funasr/datasets/dataset_jsonl.py
@@ -4,8 +4,8 @@ import numpy as np import kaldiio import librosa import torchaudio import time def load_audio(audio_path: str, fs: int=16000): audio = None @@ -17,7 +17,9 @@ if ".ark:" in audio_path: audio = kaldiio.load_mat(audio_path) else: audio, fs = librosa.load(audio_path, sr=fs) # audio, fs = librosa.load(audio_path, sr=fs) audio, fs = torchaudio.load(audio_path) audio = audio[0, :] return audio def extract_features(data, date_type: str="sound", frontend=None):
