Yishi Wang

Generalized multi-ethnic face age-estimation

Age estimation from digital pictures of the face is a very promising research field that is now receiving wide attention. As with any good research problem, face age-estimation is wrought with many challenging interactions that cannot easily be separated out. In general, aging patterns are well understood for all humans, however, these patterns become confounded by intrinsic factors of genetics, gender differences, and ethnic deviations and, equally as important, extrinsic factors of the environment and behavior choices (i.e. sun exposure, drugs, cigarettes, etc). This novel work focuses on the development of a generalized multi-ethnic age-estimation technique — the first of its kind. In addition to the novelty of this approach, the systems overall performance measure (MAE) is “on par” with algorithms that are tuned for a specific ethnic group. Further, the proposed system performance proves to be far more stable across age than the best published results.

Statistical analysis of kerf mark measurements in bone.

Saw marks on bone have been routinely reported in dismemberment cases. When saw blade teeth contact bone and the bone is not completely sawed into two parts, bone fragments are removed forming a channel or kerf. Therefore, kerf width can approximate the thickness of the saw blade. The purpose of this study is to evaluate 100 saw kerf widths in bone produced by ten saw types to determine if a saw can be eliminated based on the kerf width. Five measurements were taken from each of the 100 saw kerfs to establish an average thickness for each kerf mark. Ten cuts were made on 10 sections of bovine bone, five with human-powered saws and five with mechanical-powered saws. The cuts were examined with a stereoscopic microscope utilizing digital camera measuring software. Two statistical cumulative logistic regression models were used to analyze the saw kerf data collected. In order to estimate the prediction error, repeated stratified cross-validation was applied in analyzing the kerf mark data. Based on the two statistical models used, 70–90% of the saws could be eliminated based on kerf width.

Face age estimation using model selection

Face age estimation is a difficult problem due to the dynamics of facial aging and its complex interactions owing to genetics and behavior factors. In this work we develop a robust age estimation system tuned by model selection that outperforms all prior systems on the FG-NET face database. We study various model selection methods systematically to determine the best selection methods among Least Angle Regression (LAR), Principle Component Analysis (PCA), and Locality Preserving Projections (LPP) for age estimation. Our performance analysis on PAL and FG-NET databases suggest that age estimation with LAR or LPP outperforms the full feature model. Furthermore, this work develops a novel operator named “graph age preserving” (GAP) to build a neighborhood graph for LPP for age estimation.

Gender classification from infants to seniors

Many believe that gender classification is a solved problem, however, gender classification for children is a very difficult problem that has not been adequately addressed by the research community. In this work we demonstrate this fact and present a system that performs gender classification on children that outperforms humans. Motivated by the significant improvement in model selection for age estimation [5], we investigate a robust gender classification system via model selection and evaluate the systems using leave-one-person-out cross-validation and 5-fold cross-validation schemes on FG-NET database. Furthermore, this work develops a novel operator, graph gender preserving, to build a neighborhood graph for locality preserving projection for gender classification.

Facial feature fusion and model selection for age estimation

Automatic face age estimation is challenging due to its complexity owing to genetic difference, behavior and environmental factors, the dynamics of facial aging between different individuals, etc. In this work we propose to fuse the global facial feature extracted from Active Appearance Model (AAM) and the local facial features extracted from Local Binary Pattern (LBP), as the representation of faces. Furthermore, we introduce an advanced age estimation system combining feature fusion and model selection schemes such as Least Angle Regression (LAR) and sequential approaches. Due to the fact that different facial feature representations may come with various types of measurement scales, we compare multiple normalization schemes for both facial features. We demonstrate that the feature fusion with model selection can achieve significant improvement in age estimation over single feature representation alone. Our experiment on multi-ethnicity UIUC-PAL database suggests that age estimation with feature fusion and model selection outperforms the single feature, or the full feature model.

Learning gabor features for facial age estimation

In this work we aim to study rigorously the facial age estimation in a multiethnic environment with 39 possible combination of four feature normalization methods, two simple feature fusion methods, two feature selection methods, and three face representation methods as Gabor, AAM and LBP. First, Gabor feature is extracted as facial representation for age estimation. Inspired by [3], we further fuse the global Active Appearance Model (AAM) and the local Gabor features as the representation of faces. Combining with feature selection schemes such as Least Angle Regression (LAR) and sequential selection, an advanced age estimation system is proposed on the fused features. Systematic comparative of 39 experiments demonstrate that (1) As a single facial representation, Gabor features surprisedly outperform LBP features or even AAM features. (2) With global/local feature fusion scheme, fused Gabor and AAM or fused LBP and AAM features can achieve significant improvement in age estimation over single feature representation alone.

Feature selection for improved automatic gender classification

In this paper, we demonstrate the need for dimensionality reduction to mitigate model overfitting on the nontrivial problem of gender classification from digital images. In (his study we explore four feature selection schemes using Genetic Algorithm, Memetic Algorithms, and Random Forest, which are fed to a nonlinear support vector machine (SVM) for final classification. The performance of the model (feature) selection approaches are evaluated against two distinct datasets of facial images: FG-NET which contains toddlers to seniors and the UIUC-PAL which contains faces of adults up to seniors. This work demonstrates that feature selection can, and does, improve performance of an SVM based gender classification system significantly.

Variance estimation of the Buckley–James estimator under discrete assumptions

The Buckley–James estimator (BJE) is a widely recognized approach in dealing with right-censored linear regression models. There have been a lot of discussions in the literature on the estimation of the BJE as well as its asymptotic distribution. So far, no simulation has been done to directly estimate the asymptotic variance of the BJE. Kong and Yu [Asymptotic distributions of the Buckley–James estimator under nonstandard conditions, Statist. Sinica 17 (2007), pp. 341–360] studied the asymptotic distribution under discontinuous assumptions. Based on their methodology, we recalculate and correct some missing terms in the expression of the asymptotic variance in Theorem 2 of their work. We propose an estimator of the standard deviation of the BJE by using plug-in estimators. The estimator is shown to be consistent. The performance of the estimator is accessed through simulation studies under discrete underline distributions. We further extend our studies to several continuous underline distributions through simulation. The estimator is also applied to a real medical data set. The simulation results suggest that our estimation is a good approximation to the true standard deviation with reference to the empirical standard deviation.

Eyebrow shape analysis by using a modified functional curve procrustes distance

To tackle the problem of automatic recognition of human eyebrow, a novel approach for shape analysis based on frontal face images is proposed in this paper. First, eyebrow curves are acquired by fitting cubic splines based on landmark points. Next, we propose to use a modified functional curve procrustes distance to measure the similarities among the cubic splines, and finally a multidimensional scaling method is adopted to evaluate the effectiveness of the distance. This work extends previous work in analyzing the eyebrow for both human and machine recognition by providing a framework based on shape contours. Further this work demonstrates the effectiveness of eyebrow shape for discrimination when teamed with the appropriate metric distance.

Sensitivity analysis with cross-validation for feature selection and manifold learning

The performance of a learning algorithm is usually measured in terms of prediction error. It is important to choose an appropriate estimator of the prediction error. This paper analyzes the statistical properties of the K-fold cross-validation prediction error estimator. It investigates how to compare two algorithms statistically. It also analyzes the sensitivity to the changes in the training/test set. Our main contribution is to experimentally study the statistical property of repeated cross-validation to stabilize the prediction error estimation, and thus to reduce the variance of the prediction error estimator. Our simulation results provide an empirical evidence to this conclusion. The experimental study has been performed on PAL dataset for age estimation task.