Research On Sitting Posture Recognition With Commodity RFID

Sitting posture has a close relationship with our health,and poor sitting postures often inevitably increase the risk of modern health musculoskeletal disorders.Therefore,keeping healthy sitting posture and monitoring the sitting posture for long-term and effective monitoring are essential.Previous works either used a camera to record the image or attached wearable sensors on human body to recognize sitting postures.However,video-base approaches may face privacy issue while the wearable sensor-based approaches may cause uncomfortable to the user,so it is not suitable for long-term monitoring.Aiming at the problems of traditional sitting posture monitoring methods,this paper proposes a sitting posture recognition system based on commercial Radio Frequency Identification(RFID).This paper can successfully recognize seven habitual sitting postures with just three lightweight and low-cost RFID tags pasted to user's back.The design of this paper exploits the correlation between the phase change of RFID tags and the sitting postures.Recognition of the sitting posture through the disturbance characteristics caused by different sitting postures on the wireless signal,there is no problem of leaking privacy and affecting user experience.In this paper,the relationship between the phase change of the RF signal and the sitting posture is established,and deployment of tag is optimized to achieve the extraction of high-quality signal features related to the sitting posture,thereby achieving high-precision,and robust sitting posture recognition.The research contents and main contributions of this paper are as follows:1)This paper proposes a commercial RFID-based sitting posture recognition system,which guides the optimal design of tag deployment through the temporal and spatial propagation laws of wireless signals to achieve improved robustness and accuracy of sitting posture recognition.2)This paper proposes an effective feature extraction method for sitting posture based on radio frequency signals.This method combines the features affected by the sitting posture and the features affected by the breathing pattern in the signal phase information to perform sitting posture recognition.It is not affected by the daily small interference activities of the human body and is highly robust.3)In this paper,under the real environment,the system is tested and evaluated in terms of effectiveness and robustness.In this paper,we conducted extensive experiments and evaluations on seven habitual sitting postures in different multi-path scenarios.The results show that this paper can recognize human's sitting posture at an average accuracy of 98.83%.Besides,we evaluate the performance on different antenna gains,the material of the chair,thickness of clothes and the number of tags to validate the system robustness,which can get more than the accuracy of 98%.