Anindya Sankar Paul

Graduate Research Assistant
Adaptive Systems Lab (ADSYL)
Department of Computer Science and Electrical Engineering (CSEE)
OGI School of Science & Engineering (OGI)
Oregon Health & Science University (OHSU)
20000 N.W. Walker Road
Beaverton, OR 97006 USA
Office: CSE Central Building # 127
Telephone: (503) 748-1925 (Office)
Fax: (503) 748-1306
E-mail: anindya at csee dot ogi dot edu
URL: http://www.csee.ogi.edu/~anindya/

Education Research Interests Current Research Project Previous Research Projects Professional Activities Thesis Publications CV MISC

Education

Ph.D Candidate : Electrical Engineering
OGI School of Science & Engineering , Oregon Health and Science University, (Beaverton, OR, USA).
Degree Expected: 2008

M.Sc.Eng : Electrical Engineering
Wright State University (Dayton, OH, USA).
Degree Completed: 2003

B.Eng : Electronics and Communication Engineering
Sikkim Manipal Institute of Technology (India).
Degree Completed: 2001

Research Interests

I am interested in various Machine learning, statistical pattern recognition and optimization algorithms, neural networks, probabilistic inference and Bayesian learning including their applications on localization and tracking, signal/image processing, simultaneous localization and mapping (SLAM), automatic target detection/recognition and speech enhancement/recognition. Currently I am working under Dr. Eric Wan in Adaptive Systems Lab on Sigma-Point Kalman smoother based unobtrusive indoor location tracking using Wi-Fi received signal strength identification (RSSI) and ultrasonic sonar range sensors.

Current Research Project

My research project currently involves in developing an unobtrusive indoor location tracking system using a number of low cost noisy sensor observations. As GPS performance is limited in indoor environments, estimating the location of people and tracking them poses a fundamental challenge in ubiquitious computing. This work is supported in part by Intel as part of the Behavioral and Interventions Common (BAIC). This unique academic-industrial collaboration aims to focuss on developing a remote monitoring and assistive technologies for the elederly to keep them safe at their own home environment. Recently the potential for inhome-montoring systems for older people has grabbed the national attention. Here are the two recent articles published in NYtimes and WallStreet Journal which illustrated our contribution on building a remote monitoring system in health care.
   We developed a new system for RSSI based indoor localization and tracking. At the core of our system is a novel location and tracking algorithm using sigma-point Kalman smoother (SPKS) based Bayesian inference approach. We used our recently proposed forward-backward statistically linearized sigma-point Kalman smoother (FBSL-SPKS) and Fixed-lag (FL-SPKS) sigma-point Kalman smoother algorithms for this purpose. The person to be tracked carry a small body-borne RSSI tag which periodically measures the received signal strength from 3 or more standard Wi-Fi access points placed at predefined locations. Although RSSI tags are used as the primary sensors, augmenting the RSSI measurements are infra-red (IR) motion sensors mounted on the walls which provide a binary "on" signal when motion occures within its range. Similarly foot-switches are also placed on the floor which triggers when stepped on. We also created a room-wall model involving potential field which repels the motion away from walls. The proposed SPKS based estimator optimally fuses a model of walking motion, room-wall configurations and all sensor observations in order to track a person. Implementation and testing of our system was performed at several "living laboratories" (called Point-of-Care labs) and tracking accuracy was demonstrated to be superior to the available industry positioning engine developed by Ekahau Inc . For detailed algorithm and results, please see the Publications section.
   One of the major problems associated with tag based positioning and tracking algorithm is that the tag needs to be carried by the person while walking. From a number of trial runs at real home scenario, we observed that the people quite often forget to carry the tag and they often misplace it. Video based tracking is one solution to this problem but it concerns the users privacy. Hence the next phase of this project involved in developing a new and truely unobtrusive tracking system where there is no longer need to carry a tag. We chose ultrasonic range sensors to perform this task as they are cheap and can be easily wall mounted to a number of houses. We used the SPKS based recursive Bayesian framework for tracking and had to solve a number of data acquisition and signal processing challanges in order to use the low cost sonar sensors as a tracking device. Recently we demonstrated the feasibility of buliding a truely unobtrusive sonar based tracking system with satisfactory tracking accuracy. We will depict the performance results using sonar sensors in our next publications.
   Except the above mentioned projects, I keep myself busy with a number of theoretical analysis. We derived a new formulation for Forward-Backward and Rauch-Tung-Striebel smoothing scheme for nonlinear systems. Our approach is more accurate, direct and also bears less computational load than other nonlinear smoothers available in the literature. We have shown that our proposed smoother improves performance in various applications like multiharmonic frequency tracking and people tracking using unobtrusive sensors. Currently we are working towards a detailed journal article on this topic. Recently we proved that the estimation error of the sigma-point Kalman filter (SPKF) is bounded if certain criterias and assumptions are met. In our method, the stability and convergence of the SPKF are demonstrated from the first principle and the proof did not assume any special cases. We hope to submit this as a journal article before I graduate.

Previous Research Projects

I worked on a project which proposes developing new adaptive/statistical signal processing techniques to assist machine automated diagnosis of cardiac causes (e.g. Congestive Heart Failure) of dyspnea (shortness of breath) in the emergency department setting. This project was jointly funded by Inovise Medical Inc. and School of Medicine, OHSU. In this pilot study, both accoustic sensors data and electrocardiogram (ECG) signals were used for heart function characterization. Currently cardiologists listen the heart sound through stethoscope in order to form a diagnosis of the heart failure. As this form of diagnosis is highly subjective and depends on Cardiologist's past experience, the recent trend is to automate the heart sound diagnosis operation and make heart sound auscultation more quantitative. Inovise Medical Inc. developed a product named "Audicor" which performs an automated collection and analysis of heard sounds in the context of a standard 12 lead diagnostic ECG test. Microphones are placed at 2 ECG locations and acoustical and electrical signals are then analyzed to determine the presence of abnormal heart sounds. This analysis is highly complicated due to the presence of extraneous noise in the audio signal. The fatal heart sounds are generally low-amplitude and low-energy signals that can be easily buried under the noise floor and hence can be missed by the diagnosis. Our job on this project was to investigate advanced signal processing techniques in order to develop a robust noise reduction front end which would enhance the heart sounds with minimal signal distortion. Our approach was based on spectral domain Minimum Mean Square Error (MMSE) estimation, where an adaptive Wiener gain is applied to the spectral amplitudes. The noise spectrum was estimated online without the need for a separate reference signal and used to compute the posterior Signal to noise ratio (SNR) for adapting the Wiener gain. This MMSE based heart sound enhancement approach is shown to be effectively eliminate various interferring noise sources while keeping the critical low-amplitude heard sounds intact and hence improves diagnosis. For detail, please see the publications section.
I also worked on a project that addresses a dual kalman filter based method for autonomous terrain aided navigation in unknown environment. We recently proved that dual sigma-point kalman filter based implementation without having any priori knowledge of state and environment not only converges faster to the true state and map but also at a lower mean square error than that of the Extended Kalman filter based algorithm. Here is a short abstract of the project.
Further, I was involved in a number of projects listed below:
- Synthetic Aperture Radar (SAR) Target Detection using Subspace Filtering.
- Automatic Target Recognition (ATR) using High Range Resolution (HRR) profiles.
- Development of a New Hybridized ATR Algorithm.
- Synthetic Target and Clutter Statistical Modeling.
- Speaker Recognition using F0 Manipulation.
Brief description of the above mentioned projects can be found here. For more detail, see the relevant publications.

Professional Activities

Professional Membership

Member, IEEE Signal Processing Society, Jan 2006-present
Member, IEEE Computational Intelligence Society, Jan 2005-present
Member, IEEE Communication Society, Jan 2001-Dec 2003
Member, IEEE Computer Society, Jan 2004-Dec 2004
Student Member, IEEE, Jan 2001-present

Review Activities

Reviewer, IEEE Signal Processing Society
Reviewer, IEEE Aerospace and Electronic Systems Society
Reviewer, IEEE Potential

Thesis

Ph.D thesis
Coming soon!

MS thesis
I successfully defended my MSc thesis "Improved Target Recognition and Target Detection Algorithms using HRR Profiles and SAR Images" pdf in September, 2003. Dr. Arnab K. Shaw, Dr. Kefu Xue in Wright State University and Dr. Atindra K. Mitra in Wright-Patterson Air Force Base were my thesis advisors. Here is the slides of my thesis defense.

Publications

Peer reviewed published papers:

Sunghan Kim, Anindya S. Paul, Eric A. Wan and James Mcnames, "Multiharmonic Tracking Using Sigma-Point Kalman Filter," Accepted in 30th Annual International Conference of the IEEE Engineering in Medicine and Biology Society 2008 (IEEE EMBC 08), Vancouver, Canada, August, 2008. Full Paper
Anindya S. Paul and Eric A. Wan, "Wi-Fi Based Indoor Localization and Tracking Using Sigma-Point Kalman Filtering Methods," In proceedings of IEEE/ION Position Location and Navigation Symposium 2008 (PLANS 2008), Monterey, CA, USA, May, 2008. Full Paper
Anindya S. Paul and Eric A. Wan, "A new formulation for nonlinear forward-backward smoothing," In proceedings of IEEE International Conference on Acoustics, Speech and Siganl Processing 2008 (ICASSP 2008), pp. 3621- 3624, Las Vegas, NV, USA, March-April, 2008. Full Paper
Misha Pavel, Tamara Hayes, Ishan Tsay, Deniz Erdogmus, Anindya S. Paul, Nicole Larimer, Holly Jimision and John Nutt, "Continous Assessment of Gait Velocity in Parkinson's Disease from Unobtrusive Measurements," In proceedings of the 3rd International IEEE EMBS Conference on Neural Engineering, pp. 700-703, Kohala Coast, Hawaii, USA, May, 2007. Full Paper
Anindya S. Paul, Eric A. Wan and Alex T. Nelson, "Noise Reduction For Heart Sounds Using a Modified Minimum-Mean Squared Error Estimator with ECG Gating," In proceedings of 28th IEEE EMBS Annual International Conference, pp. 3385- 3390, New York City, USA, August-September, 2006. Full Paper
Anindya S. Paul and Arnab K. Shaw, "Subspace-based Clutter Filtering for Improved SAR Target Detection," In Proceedings of Signal and Image Processing (SIP 2006), Honolulu, Hawaii, USA, August 2006. Full Paper
Anindya S. Paul, "Dual Kalman Filter for Autonomous Terrain Aided Navigation in Unknown Environment," Technical Report, OGI School of Science and Engineering, Oregon Health and Science University, Beaverton, OR, May 2005. Full Report
Anindya S. Paul and Eric A. Wan, "Dual Kalman Filters for Autonomous Terrain Aided Navigation in Unknown Environment," In proceedings of International Joint Conference on Neural Networks (IJCNN), vol. 5, pp. 2784- 2789, Montreal, Canada, July-August, 2005. Full Paper
Anindya S. Paul, Arnab K. Shaw, Koel Das and Atindra K. Mitra, "Improved HRR-ATR Using Hybridization of HMM and Eigen Template Based Matched Filtering," In Proceedings of International Conference on Acoustics, Speech, and Signal Processing (ICASSP), Hong Kong, vol. 2, pp. 397-400, April 6-10, 2003. Full Paper
Atindra K. Mitra, Thomas L. Lewis, Anindya S. Paul and Arnab K. Shaw, "Self Training Algorithms for Ultra-Wideband Radar Target Detection," In Proceedings of Synthetic Aperture Radar Imagery X, SPIE, vol. 5095, pp. 254-264, Orlando, FL, 2003. Full Paper
Anindya S. Paul and Arnab K. Shaw, "Robust HRR Radar Target Identification by Hybridization of HMM and Eigen-Template Based Matched Filtering," In Proceedings of Automatic Target Recognition XIII, SPIE, vol. 5094, pp. 278-289, Orlando, FL, 2003. Full Paper
Atindra K. Mitra, Thomas L. Lewis, Anindya S. Paul and Arnab K. Shaw, "Ultra-Wideband Radar Data Models and Target Detection with Adaptive Rank Order Filters," In Proceedings of Synthetic Aperture Radar Imagery IX, SPIE , vol. 4727, pp. 112-121, Orlando, FL, April, 2002. Full Paper

Papers currently under review:

Sunghan Kim, Anindya S. Paul, Eric A. Wan and James Mcnames, "A New Multiharmonic Frequency Tracking Using the Sigma-Point Kalman Smoother," Under review in IEEE Transactions of Signal Processing. Abstract
Arnab K. Shaw, Anindya S. Paul and Rob Williams, "Eigen-Template based Matched Filtering (ETMF) for Robust HRR-ATR," Under review in IEEE Transactions of Aerospace and Electronic Systems (AES). Abstract
Anindya S. Paul and Arnab K. Shaw, "Robust HRR Radar Target Identification by Hybridization of HMM and Eigen-Template based Matched Filtering," Under review in IEEE Transactions of Aerospace and Electronic Systems (AES). Abstract
Atindra K. Mitra, Anindya S. Paul, Arnab K. Shaw and Thomas L. Lewis, "Improved SAR Target Detection Using Subspace Filtering," Under review in IEEE Transactions of Aerospace and Electronic Systems (AES). Abstract

Miscellaneous Stuff and Links of Interest

ReBEL: an excellent Matlab® toolkit for sequential Bayesian estimation in general state space models.
Greg Welch and Gary Bishop's excellent Kalman Filtering web-site.
Want to learn Latex: An excellent computer program for typesetting document. Visit this to learn by 15 minutes.
Darpa urban challenge was held on November 3rd, 2007. Four teams finished the course within the stipulated time. Congrats to Carnegie Mellon for being the winner.
Want to know about space explorations and our surrounding planets: see Planetary Society.
Want to be updated about Phoenix Mars Lander which is exploring the arctic plains of Mars to search for ice under the soil: visit NASA.
Have lots of questions in mind (like I do) regarding creation and evolution of the universe and our planet earth: see professor Stephen Hawking's theory.
Want to learn new subjects/concepts/ideas from MIT without even spending a penny: see MITOpenCouseware.
Kevin Murphy maintains an excellent page devoted to guidelines for prospective graduate students: visit here.
Want to know the life of a PhD student in United States: follow PhDComics. they are funny but depicts the real picture!
Bored huh! May be you are working too hard. Let's have some Jokes.
Want to be surrounded with natural beauties, visit Portland: I love to live here.
I love Portland's K103 fm: The Soft Rock Music Station. Recently I am leaning towards smooth jazz because of this.
Here is my collection of Bengali websites. You can rediscover our favorite city kolkata through these sites. Happy browsing!