Skip to main content

Csd

Namespace: UTILSLIB  ·  Library: Utilities Library

#include <utils/csd.h>

class UTILSLIB::Csd

Cross-Spectral Density (CSD) estimator.

Provides three methods for computing the CSD matrix between channels: multitaper (DPSS tapers), Fourier (Welch-style segmented), and Morlet wavelet.

src/examples/ex_dsp_spectral/main.cpp
MatrixXd twoChannels(2, n);
twoChannels.row(0) = signal;
for (int t = 0; t < n; ++t) {
    twoChannels(1, t) = std::cos(2.0 * kPi * 20.0 * t / kSFreq); // 20 Hz, 90 degrees ahead of channel 0
}
const CsdResult csd = Csd::computeMultitaper(twoChannels, kSFreq, 18.0, 22.0, 2.0);
const std::complex<double> cross = csd.matCsd(0, 1);

Static Methods​

computeMultitaper(matData, sfreq, fmin, fmax, halfBandwidth, nTapers)​

Compute CSD using multitaper (DPSS) spectral estimation.

Parameters:

  • matData : const Eigen::MatrixXd & Data matrix (n_channels × n_samples).

  • sfreq : double Sampling frequency in Hz.

  • fmin : double Minimum frequency of interest in Hz (default 0).

  • fmax : double Maximum frequency in Hz; -1 → Nyquist (default -1).

  • halfBandwidth : double Half-bandwidth parameter (NW) for DPSS tapers (default 4).

  • nTapers : int Number of tapers; -1 → as MNE-Python: those of the first floor(2*halfBandwidth) with concentration > 0.9 (default -1).

Returns:

computeFourier(matData, sfreq, fmin, fmax, nFft, overlap)​

Compute CSD using Welch-style Fourier segmented estimation.

Each segment is treated like an epoch of mne.time_frequency.csd_array_fourier (demeaned, Hann window, same scaling); DC is never returned.

Parameters:

  • matData : const Eigen::MatrixXd & Data matrix (n_channels × n_samples).

  • sfreq : double Sampling frequency in Hz.

  • fmin : double Minimum frequency of interest in Hz (default 0).

  • fmax : double Maximum frequency in Hz; -1 → Nyquist (default -1).

  • nFft : int FFT / segment length in samples (default 256).

  • overlap : double Fractional overlap between adjacent segments, in [0, 1) (default 0.5).

Returns:

computeMorlet(matData, sfreq, frequencies, nCycles)​

Compute CSD using Morlet wavelet convolution, scaled like mne.time_frequency.csd_array_morlet.

Parameters:

  • matData : const Eigen::MatrixXd & Data matrix (n_channels × n_samples).

  • sfreq : double Sampling frequency in Hz.

  • frequencies : const Eigen::RowVectorXd & Target frequencies in Hz.

  • nCycles : int Number of cycles in the Morlet wavelet (default 7).

Returns:

Example​

Authors of this file​