Quantitative Evaluation Method Of Tactile Rendering Fidelity

In recent years,tactile reproduction technology has been gradually promoted and applied in VR scenes and terminals such as mobile phones,tablets,and people's expectations for its realistic rendering of real objects are also getting higher and higher.Tactile rendering fidelity is a research hotspot in human-computer interaction,and also an important indicator of tactile representation technology,which plays an important role in improving user immersion in human-computer interaction.At present,the evaluation of tactile rendering fidelity is divided into subjective evaluation and objective evaluation.Among them,the subjective evaluation work is mainly based on people's subjective feelings,susceptible to physiological,psychological,environmental and other factors,individual differences,and not from the theoretical analysis of the evaluation of tactile rendering fidelity.Objective evaluation is to compare the data set(force,acceleration,EEG signal,etc.)of the rendered object and the real object.It is mainly divided into two aspects.One is to compare the differences between the two data sets by direct observation,and only qualitative evaluation is carried out without quantitative evaluation indicators.The second is to use Euclidean distance and root mean square error to calculate the difference between two data sets of real and virtual objects to quantitatively evaluate tactile rendering fidelity.Although the evaluation index is proposed,this evaluation method is greatly affected by the mismatch between the positions of real and virtual data points.To address these problems,this paper uses force and acceleration to evaluate the tactile rendering fidelity of 3D shapes and textures respectively,and introduces kernel principal component analysis(KPCA)and principal component analysis(PCA)to study more accurate and universal quantitative evaluation methods of tactile rendering fidelity and validate it.The main innovations of this paper are as follows:(1)The tactile presentation of 3D sinusoidal bulge based on electrostatic force and subjective reality evaluation experiments were completed.The tangential force rendering algorithm based on IP mapping is simplified and a rendering algorithm based on gradient information is proposed.The specific steps of the method are as follows: the perceptual characteristics of the Electrostatic tactile representation device are studied,and the mapping model between the driving signal and the perceptual intensity of electrostatic friction is established;According to the change of tangential force received by fingertip sliding across the 3D object surface during the interaction process,the mapping relationship between perceived intensity and fingertip tangential force was established.According to the relationship between the fingertip tangential force and the 3D shape gradient information,the mapping relationship between the driving signal and the gradient information was established.The 3D sinusoidal bulge was presented on the electrostatic tactile representation device.The scoring results of twelve participants shows that the real bulge with a height of 1.5mm had the highest similarity with the virtual bulge among the five different heights.(2)A quantitative evaluation method of 3D shape rendering fidelity based on KPCA was proposed and verified by using 3D sinusoidal bump on an electrostatic tactile representation device.The validity of the proposed method was proved by comparing the quantitative evaluation results of tactile rendering fidelity with the subjective scoring results of realism.In this method,the two original tangential force sets of real and virtual 3D objects are mapped to the kernel space,and their kernel principal feature vectors were determined by calculating the cumulative contribution ratio of kernel features,and the quantitative evaluation index was obtained by calculating the spatial similarity of the two kernel principal features.The tangential force data of fingertips interacting with virtual 3D sinusoidal bulges and real 3D sinusoidal bulges of five different heights were recorded using the self-developed fingertip acceleration acquisition platform,and the quantitative realism scores between the virtual bumps and the real bumps of five heights were calculated using the method proposed in this paper.The results show that the real bump with a height of1.5mm has the highest similarity with the virtual bump rendered in this paper,which is consistent with the subjective subjective score of realism,which proves the effectiveness of the proposed method.(3)A quantitative evaluation method of texture rendering fidelity based on PCA is proposed and verified by using anisotropic texture on a pen vibration tactile representation device.The validity of the proposed method was proved by comparing the quantitative evaluation results of tactile rendering fidelity with the subjective scoring results of realism.In this method,two acceleration set of real and virtual texture mapped to feature space,and their kernel principal feature vectors were determined by calculating the cumulative contribution ratio of kernel features vectors,and by weighting each principal component according to the square root of its corresponding eigenvalue,the similarity between the two principal feature Spaces is calculated in more detail,and the quantitative evaluation index is obtained.The four anisotropic textures were haptically rendered on a pen-type vibrotactile reproduction device using each of the three rendering methods.The acceleration data of tool pen interaction with real and virtual texture were recorded by using the self-developed acceleration acquisition platform,and the quantitative realism scores between the four real textures and their corresponding virtual textures rendered by the three rendering methods were calculated using the method proposed in this paper.The results show that the tactile rendering fidelity of the bilinear interpolation method(WST)is significantly higher than that of the other two methods,which is consistent with the subjective score of realism,and proves the effectiveness of the proposed method.