Dynamic biometrics fusion at feature level for video-based human recognition

This paper proposes a novel human recognition method in video, which combines human face and gait traits
using a dynamic multi-modal biometrics fusion scheme. The Fisherface approach is adopted to extract face
features, while for gait features, Locality Preserving Projection (LPP) is used to achieve low-dimensional
manifold embedding of the temporal silhouette data derived from image sequences. Face and gait features are
fused dynamically at feature level based on a distance-driven fusion method. Encouraging experimental results
are achieved on the video sequences containing 20 people, which show that dynamically fused features produce
a more discriminating power than any individual biometric as well as integrated features built on common static
fusion schemes.

Document clustering in correlation similarity measure space

This paper presents a new spectral clustering method called correlation preserving indexing (CPI), which is performed in the correlation similarity measure space. In this framework, the documents are projected into a low-dimensional semantic space in which the correlations between the documents in the local patches are maximized while the correlations between the documents outside these patches are minimized simultaneously. Since the intrinsic geometrical structure of the document space is often embedded in the similarities between the documents, correlation as a similarity measure is more suitable for detecting the intrinsic geometrical structure of the document space than euclidean distance. Consequently, the proposed CPI method can effectively discover the intrinsic structures embedded in high-dimensional document space. The effectiveness of the new method is demonstrated by extensive experiments conducted on various data sets and by comparison with existing document clustering methods.

Fast growing self organizing map for text clustering

This paper presents an integration of a novel document vector representation technique and a novel Growing Self Organizing Process. In this new approach, documents are represented as a low dimensional vector, which is composed of the indices and weights derived from the keywords of the document.

An index based similarity calculation method is employed on this low dimensional feature space and the growing self organizing process is modified to comply with the new feature representation model.

The initial experiments show that this novel integration outperforms the state-of-the-art Self Organizing Map based techniques of text clustering in terms of its efficiency while preserving the same accuracy level.

Face recognition with image sets using manifold density divergence

In many automatic face recognition applications, a set of a person's face images is available rather than a single image. In this paper, we describe a novel method for face recognition using image sets. We propose a flexible, semi-parametric model for learning probability densities confined to highly non-linear but intrinsically low-dimensional manifolds. The model leads to a statistical formulation of the recognition problem in terms of minimizing the divergence between densities estimated on these manifolds. The proposed method is evaluated on a large data set, acquired in realistic imaging conditions with severe illumination variation. Our algorithm is shown to match the best and outperform other state-of-the-art algorithms in the literature, achieving 94% recognition rate on average.

An information-theoretic approach to face recognition from face motion manifolds

In this work, we consider face recognition from face motion manifolds (FMMs). The use of the resistor-average distance (RAD) as a dissimilarity measure between densities confined to FMMs is motivated in the proposed information-theoretic approach to modelling face appearance. We introduce a kernel-based algorithm that makes use of the simplicity of the closed-form expression for RAD between two Gaussian densities, while allowing for modelling of complex and nonlinear, but intrinsically low-dimensional manifolds. Additionally, it is shown how geodesically local FMM structure can be modelled, naturally leading to a stochastic algorithm for generalizing to unseen modes of data variation. Recognition performance of our method is demonstrated experimentally and is shown to exceed that of state-of-the-art algorithms. Recognition rate of 98% was achieved on a database of 100 people under varying illumination

Modelling multilevel data in multimedia : A hierarchical factor analysis approach

Multimedia content understanding research requires rigorous approach to deal with the complexity of the data. At the crux of this problem is the method to deal with multilevel data whose structure exists at multiple scales and across data sources. A common example is modeling tags jointly with images to improve retrieval, classification and tag recommendation. Associated contextual observation, such as metadata, is rich that can be exploited for content analysis. A major challenge is the need for a principal approach to systematically incorporate associated media with the primary data source of interest. Taking a factor modeling approach, we propose a framework that can discover low-dimensional structures for a primary data source together with other associated information. We cast this task as a subspace learning problem under the framework of Bayesian nonparametrics and thus the subspace dimensionality and the number of clusters are automatically learnt from data instead of setting these parameters a priori. Using Beta processes as the building block, we construct random measures in a hierarchical structure to generate multiple data sources and capture their shared statistical at the same time. The model parameters are inferred efficiently using a novel combination of Gibbs and slice sampling. We demonstrate the applicability of the proposed model in three applications: image retrieval, automatic tag recommendation and image classification. Experiments using two real-world datasets show that our approach outperforms various state-of-the-art related methods.

Bayesian nonparametric multilevel clustering with group-level contexts

We present a Bayesian nonparametric framework for multilevel clustering which utilizes group- level context information to simultaneously discover low-dimensional structures of the group contents and partitions groups into clusters. Using the Dirichlet process as the building block, our model constructs a product base-measure with a nested structure to accommodate content and context observations at multiple levels. The proposed model possesses properties that link the nested Dinchiet processes (nDP) and the Dirichlet process mixture models (DPM) in an interesting way: integrating out all contents results in the DPM over contexts, whereas integrating out group-specific contexts results in the nDP mixture over content variables. We provide a Polyaurn view of the model and an efficient collapsed Gibbs inference procedure. Extensive experiments on real-world datasets demonstrate the advantage of utilizing context information via our model in both text and image domains.

Visual object clustering via mixed-norm regularization

Many vision problems deal with high-dimensional data, such as motion segmentation and face clustering. However, these high-dimensional data usually lie in a low-dimensional structure. Sparse representation is a powerful principle for solving a number of clustering problems with high-dimensional data. This principle is motivated from an ideal modeling of data points according to linear algebra theory. However, real data in computer vision are unlikely to follow the ideal model perfectly. In this paper, we exploit the mixed norm regularization for sparse subspace clustering. This regularization term is a convex combination of the l1norm, which promotes sparsity at the individual level and the block norm l2/1 which promotes group sparsity. Combining these powerful regularization terms will provide a more accurate modeling, subsequently leading to a better solution for the affinity matrix used in sparse subspace clustering. This could help us achieve better performance on motion segmentation and face clustering problems. This formulation also caters for different types of data corruptions. We derive a provably convergent algorithm based on the alternating direction method of multipliers (ADMM) framework, which is computationally efficient, to solve the formulation. We demonstrate that this formulation outperforms other state-of-arts on both motion segmentation and face clustering.

Learning vector representation of medical objects via EMR-driven nonnegative restricted Boltzmann machines (eNRBM)

Electronic medical record (EMR) offers promises for novel analytics. However, manual feature engineering from EMR is labor intensive because EMR is complex - it contains temporal, mixed-type and multimodal data packed in irregular episodes. We present a computational framework to harness EMR with minimal human supervision via restricted Boltzmann machine (RBM). The framework derives a new representation of medical objects by embedding them in a low-dimensional vector space. This new representation facilitates algebraic and statistical manipulations such as projection onto 2D plane (thereby offering intuitive visualization), object grouping (hence enabling automated phenotyping), and risk stratification. To enhance model interpretability, we introduced two constraints into model parameters: (a) nonnegative coefficients, and (b) structural smoothness. These result in a novel model called eNRBM (EMR-driven nonnegative RBM). We demonstrate the capability of the eNRBM on a cohort of 7578 mental health patients under suicide risk assessment. The derived representation not only shows clinically meaningful feature grouping but also facilitates short-term risk stratification. The F-scores, 0.21 for moderate-risk and 0.36 for high-risk, are significantly higher than those obtained by clinicians and competitive with the results obtained by support vector machines.

What shall i share and with whom? A multi-task learning formulation using multi-faceted task relationships

Multi-task learning is a learning paradigm that improves the performance of "related" tasks through their joint learning. To do this each task answers the question "Which other task should I share with"? This task relatedness can be complex - a task may be related to one set of tasks based on one subset of features and to other tasks based on other subsets. Existing multi-task learning methods do not explicitly model this reality, learning a single-faceted task relationship over all the features. This degrades performance by forcing a task to become similar to other tasks even on their unrelated features. Addressing this gap, we propose a novel multi-task learning model that leams multi-faceted task relationship, allowing tasks to collaborate differentially on different feature subsets. This is achieved by simultaneously learning a low dimensional sub-space for task parameters and inducing task groups over each latent subspace basis using a novel combination of L1 and pairwise L∞ norms. Further, our model can induce grouping across both positively and negatively related tasks, which helps towards exploiting knowledge from all types of related tasks. We validate our model on two synthetic and five real datasets, and show significant performance improvements over several state-of-the-art multi-task learning techniques. Thus our model effectively answers for each task: What shall I share and with whom?

A low-cost 3-D laser imaging system

This paper presents the design and development of a low cost three-dimensional laser imaging system for scanning suitable surfaces. A generic, low cost, off-the-shelf laser range finder is used to obtain the primary one dimensional distance measurement. The range finder’s laser beam is reflected by a twin-axis mirror assembly driven by stepper motors providing the system with two angular degrees of freedom, allowing 3-D measurements to be determined. A camera and image processing techniques are used to determine the measured 1-D range value from the generic range-finding device. A computer program then uses the obtained data to create a 3-D point cloud. An algorithm is then used to construct a 3-D wire frame mesh representing the scanned surface. The system has an angular resolution of 1.8° and the results obtained demonstrate the system to have an accuracy of approximately ± 2cm at a scanning distance of 1.0m.

Three-dimensional atom probe analysis of solute distribution in thermomechanically processed TRIP steels

Complex multiphase microstructures were obtained in transformation induced plasticity C–Mn–Si–(Nb–Al–Mo) steels by simulated controlled thermomechanical processing. These microstructures were characterized using transmission electron microscopy, X-ray diffraction and three-dimensional atom probe tomography (APT), which was used to determine the partitioning of elements between different phases and microconstituents. The measured carbon concentration (not, vert, similar0.25 at%) in the ferrite of carbide-free bainite was higher than expected from para-equilibrium between the austenite and ferrite, while the concentrations of substitutional elements were the same as in the parent austenite suggesting that incomplete bainite transformation occurred. In contrast, the distribution of substitutional elements between the ferrite lath and austenite in carbide-containing bainite indicated a complete bainite reaction. The average carbon content in the retained austenite (3.2 ± 1.6 at%) was somewhat higher than the T0 limit. On the basis of the APT measured composition, the calculated Ms temperatures for retained austenite were above room temperature, indicating its low chemical stability.

Precipitate characterisation of an advanced high-strength low-alloy (HSLA) steel using atom probe tomography

Increased fuel economy, combined with the need for the improved safety has generated the development of new hot-rolled high-strength low-alloy (HSLA) and multiphase steels such as dual-phase or transformation-induced plasticity steels with improved ductility without sacrificing strength and crash resistance. However, the modern multiphase steels with good strength-ductility balance showed deteriorated stretch-flangeability due to the stress concentration region between the soft ferrite and hard martensite phases [1]. Ferritic, hot-rolled steels can provide good local elongation and, in turn, good stretch-flangeability [2]. However, conventional HSLA ferritic steels only have a tensile strength of not, vert, similar600 MPa, while steels for the automotive industry are now required to have a high tensile strength of not, vert, similar780 MPa, with excellent elongation and stretch-flangeability [1]. This level of strength and stretch-flangeability can only be achieved by precipitation hardening of the ferrite matrix with very fine precipitates and by ferrite grain refinement. It has been suggested that Mo [3] and Ti [4] should be added to form carbides and decrease the coiling temperature to 650 °C since only a low precipitation temperature can provide the precipitation refinement [4]. These particles appeared to be (Ti, Mo)C, with a cubic lattice and a parameter of 0.433 nm, and they were aligned in rows [4]. It was reported [4] that the formation of these very fine carbides led to an increase in strength of not, vert, similar300 MPa. However, the detailed analysis of these particles has not been performed to date due to their nanoscale size. The aim of this work was to carry out a detailed investigation using atom probe tomography (APT) of precipitates formed in hot-rolled low-carbon steel containing additions Ti and Mo.

The investigated low-carbon steel, containing Fe–0.1C–1.24Mn–0.03Si–0.11Cr–0.11Mo–0.09Ti–0.091Al at.%, was produced by hot rolling. The processing route has been described in detail elsewhere [5] European Patent Application, 1616970 A1, 18.01.2006.[5]. The microstructure was characterised by transmission electron microscopy (TEM) on a Philips CM 20, operated at 200 kV using thin foil and carbon replica techniques. Qualitative energy dispersive X-ray spectroscopy (EDXS) was used to analyse the chemical composition of particles. The atomic level of particle characterisation was performed at the University of Sydney using a local electrode atom probe [6]. APT was carried out using a pulse repetition rate of 200 kHz and a 20% pulse fraction on the sample with temperature of 80 K. The extent of solute-enriched regions (radius of gyration) and the local solute concentrations in these regions were estimated using the maximum separation envelope method with a grid spacing of 0.1 nm [7]. A maximum separation distance between the atoms of interest of dmax = 1 nm was used.

The microstructure of the steel consisted of two types of fine ferrite grains: (i) small recrystallised grains with an average grain size of 1.4 ± 0.2 μm; and (ii) grains with a high dislocation density (5.8 ± 1.4 × 1014 m−2) and an average grain size of 1.9 ± 0.1 μm in thickness and 2.7 ± 0.1 μm in length (Fig. 1a). Some grains with high dislocation density displayed an elongated shape with Widmanstätten side plates and also the formation of cells and subgrains (Fig. 1a). The volume fraction of recrystallised grains was 34 ± 8%.

A haptically enabled low-cost reconnaissance platform for law enforcement

Traditionally, the control system of a modern teleoperated mobile robot consists of one or more two-dimensional joysticks placed on a control interface. While this simplistic interface allows an operator to remotely drive the platform, feedback is limited to visual information supplied by on-board cameras. Significant advances in the field of haptics have the potential to meaningfully enhance situational awareness of a remote robot. The focus of this research is the augmentation of Deakin University's OzBot trade MkIV mobile platform to include haptic control methodologies. Utilising the platform's inertial measurement unit, a remote operator has the ability to gain knowledge of the vehicle's operating performance and terrain while supplying a finer level of control to the drive motors. Our development of a generic multi-platform ActiveX allows the easy implementation of haptic force feedback to many computer based robot controllers. Furthermore, development of communication protocols has progressed with Joint Architecture for Unmanned Systems (JAUS) compliance in mind. The haptic force control algorithms are presented along with results highlighting the benefits of haptic operator feedback on the MklV OzBot trade chassis.

The dimensional and mechanical properties of wool/polyester fabrics made from vortex and ring-spun yarns

Fabric woven from wool/polyester (PES) Murata vortex spun (MVS) blend yarn is a commercially viable proposition particularly on the basis of advantageous wear-resistant properties, compared with fabric made from traditional worsted ring-spun yarn. However, in some early industrial trials with fabric made from 45/55-blend wool/PES MVS yarn, significantly greater relaxation shrinkage was found relative to comparable worsted ring-spun fabric. It was noted at the time that the amount of relaxation shrinkage in MVS fabric could be reduced to a large extent by using steamed MVS yarn.

In this study, the extent of variations in the dimensional and mechanical properties of fabric samples woven from a combination of steamed and unsteamed MVS yarn and equivalent worsted ring-spun yarn is examined. In general, greater hygral expansion and relaxation shrinkage were found in loom-state fabrics made from unsteamed MVS yarns, whereas the fabric made from steamed MVS and ring-spun yarns gave relatively low levels of relaxation shrinkage and hygral expansion. Permanent setting of fabrics, by pressure steaming, was found to be more effective than yarn pre-steaming in reducing relaxation shrinkage levels of fabrics made from unsteamed MVS yarn. After pressure steaming, all fabrics showed similar levels of relaxation shrinkage and hygral expansion.

Permanent setting of the fabrics, by pressure steaming, resulted in similar levels of relaxation shrinkage and hygral expansion, irrespective of the yarn production method; relaxation shrinkage fell to around 1% and hygral expansion increased by about 1%, relative to the loom-state samples. MVS fabrics were relatively heavier and fuller and had a firmer handle than the worsted ring-spun fabrics, reflecting the greater fabric weight, thickness and shear rigidity measured on these fabrics. These attributes are associated with different structures of the worsted ring-spun and MVS yarns used to make the fabrics.