Table of Content

    25 December 2017, Volume 15 Issue S2
    Download the whole issue (PDF)
    The whole issue of ZTE Communications December 2017, Vol. 15 No. S2
    2017, 15(S2):  0. 
    Asbtract ( )   PDF (2452KB) ( )  
    Related Articles | Metrics
    Special Topic
    Motion and Emotion Sensing Driven by Big Data
    REN Fuji, GU Yu
    2017, 15(S2):  1-2. 
    Asbtract ( )   HTML ( )   PDF (176KB) ( )  
    References | Related Articles | Metrics
    How Do Humans Perceive Emotion?
    LI Wen
    2017, 15(S2):  3-10.  doi:10.3969/j.issn.1673-5188.2017.S2.001
    Asbtract ( )   HTML ( )   PDF (443KB) ( )  
    Figures and Tables | References | Related Articles | Metrics

    Emotion carries crucial qualities of the human condition, representing one of the major challenges in artificial intelligence. Research in psychology and neuroscience in the past two to three decades has generated rich insights into the processes underlying human emotion. Cognition and emotion represent the two main pillars of the human psyche and human intelligence. While the human cognitive system and cognitive brain has inspired and informed computer science and artificial intelligence, the future is ripe for the human emotion system to be integrated into artificial intelligence and robotic systems. Here, we review behavioral and neural findings in human emotion perception, including facial emotion perception, olfactory emotion perception, multimodal emotion perception, and the time course of emotion perception. It is our hope that knowledge of how humans perceive emotion will help bring artificial intelligence strides closer to human intelligence.

    Emotion Judgment System by EEG Based on Concept Base of EEG Features
    Mayo Morimoto, Misako Imono, Seiji Tsuchiya, Hirokazu Watabe
    2017, 15(S2):  11-17.  doi:10.3969/j.issn.1673-5188.2017.S2.002
    Asbtract ( )   HTML ( )   PDF (439KB) ( )  
    Figures and Tables | References | Related Articles | Metrics

    This paper proposes an emotion judgment system by using an electroencephalogram (EEG) feature concept base with premise of noises included. This method references the word concept association system, which associates one word with other plural words and decides the relationship between several words. In this proposed emotion judgment system, the source EEG is input and 42 EEG features are constructed by EEG data; the data are then calculated by spectrum analysis and normalization. All 2945 EEG data of 4 emotions in the EEG data emotion knowledge base are calculated by the degree of association for getting the nearest EEG data from the EEG feature concept base constructed by 2844 concepts. From the experiment, the accuracy of the proposed system was 55.9%, which was higher than the support vector machine (SVM) method. As this result, the chain structured feature of the EEG feature concept base and the efficiency by the calculation of degree of association for EEG data help reduce the influence of the noise.

    Emotion and Cognitive Reappraisal Based on GSR Wearable Sensor
    LI Minjia, XIE Lun, WANG Zhiliang, REN Fuji
    2017, 15(S2):  18-22.  doi:10.3969/j.issn.1673-5188.2017.S2.003
    Asbtract ( )   HTML ( )   PDF (385KB) ( )  
    Figures and Tables | References | Related Articles | Metrics

    Various wearable equipment enables us to measure people behavior by physiological signals. In our research, we present one galvanic skin reaction (GSR) wearable sensor which can analyze human emotions based on cognition reappraisal. First, We research the factors of emotional state transition in Arousal-Valence-Stance(AVS) emotional space. Second, the influence of the cognition on emotional state transition is considered, and the reappraisal factor based on Gross regulation theory is established to correct the effectiveness from cognitive reappraisal ability to emotional state transition. Third, based on the previous work, we establish a GSR emotion sensing system for predicting emotional state transition and considering the correlation between GSR signals and emotions. Finally, an overall wearable sensor layout is built. In the experiment part, we invited 30 college students to wear our GSR sensors and watch 14 kinds of affective videos. We recorded their GSR signals while asking them to record their emotional states synchronously. The experiment results show different reappraisal factors can predict subjects’ emotional state transition well and indirectly confirm the feasibility of the Gross regulation theory.

    Multimodal Emotion Recognition with Transfer Learning of Deep Neural Network
    HUANG Jian, LI Ya, TAO Jianhua, YI Jiangyan
    2017, 15(S2):  23-29.  doi:10.3969/j.issn.1673-5188.2017.S2.004
    Asbtract ( )   HTML ( )   PDF (418KB) ( )  
    Figures and Tables | References | Related Articles | Metrics

    Due to the lack of large-scale emotion databases, it is hard to obtain comparable improvement in multimodal emotion recognition of the deep neural network by deep learning, which has made great progress in other areas. We use transfer learning to improve its performance with pre-trained models on large-scale data. Audio is encoded using deep speech recognition networks with 500 hours’ speech and video is encoded using convolutional neural networks with over 110,000 images. The extracted audio and visual features are fed into Long Short-Term Memory to train models respectively. Logistic regression and ensemble method are performed in decision level fusion. The experiment results indicate that 1) audio features extracted from deep speech recognition networks achieve better performance than handcrafted audio features; 2) the visual emotion recognition obtains better performance than audio emotion recognition; 3) the ensemble method gets better performance than logistic regression and prior knowledge from micro-F1 value further improves the performance and robustness, achieving accuracy of 67.00% for “happy”, 54.90% for “angry”, and 51.69% for “sad”.

    Emotion Analysis on Social Big Data
    REN Fuji, Kazuyuki Matsumoto
    2017, 15(S2):  30-37.  doi:10.3969/j.issn.1673-5188.2017.S2.005
    Asbtract ( )   HTML ( )   PDF (677KB) ( )  
    Figures and Tables | References | Related Articles | Metrics

    In this paper, we describe a method of emotion analysis on social big data. Social big data means text data that is emerging on Internet social networking services.We collect multilingual web corpora and annotated emotion tags to these corpora for the purpose of emotion analysis. Because these data are constructed by manual annotation, their quality is high but their quantity is low. If we create an emotion analysis model based on this corpus with high quality and use the model for the analysis of social big data, we might be able to statistically analyze emotional sensesand behavior of the people in Internet communications, which we could not know before. In this paper, we create an emotion analysis model that integrate the high-quality emotion corpus and the automatic-constructed corpus that we created in our past studies, and then analyze a large-scale corpus consisting of Twitter tweets based on the model. As the result of time-series analysis on the large-scale corpus and the result of model evaluation, we show the effectiveness of our proposed method.

    Measuring QoE of Web Service with Mining DNS Resolution Data
    LIU Yongsheng, GU Yu, WEN Xiangjiang, WANG Xiaoyan, FU Yufei
    2017, 15(S2):  38-42.  doi:10.3969/j. issn.1673-5188.2017.S2.006
    Asbtract ( )   HTML ( )   PDF (370KB) ( )  
    Figures and Tables | References | Related Articles | Metrics

    Internet service providers (ISPs) are paying more attention to the Quality of Experience (QoE) of the web service that is one of the most widely used Internet services. Measuring it with existing systems deployed in the network so far may save investment for ISPs since no additional QoE system is required. In this paper, with Domain Name System (DNS) resolution data that are available in the ISP’ network, we propose the First Webpage Time (FWT) algorithm in order to measure the QoE of the web service. The proposed FWT algorithm is analyzed in theory, which shows that its precision is guaranteed. Experiments based on the ISP’s DNS resolution data are carried out to evaluate the proposed FWT algorithm.

    An Improved K-Means Algorithm Based on Initial Clustering Center Optimization
    LI Taihao, NAREN Tuya, ZHOU Jianshe, REN Fuji, LIU Shupeng
    2017, 15(S2):  43-46.  doi:10.3969/j.issn.1673-5188.2017.S2.007
    Asbtract ( )   HTML ( )   PDF (336KB) ( )  
    Figures and Tables | References | Related Articles | Metrics

    The K-means algorithm is widely known for its simplicity and fastness in text clustering. However, the selection of the initial clustering center with the traditional K-means algorithm is some random, and therefore, the fluctuations and instability of the clustering results are strongly affected by the initial clustering center. This paper proposed an algorithm to select the initial clustering center to eliminate the uncertainty of central point selection. The experiment results show that the improved K-means clustering algorithm is superior to the traditional algorithm.

    Security Enhanced Internet of Vehicles with Cloud-Fog-Dew Computing
    MENG Ziqian, GUAN Zhi, WU Zhengang, LI Anran, CHEN Zhong
    2017, 15(S2):  47-51.  doi:10.3969/j.issn.1673-5188.2017.S2.008
    Asbtract ( )   HTML ( )   PDF (324KB) ( )  
    Figures and Tables | References | Related Articles | Metrics

    The Internet of Vehicles (IoV) is becoming an essential factor in the development of smart transportation and smart city projects. The IoV technology consists of the concepts of fog computing and dew computing, which involve on-board units and road side units in the edge network, as well as the concept of cloud computing, which involves the data center that provides service. The security issues are always an important concern in the design of IoV architecture. To achieve a secure IoV architecture, some security measures are necessary for the cloud computing and fog computing associated with the vehicular network. In this paper, we summarize some research works on the security schemes in the vehicular network and cloud-fog-dew computing platforms which the IoV depends on.

    Research Paper
    Random Forest Based Very Fast Decision Tree Algorithm for Data Stream
    DONG Zhenjiang, LUO Shengmei, WEN Tao, ZHANG Fayang, LI Lingjuan
    2017, 15(S2):  52-57.  doi:10.3969/j.issn.1673-5188.2017.S2.009
    Asbtract ( )   HTML ( )   PDF (418KB) ( )  
    Figures and Tables | References | Related Articles | Metrics

    The Very Fast Decision Tree (VFDT) algorithm is a classification algorithm for data streams. When processing large amounts of data, VFDT requires less time than traditional decision tree algorithms. However, when training samples become fewer, the label values of VFDT leaf nodes will have more errors, and the classification ability of single VFDT decision tree is limited. The Random Forest algorithm is a combinational classifier with high prediction accuracy and noise-tolerant ability. It is constituted by multiple decision trees and can make up for the shortage of single decision tree. In this paper, in order to improve the classification accuracy on data streams, the Random Forest algorithm is integrated into the process of tree building of the VFDT algorithm, and a new Random Forest Based Very Fast Decision Tree algorithm named RFVFDT is designed. The RFVFDT algorithm adopts the decision tree building criterion of a Random Forest classifier, and improves Random Forest algorithm with sliding window to meet the unboundedness of data streams and avoid process delay and data loss. Experimental results of the classification of KDD CUP data sets show that the classification accuracy of RFVFDT algorithm is higher than that of VFDT. The less the samples are, the more obvious the advantage is. RFVFDT is fast when running in the multi-thread mode.

    Deep Neural Network-Based Chinese Semantic Role Labeling
    ZHENG Xiaoqing, CHEN Jun, SHANG Guoqiang
    2017, 15(S2):  58-64.  doi:10.3969/j.issn.1673-5188.2017.S2.010
    Asbtract ( )   HTML ( )   PDF (538KB) ( )  
    Figures and Tables | References | Related Articles | Metrics

    A recent trend in machine learning is to use deep architectures to discover multiple levels of features from data, which has achieved impressive results on various natural language processing (NLP) tasks. We propose a deep neural network-based solution to Chinese semantic role labeling (SRL) with its application on message analysis. The solution adopts a six-step strategy: text normalization, named entity recognition (NER), Chinese word segmentation and part-of-speech (POS) tagging, theme classification, SRL, and slot filling. For each step, a novel deep neural network-based model is designed and optimized, particularly for smart phone applications. Experiment results on all the NLP sub-tasks of the solution show that the proposed neural networks achieve state-of-the-art performance with the minimal computational cost. The speed advantage of deep neural networks makes them more competitive for large-scale applications or applications requiring real-time response, highlighting the potential of the proposed solution for practical NLP systems.