def test_create_mel_fb(self,
n_mels=40,
sample_rate=22050,
n_fft=2048,
fmin=0.0,
fmax=8000.0,
norm=None,
mel_scale="htk"):
if (norm == "slaney" and
StrictVersion(librosa.__version__) < StrictVersion("0.7.2")):
self.skipTest(
'Test is known to fail with older versions of librosa.')
if self.device != 'cpu':
self.skipTest('No need to run this test on CUDA')
expected = librosa.filters.mel(sr=sample_rate,
n_fft=n_fft,
n_mels=n_mels,
fmax=fmax,
fmin=fmin,
htk=mel_scale == "htk",
norm=norm).T
result = F.melscale_fbanks(sample_rate=sample_rate,
n_mels=n_mels,
f_max=fmax,
f_min=fmin,
n_freqs=(n_fft // 2 + 1),
norm=norm,
mel_scale=mel_scale)
self.assertEqual(result,
torch.from_numpy(expected),
atol=7e-5,
rtol=1.3e-6)
def func(_):
n_stft = 100
f_min = 0.0
f_max = 20.0
n_mels = 10
sample_rate = 16000
norm = "slaney"
return F.melscale_fbanks(n_stft, f_min, f_max, n_mels, sample_rate, norm)
def test_melscale_fbanks_warning(self):
with warnings.catch_warnings(record=True) as w:
warnings.simplefilter("always")
F.melscale_fbanks(201, 0, 8000, 128, 16000)
assert len(w) == 1
def test_melscale_fbanks_no_warning_high_n_freq(self):
with warnings.catch_warnings(record=True) as w:
warnings.simplefilter("always")
F.melscale_fbanks(288, 0, 8000, 128, 16000)
assert len(w) == 0
# ``torchaudio.functional.melscale_fbanks`` generates the filter bank
# for converting frequency bins to mel-scale bins.
#
# Since this function does not require input audio/features, there is no
# equivalent transform in ``torchaudio.transforms``.
#
n_fft = 256
n_mels = 64
sample_rate = 6000
mel_filters = F.melscale_fbanks(
int(n_fft // 2 + 1),
n_mels=n_mels,
f_min=0.0,
f_max=sample_rate / 2.0,
sample_rate=sample_rate,
norm="slaney",
)
plot_mel_fbank(mel_filters, "Mel Filter Bank - torchaudio")
######################################################################
# Comparison against librosa
# ~~~~~~~~~~~~~~~~~~~~~~~~~~
#
# For reference, here is the equivalent way to get the mel filter bank
# with ``librosa``.
#
mel_filters_librosa = librosa.filters.mel(
