摘要
为了实现观看者无接触操作情况下的视频播放器智能控制,系统利用Kinect传感器采集彩色图像,使用FaceNet提取人脸特征向量,经支持向量机(SVM)训练后进行人脸识别,该过程在计算机中央处理器(CPU)运行环境下,利用OpenVINO实现人脸检测与识别实时运行,用于视频播放器的登录验证。系统采集音频数据使用Speech Platform Runtime v11进行中文命令识别,使用Kinect Speech Language进行英文命令识别,进而实现语音控制。采集骨骼数据,计算骨骼点之间的距离与角度进行人体姿态和手势识别,将识别结果转换为控制命令,进而实现播放器的快进、切换视频、加减音量等常用的控制功能。实验结果表明,该交互系统实现了使用者无接触全自动人体控制,为视频播放器提供了一种自然便捷的交互方式。
In order to realize the intelligent control of the video player under the non-contact operation of the viewer,the system uses Kinect sensors to collect color images,uses FaceNet to extract facial feature vectors,and performs face recognition after support vector machine(SVM) training.This process is under the computer central processing unit(CPU) operating environment.Open VINO is used to realize real-time operation of face detection and recognition,which is used for login verification of video players.The audio data collected by the system uses Speech Platform Runtime v11 for Chinese command recognition,and Kinect Speech Language for English command recognition,thereby realizing voice control.Bone data is collected,the distance and angle between the bone points are calculated to recognize human postures and gestures,and the recognition results are converted into control commands to realize the player’s common control functions such as fast forwarding,switching videos,and adding and subtracting volume.The experimental results show that the interactive system realizes the user’s contactless full-automatic human body control,and provides a natural and convenient way of interaction for the video player.
作者
李国友
王维江
李晨光
杭丙鹏
杨梦琪
Li Guoyou;Wang Weijiang;Li Chenguang;Hang Bingpeng;Yang Mengqi(School of Electrical Engineering,Yanshan University,Qinhuangdao 066004)
出处
《高技术通讯》
CAS
2021年第2期129-140,共12页
Chinese High Technology Letters
基金
国家自然科学基金(F2012203111)
河北省高等学校科学技术研究青年基金(2011139)资助项目。
