Research On Smart Wearable Sensing Device Based On Flexible Sensors

With the rapid development of Internet of things,as well as the continuous innovation of technologies such as big data and artificial intelligence,people are paying more attention on their own vital signs,sports status and surrounding environment,so the real-time acquisition of such data by wearable devices has become a lifestyle that people are keen on.Wearable devices with the advantages of portability and multifunction are widely used in our life,such as smart bracelet,smart watch.In addition,wearable devices also play an important role in fields of medical health monitoring,human-computer interaction.Therefore,sensors which is the core of wearable devices have been widely concerned in scientific research and commercial industry.At present,wearable devices mostly use hard MEMS sensors as sensing units,so there are some disadvantages such as poor wearing comfort and high cost.In recent years,flexible sensors which are light and low-cost have become a research hotspot,and it is expected to solve the shortcomings of traditional wearable devices in wearing experience when it is applied to wearable devices.However,the current research focuses on the sensing performance improvement of flexible sensors,which lacks the exploration and display of its practical application in wearable devices.Therefore,this study is focused on the design and manufacture of smart wearable devices with flexible sensors as the core sensor unit.In detail,according to the specific application scheme,a comfortable and low-cost flexible data glove is designed and fabricated based on the flexible sensor developed by our research group.At the same time,the characteristics of the sensor data obtained from the flexible sensor used for monitoring finger joint bending motion are analyzed,so as to explore the applicable data processing method and algorithm model.Finally,the data glove is used to realize accurate static gesture recognition and dynamic gesture recognition respectively,which makes the device intelligent.The relevant work and achievements of this paper are as follows:(1)We prepared a rGO-coated fiber sensor with stretchability,good sensing performance and low cost by effectively coating reduced graphene oxide(rGO)on a kind of double covering yam.Testing results show that the rGO-coated fiber sensor has good sensing stability,linearity and repeatability with the tensile strain up to 20%,and can make rapid and accurate response to the bending of finger joint,with an average angle error of 8.37°,so we considered that it is suitable as the sensing unit of data glove.(2)By sewing ten rGO-coated fiber sensors on the surface of a textile glove,with sensors intimately contacting proximal interphalangeal joint and metacarpophalangeal joint of five fingers respectively,we designed and fabricated a type of flexible data glove which can monitor the bending motion of the finger joint and possesses the distinct advantages of simplicity,low cost and good wearing comfort.And the adaptive program and the weighted filtering algorithm were proposed,which can effectively reduce the data fluctuation generated by the rGO-coated fiber in the process of monitoring the finger motion and solve the problems such as hand pattern matching and the difference of initial resistance of the sensor.(3)We accurately identified 10 static gestures representing the numbers "0"-"9" via the fabricated data glove,and the recognition accuracy of the model based on BP neural network was 98.5%in 2000 test samples.Experimental results show that rGO-coated fiber sensor can effectively collect the features of static gestures,and the DBSCAN algorithm can be used to filter outliers in the data set,so as to reduce the negative impact of unstable performance of fiber sensor in long-term use,ensuring a high recognition accuracy.(4)We used the fabricated data glove to accurately identified 9 dynamic gestures selected from common sign language,and the recognition accuracy of the model based on DTW algorithm achieved 98.3%in 180 test samples.Experimental result show that,the fabricated data glove has good real-time response ability and repeatability,which can accurately and quickly monitor the important features of dynamic gestures in the whole process of execution and will not cause serious delay and deviation due to the large changes of gestures,so it has strong practicability in dynamic scenes.