Kathleen M. Robinette

Gender Recognition Using 3-D Human Body Shapes

Gender recognition has important applications in identity recognition, demographic survey, and human-computer interaction systems. In the past, gender recognition was based on 2-D images or videos, which has many limitations and disadvantages, such as low accuracy and sensitivity to the viewpoint of the camera and lighting conditions. In this paper, we investigate gender recognition using 3-D human body shapes. The 3-D human body shapes used for gender recognition were obtained by laser scanning. Different machine-learning algorithms and feature-extraction methods are investigated and analyzed in this paper. Experimental results show that the support vector machine (SVM) is the best classification algorithm, and the features represented using distributions of normals are very effective for gender recognition. Furthermore, Fourier descriptor (FD) is a robust method to analyze the breast regions and has great potential applications in 3-D human-body-shape-based biometrics. The research demonstrates that our shape-based gender recognition has achieved a very high recognition rate. The techniques provide effective ways for gender recognition and overcome some limitations in 2-D technologies.

Anthropometry for HMD design

The importance of fit for helmet ensembles is not limited to just comfort. It impacts most other safety and performance needs of the helmets, including helmet retention, and optical and acoustical performance. The addition of optical systems to helmet ensembles increases the need for precision in fit. Helmet systems which were previously acceptable in terms of fit do not necessarily fit well enough to accommodate new performance requirements. The increased need for precision has introduced the need for better definition of human anthropometry for helmet design as well as definition of the head and helmet interface. Traditional anthropometry (human body measurements taken with calipers, or head boards, etc.) is no longer adequate. For advanced helmet systems, data on the shape, or change in the surface curvature and how this relates to helmet systems in three-dimensional space, is now a necessity. In fact, use of the old style of anthropometry can and has created problems rather than resolve them. This paper discusses some of the problems with the old methods and introduces new technologies and research which is being done to address the needs.

Gender recognition with limited feature points from 3-D human body shapes

In this paper, we investigate the possibility of using limited feature points (shape landmarks) from 3-D human body shapes to recognize the gender of human beings. Several machine learning algorithms and feature extraction algorithms (principal component analysis and linear discriminant analysis) are investigated and analyzed in this paper. Experimental results on a large dataset containing 2484 3-D shape models show that limited feature points (shape landmarks) can be used for gender recognition and can achieve high recognition rate, which provides a fast gender recognition technique. The research provides a potential research direction for gender recognition.

Static and Dynamic Human Shape Modeling

Recent developments in static human shape modeling based on range scan data and dynamic human shape modeling from video imagery are reviewed. The topics discussed include shape description, surface registration, hole filling, shape characterization, and shape reconstruction for static modeling and pose identification, skeleton modeling, shape deformation, motion tracking, dynamic shape capture and reconstruction, and animation for dynamic modeling. A new method for human shape modeling is introduced.

Saccadic eyes recognition using 3-D shape data from a 3-D near infrared sensor

Saccadic eyes are important human behaviors and have important applications in commercial and security fields. In this paper, we focus on saccadic eyes recognition from 3-D shape data acquired from a 3-D near infrared sensor. Two salient features, normal vectors of meshes and curvatures of surfaces, are extracted. The distributions of normal vectors and curvatures are computed to represent eye states. The support vector machine (SVM) is applied to classify eyes states into saccadic and non-saccadic eyes states. To verify the proposed method, we performed three groups of experiments using different strategies for samples selected from 300 3-D data, and present experimental results that demonstrate the effectiveness and robustness of the proposed algorithm.

Dynamic human shape description and characterisation

Dynamic human shape description and characterisation was investigated in this paper. The dynamic shapes of four subjects in five activities (digging, jogging, limping, throwing, and walking) were created via 3D motion replication. The Paquet shape descriptor (PSD) was used to describe subject shapes in each frame. The unique features of dynamic human shapes were revealed from the observations of the 3D plots of PSDs. The principal component analysis was performed on the calculated PSDs and principal components (PCs) were used to characterise PSDs. The PSD calculation was then reasonably approximated by its significant projections in the eigenspace formed by PCs and represented by the corresponding projection coefficients. As such, the dynamic human shapes for each activity were described by these projection coefficients. To demonstrate potential applications of shape dynamics, a preliminary study was performed on using projection coefficients for activity classification.

AN XML-BASED NETWORKING METHOD FOR CONNECTING DISTRIBUTED ANTHROPOMETRIC DATABASES †

Anthropometric data are used by numerous types of organizations for health evaluation, ergonomics, apparel sizing, fitness training, and many other applications. Data have been collected and stored in electronic databases since at least the 1940s. These databases are owned by many organizations around the world. In addition, the anthropometric studies stored in these databases often employ different standards, terminology, procedures, or measurement sets. To promote the use and sharing of these databases, the World Engineering Anthropometry Resources (WEAR) group was formed and tasked with the integration and publishing of member resources. It is easy to see that organizing worldwide anthropometric data into a single database architecture could be a daunting and expensive undertaking. The challenges of WEAR integration reflect mainly in the areas of distributed and disparate data, different standards and formats, independent memberships, and limited development resources. Fortunately, XML schema and web services provide an alternative method for networking databases, referred to as the Loosely Coupled WEAR Integration. A standard XML schema can be defined and used as a type of Rosetta stone to translate the anthropometric data into a universal format, and a web services system can be set up to link the databases to one another. In this way, the originators of the data can keep their data locally along with their own data management system and user interface, but their data can be searched and accessed as part of the larger data network, and even combined with the data of others. This paper will identify requirements for WEAR integration, review XML as the universal format, review different integration approaches, and propose a hybrid web services/data mart solution.

An Alternative 3-D Shape Descriptor for Database Mining

This research examines and compares methods for mathematically coding 3-D human shapes for shape searching in large databases. The mathematical coding is called shape description and the codes themselves are called shape descriptors. The measure of effectiveness of the shape descriptors is the ability to locate a different copy of the same person from amongst a database of thousands. The hypothesis for this study is that the automated method for deriving a compact shape descriptor developed by Paquet and Rioux (6), referred to henceforth as the Paquet Shape Descriptor (PSD), is an objective shape descriptor that can distinguish between some groups of shapes as well or better than distance type body measurement methods. The present research suggests that PSD excels in searches when body contours are the most important criterion, and offers a viable alternative when key measurement or landmark data are not available. When underlying skeletal structure is more important than surface contours, other search methods may still be preferable when available.