Table of Contents
Notes
- “If you can't explain it to a six year old, you don't understand it yourself.” — Albert Einstein
- Surely there are some simple explanations to all of this…
- BioHarness BodyTrack Site (Uploaded Data)
- -100 is on your back, +100 is on your stomach
Accelerometer
-
- Much better caloric intake estimation than FitBit because they compute activity based on footsteps from the accelerometer data. Intensity is based on speed change derived from footsteps and elevation change, which they compare a pressure sensor and GPS.
- A jog only burns <500 calories…whereas just sitting or sleeping all day/night adds up quickly(1500-2000 calories)!
- A simple method for reliable footstep detection (Ryan Libby from UbiComp group at UW)
- Uses frequency domain to adaptively find the right low-pass filter cutoff. Seems to work okay…
- Play around with in Matlab, but getting adaptive part while not using for loop is difficult. Try using for loop with FFT and see how slow it is?
- Not sure how robust it is against rotation, moving in pocket while running
EKG
- Great set of articles on reading an EKG. Mauvila
- QRS Detection Algorithms that are co-authored by Willis Tompkins.
- Open Source C QRS Detection Should be pretty fast too…Matt sent it to you.
Heart Rate Variability
Sleep
- HRV isn't helpful in distinguishing REM from awake. They don't use an accelerometer, but do use some other parameters. Anne says trying to overtake Zeo is not as important as just getting HRV.
- Anne takes naps around 4 or 5 too! Or just do something else other than coding. Or have some chocolate?
Notes
| Coherence | Corposcindosis |
|---|---|
 |  |
| Optimum state supposedly When HRV and BR are synced (5RPM) | Tries to map autonomic space. Sort of works… |
* Quantified Self Numbers from the Heart, the best overview article yet.
- Goes over different measurement methods and software
- Correlates some interesting stuff, including running styles, breathing styles, and some meditation and yoga
- Sympathetic Nervous System: Low frequency HRV, “fight-or-flight” response. Unintuitive, but that's the way it is.
- Parasympathetic Nervous System: High Freq HRV, “rest and digest” response.
- Common Measurements are
- Normalized High Frequency Variability (.05 Hz to 1 Hz???): Correlated to parasympathetic nervous system
- Normalized Low Freq Variability ( to ????): Correlated to sympathetic nervous system
- LF/HF Ratio: You want parasympathetic nervous system to dominate over sympathetic, and this ratio basically tells you the balance between them. A high ratio is worse, means you're “dead-er”.
- Should use a Lomb periodogram to get more accurate frequency representation, since “it does not require resampling of unevenly sampled signals such as RR data”
- HRV analysis should be based on respiratory data. Reading a paper now, and they use Wavelet analysis too. aysinembs06.doc
- Haha, IntelliWave references this paper and talks about how it's not as good on page 39. hm…
Clinical Tests
- Doctors usually perform an orthostatic (lying-to-standing) test and a Valsalva maneuver combined with deep breathing.
Software
FirstBeat
- Patent Listings and their white papers have some great research!
Kubios HRV
- Kubios HRV Analyzer! (from the Finns and free!) Built on Matlab and offers some cool output formatting…Descriptive help document here
- Offers 21 different parameters to analyze
IntelleWave (Nerve-Express)
- Intellewave (Nerve-Express): His theory review paper is 50 pages long?! Cites Marvin Minsky's Frame Theory as inspiration.
- His patent, Method for Quantitative Assessment of Autonomic Nervous System based on HRV Analysis, goes a little more in-depth
- $4000 for Zephyr HxM Bluetooth ($100) and his proprietary software
- Just got a patent too in case the paper doesn't spill the beans.
- Nerve-Express solved a challenging problem of subdividing the Spectral Function of R-R intervals variability into two parts:
- High Frequency (which correlates with Parasympathetic activity 95% of the time) and
- Low Frequency (correlates with Sympathetic activity up to 70% of time).
In the past, the division line (between PNS and SNS) was found at a mean Frequency of 0.15 Hz (for an average healthy person). However, for each individual, this transition point varies.
Nerve-Express was able to come up with a calculation which is unique to any given person.
In addition, Nerve-Express also dealt with the problem of reducing all possible variations of multiple parameters (20-30) into a quantitative relationship between only two parameters: SNS and PSNS. Nerve-Express is the first and only system to solve this problem of SNS-PSNS quantification. This technological breakthrough is achieved by the application of Dr. Riftine's proprietary algorithms.- From product history page
Biocom Technologies
Breathing
- Pranayama is deep breathing exercises. Anne also mentioned ayurveda being a few-millenia-old Indian medicinal/spiritual practice, her agni going crazy, –> no nightshades?!
- Starts getting into Hindu stuff?
Galvanic Skin Response
- Should provide some juicy data on stress levels, but might need some intelligent filtering to remove baseline shifts and such.
- Wiki Article, lots of links and documentation!
Filtering
- Seems like people are using slope and change in frequency of oscillation.
Location
- “Palms and soles are preferred only because they have the highest concentration of eccrine glands” (and aren't as responsible for thermoregulation, and thus less baseline drift)
Working Out
- This article talks about FirstBeat's research that shows that training programs using HRV data to not push the body too hard and rest accordingly will help you perform better than the normal exercise routine…
- You can get a very good approximation of your anaerobic threshold when you start breathing hard. It's not as sensitive as actually measuring lactic acid building and VO2 output, but close enough for 99.9% of the population. Academic PaperZephyr Technology White Paper
- Seems like aerobic is good for short activities and anaerobic is better for quick bursts. Not sure though why/how you would train knowing that!
- Zephyr's paper seems pretty good Why Conduct Fitness Testing
Firstbeat Software (HRV only)
- Works with only certain watches, but their ATHLETE version is cheap <$100. Looking at the trial now.
- They've done tons of research onto the relationship between HR/HRV and oxygen consumption, energy expenditure, training effect, etc. They're able to put your whole workout into 5 different zones, mainly based on HRV I think. However, the fact that it doesn't use respiration is kinda interesting.
- Research Papers (some research uses stuff like neural networks)
- I'd like to try this software out with some real data, which apparently can be an hrm file.
- Maybe look into buying a Garmin GPS-enabled watch? Forerunner 110 and 210 are ~$200 and do FirstBeat analysis too. Hmmm…give it a few days of thought.
Sleeping
- Since we don't have EEG yet, it might be interesting to see how well we can track sleep quality from other measures such as HR or HRV and respiration rate.
- Wiki article on Sleep. WOW. Didn't know that the human biological clock is so consistent!
- Blue light plays a rather big role too! Good article on picture:
Signal Filtering
- Jennifer Healy, Intel Researcher on mood recognition. Was an Intel “rock star” :). Her paper on and Detecting Stress During Real-World Driving Tasks
-
- GSR
- Found points of interest and prompt points when there was a large change in the signal and “specific patterns in the tonic level” >.> I think they later said peak frequency…interesting
- However, to minimize annoyance, the system prompted at most every 15 minutes even if an interesting event happened.
- Time Windows
- It seems that no one can crack the nut of knowing “mood” from GSR, so this study tried to start some of that by measuring slope and frequency change. However, know that there is a (1-3 second) delay between instantaneous “startle” events and the GSR response that follow.
-
