Back to main Workshops page

2010 Quebec China India International Workshop on Communications

SYTACom will host a workshop featuring technical presentations by communications experts from China, India and Quebec universities on May 11, 2010 at McGill University


This one-day research workshop aimed at fostering international collaborations in the area of Information and Communications Technologies (ICT), is sponsored by SYTACom. It will feature speakers from India, China and Québec, outlining their ongoing and planned international research projects. This workshop is meant as a networking opportunity for all stakeholders, from Academia, Government and Industry. Attendees are expected to be from companies in the Quebec ICT sector, government agencies and researchers from several Quebec universities.

List of speakers

  • Prof. Swades De, Department of Electrical Engineering, Indian Institute of Technology, Delhi
  • Prof. Bharadwaj Amrutur, Electrical Communication Engineering Department, Indian Institute of Science, Bangalore
  • Prof. Guixia Kang, Wireless Technology Innovation (WTI) Institute, Beijing University of Posts and Telecommunications, Beijing
  • Prof. Richard Rose, Department of Electrical and Computer Engineering, McGill University, Montreal
  • Prof. Michael Rabbat, Department of Electrical and Computer Engineering, McGill University, Montreal
  • Prof. François Gagnon, Department of Electrical Engineering, Ecole de Technologie Supérieure, Montreal

Registration:

Registration for the workshop is required and is free of charge. To register, send an email to maru.basanez@mcgill.ca and include your full name and affiliation.

Location: Trottier Building, room 2110, 3630 University St

10:00 - 10:30 Welcome  
10:30 - 11:00 Invited Talk Applications of wireless sensor networks for security and urban management
Prof. Bharadwaj Amrutur, IISC Bangalore
11:00 - 11:30 Invited Talk Gossip Algorithms for Distributed Signal Processing
Prof. Michael Rabbat, McGill University
11:30 - 11:45 Coffee Break  
11:45 - 12:15 Invited Talk Ubiquitous Sensing based on Wireless Sensor Network and its application of Wireless eHealth (WeHealth)
Prof. Guixia Kang, Beijing University of Posts and Telecommunications
12:15 - 12:45 Invited Talk Efficient Modeling of Acoustic Variability in Automatic Speech Recognition
Prof. Richard Rose, McGill University
12:45 - 14:15 Buffet Lunch  
14:15 - 14:45 Invited Talk Cross‐layer adaptive solutions to energy optimized multihop wireless ad hoc sensor networks
Prof. Swades De, IIT Delhi
14:45 - 15:15 Invited Talk Energy efficient DSP: results on novel filters as a first step
Prof. François Gagnon, Ecole de technologie supérieure
15:15 - 15:30 Coffee Break  
15:30 – 16:00 Invited Talk Delay-Tolerant Decode-and-Forward Based Cooperative Communication
Prof. Manav Bhatnagar, IIT Delhi

ABSTRACTS

Applications of wireless sensor networks for security and urban management
Prof. Bharadwaj Amrutur, IISC Bangalore

We will first describe an ongoing project on the design of a wireless fence for intrusion detection using wireless sensor networks. The fence is constructed using an array of Motes, each of which is equipped with passive infrared sensors on a low power microcontroller platform with a zigbee radio. Algorithms for ad-hoc networking, sensor signal processing, detection and fusion have been developed and integrated onto the distributed platform, achieving very high correct detection probability with very low false alarm rates. We are now looking at the development and deployment of wireless sensor networks for urban management. We will specifically target pollution monitoring as the first application. We are developing new motes which have both short-distance(zigbee) and long distance (GSM) connectivity. As a first goal, we plan to equip these with gas sensors and deploy them in critical city junctions to gather real time pollution data. We will describesome of the challenges in this effort.

Top of page

Gossip Algorithms for Distributed Signal Processing
Prof. Michael Rabbat, McGill University

Abstract: Gossip algorithms are an attractive paradigm for distributed computation in sensor networks for a number of reasons including: they run asynchronously, they do not require overhead to establish or maintain routing tables, and they are not susceptible to bottlenecks or single points of failure. Nodes iteratively exchange information with their immediate neighbors until the entire network reaches a consensus on the value being computed. Of course, robustness and autonomy come at a price.

For network topologies typically used to model connectivity in wireless networks, such as grids and random geometric graphs, the average number of messages transmitted by each node grows linearly with the size of the network, and hence does not scale. In wireless sensor networks this is especially detrimental since each transmission consumes valuable bandwidth and energy resources.

In this talk I will survey some recent advances in accelerating rates of convergence by incorporating additional memory at each node and leveraging hierarchical network structure. I will also present an example of how a variant of gossip can be used in a decentralized compression application, where one wishes to adaptively compute the most significant (largest energy) coefficients of a linear transform of distributed data without knowing their indices in advance.

Top of page

Ubiquitous Sensing based on Wireless Sensor Network and its application of Wireless eHealth (WeHealth)
Prof. Guixia Kang, Beijing University of Posts and Telecommunications

Abstract: As one of the key application areas of ubiquitous networks/ wireless sensor networks, wireless eHealth (WeHealth) was proposed by Beijing University of Posts and Telecommuncations in 2005. This presentation is to give an introduction of the background of wireless eHealth, the R&D works of BUPT on WeHealth, etc. A related international collaborative project proposed by BUPT and McGill University is sponsed by MOST and Quebec government recently.

Top of page

Efficient Modeling of Acoustic Variability in Automatic Speech Recognition
Prof. Richard Rose, McGill University

Abstract: Speech and language processing systems have matured to the point where commercial voice enabled services are familiar to all of us. Interactive human-machine dialog systems provide voice access to online information from mobile phones. Automated processing of recorded human-human interactions provides the ability to archive and search repositories of online media. However, it is generally acknowledged that the data driven statistical methods for training automatic speech recognition (ASR) systems suffer from the need for large expensive speech corpora and from their sensitivity to variations in speaker characteristics and acoustic conditions. As a result, the cost of configuring an ASR system for a given task domain can be quite high.

This presentation motivates the use of subspace methods for acoustic modeling of speaker variability and phonetic variability in ASR. Subspace methods will be presented as a means for acquiring prior knowledge about these sources of variability from training data so that their influence on acoustic model parameters can be made to lie in low dimensional subspaces. Variation due to a particular speaker or acoustic class can be described as simple projections within these subspaces.

It will be shown that incorporating these subspace parameters has the effect of reducing the amount of data required for configuring ASR system for a particular speaker population or application domain.

Top of page

Cross-layer adaptive solutions to energy optimized multi-hop wireless ad hoc sensor networks
Prof. Swades De, IIT Delhi

Abstract: In this presentation, I shall begin with outlining my research directions in wireless ad hoc networks and broadband access networks. I shall then present our recent and ongoing work on multi-criteria forwarding optimization in energy-constrained sensor networks.

First, in an ad hoc sensor network with homogeneous coverage range, to relax the constraint and overhead of transmitter controlled relaying decision, I shall present our investigation on receiver-side relay election approach with a single criteria forwarding optimization, e.g., greedy one-hop Euclidean distance progress. I shall show how a suitable mapping can be found to achieve an optimum relay election performance.

To accommodate the additional constraints in relaying decision making, such as link quality, remaining energy, I shall discuss a multi-criteria based relaying decision approach in wireless multi-hop forwarding, where a generic cost metric is introduced in the form of a multi-parameter mapping function, that aggregates all decision criteria into a single virtual criterion to rank potential relay candidates.

Taking greediness-versus-link-quality two-criteria optimization example, I shall show that a suitable mapping function can be found, which trades off greediness for link quality to obtain optimal end-to-end network performance. Further, I shall discuss two heuristics to accommodate the nodal as well as network lifetime maximization criteria.

Top of page

Energy efficient DSP: results on novel filters as a first step
Prof. François Gagnon, Ecole de technologie supérieure

Abstract: In recent years, the growing number of operation modes, module types and applications that battery-powered wireless communication systems support made low-power, high-performance digital signal processing (DSP) an important research area. In this presentation, we introduce two major contributions to design short size and low power consuming digital filters. These new circuits are unrelated to any existing canonical or organized architecture. We first introduce a new model that ensures an unique representation of any existing circuit realization.

Then, we present an evolutive metaheuristic that automatically design optimized multiplierless filters by arranging a limited quantity of component of a given circuit. This algorithm is able to produce low-pass filters that contains less than 20 adders and 20 delay elements and that treat an input signal as accurately as Infinite Impulse Response filters of degree 4 to 10. It is thus possible to attain lower power consumption, better execution speed and to occupy less chip place than the filters designed with traditional methods.

Experiments also shows that our approach can be used to design any possible type of filters: it has been used to build circuits using less than 12 adders which emulate, with a weak estimation error, 1000 randomly generated filters.

Top of page