Policy recognition in the Abstract Hidden Markov Model

In this paper, we present a method for recognising an agent's behaviour in dynamic, noisy, uncertain domains, and across multiple levels of abstraction. We term this problem on-line plan recognition under uncertainty and view it generally as probabilistic inference on the stochastic process representing the execution of the agent's plan. Our contributions in this paper are twofold. In terms of probabilistic inference, we introduce the Abstract Hidden Markov Model (AHMM), a novel type of stochastic processes, provide its dynamic Bayesian network (DBN) structure and analyse the properties of this network. We then describe an application of the Rao-Blackwellised Particle Filter to the AHMM which allows us to construct an efficient, hybrid inference method for this model. In terms of plan recognition, we propose a novel plan recognition framework based on the AHMM as the plan execution model. The Rao-Blackwellised hybrid inference for AHMM can take advantage of the independence properties inherent in a model of plan execution, leading to an algorithm for online probabilistic plan recognition that scales well with the number of levels in the plan hierarchy. This illustrates that while stochastic models for plan execution can be complex, they exhibit special structures which, if exploited, can lead to efficient plan recognition algorithms. We demonstrate the usefulness of the AHMM framework via a behaviour recognition system in a complex spatial environment using distributed video surveillance data.

Spatial reasoning via active observation

This paper presents a model for space in which an autonomous agent acquires information about its environment. The agent uses a predefined exploration strategy to build a map allowing it to navigate and deduce relationships between points in space. The shapes of objects in the environment are represented qualitatively. This shape information is deduced from the agent's motion. Normally, in a qualitative model, directional information degrades under transitive deduction. By reasoning about the shape of the environment, the agent can match visual events to points on the objects. This strengthens the model by allowing further relationships to be deduced. In particular, points that are separated by long distances, or complex surfaces, can be related by line-of-sight. These relationships are deduced without incorporating any metric information into the model. Examples are given to demonstrate the use of the model.

Learning other agents' preferences in multiagent negotiation

In multiagent systems, an agent does not usually have complete information about the preferences and decision making processes of other agents. This might prevent the agents from making coordinated choices, purely due to their ignorance of what others want. This paper describes the integration of a learning module into a communication-intensive negotiating agent architecture. The learning module gives the agents the ability to learn about other agents' preferences via past interactions. Over time, the agents can incrementally update their models of other agents' preferences and use them to make better coordinated decisions. Combining both communication and learning, as two complement knowledge acquisition methods, helps to reduce the amount of communication needed on average, and is justified in situations where communication is computationally costly or simply not desirable (e.g. to preserve the individual privacy).

Efficient cross-validation of the complete two stages in KFD classifier formulation

This paper presents an efficient evaluation algorithm for cross-validating the two-stage approach of KFD classifiers. The proposed algorithm is of the same complexity level as the existing indirect efficient cross-validation methods but it is more reliable since it is direct and constitutes exact cross-validation for the KFD classifier formulation. Simulations demonstrate that the proposed algorithm is almost as fast as the existing fast indirect evaluation algorithm and the twostage cross-validation selects better models on most of the thirteen benchmark data sets.

Hierarchical recognition of intentional human gestures for sports video annotation

We present a novel technique for the recognition of complex human gestures for video annotation using accelerometers and the hidden Markov model. Our extension to the standard hidden Markov model allows us to consider gestures at different levels of abstraction through a hierarchy of hidden states. Accelerometers in the form of wrist bands are attached to humans performing intentional gestures, such as umpires in sports. Video annotation is then performed by populating the video with time stamps indicating significant events, where a particular gesture occurs. The novelty of the technique lies in the development of a probabilistic hierarchical framework for complex gesture recognition and the use of accelerometers to extract gestures and significant events for video annotation.

Narrative structure analysis with education and training videos for e-learning

This paper deals with the problem ofstructuralizing education and training videos for high-level semantics extraction and nonlinear media presentation in e-learning applications. Drawing guidance from production knowledge in instructional media, we propose six main narrative structures employed in education and training videos for both motivation and demonstration during learning and practical training. We devise a powerful audiovisual feature set, accompanied by a hierarchical decision tree-based classification system to determine and discriminate between these structures. Based on a two-liered hierarchical model, we demonstrate that we can achieve an accuracy of 84.7% on a comprehensive set of education and training video data.

Temporal PDMs for gait classification

Gait classification is a developing research area, particularly with regards to biometrics. It aims to use the distinctive spatial and temporal characteristics of human motion to classify differing activities. As a biometric, this extends to recognising different people by the heterogeneous aspects of their gait. This research aims to use a modified deformable model, the temporal PDM, to distinguish the movements of a walking and miming person. The movement of 2D points on the moving form is used to provide input into the model and classify the type of gait present.

Adaptive model for foreground extraction in adverse lighting conditions

Background elimination models are widely used in motion tracking systems. Our aim is to develop a system that performs reliably under adverse lighting conditions. In particular, this includes indoor scenes lit partly or entirely by diffuse natural light. We present a modified "median value" model in which the detection threshold adapts to global changes in illumination. The responses of several models are compared, demonstrating the effectiveness of the new model.

Exploring Australian housing supply volatility

This study examines the volatility series of housing supply in Australia. A Generalised Autoregressive Conditional Heteroskedasticity-in-Mean (GARCH-M) model is employed to analyse the volatility series of Australian housing supply over the study period of 1974-2010. The results show the volatility of housing starts is negatively linked to housing starts, suggesting that higher uncertainty does lower housing starts. The results also reveal that the uncertainty of housing starts is also captured by the volatilities of interest rates and construction costs. Therefore policy makers should monitor and attempt to minimise the volatility of housing supply. These steps will enhance housing construction activities and increase the availability of housing supply to potential home buyers.

Investigating the housing market fluctuation under an expansionary monetary policy in Australia

This research empirically investigates the impact of monetary policy on the housing market in Australia from 1996 to 2009. Three primary variables associated with the housing sector and monetary policy, including interest rates, money supply and house prices, are estimated by a structural vector autoregression (VAR) model. Depending upon the analysis using the impulse response function, it can be identified that monetary policy significantly affects the housing market in Australia by the adjustments in interest rates and money supply. The empirical results from this study may be useful for policy makers to enact appropriate policies in relation to the infrastructure planning.

Identification of surface heterogeneity effects in cyclic voltammograms derived from analysis of an individually addressable gold array electrode

Voltammetric behavior at gold electrodes in aqueous media is known to be strongly dependent on electrode polishing and history. In this study, an electrode array consisting of 100 nominally identical and individually addressable gold disks electrodes, each with a radius of 127 µm, has been fabricated. The ability to analyze both individual electrode and total array performance enables microscopic aspects of the overall voltammetric response arising from variable levels of inhomogeneity in each electrode to be identified. The array configuration was initially employed with the reversible and hence relatively surface insensitive [Ru(NH₃)₆]^3+/2+ reaction and then with the more highly surface sensitive quasi-reversible [Fe(CN)₆]^3−/4− process. In both these cases, the reactants and products are solution soluble and, at a scan rate of 50 mV s⁻¹, each electrode in the array is assumed to behave independently, since no evidence of overlapping of the diffusion layers was detected. As would be expected, the variability of the individual electrodesʼ responses was significantly larger than found for the summed electrode behavior. In the case of cytochrome c voltammetry at a 4,4′-dipyridyl disulfide modified electrode, a far greater dependence on electrode history and electrode heterogeneity was detected. In this case, voltammograms derived from individual electrodes in the gold array electrode exhibit shape variations ranging from peak to sigmoidal. However, again the total response was always found to be well-defined. This voltammetry is consistent with a microscopic model of heterogeneity where some parts of each chemically modified electrode surface are electroactive while other parts are less active. The findings are consistent with the common existence of electrode heterogeneity in cyclic voltammetric responses at gold electrodes, that are normally difficult to detect, but fundamentally important, as electrode nonuniformity can give rise to subtle forms of kinetic and other forms of dispersion.

Characterising nonuniform electrodeposition and electrodissolution using the novel wire beam electrode method

An electrochemically integrated multi-electrode system namely the wire beam electrode (WBE) has been applied as a new method of characterising nonuniform electrodeposition and electrodissolution, by measuring and identifying characteristic patterns in electrodeposition and electrodissolution current distribution maps. Various patterns of electrodeposition current distribution have been obtained from Watts nickel plating and bright acid copper plating baths with the effects of several affecting factors such as bath concentration, temperature, agitation and electrolyte flow. Typical patterns of electrodissolution current distribution have also been detected over a WBE surface under anodic dissolution. This work suggests that the WBE method can be used as a new tool for monitoring, characterising and optimising electrodeposition and electrodissolution processes in the laboratory, and can also be applied as an experimental method to verify the accuracy and completeness of mathematical models for electrodeposition and electrodissolution.

Reducing herbicide discharge to sensitive environments using membrane bioreactors

A hybrid MBR/UV/GAC treatment system was researched to remove Ametryn, which is a commonly used herbicide in Australian farmlands, from wastewater. The research revealed that the hybrid system could be successfully used for 100% removal of Ametryn. Two mathematical models were developed to predict the frequency of chemical cleaning of MBR-membrane and the mechanism of fouling of membrane.

A marginalised Markov Chain Monte Carlo approach for model based analysis of EEG data

The work presented in this paper focuses on fitting of a neural mass model to EEG data. Neurophysiology inspired mathematical models were developed for simulating brain's electrical activity imaged through Electroencephalography (EEG) more than three decades ago. At the present well informative models which even describe the functional integration of cortical regions also exists. However, a very limited amount of work is reported in literature on the subject of model fitting to actual EEG data. Here, we present a Bayesian approach for parameter estimation of the EEG model via a marginalized Markov Chain Monte Carlo (MCMC) approach.