Skip to main content
main-content

Über dieses Buch

This two-volume book constitutes the refereed proceedings of the 3rd International Conference on Multimedia Technology and Enhanced Learning, ICMTEL 2021, held in April 2021. Due to the COVID-19 pandemic the conference was held virtually.

The 97 revised full papers have been selected from 208 submissions. They describe new learning technologies which range from smart school, smart class and smart learning at home and which have been developed from new technologies such as machine learning, multimedia and Internet of Things.

Inhaltsverzeichnis

Frontmatter

Intelligent Application in Education

Frontmatter

A Simple and Efficient Key Frame Recognition Algorithm for Sign Language Video

Sign language is an important means of social communication for hearing-impaired people, and most developed countries have established their own hand language banks. Under the guidance of the National Language Commission, China has created a national sign language corpus, which is mainly composed of video. For the database, one of the most important work is to establish the index of retrieval. For sign language videos, the most important index is the hand shape displayed in the video key frame. In this paper, a simple and efficient key frame extraction algorithm is proposed based on the video library with good consistency, namely the sign language video library, to create a fast and efficient index. At the same time, it can be used as a reference for similar video libraries.

Zhaosong Zhu, ShengWei Zhang, YunLei Zhou

Research on Dynamic Sign Language Recognition Based on Key Frame Weighted of DTW

Dynamic sign language can be described by its trajectory and key hand types. Most of the commonly used sign language can be recognized by trajectory curve matching. Therefore, In this paper, a new dynamic sign language recognition method is proposed, which uses trajectory and key hand type to extract features, adopts a key frame weighted DTW (dynamic time warping) algorithm to implement hierarchical matching strategy, and gradually matches sign language gestures from two levels of trajectory and key hand type, so as to effectively improve the accuracy and efficiency of sign language recognition.

ShengWei Zhang, ZhaoSong Zhu, RongXin Zhu

An Optimized Seven-Layer Convolutional Neural Network with Data Augmentation for Classification of Chinese Fingerspelling Sign Language

Sign language recognition especially finger language recognition facilitates the life of deaf people in China. It overcomes many difficulties and provides convenience for deaf people’s life. In this paper, we used the advanced convolutional neural network to extract the different characteristics of the input. We created an optimized seven-layer CNN, including five convolution layers for feature extraction and two fully connected layers for classification to enhance the original signal function and reduce noise after operation. Some advanced techniques such as batch normalization, ReLu and dropout were employed to optimize the neural network. Meanwhile, we adopted data augmentation technology, which not only expanded the data set and improve the performance of machine learning algorithm, but also avoided the over-fitting problem. The experimental results show that the average recognition accuracy reaches 91.99 ± 1.21%, which indicate an excellent property.

Yalan Gao, Rongxin Zhu, Ruina Gao, Yuxiang Weng, Xianwei Jiang

Similar Gesture Recognition via an Optimized Convolutional Neural Network and Adam Optimizer

The recognition significance of similar sign language (or confusing gesture) in sign language recognition is highlighted, and the goal is to realize the recognition of such gesture and sign language based on deep learning with an optimized convolutional neural network and the Adam optimizer. The convolutional layer and the pooling layer are connected alternately. The locally connected image data and parameter features are used to extract the shared pooling layer, and the image resolution reduction of image data sampling and the reducibility of iterative training are used to achieve the extraction precision requirements of feature points. In addition, the information transfer between layers is realized through convolution, the introduction of pooling layer and RELU activation function to realize nonlinear mapping and reduce the data dimension. We also use the batch normalization method for faster convergence and dropout method to reduce overfitting. Ten experiments were carried out on a nine-layer “CNN-BN-ReLU-AP-DO” method, with an average accuracy of 97.50 ± 1.65%. The overall accuracy is relatively high, and gesture recognition can be conducted effectively.

Ya Gao, Chenchong Jia, Yifei Qiao, Xi Huang, Juan Lei, Xianwei Jiang

Development and Creation of Open Online Course Resources in Tourism Colleges from the Perspective of School-Enterprise Collaborative Education

The mode of teaching and learning in the “Internet+” era is undergoing changes. In order to adapt to the development of the times, the state has promulgated corresponding informatization policies to promote the reform of college education and teaching methods. Both online learning and hybrid learning play an important role in college education and teaching, and online open course resources are indispensable to hybrid teaching and online teaching. From the perspective of school-enterprise collaborative education, the article takes “Front Office Service and Management”, an Excellent Open Online Course of China, as an example, and elaborates on online course team establishment, principles of development of online course resources, content selection, resource type creation, and creation experience summary, shedding light on how to develop and improve open online course resources in tourism colleges.

Rui Jiang, Hua Jiang

Research on the Application of MOOC in O2O Teaching Model Innovation of Aesthetic Education in Higher Vocational Colleges

The rapid development of Internet technology has promoted the era of education informatization, and the construction of an O2O teaching model with online teaching and offline communication is a central link in the reform of aesthetic education in higher vocational colleges. As a representative of the education model in the new era, the emergence of MOOCs has greatly promoted the development of online and offline hybrid teaching models. This article will discuss MOOC’s O2O teaching in aesthetic education in higher vocational colleges from four aspects: research background, current application status of MOOC in aesthetic education, implementation strategies of effective application of MOOC teaching under O2O mode, and improvement of the guarantee mechanism of MOOC teaching quality Application in model innovation.

Gege Ma

Design of Hospital Remote Consultation and Teaching System Based on Deep Learning

The application of deep learning technology makes the hospital’s remote consultation and teaching system more humane. Therefore, this research designed a hospital remote consultation and teaching system based on deep learning technology. First of all, through the analysis of the problems of system construction, clear system design objectives. With the support of the system hardware, the deep learning process is used to realize the functions of case collection, remote consultation, doctor recommendation, remote education and training, and case sharing. Using experiments to analyze the actual application performance of the hospital remote consultation and teaching system based on deep learning, and comparing it with the traditional system, it verifies that the system in this paper is more effective.

Ying Bao

A Feasibility Analysis Model for Developing Wushu Sanda Courses in Universities Based on Deep Learning

The current feasibility analysis of specific courses, expert analysis or single item analysis is usually used, and the results are relatively one-sided, so it is difficult to achieve a comprehensive feasibility analysis. Therefore, this paper designs the feasibility analysis model of Wushu Sanda course in universities based on deep learning. First of all, the relationship between curriculum system and training objectives is analyzed, and the relationship matrix between curriculum and objectives is established. Then according to the relationship between training objectives, the importance of the course is analyzed. After that, the training objective factors in the course are analyzed to realize the course process data. At last, use deep learning technology to analyze the marginal characteristics of data, and get the feasibility results of the course. The experiment is designed to analyze the feasibility of Wushu Sanda course in a university. The experimental results show that the model can analyze the feasibility of the course, and get the data for reference to meet the design requirements.

Dong-Dong Liu

Performance Evaluation Model of Wushu Sanda Athletes Based on Visual Signal Processing

In order to improve the accuracy of the performance evaluation of Wushu Sanda athletes, the performance evaluation model of Wushu Sanda athletes was designed based on visual signal processing. First, extract the contours of Wushu Sanda athletes, then collect the performance information of Wushu Sanda athletes, and finally complete the performance evaluation through the establishment of the performance evaluation index of Wushu Sanda athletes and the index processing. Experimental results show that the evaluation model designed in this study not only improves the accuracy of evaluation, but also has a higher evaluation efficiency.

Dong-dong Liu

Online Matching Method of News Communication Innovative Teaching Mode Driven by Artificial Intelligence

In order to improve the quality of training news communication talents, optimize the teaching mode of news communication innovation, combine with artificial intelligence technology, innovate the content of news communication, deeply analyze the types of students’ learning characteristics, and match the different students’ online learning content and methods according to the teaching content and the training needs of professional talents, so as to improve the teaching effect of news communication innovation, and provide a mixed teaching mode of online and offline news communication that can be discussed or criticized.

Jia Qian, Li-li Wang

Motion Recognition System of Table Tennis Players Based on MEMS Sensor

In order to improve the accuracy of table tennis players’ movement recognition, a movement recognition system based on MEMS sensor is designed. The hardware part chooses MEMS sensor chip as the core processing chip, and designs the connection circuit of sensor. In the software part, the overlapped motion recognition signals are filtered, the network camera with certain parameters is used to obtain the motion parameters, and the recognition process is constructed. Simulation results show that the system has a high recognition accuracy of up to 99.9%, which has a certain application prospect.

Wei Tang, Chonggao Chen

Open Sharing of Digital Education Training Resources Based on Machine Learning

In order to solve the problem of poor security caused by data explosion in the traditional open sharing method of digital education training resources, this paper proposes a machine learning based open sharing method. Based on the national general control protocol standard, a control protocol of education and training resource sharing service based on GCCP gateway is constructed. With the support of the control protocol, support vector machine is designed by using machine learning ability, support vector machine is used to process resource data, and resource classification is realized. With the support of cloud computing mode, resource is open and shared. Experimental results show that the proposed open sharing method based on machine learning has short response time and low memory leak probability, and the security of the method is improved.

Jichao Yan, Jingya Zheng

Design of Basketball Shot Track Recognition System Based on Machine Vision

Aiming at the positioning of the existing shooting track, which leads to the deviation of basketball shooting track, in order to maintain the basketball shooting track accurately, a basketball shooting track recognition system based on machine vision is designed. The ARM processor is used as the core application terminal of the identification component, and the high-precision orientation module and power supply conversion module are combined to build the hardware environment. Based on this, a machine vision technology model is established. Through defining the format of machine datagram, the subprogram of tilt angle sensor is optimized, and the software execution environment is realized. Experimental results show that, compared with the traditional recognition system, the physical distance between the starting position and the terminating position of the shot is greatly reduced, which can effectively solve the problem of basketball trajectory deviation caused by inaccurate shooting and meet the practical application requirements of precise maintenance of basketball trajectory.

Chonggao Chen, Wei Tang

Design and Implementation of Mobile Learning System Based on Wireless Communication Technology

Due to the lack of comprehensive definition of learning resources in traditional mobile learning systems, the transmission efficiency of learning resource packets is poor. Therefore, a mobile learning system based on wireless communication technology is proposed. In terms of hardware, it adopts SOA technology framework, optimizes the overall system architecture, adds wireless communication environment middleware, and optimizes the system communication interface and resource reading channel. In software aspect, MySQL database is established to store learning resources, and load balance control theory is used to improve the storage density of resource information. The experimental results show that the design system improves the data throughput, reduces the data transmission delay, and the actual application effect is better.

Hui-jun Wang, Ang Li

Mining Recessive Teaching Resources of University Information Based on Machine Learning

The accuracy of mining implicit teaching resources in traditional universities is low, so a method of mining implicit teaching resources in universities based on machine learning is designed. Firstly, it designs the process of data mining, define the problem, collect and preprocess the data, execute the mining algorithm and then explain and evaluate it. The classification method of data mining is optimized. In this paper, the classification technology is neural network, and the artificial neural network unit is built by biological neuron structure, and the classification is completed by biological transfer and activation function. Finally, the machine learning algorithm is improved, and the ight is updated by introducing momentum scalar factor. In the contrast experiment, it chooses the data set and train the parameters, design the process of data mining, and count the relevant parameters of the data set. The experiment results show that the accuracy of the designed method is 4.03% higher than that of the traditional method.

Zheng Jingya, Jichao Yan

Networked Teaching System of College Basketball Course Based on Virtual Reality

Aiming at the poor performance of the traditional college basketball course networked teaching system, with the purpose of improving the performance of college basketball course networked teaching system, a virtual reality-based college basketball course networked teaching system is designed. Through the user registration module design, the question answering module design and the student user management module design, the hardware design of the system is completed, and the software design of the system is completed through the management of college basketball course resources and the design of the student basketball course homework management program. The design of the networked teaching system of the course. The test results show that the network teaching system of College Basketball Course Based on virtual reality has reached the design requirements and has higher performance.

Er-wei Liu

Research on Remote Online Teaching Assistant System Based on Human-Computer Interaction

In this study, human-computer interaction is applied to the design of remote online teaching assistance system. The camera module consists of a processor, a network camera and a microphone. The system software is composed of online examination module, course resource management module, online question-answering module, client module, video teaching module and personal information management module. Among them, the online examination module is divided into question bank management, examination management and other sub-functional modules; The course resource management module mainly completes the input and output of educational resources by uploading and downloading learning materials. Online Q&A module is mainly divided into question database management, question search and other sub-functional modules; The client module is mainly composed of Flash CS5; Video teaching module is mainly used to realize remote video teaching.

Zijin Xiao, Ying Li, Hai Zhou

Towards the Automatic Generation of Pedagogical Conversational Agents from Lecture Slides

Although corresponding technological and didactical models have been known for decades, the digitization of teaching has hardly advanced beyond simple non-interactive formats (e.g. downloadable slides are provided within a learning management system). The COVID-19 crisis is changing this situation dramatically, creating a high demand for highly interactive formats and fostering exchange between conversation partners about the course content. Systems are required that are able to communicate with students verbally, to answer their questions, and to check the students’ knowledge. While technological advances have made such systems possible in principle, the game stopper is the large amount of manual work and knowledge that must be put into designing such a system and feeding it the right content.In this publication, we present a first system to overcome the aforementioned drawback by automatically generating a corresponding dialog system from slide-based presentations, such as PowerPoint, OpenOffice, or Keynote, which can be dynamically adapted to the respective students and their needs. Our first experiments confirm the proof of concept and reveal that such a system can be very handy for both respective groups, learners and lecturers, alike. The limitations of the developed system, however, also reminds us that many challenges need to be addressed to improve the feasibility and quality of such systems, in particular in the understanding of semantic knowledge.

Matthias Wölfel

Research on the Fusion Pattern Recognition System Based on the Concept of Production Education Integration and Application of Generative Countermeasure Network

In order to highlight the practical application value of network data and information fusion behavior under the background of industry-education integration, a fusion pattern recognition system that applies generative confrontation network under the concept of industry-education integration is designed. First, the cyclic neural network is used to generate independent text information packets. While establishing the generation of the confrontation network framework, various reinforcement learning parameters are adjusted to realize the construction of the hardware execution environment of the recognition system. On this basis, build an embedded network framework, with the help of EEPROM chip and LD3320 chip circuit, to supervise the fusion process of network data information identification and implementation behavior, and realize the construction of the system’s software execution environment. Combined with the related hardware equipment structure, complete the research on the fusion pattern recognition system of the application generation confrontation network under the concept of integration of production and education. Comparative experiment results show that with the application of the above system, the mean value of network data information fusion time is reduced from 17.9 s to 11.2 s, while the maximum amount of information processed by a single fusion process reaches 9.3 × 1012T which can be used in the context of the integration of production and education Effectively highlight the practical application capabilities of network data information fusion behavior.

Conggang Lv

Design of Multimedia Learning Resource Recommendation System Based on Recurrent Neural Network

The existing learning resource recommendation system has the defect that the average absolute error of the recommendation result is large due to the limitation of its own adaptation range. For this reason, this research designed a multimedia learning resource recommendation system based on recurrent neural network. Introduce the recurrent neural network to design the architecture of the multimedia learning resource recommendation system, and design the system functional modules based on this, including the learner’s demand retrieval representation module, learner preference representation module, recurrent neural network training module, database module and system management module. The simulation experiment results show that compared with the existing system, under the Gowalla data set, the average absolute error coefficient of the recommended results of this paper is reduced by 0.356; under the Yelp data set, the average absolute error coefficient of the recommended results of this paper is reduced 0.404. The above results fully show that the recommendation effect of this system is better.

Zijin Xiao, Ying Li, Hai Zhou

Human/Medical Based Data Processing and Systems

Frontmatter

Research on Constructing Regional Telemedicine Imaging Diagnosis Center Based on Ctirix Technology

Medical informatization is an important direction of medical development, and regional medical information sharing is the overall trend of medical development. Content Establish the integration of information in the radiology department of the community health center and the imaging department of the secondary hospital in the medical union. Solve the problem of ineffective sharing of clinical imaging information data caused by different systems and servers in the region. It can optimize the rapid viewing of clinical data and historical case data of referral patients in the region to achieve an efficient and fast tracking of the entire treatment process of patients And the image information of the disease development. It mainly adopts CITRIX virtualization technology, which has the advantages of small occupied bandwidth, high security, convenient deployment, remote management, convenient maintenance, stable connection, and high reliability. Through Citrix technology, the clinical data sets of the branch hospitals are effectively centralized, and a relatively unified data system is established, which is stored, managed and transferred to a standardized data model to form a data set, and integrates internal data in the region to effectively serve clinical images.

Jinshun Ding, Yu Ren, Kefeng Xu, Yixin Wang

Research on Multi-agency Data Fusion Mode Under Regional Medical Integration

2020 is not only the stage of intensive implementation of medical informatization related policies, but also a key year for the further development of regionalization of medical informatization projects. The medical community data sharing technology using multi-source heterogeneous data fusion solves the problem of different hospitals, different procedures, different database structures, and information islands in each hospital. Through ETL technology, using the SSIS tool in Microsoft SQL Server, a relatively standard data system is built for the original information system of each hospital in the medical community group to centrally convert, clean and transfer to a standardized data model to form a data set: Patient Master Index (EMPI), Master Data Management (MDM), etc., to solve the problem of reducing repeated statistics and discrepancies in various hospitals, improve data quality, complete interconnection and data sharing.

Yixin Wang, Weiqing Fang, Wei Zhu, Jinshun Ding

Research on Brain Image Segmentation Based on FCM Algorithm Optimization

Brain disease is becoming a threat to human health. Many countries begin to pay attention to the research of brain science. If brain diseases are predicted in advance, diagnosed accurately and treated with comprehensive intervention, the life expectancy of patients will be greatly improved. There are many explorations and applications in the field of computer-aided disease diagnosis, which can significantly improve the efficiency of disease diagnosis. Medical image processing is one of the medical imaging technology. It can help doctors improve the diagnosis quality by processing and analyzing the medical image data by computer. An improved FCM clustering method Sagakfcm algorithm is proposed for brain tissue segmentation in MRI images. Sagakfcm model fully combines the advantages of simulated annealing algorithm and genetic algorithm, so as to obtain the best initial clustering center, reduce the iteration times of fuzzy c-means algorithm, avoid the initial clustering falling into local optimum, and accelerate the operation speed. The algorithm combines Gaussian kernel function to improve the robustness of FCM algorithm.

Xinlei Chen, Dongming Zhao, Wei Zhong, Jiufeng Ye

Facial Expression Recognition via ResNet-18

As an important part of human-computer interaction, facial expression recognition has become a hot research topic in the fields of computer vision, pattern recognition, artificial intelligence, etc., and plays an important role in our daily life. With the development of deep learning and convolutional neural network, the research of facial expression recognition has also made great progress. Moreover, in the current face emotion recognition research, there are problems such as poor generalization ability of network model. The extraction of traditional facial expression recognition features is complex and the effect is not ideal. In order to improve the effect of facial expression recognition, we propose a feature extraction method for deep residual network, and use deep residual network ResNet-18 to extract the features of the data set. Through the experimental simulation of the specified data set, it can be proved that this model is superior to state-of-the-art methods model.

Bin Li, Runda Li, Dimas Lima

Comparison of AWS and AZURE for COVID-19 Information Retrieval

This paper deals with the most well-known and widely used technology which is cloud computing . There are some significant techniques which are needed for the implementation of the cloud computing. These techniques have also been discussed in this report. The cloud service is provided in three cloud computing model and each model will be chosen by the cloud user based on their requirement. These cloud service models has also been discussed in this report. A Covid19 information retrieval system for tracking the disease spread has been developed. This information system web application will be hosted in the AWS and Azure Cloud platform. The significant differences between the AWS and Azure cloud platform will also be discussed based on the pros and cons of each cloud platform .

Hemil Patel, Roopakala Mankaveettil, Reshmi Kanakuzhiyil Rajan, Nagamaisamma Challa, Rajeshwar Maryala, Saitheja Parsha, Pavan Kumar Bayyarapu

Expression Recognition Algorithm Based on Infrared Image

It’s important to recognize facial expressions in social communication. To solve the problem that facial expression recognition by visible light is vulnerable to interference, we built a model from the perspective of thermal infrared. Based on the distribution characteristics of thermal infrared, the face region is firstly located by building a multi-projection model toward color. Then, the level set function of the local Gaussian fitting model was optimized, the regular term was removed, and the larger iteration step size was selected to achieve accurate face segmentation on the premise of segmentation accuracy. Finally, based on the structure of traditional deep learning network, the characteristics of DPN and CBAM network are given full play to realize expression recognition by thermal infrared images.

Ying Cui, Shi Qiu

The Study About the Emotional State and Physical Activity of Adolescents During the COVID-19 Epidemic

To investigate the relationship between emotional status and physical activity in adolescents during the epidemic period of Corona Virus Disease 2019. 600 junior and senior high school students from three municipal middle schools were randomly selected as the research objects. The self-evaluation of anxiety and depression and the evaluation of physical activity were carried out in the form of questionnaire survey. A total of 600 questionnaires were put in and 562 were recovered. The scores of SDS and SAS were 49.30 ± 7.02, and 53.42 ± 5.37 respectively. According to different age groups, there was significant difference in SAS among the three groups in different age groups (P < 0.05). The total score of PA was (3.24 ± 0.98). According to different age groups, there were significant differences in PA total score, MVPA activities, physical education activities, weekend activities and one week total activities among the three groups (P < 0.05). The total score of anxiety was negatively correlated with the total score of PA (r =  −0.54, P = 0.024), MVPA (r =  −0.38, P = 0.049) and physical education (r =  −0.62, P = 0.016), and the total score of one week was negatively correlated (r =  −0.44, P = 0.041). During the period of Corona Virus Disease 2019 epidemic, the anxiety level of adolescents increases with age, while the physical activity status decreases gradually, and is negatively correlated with anxiety. It is necessary to strengthen sports activities and protect emotional health in this special period.

Runda Li, Yutong Wu, Wenxuan Zhanggu, Chihao Xu, Yuhan Gu, Shihan Yao, Hangxiao Li, Yuwei Shi, Yaojun Yang, Zhuoyang Zhen, Baijun Zhang, Chengyu Ye, Zimeng Li, Shumeng Shi, Xinyan Wang, Jingyang Chen, Jiaxi Lei

Remote Consultation Information Mobile Phone Intelligent Display System Under Augmented Reality and Human-Computer Interaction

The current mobile phone intelligent display system lacks the matching operation of display mode before the display driving operation, which leads to the large average variance of display results. Therefore, this research designs a new intelligent display system of remote consultation information based on augmented reality and human-computer interaction technology. The research and design is mainly for software design. Firstly, two-dimensional display matching algorithm is applied to complete the matching operation of display mode. At the same time, Linux is used as the system framework to be compatible with different modules. In the system, augmented reality and UI human-computer interaction module are established to increase the system functions. GPRS technology is used as the carrier in remote transmission. Finally, a new display driven algorithm is designed. In order to verify the feasibility of the intelligent display system on mobile phone, a comparative experiment was designed to display the remote consultation information, and the power consumption and mean square deviation of different systems were analyzed. The experimental results show that the power consumption and mean square deviation of the system are low, which fully proves the feasibility of the system.

Ying Bao

Design of Real Information Collection Model of Physical Fitness for the Elderly Based on Internet of Things Technology

In order to optimize the efficiency of traditional physical information collection, this research designed a physical fitness information collection model for the elderly based on the Internet of Things technology. In the Internet of Things environment, select the collection node and install sensor equipment, and use the Internet of Things technology to drive the collection program. From the three aspects of human body structure, physical fitness health parameters and fitness action posture, the collection content of physical fitness information for the elderly is determined. Then call the sensor equipment, with the support of the Internet of Things technology, through the analog-to-digital conversion and filter storage to obtain the real-state information collection results. Through comparison with traditional collection methods, it is found that the collection accuracy of the model in this paper is higher and the collection time required is shorter, so the collection efficiency of the model in this paper has been effectively improved.

Wei-Ping Cao, Yu-Shuo Tan

Track and Field Head Posture Error Correction System Based on Deep Reinforcement Learning

The problem that track and field athletes generally have non-standard postures in their playing actions, a track and field head posture error correction system based on deep reinforcement learning is designed. By optimizing the system configuration, improving the recognition accuracy, using deep reinforcement learning technology to obtain 3D deep dynamic image data of track and field sports, converting the data into quaternion format, storing the data file in VBH format, and shaping the data through deep reinforcement learning technology a dynamic three-dimensional model is used to judge whether the track and field posture is standard using the Euclidean distance comparison method. Using the powerful learning ability of deep reinforcement learning, a series of non-linear operations are performed on the input face image, the abstract features in the image are extracted layer by layer, and then the extracted features are used for classification and recognition and error correction. Finally, through actual research, the standardization of the track and field head attitude error correction system based on deep reinforcement learning is proved. The experimental results show that this method effectively improves the accuracy of attitude estimation.

Liu Er-wei

Visual Imaging Method of 3D Virtual Scene Based on VR Technology

The traditional 3D virtual scene visualization imaging method has low accuracy and serious center offset in the imaging process. Therefore, a visualization imaging method of 3D virtual scene based on virtual reality technology is designed. In order to reduce the complexity of the imaging scene and reduce the texture interval, texture mapping is used to improve the overall interaction performance of VR technology. In order to improve the reality of the scene, a 3D virtual viewpoint structure is designed by optimizing dynamic collision detection with octree. The visualization of 3D virtual scene is completed by mapping calculation. In order to verify the effectiveness of the design method, an experiment is designed. The results show that the coordinates of the center point obtained by this method are closer to the actual coordinates, indicating that the imaging process is more in line with the actual situation and the imaging accuracy is higher.

Zhao Bing, Zhou Qian

Human Centered Computing in Digital Persona Generation

Deepfake (or as we call it Digital Persona) has been very popularly used to create synthetic media in which a person in an existing image or video is replaced with someone else who is not present in that media. It refers to manipulated videos, or other digital representations produced by sophisticated artificial intelligence (AI), that yield fabricated images and sounds that appear to be real.Deepfakes generally have been used for the purpose of defaming someone, where the user experience is not much of a concern. However, our work demonstrates using this technique for a good purpose. We created a digital persona of a renowned deceased artist with the aim to bring an enriching human experience through conversing with the persona projected on a 3d holographic stage in a museum. The digital persona responds in the voice of deceased artist to any questions asked by visitors related to his art journey and artwork. To ensure that the end results would have the audience immersed or awed with the outcome a.k.a. the digital persona, we adopted the human centered computing methodology which aims at radically changing the standard computing techniques of software development. In this work, the key elements of human centered computing include: a. Technology b. Cognitive Psychology and Ergonomics c. Social and Organizational Psychology d. Design and Arts e. Interaction f. Analysis for design of systems with a human focus from beginning to the end. We present the usage, details and outcomes of the mentioned focus areas in our design of developing deepfakes for good. We also present results of a social experiment conducted with children during their interaction with digital persona.

Nisha Ramachandra, Manish Ahuja, Raghotham M. Rao, Neville Dubash

Content-Based Image Retrieval Using Local Derivative Laplacian Co-occurrence Pattern

For accessing images from huge repository in an easy manner, the images are required to be properly indexed. Content-Based Image Retrieval (CBIR) is a field which deals with finding solutions to such problems. This paper proposes a new multiresolution descriptor namely, Local Derivative Laplacian Co-occurrence Pattern (LDLCP) for CBIR. Gray level image is subjected to four-level Laplacian of Gaussian filtering in order to perform multiresolution processing of image. Local Derivative Pattern descriptors of resulting four-level filtered image is computed to extract local information from the image. Finally, the Gray-Level Co-occurrence Matrix is used for constructing feature vector. Corel-1K and Corel-5K datasets have been used to test the proposed descriptor and its performance is measured using precision and recall metrics.

Prashant Srivastava, Manish Khare, Ashish Khare

Multi-spectral Image Filtering Algorithm Based on Convolutional Neural Network

In order to solve the problem of long processing time and poor processing effect of traditional methods, a multispectral image filtering algorithm based on convolutional neural network is proposed. Based on convolution neural network, the spectrum image features are defined, and the image SNR is registered. Based on Fourier transform, the improved algorithm of multi spectrum superposition is used to realize the mean filtering of multi spectrum image. The experimental results show that this method has higher stability and effectiveness in the actual operation process, and the image filtering time is shorter. The experimental results prove the effectiveness of the algorithm.

Dan Luo, Rong Hu

Interactive Virtual Reality Indoor Space Roaming System Based on 3D Vision

Due to the low performance of traditional virtual reality roaming systems, three-dimensional vision technology is used to optimize the design of interactive virtual reality indoor space roaming systems from the three aspects of hardware, software, and database. The optimization of the hardware system is mainly aimed at the connection structure of the rover and the interactive module. Collect all relevant data of indoor space and system operation, and store them in a certain format to get the design results of the system database. With the support of hardware equipment and database, the function of interactive virtual reality indoor space roaming is realized through the steps of creating virtual scene of indoor space, determining the roaming mode of 3D virtual reality and the path planning of indoor space roaming. Through the system test experiment, it is concluded that compared with the traditional roaming system, the designed interactive virtual reality indoor space roaming system has a higher success rate and better interactivity.

Jing He

Multi-viewpoint Rendering Optimization of Indoor Scene Based on Binocular Vision

The traditional multi-viewpoint rendering method for indoor scenes has poor lighting and shadow effects, resulting in too dark or bright indoor scene multi-viewpoint rendering. Therefore, an optimization method for indoor scene multi-viewpoint rendering based on binocular vision is proposed. This research plans light and shadow effects based on the visual relationship between point light sources and indoor scenes; sets rendering points based on binocular vision; and renders indoor scenes with multiple viewpoint angles. The simulation experiment results show that compared with the traditional rendering method, the light and shadow effect of the studied method is excellent, and the rendered indoor scene is suitable for light and dark.

He Jing

A New Confidence Propagation Algorithm for Regional Image Based on Deep Learning

In order to improve the accuracy of regional image confidence propagation calculation, a regional image confidence propagation algorithm based on deep learning is designed. Firstly, the relevant information is collected, and then the data similarity is calculated. Finally, the regional image confidence propagation algorithm based on deep learning is calculated. The experimental results show that the regional image confidence propagation algorithm based on deep learning improves the calculation accuracy and reduces the calculation time.

Jia Qian, Li-li Wang, Hai-yue Huang

Feature Extraction Method of EEG Signal Based on Synchroextracting Transform

Brain-Computer Interface (BCI) can convert the electrical activity signal of the cerebral cortex into a computer or other machine language to directly control external equipment. Aiming at the problem of low recognition accuracy of visual stimulation Electroencephalogram (EEG) signals. This paper adopts a method of EEG signal feature extraction based on Synchroextracting Transform (SET). The mean value filter method is used to remove the noise in EEG signal, and the time-frequency energy of EEG signal is taken as the characteristic parameter. Finally, the signal characteristics are input into the SVM model as characteristic parameters. The experimental results show that SET can extract the characteristic energy of EEG signal well and improve the resolution of signal.

Lin Han, Liang Lu, Haoran Dong, Shuangbo Xie, Gang Yu, Tao Shen, Mingxu Sun, Tianyi Wang, Xuqun Pei

Human Cross-Border Alarm Detection Method Based on OpenPose

Cross border detection is often used to monitor the behavior of people in specific places where people often entry and exit, people may cross some unsafe or forbidden borders, thus causing dangerous behaviors, such as large power plant or electrical equipment room. In order to prevent the occurrence of dangerous behaviors, this paper proposed a method for human cross-border alarm detection. First, the camera captures the image of the scene, and design the unsafe bounding line. Second, detect the human and its foot based on OpenPose. Third, when a cross-border behavior occurs, judge whether there is an intersection between the boundary line and the line formed by human feet in two images to send an alarm signal. This method effectively saves costs, replaces artificial ways and improves detection efficiency at the same time, and can make an alarm in time when humans cross an unsafe boundary.

Hang Yu, Qinjun Zhao, Yong Zhang, Shengjun Shi

Design and Implementation of Disconnector Condition Monitoring System Based on Attitude Sensor

In order to meet the new requirements of substation for the status monitoring of disconnector switch, a disconnector status monitoring system based on MPU9250 was designed. It realizes the double confirmation of status monitoring of disconnector switch in substation. A status determination algorithm of disconnector switch was proposed.

Yueyu Du, Shubo Qiu

Backmatter

Weitere Informationen

Premium Partner