Pages: [1]   Go Down
Author Topic: Accelerometer and/or Gyroscope for gesture sensing  (Read 402 times)
0 Members and 1 Guest are viewing this topic.
Offline Offline
Newbie
*
Karma: 0
Posts: 22
View Profile
 Bigger Bigger  Smaller Smaller  Reset Reset

I want to sense 3D motion (gesture). I have 3 axis accelerometer. I read several articles about difference between accelerometer and gyroscope but I am confused if I need a gyroscope or not for orientation and rotation sensing. I don't see why accelerometer cannot detect change in orientation.

Isn't it possible to measure them from accelerometer alone? Because when the sensor rotates along one axis (say z-axis), there will be change in y-axis and x-axis output in accelerometer.

Is gyroscope really necessary or is it just easier if I have one?

Logged

MS
Offline Offline
Newbie
*
Karma: 0
Posts: 1
View Profile
 Bigger Bigger  Smaller Smaller  Reset Reset

Hi This is my first post. OK in response...Velocity is a change in position with time. If the sensor sensed velocity, you would integrate (calculus word) the velocity time series (record of velocity with time) to get position. Acceleration is the change in velocity with time. So you would have to integrate the acceleration time series to get velocity. Hence you have to integrate a time series of acceleration twice to get position. A 2-times integration involves setting some initial conditions: ex: the object was stationary at time = 0 and the position of the object was 0 at time = 0. The last condition is easy to satisfy since we can arbitrarily set the x=0 point to anything. But the initial velocity must be known at the start of the time series. "Time series" is a time record of the velocity and/or the position of the object. The only thing that I haven't defined is "integrate". That is a term used in Calculus to indicate summing over time. A book on either Calculus or Numerical Analysis will explain how to numerically integrate. If you consult one of these books, you will find out that integration results in errors that grow with time (from t=0). The errors can be reduced by reducing the time step used in the integration, but this error is the principal drawback to using accelerometers to deduce position. I hope that this brief explanation helps...
Logged

Pages: [1]   Go Up
Jump to: