Summer Intern, IBM Software Labs, Bangalore

I looked into the feasibility of using knowledge discovery and machine learning algorithms to perform data mining and knowledge discovery on clinical data which can come in structured and unstructured formats. We also developed a key word extraction algorithm for key word discovery in the data.

May '04 - July '04

Teaching and Mentoring

Publications

Conference

"BriMon: A Sensor Network System for Railway Bridge Monitoring", Kameswari Chebrolu, Bhaskaran Raman, Nilesh Mishra, Phani Kumar Valiveti and Raj Kumar WWW 2007. [under review]

"Wake-on-WLAN'', Nilesh Mishra, Kameswari Chebrolu, Bhaskaran Raman, and Abhinav Pathak, The 15th Annual Interntional World Wide Web Conference (WWW 2006), May 2006, Edinburgh, Scotland. [Acceptance rate: 11%].

"S-WOW: Signature based Wake-on-WLAN", Nilesh Mishra, Dheeraj Golchha, Akhilesh Bhadauria, Bhaskaran Raman, Kameswari Chebrolu, The First Annual Workshop on WIreless Systems: Advanced Research and Development (WISARD 2007), A COMSWARE 2007 Workshop, Jan 2007, Bangalore, India. [Acceptance rate: 50%]

Book Chapters

"Off-Line Signature Recognition" Indrani Chakravarty, Nilesh Mishra, Mayank Vatsa, Richa Singh, and P. Gupta; Vol. 2, Encyclopedia of Data Warehousing and Mining, Editor: John Wang, Idea Group Reference.

"Online Signature Recognition" Indrani Chakravarty, Nilesh Mishra, Mayank Vatsa, Richa Singh, and P. Gupta; Vol. 2, Encyclopedia of Data Warehousing and Mining, Editor: John Wang, Idea Group Reference.

Poster/Demo

"S-WOW: Signature based Wake-on-WLAN", Nilesh Mishra, Akhilesh Bhadauria, Dheeraj Golchha, The First Annual Workshop on WIreless Systems: Advanced Research and Development (WISARD 2007), A COMSWARE 2007 Workshop, Jan 2007, Bangalore, India.

Thesis

Condensed Abstract: We propose a system design for remote, on-demand railway bridge monitoring system which is easier to deploy and autonomous in its operation. The system achieves low cost using off the shelf equipments and replacing expensive sensing equipments with Micro Electronic Mechanical Systems (MEMS) based sensors and Wireless Sensor Networks (WSN) for the cabling and data logging system. Our approach using independent nodes attached with appropriate sensors and batteries makes the system easy to deploy, without hassles of cabling, as seen in existing wired solutions. Using wireless data transmission and multi-hop data transport our solution is highly scalable both in terms of span and number of sensing nodes. Use of low power devices also enable our design to have a longer maintenance cycle.

A modified implementation of the Flooding Time Synchronization Protocol (FTSP) is used. These modifications reduce the average error at nodes farther from the root as well as improve the network wide synchronization stability. A Wake-on-WLAN based event detection mechanism detects an incoming train. The train acts both as mechanical shaker for data collection as well as data transporter from remote bridge location to a data repository. The event detection method also turns on the MEMS sensor just in time for data collection thus saving power. Results reflecting the achievable distances for successful operation of Wake-on-WLAN and comparison of MEMS accelerometers with existing Forced Balance Accelerometers (FBA) show that our design is viable for future deployment. Overall our approach achieves more than 96% cost savings over currently used wired solutions.

Major Projects

BriMon: Structural Health Monitoring of Railway Bridges
September 2005 - August 2006 (Team of 2), Current [Team of 3]
Supervisor: Prof. Bhaskaran Raman, Assistant Professor, Department of CSE and Prof. Kameswari Chebrolu, Assistant Professor, Department of EE, IIT Kanpur.
We use tri-axial accelerometers and sensor motes to remotely monitor the health of a railway bridge. The motes are programmed to collect the data for free, forced and ambient vibrations and upload them to a monitoring car periodically. This data is further analyzed at the base server to identify the natural frequency of the bridge.

K-Mote: IIT-K sensor mote
August 2006 - Current [Team of 3]
Supervisor: Prof. Bhaskaran Raman, Assistant Professor, Prof. Rajat Moona, Professor, Department of CSE and Prof. Kameswari Chebrolu, Assistant Professor, Department of EE, IIT Kanpur.
We are building a sensor mote platform around the TI's MSP430 microcontroller and CC2420 radio. The mote is being built keeping in mind the requirements of current projects undergoing at our group (BriMon, WiBeaM). The target is to gain a first hand experience in building hardware platforms which will be evolved as per our requirements. We want to develop a low cost kit which can be provided to colleges and educational institutions, enabling them to work on sensor network applications as learning and knowledge gaining tools.

Wake-on-WLAN
July - December 2005 (Single)
Supervisor: Prof. Bhaskaran Raman, Assistant Professor, Department of CSE and Prof. Kameswari Chebrolu, Assistant Professor, Department of EE, IIT Kanpur.
The 802.11 (Wifi) protocol was not designed to be power efficient. The current off the shelf equipments require the whole protocol stack to be switched on even when the radio only needs to listen. Also the computing subsystems currently being used do not allow wakeup via wireless Ethernet packet. In this work we proposed a method to switch on the Soekris computing platform and the 802.11 card at remote location using a sensor mote. The mote performs carrier sense of 802.11 traffic using its 802.15.4 compliant radio and switches on the power supply for the Soekris using a relay switch.

Digital Accelerograph Generation Software
September 2005 - Present (Team of 3)
Supervisor: Dr. K. K. Bajpai, Senior Scientific Officer,  Structures Lab and Prof. S. K. Jain, Professor, Department of Civil Engineering, IIT Kanpur With the move to digital photography, photographic films are getting produced in lesser and lesser quantities. India is having more than 200 strong motion accelerometers (SMA) required to measure structural response of a building to earthquakes. They use special photographic films, which is getting obsolete and difficult to procure, to capture the events. We have digitized the image capturing module of the accelerometers using a CCD array and are able to take online data which can now be stored in databases in digital form on a computer and managed as well recalled in a much efficient fashion than the current manual method deployed.  

RFID and NFC based applications
May - September 2005 (Team of 2)
Supervisor: Prof. Bhaskaran Raman, Assistant Professor, Department of CSE and Prof. Kameswari Chebrolu, Assistant Professor, Department of EE, IIT Kanpur.
We looked into the feasible applications for RFID (Radio Frequency Identification) and its derivative latest technology NFC (Near Field Communication) in various domains. We have been successful in creating a RFID reader around ST's CRX14 reader chip and an access monitoring application around Philips NFC module. This project gave us footing in handling of PCB (Printed Circuit Board) designing, handling of SMD (Surface Mount Device) and antenna building.

Corner Point Detection using SVM
July - December 2004 (Single)
Supervisor: Prof. Pabitra Mitra, Assistant Professor, Department of CSE, IIT Kanpur.
In this work we proposed an algorithm to do corner point detection in grayscale images using a nonlinear svm classifier. It is based on computing the direction of maximum gray-level change for each edge pixel in an image, and then representing the edge pixel by a four dimensional feature vector. A support vector machine is designed using this feature vectors and the support vectors, representing critical points in a classification problem, correspond to the corner points. The algorithm is straightforward and does not involve computation of complex differential geometric operators.

Biometric Offline Signature Verification System
May 2002 - April 2005 (Single)
Supervisor: Prof. Phalguni Gupta, Professor, Department of CSE, IIT Kanpur.
We implemented an offline signature verification system which was able to give an accuracy of 93%. This system was successfully developed and deployed for both acquisitions via hand held PDA 'Simputer' and image scanners. This system is currently being used in the biometrics lab IIT Kanpur.

Medical Knowledge Discovery from Clinical Data,
May - July 2004 (Team of 2)
Mentor: Dinesh A Venkateswaran,  IBM Software Labs, Bangalore
Medical records, both in structured and un-structured form, are rich in data. A knowledge base can thus be created using the information contained in these records. In this work we looked into various techniques of machine learning and knowledge discovery useful for extracting such information. The ultimate goal of this system is to device a assistant software which can be used as a help to doctors. My work involved studying the knowledge content of such data and use of data mining techniques for recovering this knowledge.

• Stood 173^rd (99.86 percentile amongst 126,000 students) in the All India IIT Joint Entrance Examination 2001. Around 400,000 students appeared in 2006.
• Best Poster award for 'Wake-on-WLAN', at poster session TechVista 2007, Microsoft Research India's annual research symposium.
• Best Poster award for 'BRIMON: Railway BRIdge MONitoring', at WISARD 2007, a workshop in COMSWARE 2007.
• Certificate of Appreciation from Academic Senate, IIT Kanpur for my performance as a Tutor of ESC101N (Fundamentals of Computation) during the 2005-06-II Semester.
• Secured 1^st rank in 2^nd All India National Science Olympiad (12^th class PCM category) 2000.
• Was selected for National Chemistry Olympiad after securing rank in top 1% in Indian Chemistry Olympiad (IChO 2000), national examination to choose representatives of India in International Chemistry Olympiad.
• Ranked 19^th - Regional Mathematical Olympiad (RMO), state examination to choose representatives of India in International Maths Olympiad, in MP state in the year 1999.
• Ranked 31^st - MP State Science Quiz 1997 (Was selected in standard 9^th itself for a competition open for standard 9^th and 10^th)
• Placed 1^st - in the 12^th ISC 2000 board exams in the school.