Research On Texture Perception Based On Intensive Mechanical Information Of Tongue Imitating Distributed Mechanical Detection Device

Tongue is the main receptor in human oral cavity,which plays an important role in the process of food taste evaluation.It can not only sense the taste information of food,such as acid,sweet,bitter and salty,but also judge some texture characteristics of food by sensing the distributed biomechanical information produced on its surface when contacting with food.The existing electronic tongue system based on multi-sensor to simulate the mechanism of human tongue taste perception has realized the detection of food taste information.There is less research on the mechanical information of tongue touch when human tongue contacts with food.At present,some researchers study human pronunciation by constructing the corresponding tongue model.Some scholars also made a tongue model and combined with flexible array pressure sensor to explore the mechanical information of the tongue when touching food.The flexible array pressure sensor detects the mechanical information through the array of sensing sites,which is very similar to the mechanism of human tongue receptor cells to feel the texture of food.Therefore,it is reasonable and reliable to apply the flexible distributed membrane pressure sensor to simulate the flexible surface of tongue to detect the distributed mechanical information.However,the current flexible distributed thin-film pressure sensor has less mechanical information points due to the large spacing between the transverse and longitudinal arrays,which leads to the lack of measurement accuracy and the lack of comprehensive and specific reflection of mechanical information.In view of this,this paper analyzes the performance of the sensor and the characteristics of the distributed experimental data based on the contact experiment of the bionic-tongue mechanical detection device on the experimental sample.The bilinear interpolation is used to interpolate the experimental data,and the sausage pressure contact mechanical simulation model is constructed,which realizes the solution process from sparse mechanical information to dense mechanical information,and explores the relationship between stress and curvature,and the sausage texture is classified by using sparse grid data and dense grid data,which verifies the feasibility of improving the detection accuracy of sensor and providing more texture mechanics information by thickening grid data.Firstly,according to the physiological characteristics and the main movement patterns of human tongue during chewing,the corresponding parameters of tongue surface were extracted and the attributes were set,and the mechanical simulation model of indenter-food contact was established to realize the in vitro simulation of human eating process.Secondly,the distributed pressure information detected by the sensor of the bionic tongue mechanical detection device was analyzed and screened,which was used as the pressure source of the simulation model to load the sausage,and the corresponding pressure distributed cloud image and the mechanical information of each point of the tongue contact mechanical simulation model were obtained.Then,the bilinear interpolation is used to thicken the pressure data obtained by the bionic tongue mechanical detection device,and the interpolated pressure data is obtained.The thickened pressure data is compared with the actual pressure data of the point,and the similarity between the two is analyzed,so that the sparse mechanical information can be solved to the dense mechanical information.The process of solving spatial mechanical information from sparse to dense,that is,the process of mechanical simulation model mesh changing from large to small,from sparse to dense.In this process,mechanical information points will become more and more.Through the division of sparse mesh and thicken mesh,a reference value for the improvement of sensor accuracy is provided.Finally,according to the characteristics of distributed pressure data,the evaluation indices of sausage hardness and elastic texture are established.Based on bionic tongue device and simulation model,five kinds of sausages are pressed and touched respectively.Support vector machine(SVM)and random forest(RF)are applied to pattern recognition analysis of feature data in sparse grid and dense grid,and the results of the two cases are compared with those of artificial sensory evaluation.The results show that the dense grid experiment results have a higher correlation with the artificial senses,which indicates that the accuracy of sausage classification can be improved by thickening the grid through data interpolation,and the results are closer to the perception of human oral tongue.