基于卷积神经网络的运动视频可靠性评估算法  

Reliability Evaluation Algorithm of Motion Video Based on Convolution Neural Network

作　　者：李晓峰[1] 邢金明[2] LI Xiao-feng;XING Jin-ming(Department of Information Engineering,Heilongjiang International University,Harbin 150025,China;Northeast Normal University,Changchun 130024,China)

机构地区：[1]黑龙江外国语学院信息工程系,黑龙江哈尔滨150025 [2]东北师范大学,吉林长春130024

出　　处：《计算机技术与发展》2020年第9期71-76,共6页Computer Technology and Development

基　　金：教育部科技发展中心产学研创新基金(2018A01002);教育部人文社科基金项目(15YJC890004);中国博士后基金项目(2017M610852);吉林省社科基金重点项目(2016A5)。

摘　　要：为了提高对人体运动视频的自动识别和检测能力,提高可靠性能,提出一种基于卷积神经网络的人体运动视频传输可靠性评估算法。由卷积神经网络算法识别人体运动视频传输的自适应分类,提取人体运动视频的空间边缘像素点分布标,结合边缘模块特征匹配技术构建人体运动视频的分块检测模型,实现对人体运动视频的特征辨识和图像采样。采用Harris角点检测方法定位人体运动视频的分块区域,在人体运动视频的分块区域内检测人体运动视频的形体轮廓区域,构建可靠性评估均衡博弈模型完成视频干扰抑制。采用视频特征提取和自动降噪方法分离人体运动视频传输过程中的多径特征,在神经网络的隐含层引入人体运动视频的几何特征,得到人体运动视频传输的可靠性评估的学习系数,完成可靠性评估。实验结果表明,采用该方法进行人体运动视频传输的可靠性较好,对视频图像的特征分辨能力较强且视频图像传输耗时较短,降低了视频传输的误码率。In order to improve the ability of automatic recognition and detection of human motion video,as well as the reliability,we propose a reliability evaluation algorithm of human motion video transmission based on convolution neural network.The adaptive classification of human motion video transmission is recognized by convolution neural network algorithm,and the spatial edge pixel distribution mark of human motion video is extracted.Combined with the edge module feature matching technology,the block detection model of human motion video is constructed,and the feature identification and image sampling of human motion video are realized.The Harris corner detection method is used to locate the segmented area of human motion video,and the shape outline area of human motion video is detected in the segmented area of human motion video.The reliability evaluation equilibrium game model is constructed to complete the video interference suppression.The multi-path features in the process of human motion video transmission are separated by video feature extraction and automatic noise reduction.The geometric features of human motion video are introduced into the hidden layer of neural network,and the learning coefficient of reliability evaluation of human motion video transmission is obtained,and the reliability evaluation is completed.The experiment shows that the proposed method is reliable,the feature resolution of video image is strong and the transmission time of video image is short,which reduces the bit error rate of video transmission.

关 键 词：卷积神经网络 人体运动 视频传输 可靠性评估 误码率 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]