融合多特征和SVM的跌倒检测方法研究

Research on Fall Detection Method Based on Multi-Feature and SVM

为了及时发现老人跌倒并及时救助,对跌倒检测方法进行了研究。首先采用ViBe算法提取运动的人体,采用高斯滤波和形态学处理后,接着提取人体长宽比、角度、质心高度三个跌倒特征,组合成特征向量并添加保存至特征提取器,最后导入支持向量机模型进行训练。实验证明该方法能有效区分跌倒与非跌倒问题。

In order to detect and rescue the elderly fall in time, the fall detection method is studied in this paper. Firstly, the Vibe method is used to extract the moving human body. After Gaussian filtering and morphological processing, three fall features such as aspect ratio, angle, centroid height are extracted, combined into a feature vector and added to the feature extractor. Finally, the support vector machine model is introduced for training. Experiments show that this method can effectively distinguish between fall and non-fall problems.