ABSTRACT
We will first look at a LKF that uses differential state equation for computing device orientation. The orientation can be estimated using different combination of input motion sensors. The complete LKF solution for orientation estimation using the input sensor combinations: (1) accelerometer + magnetometer + gyroscope, (2) accelerometer + gyroscope, (3) accelerometer + magnetometer, and (4) only accelerometer will be described based on [1,10]. The method to estimate the orientation of the device should have very quick response and settling times for use in computation of gravity and linear acceleration. We will be describing a set of accurate equations, with rationale for calculation of linear acceleration using direct projection approach. We also provide a comparison of the accuracy of the linear acceleration components computed using different orientation estimation methods and gravity computation methods with different behaviors. In the next sections, we have a detailed look at the estimation of device orientation using different combination of input sensors (Section 12.2), the projection-based solution for computation of gravity and linear acceleration (Section 12.3) based on the orientation of the device. We provide details on the test setup and comparison of the results of various test scenarios (Section 12.4). Finally, we have a look at the results and improvements of the projection solution over existing solutions (Section 12.5).
