def _load_wav_with_speed(wav_file, speed):
""" Load the wave from file and apply speed perpturbation
Args:
wav_file: input feature, T * F 2D
Returns:
augmented feature
"""
if speed == 1.0:
return torchaudio.load_wav(wav_file)
else:
si, _ = torchaudio.info(wav_file)
# get torchaudio version
ta_no = torchaudio.__version__.split(".")
ta_version = 100 * int(ta_no[0]) + 10 * int(ta_no[1])
if ta_version < 80:
# Note: deprecated in torchaudio>=0.8.0
E = sox_effects.SoxEffectsChain()
E.append_effect_to_chain('speed', speed)
E.append_effect_to_chain("rate", si.rate)
E.set_input_file(wav_file)
wav, sr = E.sox_build_flow_effects()
else:
# Note: enable in torchaudio>=0.8.0
wav, sr = sox_effects.apply_effects_file(
wav_file,
[['speed', str(speed)], ['rate', str(si.rate)]])
# sox will normalize the waveform, scale to [-32768, 32767]
wav = wav * (1 << 15)
return wav, sr
def _load_wav_with_speed(wav_file, speed):
""" Load the wave from file and apply speed perpturbation
Args:
wav_file: input feature, T * F 2D
Returns:
augmented feature
"""
if speed == 1.0:
wav, sr = torchaudio.load(wav_file)
else:
sample_rate = torchaudio.backend.sox_io_backend.info(
wav_file).sample_rate
# get torchaudio version
ta_no = torchaudio.__version__.split(".")
ta_version = 100 * int(ta_no[0]) + 10 * int(ta_no[1])
if ta_version < 80:
# Note: deprecated in torchaudio>=0.8.0
E = sox_effects.SoxEffectsChain()
E.append_effect_to_chain('speed', speed)
E.append_effect_to_chain("rate", sample_rate)
E.set_input_file(wav_file)
wav, sr = E.sox_build_flow_effects()
else:
# Note: enable in torchaudio>=0.8.0
wav, sr = sox_effects.apply_effects_file(
wav_file,
[['speed', str(speed)], ['rate', str(sample_rate)]])
return wav, sr
def _load_wav_with_speed(wav, speed):
"""
# reference: wenet
Load the wave from file and apply speed perpturbation
"""
if speed == 1.0:
wav, sr = torchaudio.load(wav)
else:
sample_rate = torchaudio.backend.sox_io_backend.info(wav).sample_rate
ta_no = torchaudio.__version__.split(".")
ta_version = 100 * int(ta_no[0]) + 10 * int(ta_no[1])
# 0.8.0
if ta_version < 80:
E = sox_effects.SoxEffectsChain()
E.append_effect_to_chain('speed', speed)
E.append_effect_to_chain('rate', sample_rate)
E.set_input_file(wav)
wav, sr = E.sox_build_flow_effects()
else:
wav, sr = sox_effects.apply_effects_file(
wav, [['speed', str(speed)], ['rate', str(sample_rate)]])
return wav, sr