一种基于多级自适应门限的计步算法

深圳大学信息工程学院,广东深圳 518060

信息处理技术; 计步; 自适应; 多级门限; 步频特征; 三轴加速度计; 可穿戴式设备

Pedometer method based on adaptive multilevel thresholding
LI Gang,ZHENG Xinbo, and YANG Zhaocheng

College of Information Engineering, Shenzhen University, Shenzhen 518060, Guangdong Province, P.R.China

information processing technology; step counter; adaptive; multilevel thresholding; walking frequency characteristic; three-axis accelerometer; wearable device

DOI: 10.3724/SP.J.1249.2018.02158

备注

计步是智能穿戴设备和智能手机的重要功能,在健身、医疗和导航定位领域被广泛使用.针对现有计步算法中对不同运动状态变化适应性差、不能有效排除手腕抖动等非行走因素干扰的问题,提出多级自适应门限计步算法.该方法分为3级门限控制:先检测加速度信号中的峰值点和谷值点; 接着基于获得的峰值点和谷值点,计算人物的步频特征,根据所得步频特征判断人物的运动状态,如慢走、快走或跑步; 最后,针对每种运动状态,自适应分配对应的时间差阈值和动态峰谷值差值阈值,实现多级自适应门限计步.试验结果表明,所提多级自适应门限计步算法可完成对人物不同运动状态、传感器不同姿态的高精度计步,计步准确率可达95%以上.

The step counter plays an important role in the smart phone and intelligent wearable device. It has been widely used in the fields of body building, medical treatment, and navigation. In order to handle the performance degradation caused by various motion states and non-walking interference, e.g., waist's shaking, in existing pedometer methods, we propose a novel pedometer method based on the adaptive multilevel thresholding technology, which can be divided into three thresholds controlling. Firstly, it finds the peaks and valleys by using the accelerometer data. Secondly, it calculates the frequency characteristic of human's movement based on the obtained peaks and valleys, and subsequently determines the motion state(such as walking slowly, walking fast or running). Thirdly, it adaptively assigns the time threshold and peak-valley threshold according to the estimated motion states to achieve step counting. Experimental results show that the proposed algorithm can adapt to different motion states and different sensor mounting positions, and obtain an average accuracy rate above 95%.

·