Merge pull request #531 from alibaba-damo-academy/dev_new

zhifu gao

2023-05-19 7eaf608c2d4473a77bd1590f93ea9bdbedde346a

 funasr/runtime/python/websocket/ws_server_2pass.py

@@ -46,7 +46,7 @@
    inference_pipeline_punc = pipeline(
        task=Tasks.punctuation,
        model=args.punc_model,
        model_revision=None,
        model_revision="v1.0.2",
        ngpu=args.ngpu,
        ncpu=args.ncpu,
    )
@@ -74,6 +74,7 @@
    websocket.param_dict_punc = {'cache': list()}
    websocket.vad_pre_idx = 0
    speech_start = False
    speech_end_i = False
    websocket.wav_name = "microphone"
    print("new user connected", flush=True)

@@ -99,7 +100,9 @@
        
                    # asr online
                    frames_asr_online.append(message)
                    if len(frames_asr_online) % websocket.chunk_interval == 0:
                    websocket.param_dict_asr_online["is_final"] = speech_end_i
                    if len(frames_asr_online) % websocket.chunk_interval == 0 or websocket.param_dict_asr_online["is_final"]:
                        
                        audio_in = b"".join(frames_asr_online)
                        await async_asr_online(websocket, audio_in)
                        frames_asr_online = []
@@ -115,12 +118,13 @@
                        frames_asr.extend(frames_pre)
                # asr punc offline
                if speech_end_i or not websocket.is_speaking:
                    # print("vad end point")
                    audio_in = b"".join(frames_asr)
                    await async_asr(websocket, audio_in)
                    frames_asr = []
                    speech_start = False
                    frames_asr_online = []
                    websocket.param_dict_asr_online = {"cache": dict()}
                    # frames_asr_online = []
                    # websocket.param_dict_asr_online = {"cache": dict()}
                    if not websocket.is_speaking:
                        websocket.vad_pre_idx = 0
                        frames = []
@@ -173,10 +177,13 @@
async def async_asr_online(websocket, audio_in):
    if len(audio_in) > 0:
        audio_in = load_bytes(audio_in)
        # print(websocket.param_dict_asr_online.get("is_final", False))
        rec_result = inference_pipeline_asr_online(audio_in=audio_in,
                                                   param_dict=websocket.param_dict_asr_online)
        # print(rec_result)
        if websocket.param_dict_asr_online.get("is_final", False):
            websocket.param_dict_asr_online["cache"] = dict()
            return
            #     websocket.param_dict_asr_online["cache"] = dict()
        if "text" in rec_result:
            if rec_result["text"] != "sil" and rec_result["text"] != "waiting_for_more_voice":
                # print("online", rec_result)