Difference between revisions of "User:Akanter"
m (1 revision) |
|
(No difference)
| |
Latest revision as of 06:25, 12 June 2015
Contents
Personal Information
A Seattle native and senior at Oberlin College and Conservatory, attempting a B.A. in Computer Science and a B.M. in Classical Performance (Double Bass). He is working with Dr. Sheikh Iqbal Ahamed for the Summer 2011 REU at MU.
He enjoys playing/listening to music, playing games of all types (board, video, etc.), watching movies, sunshine, and long walks on the beach.
Milestones
- Week 1: Research on how to proceed with project
- Week 2: Survey state of the art, research how to write program
- Week 4: Finish development of data-collection program
- Week 5-6: Collect data
- Week 6: Analysis of Data
- Week 6-9: Further collection of data and/or analysis
- Week 9: Write paper
- Week 10: Odds 'n Ends
Week 1: Up to Speed
Tuesday, May 31
- Orientation, etc.
Wednesday, June 1
- Meeting with Dr. Iqbal
- Platform research (Android vs. iOS)
- Research on gait-matching algorithms using accelerometers and gyroscopes (IEEE/ACM articles)
Thursday, June 2
- Read more IEEE/ACM articles
- Looked at previous datasets provided by Dr. Iqbal; one requires further information because the README is in a non-latin language
- Started learning how to work with Matlab so I can use the 2nd dataset
- Looked into Android vs iOS ways of interfacing with sensors
- Research talk w/ Dr. Factor and Dr. Brylow
- Library demo/tour
Friday, June 3
- Read more articles, books
- Meeting with Dr. Iqbal re: milestones, current research, platform
Week 2: More learning/research
Monday, June 6
- Read more
- Begin looking into and working on making the Android data collection program
Tuesday, June 7
- More reading
- Finished a hopefully working data collection program; I experienced problems using the Android emulator, but the Nexus S should be arriving soon for hardware testing.
Wednesday, June 8
- More reading
- Went through some Matlab tutorials
Thursday, June 9
- More reading
- Got a semi-working version of the program on the phone, started figuring out the bugs
Friday, June 10
- More reading
- Bug ironing
Week 3: Data Collection START
Monday, June 13
- Wrote up trial methodology
- Added a timing feature (instead of a manual shutoff) for the data collection program
- Went over old resources
Tuesday, June 14
- Research paper/presentation talk w/ Dr. Brylow
- Research on some other data collection ideas based on Dr. Ahamed's ideas.
- Bug squishing
Wednesday, June 15
- Redesigned some features
- Bugs squashed!
Thursday, June 16
- Redesigned timing feature
- NSF Mandated Research Talk
- Work-in lunch
- Data collection testing
Friday, June 17
- Looking into how to most effectively use gyroscope and accelerometer data together.
Week 4: Working with Data
Monday, June 20
- Still researching best ways to use gyroscope and accelerometer data
Tuesday, June 21
- Submitted proposal for data collection
- Read/re-read IEEE/ACM articles
Wednesday, June 22
- Data collection
Thursday, June 23
- Data collection
Friday, June 24
- Researched potential feature vectors
- Worked on trying to demonstrate some of these in Matlab
Week 5: Analysis!
Monday, June 27
- Read the Griffiths book that finally came in from the library
- More vector research/getting Matlab to cooperate
Tuesday, June 28
- Worked on Powerpoint presentation
- More vector research
Wednesday, June 29
- Met with advisor
- Finished Powerpoint presentation
- Worked on manipulating data through Matlab
Thursday, June 30
- Rehearsed and put final touches on presentation
- Presentations
- More matlab
Friday, July 1
- More matlab
- IEEE article exploration
Week 6: MatLab/DSP is difficult
Tuesday, July 5
- Met with Dr. Ahamed, Dr. Corliss
- IEEE Exploration/Further reading
- Talk about how to make a good poster
Wednesday, July 6
- Went to gait analysis lab at MCW
- More reading
Thursday, July 7
- Trying to analyze data
- Reading
Friday, July 8
- Met with Marek Trawicki for help with DSP
- Met with Dr. Ahamed
- Looked more at data, especially to determine the preprocessing window for filtering
Week 7
Monday, July 11
- Looked at more data, tried different levels of filtering
- Reading
- Tried to get Fast Fourier Transforms of my data to output correctly, in hopes of finding a feature there
Tuesday, July 12
- Finished program/script to quickly produce FFT of the data
- Looked for features
- Reading
Wednesday, July 13
- Continued looking for features
- Reading, reading, reading
Thursday, July 14
- Completed NIH training
- Reading
- Feature hunting
Friday, July 15
- Met with Dr. Iqbal
- Matlab/Feature hunting
Week 8: Crunch time
Monday, July 18
- Met with Dr. Trawicki again
- More Matlab
Tuesday, July 19
- Collected some more data - Turns out I should have done a LOT more extensive data collection.
- More Matlab (neural networks? Discrete Cosine transform? xcorr?)
Wednesday, July 20
- More of the same
Thursday, July 21
- More of the same
Friday, July 22
- Still more of the same
Saturday, July 23
- Met with Tanvir
- More of the same
Sunday, July 24
- More of the same
Week 9
Monday, July 25
- Met with Dr. Johnson, an expert in DSP
- Met with Dr. Ahamed
- Started exploring some of the things Dr. Johnson mentioned
Tuesday, July 26
- Worked on poster
- Continued previous work
Wednesday, July 27
- Finished poster
- Continued previous work
Thursday, July 28
- Made some modifications to the data collection program
- Continued previous work
Sunday, July 31
- Worked on presentation
- Worked on paper
Week 10: The End
Monday, August 1
- Worked on presentation
- Worked on paper
- Met with Dr. Ahamed
Tuesday, August 2
- Worked on paper
- Poster presentation
Wednesday, August 3
- Gave presentation, went to others' presentations
Notes
Unsurprisingly, gyroscopes will be helpful in addition to accelerometers, though an accelerometer system alone could yield valuable results. The only currently available Android phone I found with a gyroscope was the Nexus S. The iPhone 4 and current generation of iPod Touches have gyroscopes as well, but I'd like to avoid Objective C if at all possible.
Some interesting articles: Spoof Attacks on Gait Based Authentication System Symbol-Based Approach to Gait-Analysis Another Algorithm (slightly older) Yet another algorithm from 2010 Someone stole our project last September?
Hey look someone wrote a book! MEMS Sensors Measuring Gait Parameters (March 2011) <---Not very useful
Analysis Stuff
Potential Features: Acceleration (Forces) Magnitude Composites: X-Y-Z, X-Y, X-Z, Y-Z Angular Velocity Composites, similar axes Hjorth Mobility & Complexity
SVM: I don't know if Principal Component Analysis counts as a Support Vector Machine, but it's a way to analyze the vectors.
Cycle detection: This algorithm (linked above as the 2010 algorithm) seems to ME to be the best way of determining cycles because it will be less error prone since it relies on the averaged data instead of just the Z axis (which is the "slightly older" algorithm above)
