def test_mbstoi(filename, regtest):
filename = os.path.dirname(os.path.abspath(__file__)) + "/data/" + filename
data = scipy.io.loadmat(filename, squeeze_me=True)
sii = MBSTOI.mbstoi(data["xl"], data["xr"], data["yl"], data["yr"])
npt.assert_allclose(sii, data["sii"], atol=ATOL, rtol=RTOL)
print(utils.hash_numpy(sii), file=regtest)
def test_FilterbankAndRecruitment(filename, regtest):
filename = os.path.dirname(os.path.abspath(__file__)) + "/data/" + filename
data = scipy.io.loadmat(filename, squeeze_me=False)
filtrecruited = filterbank_and_recruitment(
data["coch_sig"],
data["SPLequiv0dBfile"][0],
data["nchans"],
data["erbn_centfrq"][0],
data["fs"],
data["ngamma"],
data["gtn_denoms"],
data["gtn_nums"],
data["gtn_delays"][0],
data["start2poleHP"],
data["hp_denoms"],
data["hp_nums"],
data["expnsn_ratios"][0],
data["eq_loud_db"][0],
data["recombination_dB"],
)
sdr = utils.compute_siSDR(filtrecruited.flatten(), data["coch_sig_out"].flatten())
assert sdr > MIN_SDR
if all(x == 1.0 for x in data["expnsn_ratios"][0]):
# No recruitment, results should be very, very close
npt.assert_allclose(filtrecruited, data["coch_sig_out"])
else:
# With recruitment, small diffs between MATLAB and Python
# due to differences in low pass filter design.
npt.assert_allclose(filtrecruited, data["coch_sig_out"], atol=1e-8, rtol=0.02)
print(utils.hash_numpy(filtrecruited), file=regtest)
def test_audfilt(filename, regtest):
filename = os.path.dirname(os.path.abspath(__file__)) + "/data/" + filename
data = scipy.io.loadmat(filename)
audfiltered = MSBG.audfilt(data["rl"], data["ru"], data["sampfreq"],
data["asize"])
npt.assert_allclose(audfiltered, data["filter"])
print(utils.hash_numpy(audfiltered), file=regtest)
def test_ec(filename, regtest):
filename = os.path.dirname(os.path.abspath(__file__)) + "/data/" + filename
data = scipy.io.loadmat(filename, squeeze_me=True)
xl_hat = data["xl_hat"]
N = data["N"]
J = data["J"]
d = np.zeros((J, np.shape(xl_hat)[1] - N + 1))
p_ec_max = np.zeros((J, np.shape(xl_hat)[1] - N + 1))
d1, p_ec_max1 = MBSTOI.ec(
xl_hat,
data["xr_hat"],
data["yl_hat"],
data["yr_hat"],
J,
N,
data["fids"],
data["cf"],
data["taus"],
data["ntaus"],
data["gammas"],
data["ngammas"],
d,
p_ec_max,
data["sigma_epsilon"],
data["sigma_delta"],
)
npt.assert_allclose(d1, data["d"], atol=ATOL, rtol=RTOL)
npt.assert_allclose(p_ec_max1, data["p_ec_max"], atol=ATOL, rtol=RTOL)
print(utils.hash_numpy(d1), file=regtest)
def test_stft(filename, regtest):
filename = os.path.dirname(os.path.abspath(__file__)) + "/data/" + filename
data = scipy.io.loadmat(filename)
win_size = int(data["N"])
fft_size = int(data["N_fft"])
stft_out = MBSTOI.stft(data["x"], win_size, fft_size)
npt.assert_allclose(stft_out, data["x_stdft"], atol=ATOL, rtol=RTOL)
print(utils.hash_numpy(stft_out), file=regtest)
def test_generate_key_percent(filename, regtest):
filename = os.path.dirname(os.path.abspath(__file__)) + "/data/" + filename
data = scipy.io.loadmat(filename, squeeze_me=True)
key, used_thr_dB = MSBG.generate_key_percent(
data["sig"], data["thr_dB"], data["winlen"]
)
npt.assert_allclose(key, data["key"] - 1, rtol=RTOL)
npt.assert_allclose(used_thr_dB, data["used_thr_dB"], rtol=RTOL)
print(utils.hash_numpy(key), file=regtest)
print(used_thr_dB, file=regtest)
def test_gains(filename, regtest):
filename = os.path.dirname(os.path.abspath(__file__)) + "/data/" + filename
data = scipy.io.loadmat(filename)
g = GHA.gains(
data["compr_thr_inputs"],
data["compr_thr_gains"],
data["compression_ratios"],
data["levels"],
)
npt.assert_allclose(g, data["g"], atol=ATOL, rtol=RTOL)
print(utils.hash_numpy(g), file=regtest)
def test_measure_rms(filename, regtest):
filename = os.path.dirname(os.path.abspath(__file__)) + "/data/" + filename
data = scipy.io.loadmat(filename, squeeze_me=True)
rms, key, rel_dB_thresh, active = MSBG.measure_rms(data["signal"],
data["fs"],
data["dB_rel_rms"])
npt.assert_allclose(rms, data["rms"], rtol=RTOL)
npt.assert_allclose(key, data["key"] - 1,
rtol=RTOL) # index so subtract 1 OK
npt.assert_allclose(rel_dB_thresh, data["rel_dB_thresh"], rtol=RTOL)
print(rms, file=regtest)
print(utils.hash_numpy(key), file=regtest)
print(rel_dB_thresh, file=regtest)
def test_thirdoct(filename, regtest):
filename = os.path.dirname(os.path.abspath(__file__)) + "/data/" + filename
data = scipy.io.loadmat(filename)
fs = int(data["fs"])
N_fft = int(data["N_fft"])
numBands = int(data["numBands"])
mn = int(data["mn"])
obm, cf, fids, freq_low, freq_high = MBSTOI.thirdoct(
fs, N_fft, numBands, mn)
npt.assert_allclose(obm, data["A"], atol=ATOL, rtol=RTOL)
npt.assert_allclose(cf, data["cf"], atol=ATOL, rtol=RTOL)
npt.assert_allclose(fids, data["fids"], atol=ATOL, rtol=RTOL)
print(utils.hash_numpy(obm), file=regtest)
def test_gen_eh2008_speech_noise_supplied_b(filename, regtest):
np.random.seed(0)
filename = os.path.dirname(os.path.abspath(__file__)) + "/data/" + filename
data = scipy.io.loadmat(filename, squeeze_me=True)
# Use filter computed by MATLAB.
eh2008_nse = MSBG.gen_eh2008_speech_noise(data["durn"],
data["fs"],
supplied_b=data["b"])
assert len(eh2008_nse) == len(data["eh2008_nse"])
sdr = utils.compute_siSDR(eh2008_nse, data["eh2008_nse"])
assert sdr > MIN_SDR
npt.assert_allclose(eh2008_nse, data["eh2008_nse"], atol=ATOL, rtol=RTOL)
print(utils.hash_numpy(eh2008_nse), file=regtest)
def test_remove_silent_frames(filename, regtest):
filename = os.path.dirname(os.path.abspath(__file__)) + "/data/" + filename
data = scipy.io.loadmat(filename, squeeze_me=True)
xl_sil, xr_sil, yl_sil, yr_sil = MBSTOI.remove_silent_frames(
data["xl"],
data["xr"],
data["yl"],
data["yr"],
data["range"],
data["N"],
data["K"],
)
npt.assert_allclose(xl_sil, data["xl_sil"], atol=ATOL, rtol=RTOL)
npt.assert_allclose(xr_sil, data["xr_sil"], atol=ATOL, rtol=RTOL)
npt.assert_allclose(yl_sil, data["yl_sil"], atol=ATOL, rtol=RTOL)
npt.assert_allclose(xr_sil, data["xr_sil"], atol=ATOL, rtol=RTOL)
print(utils.hash_numpy(xl_sil), file=regtest)
def test_gainrule_camfit_compr2(filename, regtest):
filename = os.path.dirname(os.path.abspath(__file__)) + "/data/" + filename
data = scipy.io.loadmat(filename)
sFitModel = {}
sFitModel["frequencies"] = np.array([
177.0, 297.0, 500.0, 841.0, 1414.0, 2378.0, 3999.99976, 6726.99951,
11314.0
])
sFitModel["edge_frequencies"] = np.array([
9.99999994e-09,
2.29279297e02,
3.85356964e02,
6.48459717e02,
1.09049243e03,
1.83370972e03,
3.08415283e03,
5.18729199e03,
8.72406250e03,
1.00000010e07,
])
sFitModel["levels"] = np.linspace(-10, 110, 121)
l = np.array([45, 45, 35, 45, 60, 65, 70, 65])
r = np.array([40, 40, 45, 45, 60, 65, 80, 80])
audf = np.array([250, 500, 1000, 2000, 3000, 4000, 6000, 8000])
noisegatelevels = np.array([38, 38, 36, 37, 32, 26, 23, 22, 8])
noisegateslope = 0
level = 0 # variable
audiogram = GHA.Audiogram(levels_l=l, levels_r=r, cfs=audf)
sGt = GHA.gainrule_camfit_compr(audiogram,
sFitModel,
noisegatelevels,
noisegateslope,
level=level)
newl = sGt["sGt"][:, :, 0]
newr = sGt["sGt"][:, :, 1]
npt.assert_allclose(newl, data["l"], atol=ATOL, rtol=RTOL)
npt.assert_allclose(newr, data["r"], atol=ATOL, rtol=RTOL)
print(utils.hash_numpy(newl), file=regtest)
def test_get_gaintable(filename, regtest):
filename = os.path.dirname(
os.path.abspath(__file__)) + "/auddata/" + filename
data = scipy.io.loadmat(filename)
audfn = filename.split("/")[-1]
audfn = "aud" + audfn[3:]
audfilename = os.path.dirname(filename) + "/" + audfn
# Set audiogram frequencies
if "ex" in audfilename:
audf = [125, 250, 500, 1000, 2000, 4000, 6000, 8000]
else:
audf = [125, 250, 500, 750, 1000, 1500, 2000, 3000, 4000, 6000, 8000]
# If importing audiogram from mat file
dirname = os.path.dirname(__file__)
dirname = os.path.split(dirname)[0] # go to parent folder GHA
filename = os.path.join(dirname, audfilename)
auddata = spio.loadmat(filename)
audiogram = GHA.Audiogram(levels_l=auddata["L"],
levels_r=auddata["R"],
cfs=np.array(audf))
noisegatelevels = 45
noisegateslope = 1
output = GHA.get_gaintable(audiogram,
noisegatelevels,
noisegateslope,
cr_level=110,
max_output_level=100)
npt.assert_allclose(output["sGt"][0:9, :],
data["gaintable1"][0:9, :],
atol=ATOL,
rtol=RTOL)
npt.assert_allclose(output["sGt"][9:, :],
data["gaintable1"][9:, :],
atol=ATOL,
rtol=RTOL)
print(utils.hash_numpy(output["sGt"]), file=regtest)
def test_multifit_apply_noisegate(filename, regtest):
filename = os.path.dirname(os.path.abspath(__file__)) + "/data/" + filename
data = scipy.io.loadmat(filename)
noisegatelevel = np.transpose(
np.tile(np.array(data["sGt"].item()[1][0][0][0][0][0][0]), (2, 1)))
noisegateslope = np.transpose(
np.tile(np.array(data["sGt"].item()[1][0][0][0][0][0][1]), (2, 1)))
l = data["sGt"]["l"].item()
sGt = np.zeros((np.shape(l)[0], np.shape(l)[1], 2))
sGt[:, :, 0] = l
sGt[:, :, 1] = data["sGt"]["r"].item()
output = GHA.multifit_apply_noisegate(
sGt,
data["sGt"]["frequencies"][0][0],
data["sGt"]["levels"][0][0][0],
noisegatelevel,
noisegateslope,
)
npt.assert_allclose(output["sGt"], sGt, atol=ATOL, rtol=RTOL)
print(utils.hash_numpy(output["sGt"]), file=regtest)
def test_smear3(filename, regtest):
filename = os.path.dirname(os.path.abspath(__file__)) + "/data/" + filename
data = scipy.io.loadmat(filename, squeeze_me=True)
outbuffer = MSBG.smear3(data["fsmear"], data["inbuffer"])
npt.assert_allclose(outbuffer, data["outbuffer"], rtol=RTOL)
print(utils.hash_numpy(outbuffer), file=regtest)
def test_isothr(filename, regtest):
filename = os.path.dirname(os.path.abspath(__file__)) + "/data/" + filename
data = scipy.io.loadmat(filename)
vIsoThrDB = GHA.isothr(data["vsDesF"])
npt.assert_allclose(vIsoThrDB, data["vIsoThrDB"], atol=ATOL, rtol=RTOL)
print(utils.hash_numpy(vIsoThrDB), file=regtest)
def test_freq_interp_sh(filename, regtest):
filename = os.path.dirname(os.path.abspath(__file__)) + "/data/" + filename
data = scipy.io.loadmat(filename)
y = GHA.freq_interp_sh(data["f_in"], data["y_in"], data["f"])
npt.assert_allclose(y, data["y"], atol=ATOL, rtol=RTOL)
print(utils.hash_numpy(y), file=regtest)
