Signal segmentation

[Bob_Schor]

I've spent many years working on an inertial guidance system that has linear and angular acclerometers (look at my Avatar).  I think you might not be clearly distinguishing between angular and linear motion.  Moving in a linear way (say, by squatting) will induce a small linear acceleration going down, but potentially a larger one going up (as you can "jump" and potentially achieve "lift-off", meaning your acceleration is > 1g; it's very difficult (= impossible) to squat that fast.  Another way to achieve significant acceleration with little motion is to tilt (slowly), changing the 1g gravity vector alignment with the accelerometer.  Note that if you rotate quickly, you have 3(!) accelerations to consider -- tangential acceleration, centripetal acceleration, and gravitational acceleration.

Maybe it's time to learn some Mechanics ...

Bob Schor

[LaBoris]

I was wondering what that avatar is, however, regarding the squat experiment was done just to find that it can not be properly classified.

So the point is that yaw pitch roll analysis is not appropriate but that it would be better to analyze raw data from an accelerometer, gyroscope and magnetometer?

[Bob_Schor]

The point is that when trying to instrument a physical system (such as a human) with a sensor (such as a multi-degree-of-freedom movement/orientation device), you need to clearly define how the sensor is mounted on the physical system, the behavior of the physical system (which involves how it moves, including rotations and translations), and the questions being asked.  You still haven't said what you are trying to measure and what questions you are trying to ask/answer.

Bob Schor

[DelWilson]

Go to the functions menu, select Signal Processing, then Signal Analysis.  You should find the tools you need.  Use Auto Power Spectrum or Power Spectrum for energy for openers.  Use curser control for finding peaks in graphic presentations...