Somesh Srivastava
1813 Plymouth Rd, Apt#A1
Ann Arbor, MI. 48105
Phone(off): (734) 763 4497
Phone(res): (734) 995 1376
E-mail: someshs AT umich DOT edu
 

Somesh Srivastava


 

Current Status

Graduate Student (Candidate) in the Electrical Engineering (Systems) Department at University of Michigan, Ann Arbor.


Research Interests

Statisical image reconstruction algorithms, X-ray computed tomography, and large-scale constrained and unconstrained optimization methods.


Work Experience

July 2001 - June 2002: Software Engineer in Motorola India DSP Group.
July 2000 - June 2001: Software Engineer in Motorola India Network Management – Wireless Infrastructure(NM-WI) Group.


Signal Processing and related Engineering Projects

A Comparative Study of Input Feature Selection using PWFS and MIFS (based on Mutual Information)
When : February, 2002 – March, 2002
Where: Course Project for EECS 559, EECS Deptt., University of Michigan, Ann Arbor
The project involved the study of the data set and the Neural Networks based classfication algorithm as described in "Analysis of Hidden Units in a Layered Network Trained to Classify Sonar Targets" by Paul Gorman and Terrence J. Sejnowski. Then, a study of two Input Feature Selection algorithms (PWFS and MIFS) based on the Mutual Information criterion was carried out. MATLAB's Neural Network toolbox was studied and an attempt to simulate the network was made. The aforesaid input feature selection algorithms were implemented and their performance was tested using the Fisher Linear Discriminant Criterion, a k-Nearest Neighbor Classifier and a Linear Discriminant Classifier. MATLAB was used in this project.

Wavelet Based Image Restoration using EM Algorithm
When : November, 2002 – December, 2002
Where: Course Project for EECS 551, EECS Deptt., University of Michigan, Ann Arbor
The project involved study and verification of the results of the paper "Image Restoration Using the EM Algoritm and Wavelet-Based Complexity Regularisation" by M.A.Figueiredo and Robert Novak. A back-ground study of Image Restoration, Optimization Transfer and EM-Algorithm was done. Then, the Wavelet-Based Image restoration method presented in the above paper was studied and implemented. It was shown that the above algorithm performed better than Quadratic and Non-Quadratic Penalized Least Squares using Separable Paraboloidal Surrogates Algorithms(as implemented by Dr.Jeffrey Fessler's Group. Code was downloaded from his web-site). The project also involved implementation of 2-D Translation-Invariant Discrete Wavelet Transforms. MATLAB was used in this project.

Adaptive Signal Processing Techniques For Noise Suppression Using TMS320C30 DSP Kit
When : October, 1999 – April, 2000
Where: Department of Electronics and Computer Engineering, IIT Roorkee (then, University of Roorkee)
The project involved real-time experiments on noise suppression using Adaptive Filters. A proof-of-concept Active Noise Control System was developed using the Texas Instruments (TI), TMS320C30 DSP kit. It used the Filtered-X Least Mean Square (FX-LMS) algorithm and secondary path modeling using an internal noise sequence generator. Detailed studies on the Convergence Rates of various Adaptive Filtering algorithms, Polyphase FFT techniques and Delayless Sub-band Adaptive Filter Architecture were also carried out. Simulation software was developed for the Delayless Sub-band Adaptive Filter Architecture using C and MATLAB. The real-time experiments and the Active Noise Control System were developed using the TMS320C30 assembly language and TI DSP tools. This project is a compulsory part of the Bachelor of Engineering Degree and is done in the Senior Year of the course.

MPEG-2 AAC Decoder
When : September, 2001 – December, 2001
Where: Motorola India DSP Group
The project involves developing and optimizing the MPEG-2 Advanced Audio Coding (AAC) decoder for Motorola's HawkV2 Digital Signal Processor (DSP) chip. DSP algorithms for the various components of the AAC decoder are chosen after a thorough investigation of the various alternatives. The chosen ones are then implemented on the HawkV2 processor keeping it’s architecture in mind so as to minimize the processor time and memory requirement. Motorola HawkV2(DSP56800E) Assembly language and C are used in the Metrowerks Code Warrior IDE in it.


Software Projects

MIB-Parser
The project involved the upgradation of a parser for MIBs(Management Information Bases) from Simple Network Management Protocol (SNMP) version 1(SNMPv1) to SNMP version 2(SNMPv2c). Requirements for the MIB-Parser were identified after a study of the various Request For Comments (RFC) for SNMP. Theory behind parsers and lexers was studied and the parser generator tool ANTLR was learnt. The complete test case and test tool development for the MIB-parser was carried out. Duration: August, 2000 – January, 2001

PC-Evaluation Module (EVM) Universal Serial Bus (USB) Interfacing
The project involved development of PC and EVM software to create a full speed USB interface. The technical specification of USB 2.0 was studied and Windows Driver Model (WDM), VC++, Win32, MFC Programming was learnt. The development of the interface involved from-the-scratch development work on both the PC and EVM. A throughput of 768kbps was achieved and designs for higher bit-rates were drawn up.

MIB-Browser
The project involved the design and coding of a Graphical User Interface (GUI) to browse a MIB (Management Information Base) and execute SNMP queries. Designs were done in the Universal Markup Language (UML) and code was written in JAVA. JAVA Swing was used for the GUI.

SNMPv3 Stack Test Tools
The project involved the development of test tools to test a SNMPv3 stack for functionality and security features. JAVA and shell-scripting were used on Windows and Solaris platforms.

SNMP Agent Maintenance Project
The project involved the maintenance of SNMP Agents for 3G equipment. C, Unix Sockets, shell scripting and an SNMP agent framework were used to develop code for agents running under various Operating Systems like Solaris, Lynx and VxWorks.


Computer Networking Projects

Performance Analysis Of Centralized Heuristics For Multicast Path Setup In Data Networks
The project was done as a part of the compulsory two month long Summer Training for the Bachelor of Engineering degree. It involved development of a simulation software for comparing the performance of two heuristics for multicast path setup, namely Shortest Path Heuristic (SPH) and Kruskal Shortest Path Heuristic (K-SPH). It is an application of Graph Theory in multicast routing in Asynchronous Transfer Mode (ATM) networks. It was an extension of the work done for the publication Routing Techniques In IRIDIUM Like Satellite Networks (see Publications below). C and Turbo-C graphics were used in it.

Routing Techniques In IRIDIUM Like Satellite Networks
See Publications below


Publications

Paper titled Routing Techniques In IRIDIUM Like Satellite Networks was presented at INFLUX 99, the technical festival of Electrical Engineering Department at University of Roorkee (now IIT Roorkee) held from 29th to 31st January, 1999. The paper discussed the issues involved in routing in satellite networks like IRIDIUM which are low earth orbit satellites and use Asynchronous Transfer Mode(ATM). The simulation software developed for the paper involved the simulation of the satellite constellation geometry and computation of path sequence between any two satellites over the entire system-period. It is an application of Graph Theory for point-to-point routing in ATM Networks. This work was extended for multicast routing in the Summer training project titled Performance Analysis Of Centralized Heuristics For Multicast Path Setup In Data Networks (see Computer Networking Projects above). The computational and visualization programs were written in C and Turbo-C graphics was used.


Practical Training

Performance Analysis Of Centralized Heuristics For Multicast Path Setup In Data Networks
See Computer Networking Projects above.

Object Oriented Analysis and Design (OOAD) using Universal Markup Language (UML)
It was a 5 day course conducted by Rational Inc. in association with Motorola. Object Oriented Analysis concepts and UML specification were covered and a design of a “University Report Card System” was carried out using OOAD and UML.


Technology and Software Skills

Languages
Pascal, C, MATLAB, C++, VC++, JAVA, HTML, Assembly(80x86,TI DSP TMS320C30, Motorola DSP56800E), UML, Shell Scripting and Lisp.

Technologies
Lexers and Parsers, ANTLR, OOAD, WDM(Windows Driver Model), Graphical User Interface development using JAVA Swing and VC++, XML and socket programming over UNIX.

Protocols
Simple Network Management Protocol (SNMP) v1,v2c and v3, Universal serial Bus (USB) Protocol

Operating Systems
Unix (Solaris), Windows, Lynx and VxWorks


Education

Year Degree Details
2002- Master of Science Currently pursuing Master of Science followed by Ph.D. at Deptartment of Electrical Engineering(Systems), University of Michigan, Ann Arbor.
1996 - 2000 Bachelor of Engineering Bachelor of Engineering(Electronics and Communiations Engg) Degree from IIT, Roorkee (then, University of Roorkee). Stood in First Division securing 74% marks.
1995 - 1996 AISSCE All India Senior Secondary Certificate Examination. National Level examination conducted by the Central Board of Secondary Education, India to certify a student as having passed the 12th Grade successfully. Secured 94% marks and stood First in school.
1993-1994 AISSE All India Secondary School Examination. National Level examination conducted by the Central Board of Secondary Education, India to certify a student as having passed the 10th Grade successfully. Secured 85% Marks.

Courses Taken at University of Michigan, Ann Arbor
Probability and Random Processes (EECS 501)
Wavelets and Time Frequency Methods (EECS 551)
Medical Imaging Systems (EECS 516)
Image Processing (EECS 556)
Advanced Signal Processing (EECS 559)
Estimation, Filtering and Detection (EECS 564)
Adaptive Signal Processing (EECS 659)
Special topics in signal processing - Advanced topics in imaging (EECS 755)
Advanced calculus (MATH 451)
Analysis II (MATH 597)
 
 
Function space methods in systems theory (EECS 600)
 
Fourier optics (EECS 435)
Topics in applied statistics (STAT 701)
Functional analysis (MATH 602)  

Scholarships

National talent Search Exam (NTSE) Scholarship
This scholarship is awarded by the National Council for Educational Research and Training (NCERT) to students completing their 10th Grade. 750 students are chosen after a 3-stage national level test which includes an interview at the final stage.

Certificate of Merit in Physics
This certificate is awarded by Central Board of Secondary Education (CBSE) to students completing their 12th Grade. It is awarded to the students placed in top 0.1% in the country.

University Merit Scholarship
This scholarship is awarded by the University of Roorkee (now IIT, Roorkee) to the top 25% students of a class.

Fellowship Award
This scholarship is awarded by the Department of Electrical Engg and Comp Science, University of Michigan, Ann Arbor to new Graduate Students.


Other Academic Achievements

Achieved a national rank of 47 in the Roorkee Entrance Exam (REE). REE was a nation-wide entrance exam conducted by the University of Roorkee(now IIT, Roorkee) for admission into the Under-Graduate Engineering courses. About 80,000 students appear in the exam every year.