brige/human
1
0
Fork 0
Code Issues Pull Requests 1 Actions Packages Projects Releases 2 Wiki Activity

add temp commit

Browse Source
This commit is contained in:
jiegeaiai 2024-10-31 08:15:49 +08:00
parent dc5068b237
commit 72833a9b7b
2 changed files with 9 additions and 4 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

8
human/audio_inference_handler.py
View File

@ -34,7 +34,7 @@ class AudioInferenceHandler(AudioHandler):
self._mal_queue.put(stream) self._mal_queue.put(stream)
elif type_ == 0: elif type_ == 0:
self._audio_queue.put(stream) self._audio_queue.put(stream)
print('AudioInferenceHandler on_handle', type_) print('AudioInferenceHandler on_handle', type_, self._audio_queue.size())
def on_message(self, message): def on_message(self, message):
super().on_message(message) super().on_message(message)
@ -70,6 +70,7 @@ class AudioInferenceHandler(AudioHandler):
audio_frames = [] audio_frames = []
for _ in range(batch_size * 2): for _ in range(batch_size * 2):
frame, type_ = self._audio_queue.get() frame, type_ = self._audio_queue.get()
print('AudioInferenceHandler type_', type_)
audio_frames.append((frame, type_)) audio_frames.append((frame, type_))
if type_ == 0: if type_ == 0:
is_all_silence = False is_all_silence = False
@ -129,5 +130,6 @@ class AudioInferenceHandler(AudioHandler):
self._run_thread.join() self._run_thread.join()
def pause_talk(self): def pause_talk(self):
self._audio_queue.clear() pass
self._mal_queue.queue.clear() # self._audio_queue.clear()
# self._mal_queue.queue.clear()

5
human/audio_mal_handler.py
View File

@ -44,15 +44,18 @@ class AudioMalHandler(AudioHandler):
logging.info('chunk2mal exit') logging.info('chunk2mal exit')
def _run_step(self): def _run_step(self):
count = 0
for _ in range(self._context.batch_size * 2): for _ in range(self._context.batch_size * 2):
frame, _type = self.get_audio_frame() frame, _type = self.get_audio_frame()
self.frames.append(frame) self.frames.append(frame)
self.on_next_handle((frame, _type), 0) self.on_next_handle((frame, _type), 0)
print("AudioMalHandler _type", _type)
count = count + 1
# context not enough, do not run network. # context not enough, do not run network.
if len(self.frames) <= self._context.stride_left_size + self._context.stride_right_size: if len(self.frames) <= self._context.stride_left_size + self._context.stride_right_size:
return return
print('AudioMalHandler _run_step') print('AudioMalHandler _run_step', count)
inputs = np.concatenate(self.frames) # [N * chunk] inputs = np.concatenate(self.frames) # [N * chunk]
mel = melspectrogram(inputs) mel = melspectrogram(inputs)
# print(mel.shape[0],mel.shape,len(mel[0]),len(self.frames)) # print(mel.shape[0],mel.shape,len(mel[0]),len(self.frames))