Research On Realism Assessment Methodology For Haptic Rendering Based On Haptic Perception Mechanism

Haptic rendering technology provides various haptic feedback in virtual environments through haptic devices.Since haptic rendering technology is being increasingly used in computer games,surgical simulators,mobile phones etc.,there is a need for defining standards for evaluating rendering performance of haptic applications.The objective assessment methodology for haptic rendering is to describe the performance of rendering effect qualitatively and quantitatively through selected parameters and metrics based on comparing real and virtual haptic interaction.Present assessment methodologies have various shortcomings such as insufficient index,complicated procedure and ambiguous evaluating results.The main problems leading to these results are lack of considering haptic perception characteristics,as well as the interaction with objects.In order to realize the objective assessment taking haptic perceptual mechanism in to consideration,this article focuses on the key problems of acquirement of source,building of haptic perceptual model and construction of metrics during the assessment process.The main work and innovation points are as follows:(1)An objective assessment methodology for haptic rendering based on the perception mechanism is proposed in this article.The basic idea is using perceptual models of objective stimuli inputs and subjective perceptual results based on the perception mechanism to perceive and assessment the haptic realism instead of human.Testing results show that,the assessment results of our method have similar variation trend with assessment results from subjects and smaller forecast error and standard deviation,which indicates that our method has better repeatability and stability,and closer to human subjective feelings.(2)The methods of quantitative description of inputs and outputs for the perception model are studied.For inputs,interactive data are recorded and processed to construct the “haptic image” to describe the dynamic haptic interaction.Steps of standardized data acquisition process,and perceptual filtering are conducted to extract effective features based on perceptual characteristics as model inputs.For outputs,the method of mapping stimuli into perceptual space is proposed to describe subjective results completely and quantitatively.Basic dimensions and their relationships are analyzed to determine the distribution of stimuli in perception space as model outputs.Quantitative description and analysis of input and output sets improves the problem that input and output sets are difficult to normalize and quantize,which affects the prediction performance of perception model.(3)Perceptual models are built to estimate subjective perceptual results from objective stimuli,taking the texture perception as an example.Compared with the multilayer regression and other similar machine learning models,the machine learning model based on sparse expression and SVM in this article is built with extra steps of supervised feature extraction,consideration of derived sequences and screen of dictionaries and features.Testing on the model shows better performance on estimating perceptual results.(4)A method of constructing assessing metrics based on four basic psychophysical experimental types and corresponding measuring parameters is proposed to assess the realism.The rendering and perception of texture surfaces are taking as example to implement the whole process of the assessment method proposed in this paper.Compared with subjective evaluation,results of our method have better consistency,higher correlation coefficient,smaller mean square error and average absolute error.The evaluation cost is shortened by nearly 80%,which is more efficient,and the standard deviation of the evaluation result is greatly reduced and more stable.Results of this paper provide new ideas for the study of realism assessment methodology,have important theoretical significance and practical value.