Recognising behaviours of multiple people with hierarchical probabilistic model and statistical data association

Recognising behaviours of multiple people, especially high-level behaviours, is an important task in surveillance systems. When the reliable assignment of people to the set of observations is unavailable, this task becomes complicated. To solve this task, we present an approach, in which the hierarchical hidden Markov model (HHMM) is used for modeling the behaviour of each person and the joint probabilistic data association filters (JPDAF) is applied for data association. The main contributions of this paper lie in the integration of multiple HHMMs for recognising high-level behaviours of multiple people and the construction of the Rao-Blackwellised particle filters (RBPF) for approximate inference. Preliminary experimental results in a real environment show the robustness of our integrated method in behaviour recognition and its advantage over the use of Kalman filter in tracking people.

Discovery of activity structures using the hierarchical hidden Markov model

In building a surveillance system for monitoring people behaviours, it is important to understand the typical patterns of people's movement in the environment. This task is difficult when dealing with high-level behaviours. The flat model such as the hidden Markov model (HMM) is inefficient in differentiating between signatures of such behaviours. This paper examines structure learning for high-level behaviours using the hierarchical hidden Markov model (HHMM).We propose a two-phase learning algorithm in which the parameters of the behaviours at low levels are estimated first and then the structures and parameters of the behaviours at high levels are learned from multi-camera training data. Our algorithm is then evaluated using data from a real environment, demonstrating the robustness of the learned structure in recognising people's behaviour.

Learning environment stimulation in the real world

This paper analyses some of the unique circumstances which confronted a member of the lecturing staff in the Real Estate Valuation course at the Gatton campus of the University of Queensland during the late 1990's. Due to the rapidly changing educational needs of valuation students, and requirements of the respective accreditation bodies, the location of the course in Gatton (a rural town) was recognised as being remote from a major city and although relocating the course was inevitable, this could not happen overnight. From Semester 1, 2000 after several years of planning, the degree course relocates to the Department of Geographical Sciences and Planning, St Lucia. This paper discusses some of the challenges confronted and the strategies employed to simulate a city environment in the bush whilst waiting for the move.

Would the 'real' girl gamer please stand up? Gender, LAN cafes and the reformulation of the 'girl' gamer.

In this paper we consider the significance of cyber 'LAN' cafeacutes as sites where on and off-line practices meet in way that complicates binary notions of the gendered gamer. Existing research into computer games culture suggests a male dominated environment and points to girls' lower levels of competence and participation in games. Building on recent studies interested in the constitution of gender through engagement with online technologies, we draw on Judith Butler's politics of performative resignification, and conceptualise digital culture as a resource through which 'girl' gamers are mobilised and potentially reformulated, experiencing their gaming identities in contradictory ways, and fragmenting the category 'girl' in the very act of articulating their place in a male dominated gaming culture. It is argued that through the meeting of on and off-line practices, LAN cafeacutes operate as a location that is particularly amenable to reformulative work in relation to gendered gaming identities.

Towards optimization of a real-world robotic-sensor system of systems

The problem of threat detection in an unstructured environment is considered. Three systems, comprising of robots and sensors, are proposed to form a system of systems (SoS) to find a solution to the problem. System interactions are defined to provide a framework for formulation as an SoS optimization problem. Different cost and objective functions are introduced for optimization of local criteria. Using different weights, a linear combination of the local cost and objective functions is obtained to propose a global objective function. An algorithm is suggested to find an optimum value for the global objective function leading towards optimization of the SoS.

Socioeconomic variations in women's diets : what is the role of perceptions of the local food environment

Objectives : To test the contribution of perceived environmental factors (food availability, accessibility and affordability) to mediating socioeconomic variations in women’s fruit, vegetable and fast food consumption.

Methods : A community sample of 1580 women from 45 neighbourhoods provided survey data on their socioeconomic position (SEP) (education and income); diet (fruit, vegetable and fast food consumption); and the perceived availability of, access to and cost of healthy food in their local area.

Results : Once perceived environmental variables were considered, the associations between SEP and diet were weak and non-significant, suggesting that socioeconomic differences in diet were almost wholly explained by perceptions of food availability, accessibility and affordability.

Conclusions : Strategies to decrease socioeconomic inequalities in diet could involve promoting inexpensive ways to increase fruit and vegetable consumption, and ensuring that people of low SEP are aware that many healthy foods are available at relatively low cost. Future research should also confirm if perceptions match objective measures of food availability, accessibility and affordability, in order to address the real and/or perceived lack of healthy options in low SEP neighbourhoods.

A time for real building performance measurement

The Mobile Architecture and Built Environment Laboratory (MABEL) was conceived upon the principle of investigating building environmental performance in situ. MABEL provides the first means of integrated, on-site measurement of the key aspects of internal built environments; energy, lighting, air quality, ventilation, acoustics and comfort using state-ofthe- art technology and instrumentation. The intention of this paper is to explain the how and what need to be measured in our buildings if we are to search of a genuine performance answer as well as the information to provide a solution. Several results of real building measurement are provided here, suggesting that a national program on 'as performing' is required if we are to proceed in a sustainable manner.

Office buildings and the environment : the increasing importance of ESD

The links between the built environment and sustainability issues such as fossil fuel consumption and climate change is clear. In developed countries buildings contribute around half of all carbon dioxide emissions and offer considerable scope for a significant contribution to sustainability through ecologically aware design and increased energy efficiency (BRE, 1996). The Australian commercial stock emits 12% of all greenhouse gas emissions however the commercial property market has some inherent barriers to sustainability (DSE, 2005). A substantial proportion of the stock is owned by institutional investors who are unconvinced by the need to improve their stock and pass on running costs to tenants (Callender & Key, 1997). As capital values are not greatly affected by sustainability, owners react by doing little or nothing and the effect is to limit sustainability related investment and undermine efforts to deliver sustainability in the sector.

Furthermore the efficiency of buildings declines over time and whilst energy efficiency is important to new design, the existing stock must be improved if urban built environment greenhouse gas emissions are to be reduced. Much of the property and surveying research has previously adopted an illustrative case study approach advocating the benefits of ESD and energy efficiency in existing buildings. This research adopts a radically different approach and profiles the entire office stock of a global CBD, namely Melbourne, which is seeking to become a carbon neutral city by 2020. The research also employs scenario forecasting to model future changes to the stock over a fifteen year period. This paper sets out the rationale for the research and establishes the methodological approach adopted by the research team.

Kernel performance in meeting hard real-time system characteristics (the ARTOS real-time kernel)

Describes the design and implementation of an operating system kernel specifically designed to support real-time applications. It emphasises portability and aims to support state-of-the-art concepts in real-time programming. Discusses architectural aspects of the ARTOS kernel, and introduces new concepts on the areas of interrupt processing, scheduling, mutual exclusion and inter-task communication. Also explains the programming environment of ARTOS kernal and its task model, defines the real-time task states and system data structures and discusses exception handling mechanisms which are used to detect missed deadlines and take corrective action.

Autonomous underwater vehicles in a hazardous non-deterministic environment

This research investigated the cooperation of multi underwater robots to perform a task. This combined engineering design, electronics and consensus control to create systems capable of achieving the task. Challenges such as underwater radio communications were researched and a simulation framework was created and tested on virtual and real systems.

Understanding the context of distance students : differences in on- and off-campus engagement with an online learning environment

 While a growing number of higher education institutions are providing online modes of study for both on- and off-campus students, there are very real differences in demography, technology experiences, reasons for study, etc. between on- and off-campus students, and research into engagement with online learning environments (OLEs) indicate differences in the way that on- and off-campus students interact with OLEs. In Australia, Deakin University is a major provider of distance and online education, and provides a case study of a higher education institution with a mature and large-scale OLE implementation providing support to both on- and off-campus students. Deakin Studies Online (DSO) is Deakin University’s OLE. Based on a representative sample of 1322 responses to the 2011 DSO evaluation survey, this paper presents a large-scale, up-to-date and fine-grained investigation of impact of mode of study on the student experience of using an OLE. It was found that the primary place of access to DSO for both groups was home, mobile access to DSO seems likely to be of growing importance to both groups, and there was no statistically significant difference in the mean satisfaction ratings between on- and off-campus students for virtually all DSO functions. Off-campus students gave significantly higher mean ratings of importance (though not satisfaction) to a range of DSO functions that could be viewed as ‘value adders’ by off-campus students, enhancing their overall learning experience. For more than half of the DSO functions surveyed, on-campus students reported statistically significant higher mean frequency of access than off-campus students. The finding that elements of the institutional OLE are not universally perceived and used the same way by all students groups challenges the value of standard, one-size-fits-all institutional policies and templates relating to the use of OLEs.

A probabilistic model with parsinomious representation for sensor fusion in recognizing activity in pervasive environment

To tackle the problem of increasing numbers of state transition parameters when the number of sensors increases, we present a probabilistic model together with several parsinomious representations for sensor fusion. These include context specific independence (CSI), mixtures of smaller multinomials and softmax function representations to compactly represent the state transitions of a large number of sensors. The model is evaluated on real-world data acquired through ubiquitous sensors in recognizing daily morning activities. The results show that the combination of CSI and mixtures of smaller multinomials achieves comparable performance with much fewer parameters.