多行为模式高频切换下尺度指纹步态算法

Scale Fingerprint Gait Algorithm of Multi Patterns of Behavior Under High Frequency Switching

为了实现基于微惯性加速度计的多行为模式高频切换下步态信息跟踪,提出一种尺度指纹步态算法,并通过RP活动区域聚类的方法减小算法开销。在离线阶段建立多行为模式的指纹数据库,根据行为模式分类算法将指纹数据库分成2个区域类,即区域类Ⅰ和区域类Ⅱ。在线运行阶段,将实时计算出的尺度特征值先进行区域类匹配,然后在对应的区域类内进行指纹点匹配。该算法经Android手机平台验证表明,在多行为模式高频切换条件下,步态跟踪精度至少提高12%以上,基本满足行人对行为模式切换的要求,具有较大的工程实用价值。

In order to track the gait information of multi patterns of behavior based on MEMS inertial accelerometers under high frequency switching,a scale fingerprint gait algorithm was proposed,and a RP activity regional clustering method was presented to reduce the algorithm overhead.First,a fingerprint database of multi patterns of behavior was created in the offline stage.Then,the fingerprint database was divided into two large regions（regionⅠand regionⅡ）based on classification algorithm.Last,in the online stage,the scale characteristic values which were calculated in real-time were matched to correspond regions,then these scale characteristic values were matched by fingerprint points in correspond regions.The proposed algorithm was verified by Android mobiles.The results showed that the accuracy of tracking was improved by 12% or more at high frequency switching.It could meet the requirements of pedestrian behavior pattern＇s switching,so it had a high practical engineering value.