| Smart wearables with various functions are embedded in various domains in our daily life.Wearing comfort rather than the basic function of these products is imperative to the consumers’ demands,which requires a close and accurate fit to the body of the users.However,the complexity of the external ear brings great challenges to the design of ear related wearables,the proper fit of which is essential to their function and comfort.In this paper,a K-means++ clustering algorithm was performed for the classification of the external ear based on the statistical shape models that take account of full shape information of concha cavum and external ear canal,aiming to investigate the ear shape variation and provide accurate reference for the ergonomic design of TWS earphones.The paper is organized as follows:(1)A template-based registration method of the external ear was proposed,through which 1157 shape models with corresponding orientation and points of ear concha and external ear canal were generated.In this method,a rigid transformation of all samples was first performed to filter out the location and rotational effects.Next,a template with proper shape and size was selected and reconstructed,with 6657 vertices distributed uniformly on the surface,which consists of 3D information of the concha cavum and external ear canal.A nonrigid registration was then performed in two steps,using the linear blending skinning technology with biharmonic bounded weights and non-rigid iterative closest point algorithm.This dense point-to-point correspondence in the same coordinate system enables the statistical analysis of 3D ear shape.(2)A collection of SSM was established to capture the morphological ear shape variability present in Chinese population,by applying a principal component analysis(PCA)on these corresponded ear models.15 principal components(PC)were extracted with 90% cumulative contribution rate to represent the shape information in the subsequent clustering process instead of the high-dimensional coordinates.SSM were calculated accordingly using the 4 top ranked PC eigenvalues and average ear shape.The variation mode was explained from both qualitative and quantitative perspectives using deviation analysis with the mean shape and correlation analysis with the anthropometric measurements.The statistical analysis of external ear shape results in a complete and accurate representation of the local and global shape variation,as well as provides the deformable and extreme ear models for the design of ear-related wearables.(3)K-means++ algorithm was used to perform cluster analysis based on the 15 PC scores of each individual.According to the clustering validity index DBI and SSE together with the needs of large-scale manufacturing of ear-worn products,a total of 1157 samples of Chinese adults aged 18-65 were divided into 4 categories.Representative 3D model and anthropometric measurements of each cluster were generated through the first five nearest models with the corresponding cluster centers.The reliability of the classification result was verified by the deviation analysis of models within-group and between-group.(4)The design of in-ear earphone is taken as an example using the results of morphological research of external ear.According to the characteristics of in-ear earphones,a targeted design research was conducted.Firstly,the ear canal entrance of the average model were extracted for fitting and data analysis,which provides a standard and basis for the earplug design of in-ear earphone;then the representative model of clustering and SSM were used as reference to carry out in-ear earphone optimization design and wearing experiments.This paper is the first study to analyze the external ear shape variation and classification considering the full 3D shape information of concha cavum and external ear canal.Besides,the sample size and sampling accuracy of the models ensure the representativeness of the Chinese population.The results have broad applicability in TWS earphone design. |
PDF Full Text Request