ESP32 Forum

Hello, I only have a conceptual question.

Suppose I am recording wifi CSI data where I keep the 100 most recents frames of wifi CSI data in an array.

So row 1 represents wifi CSI at frame 't' (now), row 2 is the CSI data previously recorded so on and so forth. How would one get CIR and CFR from this CSI matrix. I read that FFT(CSI) is CFR and IFFT(CSI) is CIR, but there is no example of code on the ESP32 for me to dissect and understand. I have heard some even say that the matrix itself is CIR since I am reading the CSI over time, but I feel like IFFT is a mandatory step and the logic presented is incorrect.

Can someone explain to me how I am to derive CFR and CIR? I think the idea of matrix itself is CIR, and to obtain CFR, all I have to do is take a FFT of the matrix for each column.

I think that with CSI matrix I have described above,

I am just trying to build a neural network that takes the CSI data from 2 ESP32s and predicts if there is a line of sight present between the 2 boards, so this is a necessary step for my feature extraction.

Regards, JJ Satti

Ok, I think I know what to do now.

An array of consecutive CSI measurements by itself carries no meaning. We only see each carrier's
performance over time as term CSI(t). We can then apply FFT to each carrier to get the CFR matrix;
The CFR matrix is a frequency domain representation of our CSI data. To obtain the CIR matrix, we
can then apply IFFT to each column in our CFR matrix.

If I am wrong, please correct me!!!