question about 'PCA_STFT_visualize.ipynb'

[thinszx]

Hi, thank you for sharing the code. I'm trying to implement CARM and your code helps a lot.

I'm a little bit confused about the data processing codes in PCA_STFT_visualize.ipynb. Does the data processing method in this script totally equal to the method in CARM paper? np.convolve is used to calculate the constant offset in the script , while the paper says it is calculated by average CSI amplitude for 4 seconds, so I think it refers to simply averaging the amplitude for 4 sec? Besides the H matrix in the paper is built by chunks obtained in 1-second interval which are cut from CSI streams, while the script used another np.convolve whose window_size is 0.1s to build this matrix (or maybe just to smooth the signal, I'm not sure). Could you help explain why these np.convolve are used? Thanks a lot.