Date: March 11, 2005

Time: 3:00 pm

Location: GCATT Room 325

Speaker(s): Rajbabu Velmurugan

Title: A Particle Filter Solution to Multiple Target Tracking using Acoustic Sensor Array

Abstract:
In this talk, we present various methods to track maneuvering targets, using their acoustic signals, received at an array of microphones. Our approach is based on a state-space formulation. The state update is formulated using a locally linear motion model. The observations are a batch of DOA estimates calculated by a beamformer that pre-processes the received acoustic data. The beamformer output is statistically modelled to account for spurious as well as missing DOA estimates. The advantages of using this specific model is presented. Next we present a particle filter formulation to solve the multi target tracking problem. This formulation is compared with (1) extended Kalman filter (EKF) based solution, and (2) a Bayesian approach that uses Laplace's method to approximate the posterior density by a Gaussian. The performances of the methods are compared using simulations on synthetic data. We conclude that the particle filter based approach performs better compared to the other two methods for the acoustic tracker.

Work performed in collaboration with Dr. Volkan Cevher and Prof. James H. McClellan.

Bio:
Rajbabu Velmurugan received the B.E. degree from Government College of Technology, Coimbatore, India in 1995 and the M.S. degree from Clarkson University, Potsdam, USA in 1998, both in electrical engineering. He worked in Larsen & Toubro Ltd., India before joining Clarkson University and in The MathWorks from 1998 to 2001. He is currently pursuing Ph.D. degree at Georgia Institute of Technology under the advisement of Prof.James H McClellan. His research interests include array signal processing, particle filtering techniques and real-time signal processing for target-tracking applications. He is interested in teaching and involved in educational projects related to signal processing.

Date:	March 11, 2005
Time:	3:00 pm
Location:	GCATT Room 325
Speaker(s):	Rajbabu Velmurugan
Title:	A Particle Filter Solution to Multiple Target Tracking using Acoustic Sensor Array
Abstract:	In this talk, we present various methods to track maneuvering targets, using their acoustic signals, received at an array of microphones. Our approach is based on a state-space formulation. The state update is formulated using a locally linear motion model. The observations are a batch of DOA estimates calculated by a beamformer that pre-processes the received acoustic data. The beamformer output is statistically modelled to account for spurious as well as missing DOA estimates. The advantages of using this specific model is presented. Next we present a particle filter formulation to solve the multi target tracking problem. This formulation is compared with (1) extended Kalman filter (EKF) based solution, and (2) a Bayesian approach that uses Laplace's method to approximate the posterior density by a Gaussian. The performances of the methods are compared using simulations on synthetic data. We conclude that the particle filter based approach performs better compared to the other two methods for the acoustic tracker. Work performed in collaboration with Dr. Volkan Cevher and Prof. James H. McClellan.
Bio:	Rajbabu Velmurugan received the B.E. degree from Government College of Technology, Coimbatore, India in 1995 and the M.S. degree from Clarkson University, Potsdam, USA in 1998, both in electrical engineering. He worked in Larsen & Toubro Ltd., India before joining Clarkson University and in The MathWorks from 1998 to 2001. He is currently pursuing Ph.D. degree at Georgia Institute of Technology under the advisement of Prof.James H McClellan. His research interests include array signal processing, particle filtering techniques and real-time signal processing for target-tracking applications. He is interested in teaching and involved in educational projects related to signal processing.