利用模糊聚合特征向量的视频人体识别方法

A Video Human Recognition Method Using Fuzzy Aggregation Feature Vector

针对传统视频人体动作识别中的错误警报和冗余计算问题,提出了一种模糊聚合特征向量融合隐马尔可夫模型(HMM)的识别方法。首先,利用模糊推理系统检测前景目标轮廓,使用边缘检测算法获得轮廓边缘;然后,利用特征提取技术获取距离特征、角度特征和CA比率特征,并将其聚合为一种特征;最后,通过矢量量化将聚合特征量化为相应的符号,并利用HMM完成人体动作识别。实验结果表明,提出的方法对近目标的检测精度可达99.8%,相比其他几种较新的方法,提出的方法取得了更好的识别性能,表明多特征聚合可有效解决视频人体识别问题。

For the issue of false alarms and redundant computation, a method based on the combination of fuzzy aggregation feature vector and HMM is proposed. Firstly, the prospect target outline is detected by the fuzzy inference system, and the edge detection algorithm is used to get contour edges. Then, the feature extraction technology is used to extract the distance feature, the angle feature and the CA ratio feature, which is aggregated as a single feature. Finally, the vector quantization is used to quantize the aggregation feature to be corresponding symbols, and HMM is used to recognize human actions. Experimental results show that the recognition accuracy of the proposed method can achieve 99.8 %, and it has better recognition performance than several other advanced recognition methods, which indicates that multiple feature aggregation is helpful to human recognition from videos.