update wss_client_asr.py

雾聪

2023-06-13 59113b50b578e097be5656c4044a58282b4d0b9e

 funasr/runtime/python/websocket/wss_client_asr.py

@@ -12,6 +12,7 @@

import logging

SUPPORT_AUDIO_TYPE_SETS = ['.wav', '.pcm']
logging.basicConfig(level=logging.ERROR)

parser = argparse.ArgumentParser()
@@ -68,12 +69,14 @@
print(args)
# voices = asyncio.Queue()
from queue import Queue

voices = Queue()

ibest_writer = None
if args.output_dir is not None:
    writer = DatadirWriter(args.output_dir)
    ibest_writer = writer[f"1best_recog"]


async def record_microphone():
    is_finished = False
@@ -94,19 +97,16 @@
                    input=True,
                    frames_per_buffer=CHUNK)

    message = json.dumps({"mode": args.mode, "chunk_size": args.chunk_size, "chunk_interval": args.chunk_interval, "wav_name": "microphone", "is_speaking": True})
    message = json.dumps({"mode": args.mode, "chunk_size": args.chunk_size, "chunk_interval": args.chunk_interval,
                          "wav_name": "microphone", "is_speaking": True})
    voices.put(message)
    while True:

        data = stream.read(CHUNK)
        message = data  
        
        voices.put(message)

        await asyncio.sleep(0.005)

async def record_from_scp(chunk_begin,chunk_size):
    import wave
    global voices
    is_finished = False
    if args.audio_in.endswith(".scp"):
@@ -121,22 +121,27 @@
        wav_name = wav_splits[0] if len(wav_splits) > 1 else "demo"
        wav_path = wav_splits[1] if len(wav_splits) > 1 else wav_splits[0]
        
        # bytes_f = open(wav_path, "rb")
        # bytes_data = bytes_f.read()
        if wav_path.endswith(".pcm"):
            with open(wav_path, "rb") as f:
                audio_bytes = f.read()
        elif wav_path.endswith(".wav"):
            import wave
        with wave.open(wav_path, "rb") as wav_file:
            params = wav_file.getparams()
            # header_length = wav_file.getheaders()[0][1]
            # wav_file.setpos(header_length)
            frames = wav_file.readframes(wav_file.getnframes())

        audio_bytes = bytes(frames)
        else:
            raise NotImplementedError(
                f'Not supported audio type')

        # stride = int(args.chunk_size/1000*16000*2)
        stride = int(60*args.chunk_size[1]/args.chunk_interval/1000*16000*2)
        chunk_num = (len(audio_bytes)-1)//stride + 1
        # print(stride)
        
        # send first time
        message = json.dumps({"mode": args.mode, "chunk_size": args.chunk_size, "chunk_interval": args.chunk_interval, "wav_name": wav_name,"is_speaking": True})
        message = json.dumps({"mode": args.mode, "chunk_size": args.chunk_size, "chunk_interval": args.chunk_interval,
                              "wav_name": wav_name, "is_speaking": True})
        voices.put(message)
        is_speaking = True
        for i in range(chunk_num):
@@ -154,7 +159,6 @@
            sleep_duration = 0.001 if args.send_without_sleep else 60*args.chunk_size[1]/args.chunk_interval/1000
            await asyncio.sleep(sleep_duration)


async def ws_send():
    global voices
    global websocket
@@ -171,8 +175,6 @@
                exit(0)
            await asyncio.sleep(0.005)
        await asyncio.sleep(0.005)



async def message(id):
    global websocket
@@ -216,6 +218,7 @@
            traceback.print_exc()
            exit(0)


async def print_messge():
    global websocket
    while True:
@@ -252,7 +255,6 @@
   asyncio.get_event_loop().run_until_complete(ws_client(id,chunk_begin,chunk_size))
   asyncio.get_event_loop().run_forever()


if __name__ == '__main__':
   # for microphone 
   if  args.audio_in is  None:
@@ -267,9 +269,18 @@
         wavs = f_scp.readlines()
     else:
         wavs = [args.audio_in]
        for wav in wavs:
            wav_splits = wav.strip().split()
            wav_name = wav_splits[0] if len(wav_splits) > 1 else "demo"
            wav_path = wav_splits[1] if len(wav_splits) > 1 else wav_splits[0]
            audio_type = os.path.splitext(wav_path)[-1].lower()
            if audio_type not in SUPPORT_AUDIO_TYPE_SETS:
                raise NotImplementedError(
                    f'Not supported audio type: {audio_type}')

     total_len=len(wavs)
     if total_len>=args.test_thread_num:
          chunk_size=int((total_len)/args.test_thread_num)
            chunk_size = int(total_len / args.test_thread_num)
          remain_wavs=total_len-chunk_size*args.test_thread_num
     else:
          chunk_size=1
@@ -292,5 +303,3 @@
         p.join()

     print('end')