Research On Device-free Unarmed Fitness Activity Recognition And Assessment With Commodity RFID

With the development of the domestic economy and the country's advocacy of “national fitness programme”,the idea of “harvesting health by exercising” has become more and more popular.As a result,an increasing number of people are joining the fitness team.Compared with other fitness methods,unarmed fitness is favored by more and more people because it sets no limit to age,gender or physical strength,and requires no additional equipment.However,for ordinary unarmed exercisers,due to the lack of corresponding supervision and assessment feedback,many fitness workers are unaware of long-term,erroneous exercise.In this way,not only can they fail to achieve the expected training goals,but may cause serious physical injuries.The main cause of this problem is the lack of corresponding low-cost and automated methods for identifying and assessing unarmed fitness activities.In this regard,this paper proposes a research of recognizing and assessing unarmed fitness activities based on commodity device-free RFID equipment.The key research contents is analyzed from the following tow aspects:1)Fingerprint library set up with low deployment cost on low-cost devices.This paper proposes a multi-channel approach to obtain phase information that is least affected by environmental multipath effects.Then the method calculates the corresponding angular frequency characteristics for the establishment of the fingerprint library through the phase information of the signal.This reduces the impact of environmental multipath and reduces deployment costs.2)To achieve efficient and reliable activity identification and assessment.This paper designs an active segmentation method with different granularities to improve the efficiency of active segmentation,and normalizes the existing fast DTW algorithm to reduce the impact of activity speed.Finally led to a quantitative assessment of each unarmed fitness exercise in groups.Finally,a series of experiments under three scenarios(gym,office,home scene)were used to verify the method.Experimental results show that the true positive rates of motion recognition in three different scenarios are: 91.8%,89.7%,and 89.1%,respectively.Thus,it can be seen that the method can effectively identify unarmed fitness activities on the premise of low-cost deployment costs,and provide valuable evaluation feedback for unarmed fitness trainers.