aspcol.soundfieldestimation_jax.estimate_from_regressor

aspcol.soundfieldestimation_jax.estimate_from_regressor(regressor, pos, pos_eval, wave_num, dir_coeffs=None)

Takes the regressor from inf_dimensional_shd_dynamic, and gives back a sound field estimate.

Gives the same result as inf_dimensional_shd_dynamic, but is much faster since computing the regressor is the primary computational cost. Implements the method in J. Brunnström, M.B. Moeller, M. Moonen, “Bayesian sound field estimation using moving microphones”

Parameters:

  • regressor (ndarray of shape (num_real_freqs, N)) – The regressor calculated by inf_dimensional_shd_dynamic. Represents Phi* v in eq (31) from [brunnstromBayesian2024]

  • pos (ndarray of shape (N, 3)) – position of the trajectory for each sample

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

  • samplerate (int)

  • c (float) – speed of sound

  • dir_coeffs (ndarray of shape (N, num_coeffs), optional) – harmonic coefficients of microphone directivity, Note that a higher number of coefficients will drastically increase the computational cost If not provided, the microphones are assumed to be omnidirectional.

Returns:

est_sound_pressure – time-domain harmonic coefficients of the estimated sound field

Return type:

ndarray of shape (num_real_freqs, num_eval)

Notes

Assumptions: The data is measured over an integer number of periods of the sequence N = seq_len * M, where M is the number of periods that was measured The length of sequence is the length of the estimated RIR