def read_kaldi(fd, endian='<', return_size=False):
"""Load kaldi
Args:
fd (file): Binary mode file object. Cannot input string
endian (str):
return_size (bool):
"""
assert endian in ('<', '>'), endian
binary_flag = fd.read(4)
assert isinstance(binary_flag, binary_type), type(binary_flag)
if seekable(fd):
fd.seek(-4, 1)
else:
fd = MultiFileDescriptor(BytesIO(binary_flag), fd)
if binary_flag[:4] == b'RIFF':
# array: Tuple[int, np.ndarray]
array, size = read_wav(fd, return_size=True)
# Load as binary
elif binary_flag[:2] == b'\0B':
if binary_flag[2:3] == b'\4': # This is int32Vector
array, size = read_int32vector(fd, endian, return_size=True)
else:
array, size = read_matrix_or_vector(fd, endian, return_size=True)
# Load as ascii
else:
array, size = read_ascii_mat(fd, return_size=True)
if return_size:
return array, size
else:
return array
def read_wav_scipy(fd, return_size=False):
if not seekable(fd):
# scipy.io.wavfile doesn't support unseekable fd
data = fd.read()
fd = BytesIO(data)
offset = None
else:
offset = fd.tell()
rate, array = wavfile.read(fd)
size = 44 + array.nbytes
if offset is not None:
fd.seek(size + offset)
if return_size:
return (rate, array), size
else:
return rate, array
def read_kaldi(fd, endian="<", audio_loader="soundfile"):
"""Load kaldi
Args:
fd (file): Binary mode file object. Cannot input string
endian (str):
audio_loader: (Union[str, callable]):
"""
assert endian in ("<", ">"), endian
max_flag_length = len(b"AUDIO")
binary_flag = fd.read(max_flag_length)
assert isinstance(binary_flag, binary_type), type(binary_flag)
if seekable(fd):
fd.seek(-max_flag_length, 1)
else:
fd = MultiFileDescriptor(BytesIO(binary_flag), fd)
if binary_flag[:4] == b"RIFF":
# array: Tuple[int, np.ndarray]
array = read_wav(fd)
elif binary_flag[:3] == b"NPY":
fd.read(3)
length_ = _read_length_header(fd)
buf = fd.read(length_)
_fd = BytesIO(buf)
array = np.load(_fd)
elif binary_flag[:3] == b"PKL":
fd.read(3)
array = pickle.load(fd)
elif binary_flag[:5] == b"AUDIO":
fd.read(5)
length_ = _read_length_header(fd)
buf = fd.read(length_)
_fd = BytesIO(buf)
if audio_loader == "soundfile":
import soundfile
audio_loader = soundfile.read
else:
raise ValueError(
"Not supported: audio_loader={}".format(audio_loader))
x1, x2 = audio_loader(_fd)
# array: Tuple[int, np.ndarray] according to scipy wav read
if isinstance(x1, int) and isinstance(x2, np.ndarray):
array = (x1, x2)
elif isinstance(x1, np.ndarray) and isinstance(x2, int):
array = (x2, x1)
else:
raise RuntimeError(
"Got unexpected type from audio_loader: ({}, {})".format(
type(x1), type(x2)))
# Load as binary
elif binary_flag[:2] == b"\0B":
if binary_flag[2:3] == b"\4": # This is int32Vector
array = read_int32vector(fd, endian)
else:
array = read_matrix_or_vector(fd, endian)
# Load as ascii
else:
array = read_ascii_mat(fd)
return array