CECS Upcoming Seminars

Thursday 04 March, 2010 Sandun Kodituwakku - TBA

2010-03-04 11:00:00

11:00:00 - 12:00:00 - RSISE Seminar Room, ... Applied Signal Processing Series

TBA

Thursday 25 February, 2010 Kim Blackmore - TBA

2010-02-25 11:00:00

11:00:00 - 12:00:00 - RSISE Seminar Room, ... Applied Signal Processing Series

TBA

Thursday 04 February, 2010 Oday Jerew - TBA

2010-02-04 11:00:00

11:00:00 - 12:00:00 - RSISE Seminar Room, ... Applied Signal Processing Series

TBA

Thursday 28 January, 2010 Chris J Baker - TBA

2010-01-28 11:00:00

11:00:00 - 12:00:00 - RSISE Seminar Room, ... Applied Signal Processing Series

TBA

Thursday 17 December, 2009 Thushara Abhayapala - TBA

2009-12-17 11:00:00

11:00:00 - 12:00:00 - RSISE Seminar Room, ... Applied Signal Processing Series

TBA

Monday 14 December, 2009 Prof. Anthony Ephremides - Cooperation at the Network Level

2009-12-14 11:00:00

11:00:00 - 12:00:00 - RSISE Seminar Room, ... CECS Seminar Series

The concept of cooperative transmissions in a wireless network evolved from the notion of the relay channel and the MIMO technology. Most of the work to date has focused on physical layer techniques (such as decode-and-forward and the like) and aimed at characterizing the source-to-destination Information-theoretic Capacity.We take a different view and exploit the possibility of relaying in a cognitive fashion. That is, we sense unused resources (time-slots and/or frequency bamds) and make opportunistic use of these resources for relaying. We consider packets and slots rather than bits and seconds and thus our approach focuses on computing throughput, rather than capacity. In fact, we consider the realistic and practical case of finite delays, and, hence, bursty source traffic; thus we focus on the "stable throughput" region that can be achieved in a network with or without relaying. We find that significant improvements can be achieved if such cooperative/cognitive methods are used. And in fact we also consider superposing on these methods the more familiar physical layer schemes (such as combining, decode-and-forward, etc) and show that we can achieve, as expected, even further gains.We then look at cooperative routing in sensor networks (a totally different concept of cooperation, again at the network level) and identify some remarkable consequences in that case.

These novel ideas and methods are the beginning of what we call "cooperation at the network level"; the potential exists that, through such cooperation, far-reaching implications may follow with regard to achievable transmission rates under bursty traffic conditions.

Thursday 03 December, 2009 Yan Jennifer Wu - Multizone 2D Soundfield Reproduction Via Spatial Band Stop Filters

2009-12-03 11:00:00

11:00:00 - 12:00:00 - RSISE Seminar Room, ... Applied Signal Processing Series

Any attempt to create multiple independent soundfields in separate zones over an extended region of open space results in unintended interference in a given zones from other zones. For a multizone soundfield system, there are undesirable effects inherently produced on one zone which are caused by the lower order terms from the other zones. In this seminar, we present a spatial band stop filters to suppress interzone interference in the regions of interests and pass the desired soundfields with no distortion. This is achieved by using the higher order spatial harmonics of one zone to cancel the undesirable effects of the lower order harmonics of the same zone on the other zones. We illustrate the work by designing and simulating a 2D two-zone soundfield.

Friday 27 November, 2009 Dr. Sergej Celikovsky - Nonlinear techniques for the Acrobot tracking with application to robot walking

2009-11-27 11:00:00

11:00:00 - 12:00:00 - RSISE Seminar Room, ... Systems and Control Series

This talk aims to compare the performance of various techniques for the stabilization of the error dynamics of the Acrobot's walking like reference trajectory. Both the walking reference planning and the tracking feedback design are based on the Acrobot's model partial exact feedback linearization of order 3. Namely, such an exact system transformation leads to an almost linear system where error dynamics along trajectory to be tracked is a 4 dimensionallinear time varying system having 3 time varying entries only, the remaining entries are either zero or equal to one.

Three techniques to stabilize asymptotically such an error dynamics are presented and compared. All of them are based on the robust control approaches viewing the abovementioned time-dependent terms as some partially unknown disturbances. First technique combines high-gain feedback with Lyapunovs analysis, the latter one enables to obtain some reasonable gain values, yet resulting in an unrealistically large actuator torque. The second technique relies on the fact that the desirable exponentially stable tracking can be obtained by solving quadratic stability of a linear system with polytopic uncertainty. To do so, LMI methods are engaged to solve this problem numerically. The results of this careful analysis are shown to be a significant improvement of previously known approaches, especially in the case of nonrectangular convex polytopic uncertainty set. Here, the advantage is taken of the fact that "uncertainty" is, in fact, a triple of the known scalar time functions, so that one can find significantly smaller bounding convex set than just a large rectangular box. The last technique uses the time-dependent transformation of the error dynamics resulting in much better results when the high-gain approach is applied. Similarly as with the first technique, the corresponding results are in a closed form, rather than in a numerical one. Again, the advantage is taken of the fact that the time dependent error dynamics parameters are, in fact, known and differentiable.

Numerical simulations and animations of the Acrobot walking based on all above mentioned techniques will be demonstated as well. Finally, some outlooks will be presented to extend these ideas to the walking design for the general n-link systems with n-1 actuators, underactuated at the pivot point.

Thursday 26 November, 2009 Parastoo Sadeghi - TBA

2009-11-26 11:00:00

11:00:00 - 12:00:00 - RSISE Seminar Room, ... Applied Signal Processing Series

TBA

Monday 23 November, 2009 Prof. Koichi Asatani - Trends of NGN and Its Issues

2009-11-23 11:00:00

11:00:00 - 12:00:00 - RSISE Seminar Room, ... CECS Seminar Series

Next Generation Network (NGN) is designed to be capable of QoS management and controls like in traditional telecommunication networks and to support economical, versatile multi-media applications like those on the current Internet. NGN also provides fixed-mobile convergence (FMC) with generalized mobility, and horizontal and vertical roaming as well as improved security.

The concepts and architecture of the Next Generation Networks (NGN) are described. NGN voice and Internet services are attractive from the view points of service flexibility and cost effectiveness and the capability of integrating third-party applications with high dependability and high security. The current status of NGN implementation in a commercial offer by NTT (Nippon Telegraph and Telephone Corporation) is touched upon.

Issues for the global evolution of NGN are also described, such as technical key issues, global standards and regulations.

Thursday 19 November, 2009 Mengqiu (Karan) Zhang - HRTF Measurement on a KEMAR Manikin

2009-11-19 11:00:00

11:00:00 - 12:00:00 - RSISE Seminar Room, ... Applied Signal Processing Series

Presenting sounds to humans in virtual environments requires convolving the free field signals with a head related transfer function (HRTF), which is a frequency response describing the filtering effects of the pinna, head and torso of a human. Sets of HRTFs are usually measured on the dummy head or human subjects at different directions in an anechoic room. This paper describes the details of an experimental HRTF measurement procedure with emphasis on the design of the test signal and the post-processing to extract HRTFs. We construct a pre-emphasized logarithmic sweep as the excitation signal which separates the nonlinear and time variant distortions from the main desired response. For the received raw data, a series of signal processing techniques are applied to determine the timing offset when the head response begins, to mitigate room reverberation, and to equalize the HRTF measurements. One of the goals of this paper is to provide details of the experimental setup. Also, we intend to publish our processed measurements in the form of a HRTF data base as a resource for the acoustic community.

Tuesday 17 November, 2009 Dr Matthias Kaiserswerth - Innovating to Create a Smarter Planet

2009-11-17 12:30:00

12:30:00 - 13:30:00 - NICTA - 7 London Cir... NICTA Seminar

The world is getting increasingly intelligent, thanks to new technologies, the Internet and the enormous computing power of modern PCs. But a more intelligent world does not happen all by itself. Business enterprises, governments and the scientific community must share the responsibility for ensuring that all this potential can develop and be used in the right way.

At the moment we are still wasting energy, our healthcare systems are not working efficiently enough and the economy is overtaken by one crisis after another. What strategies and solutions are available for dealing with basic challenges such as these? How can a business enterprise use information more systematically and more intelligently? How can the natural resources of our planet be put to better and more efficient use? With the technologies and solutions available to us today we can do more than we have ever been able to do in the past to link together people, things, processes and systems, and to make the world in general a more intelligent place. That is our vision for a smarter planet.

Monday 16 November, 2009 Simon Lucey - Relaxation Methods in Vision: Convex or just plain vexing?

2009-11-16 14:00:00

14:00:00 - 15:00:00 - CSIT Seminar Room, N... CSIRO ICT

Convex quadratic objective functions are an attractive means of expressing some goal/task in vision like alignment or classification as: (i) local minima = global minimum, (ii) the sum of N convex quadratics is itself a convex quadratic, and (iii) they offer computationally efficient solutions. Famous algorithms in vision such as the Lucas-Kanade (LK) algorithm (alignment), and the Support Vector Machines (classification) can both be viewed as minimizing a convex quadratic objective function.

In the first part of this talk, I will be exploring problems/applications in vision where the original objective function can be suitably relaxed to take advantage of the convex quadratic form. We will introduce two examples of such successful relaxations: (i) Convex Quadratic Fitting (CQF) for non-rigid face alignment with local-experts, and (ii) Least-Squares Congealing (LSC) for the task of unsupervised image ensemble alignment. Both examples, at the time of writing, exhibit superior performance to current state of the art performance.

In the second part of this talk, I will explore the concept that if our learning goal can be expressed as a convex quadratic, and our feature extraction step linear, then the sequential feature extraction and optimization steps can be re-interpreted within a single learning goal. This alternate view of linear feature extraction with respect to a convex quadratic learning goal has a number of advantages. First, for the case of classification within the well known linear support vector machine (SVM) framework the memory and computational overheads, typically occurring due to the high dimensionality of the feature extraction process now disappear. From a theoretical perspective the feature extraction step can now be viewed alternately as manipulating the margin of the SVM. This insight is synergetic with recent work in learning that has demonstrated that the choice of margin employed while learning a SVM is critical for high classification performance in many circumstances. Second, for the case of alignment we demonstrate that a similar approach can be applied when employing linear feature extraction in conjunction with the LK algorithm for alignment. By framing the LK algorithm within the Fourier- domain, an algorithm we refer to as Fourier-LK (FLK), we demonstrate superior alignment performance with nearly no additional computational overhead.

Monday 16 November, 2009 Prof. Shoji Makino - Blind Audio Source Separation based on Independent Component Analysis

2009-11-16 11:00:00

11:00:00 - 12:00:00 - RSISE Seminar Room, ... CECS Seminar Series

This talk describes a state-of-the-art method for the blind source separation (BSS) of convolutive mixtures of audio signals, especially speech. A statistical and computational technique, called independent component analysis (ICA), is examined. We provide examples to show how ICA criteria change as the number of audio sources increases. By achieving nonlinear decorrelation, nonstationary decorrelation, or time-delayed decorrelation, we can find source signals only from observed mixed signals. We then discuss a frequency-domain approach where simple instantaneous ICA is employed in each frequency bin. Particular attention is paid to the physical interpretation of BSS from the acoustical signal processing point of view. Frequency-domain BSS is shown to be equivalent to two sets of frequency domain adaptive microphone arrays, i.e., adaptive beamformers (ABFs). Although BSS makes spatial notch to a jammer as well as ABF, BSS has a strong advantage over ABF. BSS can be regarded as an intelligent version of ABF in the sense that it can adapt without any information on the array manifold or the target direction, and sources can be simultaneously active in BSS.

Friday 13 November, 2009 Lachlan Blackhall - Recursive Sparse Estimation using a Gaussian Sum Filter

2009-11-13 11:00:00

11:00:00 - 12:00:00 - RSISE Seminar Room, ... Systems and Control Series

In many linear regression problems the number of possible explanatory parameters is often much greater than the number necessary to explain the observations. This corresponds to a sparse solution of the estimation problem and is of significant interest as a high degree of sparsity corresponds to simpler models. This seminar will detail recent investigations into a sparse estimator that can be implemented recursively, like the Kalman filter, while systematically producing appropriately sparse parameter estimates. The relationship between this work and compressive sensing will also be outlined.

Thursday 12 November, 2009 Gerard Borg - Coordinated Distributed MIMO: turning wireless networks on their head(s?)

2009-11-12 11:00:00

11:00:00 - 12:00:00 - RSISE Seminar Room, ... Applied Signal Processing Series

The conventional approach to infrastructure wireless networking employs a single antenna or single-user MIMO base station to transmit and receive information to and from multiple mobiles. Information flow is from a one to many and all links are effectively SISO in nature. In this talk we take a look at networks in which we deliberately deploy many more base stations than mobiles. Such networks are a branch of distributed MIMO networks: a topic of considerable interest in current wireless research. In this talk we describe the many advantages of coordinated distributed MIMO networks in the context of a simple proposed implemention. We show that challenging problems such as coordination of information, carrier and timing synchronisation and scalability of physical layer processing can largely be dealt with using techniques already applied in SISO wireless communications.

Wednesday 11 November, 2009 Charles Featherston - A Technology Roadmap for the Australian Automotive Industry

2009-11-11 12:00:00

12:00:00 - 12:30:00 - Graduate Teaching Ro... Materials and Manufacturing

Tuesday 10 November, 2009 Mr Debdeep Banerjee - Integrating Planning and Scheduling in a CP Framework: A Resource Oriented Approach

2009-11-10 15:00:00

15:00:00 - 15:30:00 - RSISE Seminar Room, ... CS PhD Monitoring

Many potential real-world planning applications are on the border of planning and scheduling. To handle the complex choices of actions and temporal and resource constraints of these problems we need to integrate planning and scheduling techniques. Here we propose a transition-based formulation for temporal and resource constraint planning problems, that enables us to represent features like deadlines, time windows, release times, unary resources etc in a simple way. In this talk we will describe a CSP encoding of the transition-based formulation and propagation techniques for time and resource constraints. We will also show how different problems, that are in between planning and scheduling, can be modeled within this framework.

Tuesday 10 November, 2009 Ms Cindy Wang - Tractable Multi-Agent Path Planning on Gird Maps: Initial Experimental Results

2009-11-10 14:30:00

14:30:00 - 15:00:00 - RSISE Seminar Room, ... CS PhD Monitoring

Navigating multiple mobile units on grid maps is a challenging problem with many real-life applications. We introduced a tractable multi-agent path planning algorithm, MAPP [Wang and Botea, 2009], to address the limitations inherent in previous approaches. In particular, poor scalability of existing centralised methods, and lack of completeness, runtime and solution quality guarantees for the decoupled methods.

In this talk, I will show initial experimental results with the basic MAPP algorithm. These results indicate that the SLIDEABLE class, for which MAPP is guaranteed to solve in low polynomial time, occurs quite often in practice. I will also discuss extensions to more general classes of problems, and heuristic improvements that preserve the algorithm's completeness.

Tuesday 10 November, 2009 Ms Joanna Cheng - Vehicle Routing Using Seeding

2009-11-10 14:00:00

14:00:00 - 14:30:00 - RSISE Seminar Room, ... CS PhD Monitoring

The Vehicle Routing Problem (VRP) refers to finding optimal paths for anagent, subject to "stops" that must be visited along the way. We focuson the construction part of a solution algorithm, so as to ensure westart the optimisation/local search phase with a reasonably goodsolution. We aim to use seeding to guide the construction phase, and asa method to deal with dynamic or uncertain data.

Tuesday 10 November, 2009 Mr Daniel Harabor - Ejection Chain Methods For Solving TSPs

2009-11-10 13:30:00

13:30:00 - 14:00:00 - RSISE Seminar Room, ... CS PhD Monitoring

An ejection chain is a special kind of move operator used in local search to explore a neighbourhood.Ejection chains are interesting because of their ability to cross infeasible regions of the search spaceand connect otherwise non-adjacent neighbourhoods. For this reason, ejection chain methods are among the most competitive algorithms for solving a wide range of classical computer science and operations research problems.

In this talk I will provide an introduction to ejection chain methods with particular emphasis on their application to solving the Traveling Salesman Problem (TSP). I will then outline some proposed research directions which involve using a spanning-tree heuristic to guide the construction of ejection chains.