针对基于手部几何特征的手形识别方法可利用的个体信息有限的问题, 提出了一种将手指轮廓特征与几何特征相融合的身份识别方法。该算法首先分离手指, 采用曲线拟合算法定位手指中轴线; 然后采用分步对齐方法规范化手指, 并提取手指轮廓特征和几何特征; 最后采用粒子群算法对手指截取系数和权值系数进行优化, 以进一步提高识别准确率。实验结果表明: 采用该方法后, 识别率可达98.61%。该方法手指定位更准确, 更充分地利用了手部信息, 且避免了特征点定位不准及手指根部不稳定轮廓特征对识别准确率的影响, 具有较高的识别率和良好的鲁棒性。

As the hand shape recognition method based on geometric features is limited by its individual information, this paper presents a new hand shape recognition method based on the fusion of finger contour features and geometric features. A curve fitting method was proposed to position the axis of finger after four fingers were separated. Then the matched fingers were normalized by stepwise alignment method, and the contour features and geometric features were extracted. Finally, the Particle Swarm Optimization (PSO) was used to optimize the cut-off coefficients and the weight values of fingers to further improve the recognition rate. Experimental results show that the recognition rate by proposed method is 98.61%. As the four fingers are separated, the proposed method reduces the computing consumption and gets more accurate located hand shapes. With make full use of the hand information, it also avoids the influence of inaccurate feature point location and instable contour around finger valleys on the recognition accuracy, and has high recognition rate and good robustness.