基于区间分布概率矩阵模型的动态手势识别方法
TP391.4; 针对目前基于加速度传感器的手势识别算法的动态实时性与识别率的相互矛盾性,提出一种区间分布概率矩阵模型及动态手势识别方法.将手势动作的三维加速度信号进行动作数据自动检测、归一化和三次样条插值预处理,再根据信号分布特征,确定数据观测点,构造各观测点处的区间分布概率矩阵,优化矩阵,实现在线快速手势识别.该方法对手指可穿戴设备得到的真实数据集中进行了评估.结果显示其实时效果好,识别率高,实用性强....
Saved in:
| Published in | 电子技术应用 Vol. 39; no. 1; pp. 72 - 75 |
|---|---|
| Main Author | |
| Format | Journal Article |
| Language | Chinese |
| Published |
燕山大学信息科学与工程学院,河北秦皇岛,066004
2013
|
| Subjects | |
| Online Access | Get full text |
| ISSN | 0258-7998 |
Cover
| Summary: | TP391.4; 针对目前基于加速度传感器的手势识别算法的动态实时性与识别率的相互矛盾性,提出一种区间分布概率矩阵模型及动态手势识别方法.将手势动作的三维加速度信号进行动作数据自动检测、归一化和三次样条插值预处理,再根据信号分布特征,确定数据观测点,构造各观测点处的区间分布概率矩阵,优化矩阵,实现在线快速手势识别.该方法对手指可穿戴设备得到的真实数据集中进行了评估.结果显示其实时效果好,识别率高,实用性强. |
|---|---|
| Bibliography: | For the present, gesture recognition algorithms based on the accelerometer have contradiction between dynamic real- time and recognition rate. Aiming at this problem, a probability matrix model of interval distribution and dynamic gesture recognition method is proposed. The gestural signal from the three dimension accelerometer is preprocessed through a series of methods includ- ing automatically action data detection algorithm, normalization, and cubic spline interpolation. Moreover, according to the eharaeter- istics on signal distribution, the observation points in each axis are determined and get the probability matrix of interval distribution on each observation point. Further, the on-line and fast gesture discrimination algorithm is realized on the optimized matrixes. The method is evaluated on real data set from a finger-mount wearable device. The result shows that it has good real time effect and high recognition rate. accelerometer; signal preprocessing; probability matrix of interval distribution; dy |
| ISSN: | 0258-7998 |