针对人脸表情时空域特征信息的有效提取,提出了一种CBP-TOP(centralized binary patterns from threeorthogonal panels)特征和SVM分类器相结合的人脸表情识别新方法。该方法首先将原始图像序列进行图像预处理,包括人脸检测、图像截取和图像尺度归一化,然后用CBP-TOP算子对图像序列进行分块提取特征,最后采用SVM分类器进行表情识别。实验结果表明,该方法能更有效地提取图像序列的运动特征和动态纹理信息,提高了表情识别的准确率。与VLBP(volume local binary pattern)特征相比,CBP-TOP特征在表情识别中具有更高的识别率和更快的识别速度。
Traditional locking mechanism is fit for the concurrency controlling and data consistency maintenance undercentralized architecture, but its response is slow, so it can not be used as concurrency controlling strategy for real-time distributed cooperative editing system. Moreover, operation transformation and data consistency maintenancetechnology are able to assure of quick response and unconstraint, but can't solve the context-specific inconsistencyproblems. Enlightened by the idea of operation transformation and optimistic locking mechanism, we bring forwardthe concurrency control algorithm of optimistic locking based on relative position. In this algorithm, the start positionof locking region and the position of operation are relative, and they are not transformed into absolute position untiloperations are sent to cooperative sites. Furthermore, any coeditor can edit in advance before his/her locking is con-firmed. If his/her locking is successful, the previous operations go into effect, or else undo these previous operations.We have analyzed the actual applications and can find that the possibility of undoing the previous editing operationsbecause of locking conflict is very little. So this concurrency controlling algorithm has virtues of quick response, nicedata consistency maintenance etc.
