复杂背景下基于肤色检测的动态手势分割与识别

在类肤色的复杂背景下,基于肤色检测的动态手势识别会因肤色干扰导致识别效率较低。提出了一种基于YCbCr颜色空间的改进三帧差分法的动态手势识别方法。首先利用改进的三帧差分法对动态手势进行分割,有效去除类肤色背景;然后根据人体肤色在YCbCr颜色空间中的聚类效果,采用基于椭圆模型的肤色检测方法有效去除非肤色背景,分割出手势区域。通过双特征提取,有效去除大范围的肤色背景,最终得到完整的手势;最后利用BP神经网络较强的自学习能力,对分割的动态手势进行检测识别。实验结果表明,此方法在应对环境变化时具有较好的实时性和抗干扰能力,拥有较高的识别率。

Research on Dynamic Gesture Segmentation and Recognition Based on Skin Color Detection in Complex Background

In complex skin-like background, dynamic gesture recognition based on skin color detection will lead to low recognition efficiency due to skin color interference. In this paper, an improved three-frame difference method for dynamic gesture recognition based on elliptic skin color model is proposed. Firstly, the improved three-frame difference method was used to segment the dynamic gesture, and the skin-like background was effectively removed. Then, according to the clustering effect of human skin color in YCbCr color space, the skin color detection method based on ellipse model was used to effectively remove the non-skin background and segment the gesture region. Through double feature extraction, the large-scale skin color background was effectively removed to obtain complete gestures. Finally, the BP neural network was used to detect and recognize the segmented dynamic gestures with strong self-learning ability. The experimental results show that this method has good real-time and anti-jamming ability and high recognition rate when dealing with environmental changes.