speech_recognition/FunASR
1
0
Fork 0
mirror of https://github.com/modelscope/FunASR synced 2025-09-15 14:48:36 +08:00
Code Issues Actions Packages Projects Releases Wiki Activity

support pcm audio format

Browse Source
This commit is contained in:
hnluo 2023-02-06 17:03:38 +08:00 committed by GitHub
parent c14169f374
commit ade51b3a12
No known key found for this signature in database
GPG Key ID: 4AEE18F83AFDEB23
1 changed files with 11 additions and 1 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

12
funasr/datasets/iterable_dataset.py
View File

@ -20,7 +20,7 @@ import os.path
from funasr.datasets.dataset import ESPnetDataset from funasr.datasets.dataset import ESPnetDataset
SUPPORT_AUDIO_TYPE_SETS = ['flac', 'mp3', 'm4a', 'ogg', 'opus', 'wav', 'wma'] SUPPORT_AUDIO_TYPE_SETS = ['flac', 'mp3', 'ogg', 'opus', 'wav', 'pcm']
def load_kaldi(input): def load_kaldi(input):
retval = kaldiio.load_mat(input) retval = kaldiio.load_mat(input)
@ -60,9 +60,14 @@ def load_bytes(input):
array = np.frombuffer((middle_data.astype(dtype) - offset) / abs_max, dtype=np.float32) array = np.frombuffer((middle_data.astype(dtype) - offset) / abs_max, dtype=np.float32)
return array return array
def load_pcm(input):
with open(input,"rb") as f:
bytes = f.read()
return load_bytes(bytes)
DATA_TYPES = { DATA_TYPES = {
"sound": lambda x: torchaudio.load(x)[0][0].numpy(), "sound": lambda x: torchaudio.load(x)[0][0].numpy(),
"pcm": load_pcm,
"kaldi_ark": load_kaldi, "kaldi_ark": load_kaldi,
"bytes": load_bytes, "bytes": load_bytes,
"waveform": lambda x: x, "waveform": lambda x: x,
@ -219,6 +224,9 @@ class IterableESPnetDataset(IterableDataset):
if audio_type not in SUPPORT_AUDIO_TYPE_SETS: if audio_type not in SUPPORT_AUDIO_TYPE_SETS:
raise NotImplementedError( raise NotImplementedError(
f'Not supported audio type: {audio_type}') f'Not supported audio type: {audio_type}')
if audio_type == "pcm":
_type = "pcm"
func = DATA_TYPES[_type] func = DATA_TYPES[_type]
array = func(value) array = func(value)
if self.fs is not None and name == "speech": if self.fs is not None and name == "speech":
@ -318,6 +326,8 @@ class IterableESPnetDataset(IterableDataset):
if audio_type not in SUPPORT_AUDIO_TYPE_SETS: if audio_type not in SUPPORT_AUDIO_TYPE_SETS:
raise NotImplementedError( raise NotImplementedError(
f'Not supported audio type: {audio_type}') f'Not supported audio type: {audio_type}')
if audio_type == "pcm":
_type = "pcm"
func = DATA_TYPES[_type] func = DATA_TYPES[_type]
# Load entry # Load entry
array = func(value) array = func(value)