欢迎您！东篱公司
退出
-
申报数据库
-
立项数据库
-
成果数据库
-
项目获奖数据库
中文摘要：
以3D梯度描述为依据,提出了一种基于3D梯度投影描述捕捉微表情关键帧的方法.首先,通过对视频流中面部特征区域的投影梯度方向直方图的直观描述来分析面部表情动作趋势,进而通过直方图的峰值区域捕捉微表情所在的关键帧;然后,运用多尺度多方向的Gabor滤波器组提取微表情特征区域的Gabor图谱,并引入局部二值模式进行特征降维;最后,通过基于梯度量级加权的最近邻算法进行微表情的识别与分类.实验结果表明:该方法摆脱了传统视频流表情分析系统对于动态图像序列进行逐帧检测识别的不足,较为有效地实现了图像序列中微表情关键帧的捕捉与识别,提高了系统的实时性与准确性,基本满足微表情对于系统强实时性的需求.
英文摘要：
A 3D‐gradient projection descriptor was proposed to capture the micro‐expression key frames in the video stream .The expression tendency was analyzed by the intuitive description of the projection gradient direction histogram for facial feature area ,and the micro‐expression key frames were found by the peak area of the histogram .Then ,the Gabor spectrum of the feature area was ex‐tracted via the multi‐scale and multi‐direction Gabor filter bank ,and the feature dimension w as re‐duced by local binary pattern .Finally ,the classification and recognition of micro‐expressions was real‐ized by the gradient magnitude weighted nearest neighbor algorithm .The experiment results show that this method gets rid of the frame by frame detection and recognition in the traditional expression analysis system to greatly improve the real‐time and accuracy .This system effectively achieves the purpose that the key frames capture and recognition in the facial image sequence ,and basically meets the requirements of micro‐expression for strong real‐time .
同期刊论文项目
同项目期刊论文
期刊信息
位置：