aspcol.soundfieldestimation.est_ki_freq_rff

aspcol.soundfieldestimation.est_ki_freq_rff(p_freq, pos, pos_eval, k, reg_param, num_basis=64, rng=None, direction=None, beta=None)

Estimates the RIR in the frequency domain using random Fourier features

Uses the frequency domain sound pressure as input

Parameters:

  • p_freq (ndarray of shape (num_real_freqs, num_mics)) – sound pressure in frequency domain at num_mic microphone positions

  • pos (ndarray of shape (num_mic, 3)) – positions of the microphones

  • pos_eval (ndarray of shape (num_eval, 3)) – positions of the evaluation points

  • k (ndarray of shape (num_freq)) – wavenumbers

  • reg_param (float) – regularization parameter for kernel interpolation

  • num_basis (int) – number of basis directions to use for the random fourier features

  • direction (ndarray of shape (3,)) – direction of the directional weighting. This should be towards the source, i.e. in the opposite of the propagation direction

  • beta (float) – strength of the directional component

Returns:

est_sound_pressure – estimated RIR per frequency at the evaluation points

Return type:

ndarray of shape (num_real_freqs, num_eval)

Notes

If direction is set, the basis directions are sampled from a von Mises-Fisher distribution, with pdf p(x) = e^{-beta * direction^T x}.

References

[uenoKernel2018]