面向多层建筑的人员室内运动状态识别方法

Indoor movement state recognition method for multi-storey buildings

针对人员运动状态识别在复杂室内环境下准确性及稳定性方面的不足,提出一种面向多层建筑的人员室内运动状态识别方法。对于上下电梯、上下楼梯等垂直方向上的运动状态,在传统惯性传感器数据的基础上,融入气压传感器数据。气压传感器可以有效表征室内人员在垂直方向上运动状态的特征,显著提升了人员室内在垂直方向上运动状态的识别准确率。使用滑窗理论提取的传感器数据动态时序特征能比瞬时数据更有效地表示人员运动时的过程,并利用多层神经网络充分提取了人员室内运动状态更高层次的特征,提升了算法的准确及稳定性。实验结果表明,该方法相比于仅使用惯性传感器的人员运动状态识别方法,识别率提高了约10%,垂直方向上的运动状态识别率提高了20%。

Considering the inadequacy of the accuracy and stability of personnel movement state recognition in complex indoor environment,we have proposed a method of personnel movement state recognition for multi-storey buildings.For the vertical movement state of elevators and persons up-down stairs,based on the traditional inertial sensor data,we integrated the air pressure sensor data,for the air pressure sensor can effectively characterize the indoor personnel movement state in the vertical direction which significantly improves the recognition accuracy of indoor personnel movement state in the vertical direction.The dynamic time series feature extracted by sliding window can more effectively represent the process of human movement than instantaneous data.The multi-layer neural network can fully extract the higher level features of human indoor motion state,which improves the accuracy and stability of the algorithm.The experimental results show that the recognition rate of this method is about 10%higher than that of the method using only inertial sensors,and that the recognition rate in vertical direction is 20%higher.