基于轮廓PCA的字母手势识别算法研究

Method of gesture recognition based on contour PCA

提出了一种改进的手势检测识别算法。首先对采集的手势图像综合运用Krisch算子的背景差分与肤色分割等算法进行预处理,然后根据手势的方向去除胳膊等非目标,截取手型区域。最后对手型图像进行轮廓检测,根据手型轮廓的二值图像进行PCA降维,并与投影降维后的样本计算空间距离,从而进行手势分类。通过此方法可以快速、准确地检测到手型区域,识别手型含义,且具有实时性。

This paper gives an improved gesture recognition algorithm.First,the gesture images are processed through a combination of background subtraction method with Krisch and color segmentation to get the hand area.Then the non-target region such as arm which depending on the direction of the gesture is removed.Finally,the images are processed through contour detection and the dimension is reduced by PCA,the spatial distance with the sample after the projection is calculated for classification.The method can detect hand region and identify the meaning of gestures quickly and accurately,and can be used in real-time recognition.