Hykso punch trackers use an accelerometer and a gyroscope to do full 3D motion tracking to fully categorize the acceleration, velocity and position of each punch at each moment.
Boxing Specific Algorithms
Our motion tracking algorithm uses 3 sub-algorithms developed specifically with the punching motion in mind.
Their cumulative goal is to:
- Calculate the orientation of each hand in space.
- Translate the motion to a fixed reference frame (a.k.a. the frame everyone is used to, the earth’s frame) from the complex frame of a fist in motion.
- Remove the pesky gravitational readings from the accelerometer (this is actually one of the harder things to do, and we’ve developed a better way to remove it compared to traditional methods).
Our punch recognition and detection algorithms are the most advanced by far, as they are the subject of over a year of research and development to specifically tune them for boxing at all levels.
The first part to our motion recognition was to distinguish between the types of punches at a very high accuracy, and to do this we used something called Machine Learning, which essentially teaches a computer to think like a human being (just a lot faster than a human can think) and judge what type of punch was thrown in real time. In order to be as accurate as possible we collected tens of thousands of punches from several hundred Professional, Olympic and amateur boxers, to make sure our system is incredibly accurate and works with all levels of user.
The next part of the algorithm is to ignore all the other undesired motions, such as blocks and general hand motion (we call them false-positive punches). Boxing is super dynamic and the hands never stop moving, which made it a complex problem to remove all the unwanted motions and only truly recognize actual punches.
Our system has accuracies of over 95% in punch recognition, motion tracking, false positive removal and velocity calculations. It is trusted for use by Olympic teams and top professionals.