volume data of the F3 work area. Experimental results show that the proposed method performs higher accuracy, finer size and better continuity in crack identification.

Cascaded Separable and Dilated Residual U-Net for Liver Tumor Segmentation

YU Qun, ZHANG Jianxin, WEI Xiaopeng, ZHANG Qiang

2021, 39(3):  378-377.  doi:10.3969/j.issn.0255-8297.2021.03.004

Asbtract ( 1761 )   PDF (8083KB) ( 299 )  

References | Related Articles | Metrics

Computer-aided liver tumor segmentation can effectively reduce the workload of doctors and improve the success rate of surgery, and it has important clinical diagnosis and treatment value. Meanwhile, recently proposed U-Net model has achieved great success in the field of medical image segmentation. To obtain more accurate liver tumor segmentation results, this paper proposed an improved U-net model, i.e., cascaded separable and dilated residual U-Net (CSDResU-Net), for this medical application. CSDResU-Net utilizes cascade operation to solve the problem of unbalanced data in tumor segmentation due to the small proportion of tumors in the whole image. Besides, residual unit, depthwise separable convolution and dilated convolution are integrated into a single network to increase the convolution kernel receptive field, which can quickly extract more discriminative liver image features and lead to the performance improvement of liver tumor segmentation. Experimental results on the International Conference on Medical Image Computing and Computer Assisted Intervention (MICCAI) liver tumor segmentation (LiTS) benchmark dataset show that CSDResU-Net is relative to the baseline. The method improves the performance of the Dice coefficient by 1.3%, and at the same time proves that different void ratios have a greater impact on the performance of the segmentation network.

Tracking Interdisciplinary Memes in Physics

ZHOU Yi, YAN Guanghui, LU Binwei, WANG Shan, LI Shikui, WEI Xiang, YANG Shibo, JIN Dan

2021, 39(3):  387-386.  doi:10.3969/j.issn.0255-8297.2021.03.005

Asbtract ( 1630 )   PDF (9348KB) ( 60 )  

References | Related Articles | Metrics

Based on the papers published in American Physical Society (APS) journals from 2000 to 2019 and abstracts from Web of Science, we firstly use meme phrases to describe knowledge and construct meme correlation network (MCN) to track interdisciplinary memes. And then we use Rao-Stirling index to calculate the interdisciplinary scores of memes, and track interdisciplinary memes in physics. The verification experiment was carried out from three perspectives of network topology index, interdisciplinary measurement index and professional terminology comparison. It proved that the meme correlation network (MCN) and the interdisciplinary score of memes in this paper can effectively reflect the diffusion of knowledge in different domains.

Communication Engineering

Path Loss Prediction for UAV-to-Ground Millimeter Wave Communications

YANG Jingwen, ZHU Qiuming, WANG Jian, YAO Mengtian, CHEN Xiaomin, ZHONG Weizhi

2021, 39(3):  398-397.  doi:10.3969/j.issn.0255-8297.2021.03.006

Asbtract ( 1846 )   PDF (10079KB) ( 259 )  

References | Related Articles | Metrics

For the complex propagation environment experienced by millimeter wave (mmWave) communications between unmanned aerial vehicle (UAV) to ground, considering the factors of outdoor path loss, indoor path loss, penetration loss, and rain attenuation, this paper presents a novel UAV-to-ground composite propagation loss prediction model. In the model, the density and height of buildings are adopted to depict different scenarios, such as urban, suburban, and rural areas. Further, light-of-sight (LoS) probability with respect to the UAV height is adopted to get the statistical average value of path loss. The proposed model is simulated and validated onto a campus scene at different distances and heights by using a ray tracing method. Comparison and analysis show that the results of the proposed model agree well with those of ray-tracing method, and the maximum errors of three specific scenarios are 5.2 dB, 4.8 dB, and 7.2 dB, respectively. This proves that the proposed model is potential in the prediction of propagation loss and the performance evaluation of UAV-to-ground mmWave communication systems.

Signal and Information Processing

Research on the Influence of Network Externalities of Rail Transit Network Platform on Passenger Flow

CHEN Yuegang, DU Xinglong

2021, 39(3):  409-408.  doi:10.3969/j.issn.0255-8297.2021.03.007

Asbtract ( 1698 )   PDF (234KB) ( 110 )  

References | Related Articles | Metrics

Rail transit network is able to attract passenger flow from road network so as to help alleviating urban problems such as traffic congestion and environmental pollution. From the perspective of network platform, this paper describes the state of passenger flow on rail transit network platform by constructing platform passenger flow model with utility function. Based on traffic data tracked in Shanghai, we have conducted the comparative analysis between direct-network and cross-network externalities and studied their impacts on the passenger flow of the platform. In the study, analytical methods including ordinary least squares, instrumental variable-two-stage least squares, heterogeneity analysis and robustness test are utilized. Research results show that the increase of station number exerts limited effects on passengers’ willingness to choose rail transit, and the current scale of passenger flow carried on Shanghai rail transit network platform has exceeded the optimal load of the platform, thus leading to a continuous loss of passengers.

Queue Model of Contact Center for Customer Abandon and Variable Number of Reception Desks

ZANG Wanbin, LI Junxiang

2021, 39(3):  419-418.  doi:10.3969/j.issn.0255-8297.2021.03.008

Asbtract ( 1566 )   PDF (734KB) ( 170 )  

References | Related Articles | Metrics

With the strategy that new service desks can be applied for opening when the queuing number of customers exceeds the limited bottleneck threshold in peak hours or emergencies, and considering constraints such as increased staff fatigue and customer patience in actual situations, a contact center queue model with the objective of cost optimization is established based on parameters of customers abandon, multi-channel service and variable number of reception desks. The performance of the contact center queue model is simulated and analyzed by using ProModel software, and its effectiveness is verified by comparing with the traditional call center queue model. Simulation results show that, when the number of customers waiting for service reaches a peak, the contact center queue model based on queue threshold condition to control the number of reception desks not only can improve the operating efficiency of the system as a whole, but also can effectively keep the staffing number in a reasonable range in order to control the loss of staffing costs, reduce the loss cost and loss rate of customers, and improve the service level of staff.

FSCD-CNN Based Fast Mode Decision Algorithm for Intra-prediction in Depth Map Coding

CUI Pengtao, ZHANG Qian, LIU Jinghuai, ZHOU Chao, WANG Bin, SI Wen

2021, 39(3):  433-432.  doi:10.3969/j.issn.0255-8297.2021.03.009

Asbtract ( 1589 )   PDF (636KB) ( 260 )  

References | Related Articles | Metrics

In view of the coding complexity caused by the encoding format of multiview video plus depth map and the quadtree coding structure in 3D-HEVC, a fast intra prediction mode selection algorithm for depth images based on FSCD-CNN (fast selecting cu’s depth-convolutional neural network) is proposed. First, a training set is obtained by dividing the depth of the optimal depth map LCU (largest coding unit) of a depth video sequence as labels. Second, a FSCD-CNN network is constructed, which is suitable for deep decision-making of LCU. At last, the optimal division of LCU is achieved by carrying out the depth-division prediction of depth map LCU and skipping some coding mode decisions. Experimental results show that the proposed algorithm could reduce the coding time by 15% on average while maintaining the same coding performance as other relevant literatures, and verify the effectiveness and reliability of this method.

News Summarization Extracting Method Based on Improved MMR Algorithm

CHENG Kun, LI Chuanyi, JIA Xinxin, GE Jidong, LUO Bin

2021, 39(3):  443-442.  doi:10.3969/j.issn.0255-8297.2021.03.010

Asbtract ( 1902 )   PDF (677KB) ( 360 )  

References | Related Articles | Metrics

This paper proposes a news extraction method based on maximal marginal relevance (MMR) and a news extraction method based on support vector machine and maximal marginal relevance (SVM-MMR). The first method improves the traditional MMR news extraction method, and the second one uses the improved MMR news extraction method to make a second choice of the SVM classification results. Compared with the traditional MMR news extraction method, the average precision of MMR-based and SVMMMR-based news extraction methods are improved by 0.148 and 0.204, respectively. And the extraction efficiency of the MMR-based method is about 3 times of that of the SVMMMR method. The augmented MMR algorithm is more suitable for application scenarios that require high summarization efficiency, especially for long text summarization, while the SVM-MMR method is more suitable for generating a more comprehensive summary of the text content.

Spatial-Temporal Evolution Analysis of Urban Development in the Western Taiwan Straits Economic Zone Using Night-Time Light Imagery

XIE Jinlong, LI Xi, XU Humin

2021, 39(3):  456-455.  doi:10.3969/j.issn.0255-8297.2021.03.011

Asbtract ( 1716 )   PDF (6229KB) ( 179 )  

References | Related Articles | Metrics

Based on the time series of VIIRS night-time light remote sensing images, this study analyzed the spatial and temporal patterns of urban development in Western Taiwan Straits Economic Zone (WTSEZ) at prefecture-level and county-level from 2012 to 2019, by utilizing a spatial statistics method, rank-size distribution method, and the Night-time Light Development Index (NLDI). Study results show that from 2012 to 2019, the total amount of night time light in WTSEZ presents an unbalanced distribution of higher level in coastal areas and lower level in inland areas, while the growth rate of night time light in inland areas is more significant than that in coastal areas. According to the result of the rank-size distribution analysis, the city sizes of the top cities both at prefecture-level and at county-level are gradually increasing. The Zipf index of prefecture-level remains around 0.95, whereas the Zipf index at county-level shows a trend of gradual decrease. The development among cities is becoming more balanced. According to the analysis of NLDI, the night-time light levels are gradually becoming proportional to the population distribution levels in WTSEZ, and the NLDI of coastal cities, especially in Shantou and Jieyang, decreases more pronouncedly than that of inland cities. There is a balanced development trend within each city. All above show that in the future development planning, it is necessary to take a full consideration of the regional and ecological advantages and disadvantages of WTSEZ, so as to promote the coordinated development of coastal and inland areas; to strengthen intercity transportation, so as to promote the integration of regional development; and to further explore the developing potential of eastern GuangdongHongkong-Macao region, so as to accelerate communication between urban agglomerations and inject new external forces into WTSEZ.

Risk Assessment for Land Use Transition in the Middle of the Yangtze River Economic Belt

WANG Kai, ZHANG Xubing, ZHUO Chenggang, HU Shougeng

2021, 39(3):  469-468.  doi:10.3969/j.issn.0255-8297.2021.03.012

Asbtract ( 1610 )   PDF (2051KB) ( 69 )  

References | Related Articles | Metrics

The potential economic, social and ecological risks existing in land use transitions will affect the sustainable development of the Middle of the Yangtze River Economic Belt, thus it is critical to develop a method of assessing the risks of land use transition. In this paper, firstly, a risk assessment index system was built based on pressure-state-response (PSR) model. Then entropy-TOPSIS (technique for order preference by similarity to ideal solution) was adopted to assess the current risks of land use transitions. Subsequently, the critical factors of the risks were identified by means of obstacle degree analysis. Finally, the trend of economic, social and ecological risks in the Middle of the Yangtze River Economic Belt from 1990 to 2015 was assessed and analyzed, and the risk factors in each risk stage were found out. The results of the present study could promote the research progress of the risk assessment of land use transitions in the Middle of the Yangtze River Economic Belt.

Computer Science and Applications

Classification Method of Improved Support Vector Machine and Its Application in Early Detection of Primary Liver Cancer

CAO Guogang, LI Mengxue, CHEN Ying, CAO Cong, WANG Ziyi, FANG Meng, GAO Chunfang, LIU Yunxiang

2021, 39(3):  481-480.  doi:10.3969/j.issn.0255-8297.2021.03.013

Asbtract ( 1676 )   PDF (316KB) ( 192 )  

References | Related Articles | Metrics

To screen out primary liver cancer patients as early as possible, assist doctors to make better decisions and improve treatment effects, an early screening method based on routine laboratory data is proposed. To test the classification of healthy, benign lesions or primary liver cancer, a support vector machine method is optimized by using a differential evolution algorithm, in which the evaluation cost is the area under the ROC (receiver operating characteristic) curve. Moreover, to satisfy different clinical requirements, performance index curves and cut-off lookup tables of the training model are built, then cut-off values are selected by users to further improve the prediction performance. Compared with other 5 state-of-the-art methods, the proposed methods have better classification performance, of which the accuracy reaches 0.944 1, and the Kappa coefficient reaches 0.903 1. The research results can assist doctors to screen out the primary liver cancer early and improve the long-term survival rate of patients.

TF-Ranking Recommendation Method Based on User Session

JIA Dan, SUN Jingyu

2021, 39(3):  495-494.  doi:10.3969/j.issn.0255-8297.2021.03.014

Asbtract ( 1640 )   PDF (357KB) ( 157 )  

References | Related Articles | Metrics

Based on users’ session logs, a TF-Ranking recommendation method that integrated XGBoost and gated recurrent unit was proposed. This method used gated recurrent unit to learn user session data. Firstly, XGBoost was used to extract features, and experimental results showed that XGBoost could overcome the defects of traditional data model and greatly reduced the complexity of data models while maintaining their original attributes. Secondly, an improved Dropout network was used to process data, leading to a recall rate improved by 1.32%. Finally, by training data based on Learning to Rank method in combination with pairwise method, a positive sort recommendation list with strong relevance to query contents was provided for users. Experiments were conducted on the data set of Trivago Recsys Challenge 2019. The results show that the proposed algorithm could improve the recall rate and the average reciprocal ranking, and could be applied to large-scale data recommendation.