基于历史运动轨迹的自适应手势识别算法

Adaptive Gesture Recognition Algorithm Based on Motion History Image

针对复杂背景严重影响手势识别准确率的问题,提出了一种基于历史运动轨迹的自适应手势识别算法.首先采用基于H值肤色模型进行肤色侦测,并采用手部分割和自适应检测提取手部,然后对手部进行快速跟踪得到其历史运动轨迹,最后对运动方向进行辨识.实验结果表明,提出的算法复杂度较低,每一帧图像的平均处理时间为21ms,能实时有效地对手部运动进行追踪,在不同背景下平均准确率高达97.1%,有较好地鲁棒性.

Currently human-computer interaction （HCI） is the hottest topic, and hand gesture recognition is the key technology of HCI. Due to the effect of complex background, an adaptive hand gesture recognition algorithm based on motion history image is proposed. Firstly, a skin color model using H threshold is utilized to detect skin color regions like hands, and using hand segmentation and adaptive hand detection to detect hand. Then a simple and fast motion history image is developed to record the trajectory of hand-movement. Finally, the motion history is used to classify the motion direction. Experiment results show that this algorithm is simple, and the average processing time is 21 ms per frame, meeting the real-time tracking. The accuracy of recognition is as high as 97.1% in average, showing the strong robustness. These demonstrated the feasibility of the proposed algorithm.