Comment by rapjr9 | Hacker News Reader

Comment by rapjr9 | Hacker News Reader

I wrote software for the Amiga that did something like this for golf swings a few decades ago:

http://zoom.interoscitor.com/PetersonEnterprises/resume/logo...

http://zoom.interoscitor.com/PetersonEnterprises/Consulting/...

http://zoom.interoscitor.com/PetersonEnterprises/resume/down...

It could adjust the stick figure size to match the persons height and controlled a video tape deck via RS-232 to single step the video. The stick figure data was digitized from films of golf pro's at a high frame rate and the frames that best matched the 1/30 second frame rate of video were used. A genlock superimposed the graphics over the video. Wrote the software under contract and got a percentage, but the business never took off. It was somewhat cumbersome to use (roll the big rack shown in the picture out onto a golf course), they charged a lot, and they had no proof it actually helped improve golf swings, though they gathered some testimonials. The results of an analysis were difficult for a customer to use as well, for example "slow down your swing in the early 8 o'clock part of the swing", how do you do that? How much do you do it? It was difficult to translate the differences in swings into a kinesthetic sense of what your muscles should be doing. Perhaps adding an accelerometer wrist/ankle strap could provide additional data that would help with that.

Wooowwww! It's awesome to see vision tech being used like this! How long did this take you?? Also, is the club being used at all or just the person's stick figure? Oh and did adjusting the stick figure to match the person's height affect the feedback given? Like if someone was shorter, would it trigger cues like "Bring your club up higher on the backswing" compared to a taller person?

We talked with a baseball academy facing a similar issue where their previous software analyzed recorded videos really good, but the analytics were too confusing for coaches to understand and took too long to set up.

From what we learned, form is the main aspect. Regardless of speed, form varies by individual, and any deviations can impact the results. In our case, we fire a feedback like "Bring your hips down lower" after every rep attempt until the user completes their workout. From working with martial arts instructors and their students, short and direct feedback that guide you work the best!

And yes, would love to integrate more data for the user! Maybe using the user's apple watch features! Thank you for sharing!

It took a few months to write, but about a year to fine tune and bring it all together. The guy who hired me did the digitization of the pro golfers swings using a Sun workstation, I did the Amiga graphics and tape deck control. He digitized four pro swings, just for one club, a driver, with plans to expand later. One of the pro's was shorter and female, and there were three males of varying height. The height adjustment was usually small if you chose the right model, so we assumed any distortions didn't matter. The club is used, but there is no digital comparison of the swings, no automated analysis, it's all just done visually by the operator of the machine, and the person whose swing is being analyzed just looks at it also, frame by frame, and can see where the two don't match in limb and club positions, and hip, knee, and elbow angle. With a little training the operator knows what areas to focus on and point out to the customer. It was not much different than what trainers were already doing visually, but it greatly improved interactions with customers since they could directly see the mistakes they were making and by how much they were off. Putting a sensor on the club/bat/etc might be something to consider also. A camera would be ideal so you could see how accurately they were hitting the ball and whether they had a consistent bias that needed correcting.

Reply on news.ycombinator.com