Anthropomorphic Classification Of Tactile Qualities Of Fabrics Based On Skin/Textile Friction-induced Vibrations

With the development of technology and the improvement of living standard,people’s demands for textiles are not limited in taking covers and keeping warm.Today, more attention is paid on wear comfort. And the wear comfort is decided bythe tactile quality of textiles to a large extent. In daily life, consumers often choose thesuitable garment by touching it, and this process is actually a classification of fabrictactile quality.Fabric tactile textures are human’s sensory responses to the interaction betweenhands and fibers or fabrics. The existing classification methods mainly include KESfabric evaluation system, FAST garment instrument and PhabrOmeter fabric handevaluation system. Essentially, the above three methods are based on the basicphysical and mechanical properties of fabrics. By using statistic methods, the statisticrelationship is obtained between the basic physical and mechanical qualities andfabric tactile qualities such as drape, crease resistance and pilling, eventuallycontributing to the classification of fabric tactile qualities.However, the existing methods are lack of universality. That’s mainly because ofthe neglect of physiological processes and the essential meaning of tactile textureproduction. When human skin slides across fabric surfaces, the friction interactionbetween fabrics and skin will occur and trigger the cutaneous tactile receptors, whichare responsible for perceived tactile sensation. Surveys have shown that differentsurface textures create multi-frequency vibration during finger sliding, which is therelevant stimulus to trigger the cutaneous tactile receptors. There are two mechanismsabout tactile texture coding: temporal coding and spatial encoding mechanism. Theyare responsible for encoding stimulation signals with different frequencies and phases,respectively.Given the physiological processes producing tactile texture, it is of vitalimportance to study the multi-frequency vibration signals. As fabric is a flexible fiberassembly, has rich surface texture, and composed of hairiness, yarn and fabric.Complex deformation will occur during the interaction with fingers. At the same time,the skin itself will have viscoelastic deformation, which will not be characterized bythe pure fabric surface properties parameters. As a result, the mechanical and physicalqualities of fabric cannot represent human tactile sensation. In the contact interactionof finger and fabrics during the sliding, different fabric qualities generated various vibration signals. These signals not only include the information about textures andmaterials, but also reflect the interaction between fingers and fabrics. Therefore,tactile quality classification by the vibration signal will be closer to human’s realfeelings.Hence, this paper is on the basis of the amount of physiological stimuli on thecutaneous tactile receptors, seeking for an objective classification method of fabrictactile. Detailed research work is as follows:(1) Record the vibration signal during the interaction between finger and fabric,and extract the feature space by the response characteristics of cutaneous tactilereceptor to vibration stimuli. Analyses show that the extracted feature space canefficiently characterize the vibration stimuli received by finger.Four groups of woven fabrics by typical construction parameter are chosen as thetest samples for tactile classification. The contact vibration signal during fingersliding across the sample is measured at a constant speed. After subtracting the noisesignal, Fast Fourier Transform (FFT) is applied to turn the data into power spectrumin frequency domain. Different cutaneous tactile receptors are sensitive to differentfrequencies of vibration stimuli, and each one has a preference stimulation threshold.Here we subdivide the frequency range0-500Hz, and they are classified into16frequency bands. Then the area between power spectrum and threshold curve of eachband is calculated, and it is defined as the vibration intensity value. In order to avoidthe information redundancy and enhance the efficiency, Principle ComponentAnalysis(PCA) is used on vibration intensity values to reduce dimension. Theresulting feature space is5dimensions to represent mechanical stimulus generatingfabric tactile sensation.(2) By setting the extracted feature spaces of vibration stimuli as independentvariables, fabric tactile textures are classified by different supervised learningalgorithms. The result showed that fabric tactile quality can be classified effectivelyby the extracted feature space of vibration stimuli. And Radial Basis Function NeuralNetwork has highest efficiency among three chosen supervised learning algorithms.The extracted feature with5dimensions are used as inputs, the Cluster Analysis,the K-Nearest Neighbor Classification, Multilayer Perceptron Neural Network andRadial Basis Function Neural Network are employed to classify the roughness andsoftness of fabric, respectively. And meanwhile, sensory evaluation is conducted andcompared with the accuracy of the objective classification method. Analysis showsthat the classification result of the Clustering Analysis is the least consistent with sensory evaluation test, for its simple data processing. And Radial Basis FunctionNeural Network gets the most accurate result close to human sense for its mimicryabout human mind.(3) To ensure the feasibility and reliability of the proposed objectiveclassification method in this study, another group of woven fabric samples are tested.After comparing the classification correction of sensory evaluation with the newestablished classification method, the classification of roughness is robust, and theclassification of softness needs to be improved.The tactile evaluation test and the objective classification are performed on agroup of randomly chosen woven fabrics with different surface properties. The resultsshowed that, for the same subject, the classification results of both roughness andsoftness are quite close to those of sensory evaluation. For different subjects, theclassifications of roughness sensation are of high consistency, and the classification ofsoftness has little difference. The difference is due to the way perceiving roughness,which is not so efficient to extract all the information responsible for tactile softnesssensation.Above all, when the contact vibration signal induced by the interaction of fingerand fabric is combined with the physiological mechanism of human producing tactile,it is feasible to extract the stimulus feature space responsible for tactile textures offabric. As for the classification algorithm, the Radial Basis Function Neural Networkis able to achieve the most accurate results, which are close to sensory evaluation. Yet,the robustness of the newly established classification method among different peopleis needed to be improved. Specifically, it matters a lot of the difference of individualfingerprints.