EEG feature extraction

💬 Support & Discussions
1

Members of the community,
With mne features, I’m extracting band power features from 19 channel eeg data. We can define different frequency ranges when using all channels to calculate a feature like ‘pow freq bands.’ But what is the output order?
Consider the following:
[freq band1, freq band2, freq band3] and [ch1, ch2, ch3, ch4].
Which one the output order?
[pow_freq_band1 of ch1, pow_freq_band1 of ch2, pow_freq_band1 of ch3, … ]
or
[pow_freq_band1 of ch1, pow_freq_band2 of ch1, pow_freq_band3 of ch1, … ]

2

Hi @FRSHT,
could you specify exactly what function you are using? If you are using compute_pow_freq_bands from mne-features then the documentation gives you the answer:

Returns:
    output: ndarray, shape (n_channels * (n_freqs - 1),)

I understand this as channels-first organisation (channels x frequencies matrix unrolled into a vector), so (ch0_freq0, ch0_freq1, ..., chN_freq0, chN_freq1).