update

audio_decoder.py

@@ -1,10 +1,15 @@
+import soundfile as sf
+import numpy as np
 from typing import Generator, NamedTuple
 
 import av
 
 from towhee.operator.base import Operator
+from torch_vggish import TorchVggish
 
-AudioOutput = NamedTuple("Outputs", [("audio_frame", 'ndarray'), ('simple_rate', 'int'), ("TIMESTAMP", 'int')])
+# from torch_vggish import TorchVggish
+
+AudioOutput = NamedTuple("Outputs", [("audio_frame", 'ndarray'), ('sample_rate', 'int'), ("TIMESTAMP", 'int')])
 
 
 class AudioDecoder(Operator):
@@ -18,15 +23,7 @@ class AudioDecoder(Operator):
         in_container = av.open(audio_path)
         stream = in_container.streams.get(audio=0)[0]
         for frame in in_container.decode(stream):
-            timestamp = frame.time * 1000
+            timestamp = int(frame.time * 1000)
             sample_rate = frame.sample_rate
             ndarray = frame.to_ndarray()
             yield AudioOutput(ndarray, sample_rate, timestamp)
-
-
-if __name__ == '__main__':
-    d = AudioDecoder()
-    audio_path = '/Users/jiangjunjie/WorkSpace/audio_data/stereo.wav'
-    it = d(audio_path)
-    for data in it:
-        print(data._asdict())