I've been working on 3-axis accelerometer data to predict the gesture of an user.
Lot of literature pointed me to consider Frequency domain features for my problem and i'm using
- Mean of the signal on each axis first
- 'k' coefficients of DFT of the signal.
Since X(0) is essentially the DC component of the signal, i tried checking the DC component with the Mean of the signal and observed they differ in sign because of the
sqrt(X(0)*conj(X(0)))/N gives only positive value.
Will it affect my problem in any means?