David Smith

Working on using a Kalman filter to improve my distance calculations during GPS tracked workouts. Two observations:

A) I've never felt more like the “I have no idea what I'm doing" dog meme. Math can be very difficult.

B) Having spent all day looking at GPS traces from various devices, it is striking just how good the dual band receiver in an Apple Watch Ultra is compared to older iPhones. My processing has almost affect on most Ultra traces, but can massively improve iPhone traces.

Nov 14, 2023 at 2:20pmIvory for Mac
Show threadveaderNov 14, 2023
@_Davidsmith The accuracy of the Ultra was what made me pickup one. As a runner I still had a dedicated running watch until the Ultra allowed me to retire it for the one-watch lifestyle.
Show threadAlex HallNov 14, 2023
@_Davidsmith Pawject... I love that you put a pun in the image description.
Show threadDavid SmithNov 14, 2023
@alexhall 😁 Hehe. I wasn't sure how many folks would read it, but seemed like a funny little "easter egg" to hide.
Show threadYadiel SotomayorNov 14, 2023
@_Davidsmith Pawject... Going to steal that one.
Show threadBenjamin HerrinNov 14, 2023
@_Davidsmith if you want some data I’m happy to share some of mine with my iPhone 15 and Ultra (1st gen) I’m in Orlando the next few days for a conference then going camping and hiking over the weekend.
Show threadDavid SmithNov 14, 2023
@benjaminherrin Thanks! but the nice thing about the way Health stores past workout routes is that I actually have a massive compendium of testing data which I've bee inadvertently collecting for years. Which I think provides me enough data to work from.
Show threadBenjamin HerrinNov 14, 2023
@_Davidsmith oh that’s fascinating! Super cool that you can dive back into that data!
Show threadPaulNov 14, 2023
@_Davidsmith Fun fact. Kalman filters are also used to track particles in the detectors used in particle physics experiments.
Show threadJakob KirkegaardNov 14, 2023
@_Davidsmith Check out the following for a practical hands-on introduction https://github.com/rlabbe/Kalman-and-Bayesian-Filters-in-Python
GitHub - rlabbe/Kalman-and-Bayesian-Filters-in-Python: Kalman Filter book using Jupyter Notebook. Focuses on building intuition and experience, not formal proofs. Includes Kalman filters,extended Kalman filters, unscented Kalman filters, particle filters, and more. All exercises include solutions.

Kalman Filter book using Jupyter Notebook. Focuses on building intuition and experience, not formal proofs. Includes Kalman filters,extended Kalman filters, unscented Kalman filters, particle filt...

GitHub
Show threadDavid SmithNov 14, 2023
@perceptoid Thank you, I'll take a look!
Show threadMark AlldrittNov 14, 2023
@_Davidsmith I can appreciate how you are feeling. Lately, I’ve been working on real-time EEG processing on the Apple Watch and the math very quickly moves past what I understand.
Show threadCat Lady Kydia Music 🇺🇦Nov 17, 2023
@_Davidsmith Yay, I found the pun you were referring to in your toot about ALTtext! 
Show threadJukka PartanenNov 18, 2023
@_Davidsmith Other really problematic cases are the routes in urban environment, which contain a lot of jumps even with an iPhone 14 Pro. I've been using a simple median filter, but Kalman seems to produce smoother results. Thanks for giving a push to try a Kalman filter! I can really relate to A… 😄