comnum.classify_modulation#

comnum.classify_modulation(x, candidates, method='ml', *, noise_var=None)#

Blind modulation classification at one sample per symbol, timing and carrier recovered.

Methods: "ml" (the likelihood, the scale and the noise estimated by M2M4 under every candidate), "ml_known" (the likelihood with the noise variance noise_var known and the symbols of unit energy, Wei & Mendel 2000), "cumulants" (the nearest fourth-order features |C40|/C21², C42/C21², Swami & Sadler 2000; with noise_var, normalised by C21 - noise_var).

Parameters:

  • x (array_like) – the samples (complex).

  • candidates (list[Constellation]) – the candidate alphabets (scaled to unit energy).

  • method (str) – the classifier.

  • noise_var (float, optional) – the noise variance, for "ml_known" and "cumulants".

Returns:

index (the decided candidate) and scores (one per candidate, the largest wins).

Return type:

dict

Raises:

ValueError – for an invalid parameter or input (the message names it).

Examples

>>> import comnum, numpy as np
>>> rng = comnum.Rng(1)
>>> cands = [comnum.Constellation(k, m).normalized() for k, m in [("psk", 2), ("qam", 4), ("qam", 16)]]
>>> x = cands[2].points[rng.integers(16, 2000)] + rng.complex_noise(2000, 0.02)
>>> comnum.classify_modulation(x, cands)["index"]
2