How to remove DC Offset for fft calculations in python?

1 Like

I recommend asking this question of the numpy/scipy communities, or in dsp.stackexchange.com.

Assuming you’re using numpy for fft calculations, removing the DC offset is a matter of subtracting the mean. For a 1-D signal, it would look something like this

```
import numpy as np
balanced_signal = biased_signal - np.mean(biased_signal)
spectrum = np.fft.rfft(balanced_signal)
```

For a windowed signal stored as a 2-D ndarray, it isn’t very different. Assuming you have a time axis as the 0th axis and frames as the 1st axis, you could do

```
import numpy as np
balanced_signal = biased_signal - np.mean(biased_signal, axis=-1)
spectrogram = np.fft.rfft(balanced_signal, axis=-1)
```

This of course assumes you’ve already made a lot of decisions about how to window the signal, etc.

1 Like

Dear Stefan,

Thank you so much for your suggestions.