Multi-floor indoor pedestrian pattern recognition based on improved LSTM network
Yaxin Li, Guoyu Liu, Zeping Rui · 2025
The appearance of many auxiliary devices in multi-story buildings enriches the movement mode of indoor pedestrians, making the object of indoor pedestrian pattern recognition not only limited to the common three modes of going up, down and walking, but also taking elevator, escalator and automatic sidewalk are new indoor pedestrian movement modes that need to be taken into consideration. In order to recognize the movement pattern of indoor pedestrians in multi-story buildings as precise comprehensively as possible, this paper proposes a new method of movement pattern recognition based on smart phone. Firstly, the relationship between the indoor pedestrian movement pattern and the changes of air pressure, acceleration data obtained by mobile phone was analyzed, and then nine movement patterns including walking, going up and down stairs, lift, escalator and automatic sidewalk were identified by the CNN-LSTM-BO after Bayesian optimization. The experimental results show that the accuracy can reach 0.96.