def _MakeLogMelFromTensorflowBuiltin(tf_wav_bytes):
sample_rate, audio = audio_lib.DecodeWav(tf_wav_bytes)
static_sample_rate = 16000
with tf.control_dependencies(
[tf.assert_equal(sample_rate, static_sample_rate)]):
log_mel = audio_lib.AudioToMfcc(static_sample_rate, audio, 25, 25, 40)
return log_mel
def testAudioToMfcc(self):
with open(
test_helper.test_src_dir_path('tools/testdata/gan_or_vae.wav'),
'rb') as f:
wav = f.read()
sample_rate, audio = audio_lib.DecodeWav(wav)
static_sample_rate = 24000
mfcc = audio_lib.AudioToMfcc(static_sample_rate, audio, 32, 25, 40)
with self.session() as sess:
audio_sample_rate, mfcc = sess.run([sample_rate, mfcc])
assert audio_sample_rate == static_sample_rate
self.assertAllEqual(mfcc.shape, [1, 126, 40])
