Subtract the mean of each signal to center it around zero. This reduces the sharpening of the signal.
Apply a Butterworth filter to retain frequencies between 0.5 Hz and 20 Hz. This removes noise and irrelevant components.
Scale values from 0 to 1.
Downsample the signal. Remove high frequencies before downsampling using a low-pass filter.
Decompose signals into components at different frequency bands.
Use the Daubechies wavelet (db4).
Apply four levels of decomposition to capture frequencies ranging from 0.5 Hz to 20 Hz.
- Training Accuracy: 100%
- Test Accuracy: 90%