基于递归多尺度卷积网络的图像超分辨率重建

不断加深网络的深度可提高网络的超分辨率重建效果,但是网络的加深会导致网络参数量急速增加,难以进行网络训练和内存存储.为了减小深度网络的参数规模并尽量保持网络的重建性能,基于递归和多尺度的思想,文中提出精简的基于递归多尺度卷积网络的图像超分辨率重建方法.首先利用多尺度模块充分提取图像在不同尺度下的特征信息,再通过递归操作实现网络规模的加深而不增加网络的参数量,最后将每次递归操作的输出进行特征融合,作为高分辨率图像重建的输入.实验表明,文中方法在网络参数量较少时重建效果较优.     

基于SPOC视频学习数据的学习行为特征分析方法

在基于小规模限制性在线课程(Small Private Online Course,SPOC)平台的学习过程中,教育数据为学习分析提供了依据。为了帮助教师改进教学策略,帮助学生改进学习效果,本文提出了一种综合考虑学生分类和视频学习时间的学习行为分析方法,旨在发现学生潜在的学习现象,并根据该现象对不同类型学生的学习行为之间的相似性和差异性进行分析和总结,发现不同类型学生的学习习惯以及影响学生学习效果的时间因素。通过案例分析表明,该方法可以为教师教学和学生学习提供反馈,辅助学生的自我调节学习和教师的个性化教学。     

抗扰的板球控制系统的设计与应用

设计与制作集机械设计、控制理论应用、软硬件控制于一身的机电一体化高精度板球控制系统。系统采用微处理器Kinetis K60单片机为主控芯片,选用灰度摄像头OV5116作为滚球位置坐标的反馈模块,使用灰度转化二值图像以及卡尔曼滤波处理,并以MG995舵机作为系统的执行部件。运用位置式PID 闭环控制算法,控制光滑平板的倾斜角度使滚球能在平板上按可设路径滚动,并实现任意可设坐标的定点停留。实验结果表明,板球控制精度高,响应速度快,滚球停于指定位置的误差小于4mm,其滚动路径偏离误差小于4mm,同时系统具有较强的抗干扰能力。     

Predicting students’ learning style using learning analytics: a case study of business management students from India

ABSTRACT

Business Management Education in India has shown an upward growth trend in the last couple of decades. Due to the diverse nature of the course, students from diverse academic backgrounds are being admitted to the course. Therefore, differences in students’ abilities and their learning styles have a significant effect on their learning outcomes. Meanwhile, with the development of learning technologies, learners can be provided a more effective learning environment to optimise their learning. The purpose of this study was to develop a model to automatically detect the students’ learning styles from their personal, academic and social media data and make recommendations for students, teachers, educators and administrators for overall improvement of learning outcomes. Data analysis in this research was represented using data collected from post-graduate business management students in India. A 10-fold cross-validation was used to create and test the models. The data were analysed by R and R-Studio. Classification accuracy, Precision, Recall, Kappa, ROC curve and F measure were observed. The results showed that the accuracy of classification by the C4.5 technique had the highest value at 95.7%, and it could be applied to develop Felder–Silverman’s learning style while taking into consideration students’ academic, personal information and social media preferences.     

Learned Action SLAM: Sharing SLAM through learned path planning information between heterogeneous robotic platforms

Currently when path planning is used in SLAM it is to benefit SLAM only, with no mutual benefit for path planning. Furthermore, SLAM algorithms are generally implemented and modified for individual heterogeneous robotic platforms without autonomous means of sharing navigation information. This limits the ability for robot platforms to share navigation information and can require heterogeneous robot platforms to generate individual maps within the same environment. This paper introduces Learned Action SLAM, which for the first time autonomously combines path-planning with SLAM such that heterogeneous robots can share learnt knowledge through Learning Classifier Systems (LCS). This is in contrast to Active SLAM, where path-planning is used to benefit SLAM only. Results from testing LA-SLAM on robots in the real world have shown; promise for use on teams of robots with various sensor morphologies, implications for scaling to associated domains, and ability to share maps taken from less capable to more advanced robots.     

Automatic data clustering using continuous action-set learning automata and its application in segmentation of images

Most of the proposed algorithms to solve the dynamic clustering problem are based on nature inspired meta-heuristic algorithms. In this paper a different reinforcement based optimization approach called continuous action-set learning automata (CALA) is used and a novel dynamic clustering approach called ACCALA is proposed. CALA is an optimization tool interacting with a random environment and learn the optimal action from the environment feedbacks. In this paper the dynamic clustering problem considered as a noisy optimization problem and the team of CALAs is used to solve this noisy optimization problem. To build such a team of CALAs this paper proposed a new representation of CALAs. Each automaton in this team uses its continuous action-set and defining a suitable action-set for each automaton has a great impact on the CALAs search behavior. In this paper we used the statistical property of data-sets and proposed a new method to automatically find an action-set for each automaton. The performance of ACCALA is evaluated and the results are compared with seven well-known automatic clustering techniques. Also ACCALA is used to perform automatic segmentation. The experimental results are promising and show that the proposed algorithm produced compact and well-separated clusters.     

A robust incremental clustering-based facial feature tracking

Emerging significance of person-independent, emotion specific facial feature tracking has been actively tracked in the machine vision society for decades. Among distinct methods, the Constrained Local Model (CLM) has shown significant results in person-independent feature tracking. In this paper, we propose an automatic, efficient, and robust method for emotion specific facial feature detection and tracking from image sequences. A novel tracking system along with 17-point feature model on the frontal face region has also been proposed to facilitate the tracking of human basic facial expressions. The proposed feature tracking system keeps patch images and face shapes till certain number of key frames incorporating CLM-based tracker. After that, incremental patch and shape clustering algorithms is applied to build appearance model and structure model of similar patches and similar shapes respectively. The clusters in each model are built and updated incrementally and online, controlled by amount of facial muscle movement. The overall performance of the proposed Robust Incremental Clustering-based Facial Feature Tracking (RICFFT) is evaluated on the FGnet database and the Extended Cohn-Kanade (CK+) database. RICFFT demonstrates mean tracking accuracy of 97.45% and 96.64% for FGnet and CK+ database respectively. Also, RICFFT is more robust by minimizing average shape distortion error of 0.20% and 1.86% for FGnet and CK+ (apex frame) database, as compared with classic method CLM.     

A software architecture for virtualized educational multimedia games in IPTV environments

T‐Learning makes it possible to deliver educational content to home TVs. TV operators, which manage huge populations of devices such as set‐top‐boxes (STBs) in user homes, are considering t‐Learning as an interesting option for expanding the service they offer. However, typical STB hardware configurations are limited in terms of satisfying operator needs and do not easily support all types of applications or content. In this work, we consider graphic < e‐Adventure > educational games, which are not directly executable on typical STBs. To cover this gap and guarantee an enjoyable user experience, we present an architecture based on a combination of streaming and remote desktop protocols that relies on virtualized servers deployed in a cloud computing infrastructure. It features an original image‐encoding signalling mechanism that identifies multimedia content in educational games and permits seamless protocol switching at the client side. This architecture is a complete technological solution to virtualize heavy educational games and execute them smoothly on STB light clients over Internet Protocol Television networks. We present performance results that show that our proposal is an efficient scalable solution to deliver t‐Learning to home environments. Copyright © 2013 John Wiley & Sons, Ltd.     

Moderating effect of learning styles on a learning management system’s success

This study uses the DeLone and McLean model to determine the moderating impact of learning styles on the success of learning management systems from a student’s point of view. The main objectives of this research are: (1) to evaluate the Delone and McLean model of information system success in the context of learning management systems and, (2) to determine the effect of the learning styles of students on this model. An in-person survey of 258 engineering students was used to evaluate the research model. The analysis is based on structural equation modelling, specifically partial least squares. The results indicate that the research model explains use, user satisfaction, and perceived benefits of a learning management system. In addition, the Felder-Silverman learning styles (sensing-intuitive, visual-verbal; active-reflective; sequential-global) modify the strength of the relationships between the variables of the success model.     

基于属性权值多级分类的测试样本数据预处理方法

针对算法测试在工业生产实际操作应用中,经常面临样本数据迅速大量增加,测试执行时间过长、测试效率低下、算法改进周期过长的问题.文章提出一种精简测试样本的算法,快速实时处理新增样本,去除冗余数据.首先对新增的样本数据按照分类阈值分配到各子数据集中,然后在每一个子数据集中根据数据的重要程度属性计算相似度值,为减少计算量,先对子数据集进行排序处理,最后根据相似性的阈值判断去除冗余数据.实验利用所得样本库不断对算法进行回归测试,分析表明算法的测试效率及测试充分性都有明显提升,并提高了算法在工业生产实际应用过程中的稳定性和适应性,有力的保证了算法的质量.