Jun 2020 - Jun 2022
Jun 2020 - Jun 2022
Me - at MEDICA 2021, Düsseldorf, Germany
Eitan and Ron showing the Balance system
I did a lot of UI work.
As part of my job, I also did the interesting algorithms.
The plane of motion is identified by determining the resting orientation and initiating movement in the intended exercise direction. Subsequent orientations are linearly projected onto this plane, enabling the calculation of the angle of deviation from the resting position. The resting orientation is determined by averaging orientation data during calibration, reducing noise, and ensuring maximum accuracy.
The device has two sensors: an accelerometer and a gyroscope.
The accelerometer is used for low-frequency measurements, and the gyroscope for high-frequency measurements. The Hysteresis Threshold is used to switch between the calculations.
The movement is circular, so it is a sine wave in time.
The midline is determined by averaging all the last data points.
The amplitude is calculated when the user crosses the midline by identifying the point closest to the peak. Using its time, we can determine its position on the graph and normalize the value accordingly.
We have some types of frequencies:
Midline crossing time frequency - the best one. Exactly half a revolution passes between two crossings. Therefore, the frequency is equal to the reciprocal of two times the time difference between them.
Soft frequency - Calculated based on the estimated distance and time elapsed since crossing the midline one before the last.
Smart frequency - Calculated by Linear interpolation between the midline crossing time frequency and the soft frequency. Based on the distance of the data from the last midline crossing.