Abstract: To solve the problem that multiscale Gabor features are unsuitable for ARM because of huge data and dimensions in the embedded face recognition system, a multiscale Gabor feature weight fusion method is proposed, including multiscale Gabor features extraction, the weight of features calculating and weight fusion. The embedded face recognition system detects face by using Haar-like features of face, and reduces dimensions by using 2DPCA algorithm. The system is implemented based on EELiod 270 development board, experimental results on ORL face database, Yale face database and actual images verify the recognition time is reduced when the system guarantees the adequate rate of recognition.

Key words: Gabor, weight features fusion, Haar, 2 Dimensional Principal Component Analysis（2DPCA）, embedded system

摘要： 在嵌入式人脸识别系统中，由于多尺度Gabor抽取特征的维数和数据量过大，不适合在ARM板上实现，提出了多尺度Gabor特征加权融合的方法，很好地解决了图像维数和数据量过大的难点。加权融合过程包括多尺度Gabor特征的提取、特征权值的计算和加权融合过程。同时使用了类Haar特征提取人脸、2DPCA对人脸图像进行降维。基于EELiod 270嵌入式开发平台实现了一个嵌入式系统，结合典型图片库和实际图片进行了人脸识别测试，实践结果表明，系统在保证一定的识别率的同时，大幅度降低了运行时间，实时识别效果良好。

关键词: Gabor, 加权特征融合, Haar, 二维主成分分析（2DPCA）, 嵌入式系统