def call(self, audio_data, sample_rate=None):
"""
Caculate pitch features of audio data.
:param audio_data: the audio signal from which to compute spectrum. Should be an (1, N) tensor.
:param sample_rate: [option]the samplerate of the signal we working with, default is 16kHz.
:return: A float tensor of size (1, num_frames) containing pitch features of every frame in speech.
"""
p = self.config
with tf.name_scope('pitch'):
if sample_rate == None:
sample_rate = tf.constant(p.sample_rate, dtype=float)
assert_op = tf.compat.v1.assert_equal(
tf.constant(p.sample_rate), tf.cast(sample_rate, dtype=float))
with tf.control_dependencies([assert_op]):
pitch = py_x_ops.pitch(
audio_data,
sample_rate,
window_length=p.window_length,
frame_length=p.frame_length,
thres_autoc=p.thres_autoc)
pitch = tf.squeeze(pitch)
pitch = tf.transpose(pitch[None, :])
return pitch
def test_pitch(self):
''' test pitch op'''
with self.session():
# read wave
sample_rate, input_data = feat_lib.load_wav(self.wavpath, sr=16000)
output = py_x_ops.pitch(input_data, sample_rate)
output_true = np.array([
0.000000, 0.000000, 0.000000, 0.000000, 0.000000, 0.000000, 0.000000,
0.000000, 0.000000, 0.000000, 0.000000, 0.000000, 0.000000, 0.000000,
0.000000, 0.000000, 0.000000, 0.000000, 0.000000, 0.000000,
122.823532, 117.647057, 116.788322, 116.788322, 119.402985,
119.402985, 119.402985, 119.402985, 119.402985, 123.076920,
124.031006, 125.000000, 132.065216, 139.130432, 139.130432,
137.931030, 126.108368, 114.285713, 115.107910, 122.070084,
129.032257, 130.081299, 130.081299, 129.032257, 130.081299,
131.147537, 129.032257, 125.000000, 120.300751, 115.107910
])
self.assertEqual(tf.rank(output).eval(), 1)
self.assertAllClose(output.eval().flatten()[:50], output_true)