def test_lifter(self):
fake_cepstra = np.ones((20, 13))
lifted_ceps = Amfcc.lifter(fake_cepstra, L=22)
self.assertEqual(fake_cepstra.shape, lifted_ceps.shape)
# if L negative, no lifting applied
lifted_ceps = Amfcc.lifter(fake_cepstra, L=-3)
self.assertTrue(np.array_equal(lifted_ceps, fake_cepstra))
def test_construct(self):
# If x is asig, it will ignore sr but use x.sr instead.
amfcc = Amfcc(self.test_asig, sr=45687)
self.assertEqual(amfcc.sr, 8000, msg='sr does not match.')
# if x is ndarray and sr is not given
with self.assertRaises(AttributeError):
_ = Amfcc(self.test_sig)
# x is ndarray and sr is given.
amfcc = Amfcc(self.test_sig, sr=1000)
self.assertEqual(amfcc.sr, 1000, msg="sr does not match.")
# Unsupported input type
with self.assertRaises(TypeError):
_ = Amfcc(x="String")
def test_melfb(self):
fb = Amfcc.mel_filterbanks(8000) # Using default
self.assertEqual(fb.shape, (26, 257)) # nfilters, NFFT // 2 + 1
def test_preemphasis(self):
self.assertTrue(
np.array_equal(np.array([0., 1., 1.5, 2., 2.5]),
Amfcc.preemphasis(np.arange(5), coeff=0.5)))
def test_nowindowing(self):
amfcc = Amfcc(self.test_asig, window=False)
result = np.ones((amfcc.n_per_frame, ))
self.assertTrue(np.array_equal(result, amfcc.window))
def test_nfft_not_pow2(self):
with warnings.catch_warnings(record=True):
amfcc = Amfcc(self.test_asig, nfft=23)
def test_hopsize_greater_than_npframe(self):
with warnings.catch_warnings(record=True):
amfcc = Amfcc(self.test_asig, hopsize=100, n_per_frame=50)
def test_get_attributes(self):
amfcc = Amfcc(self.test_asig)
self.assertTrue(amfcc.timestamp is not None)
self.assertTrue(amfcc.features is not None)