From b0887f17678e0e5c4bd895e36695b242f2f1aee8 Mon Sep 17 00:00:00 2001
From: mengzhe.cmz <mengzhe.cmz@alibaba-inc.com>
Date: 星期四, 23 三月 2023 19:59:28 +0800
Subject: [PATCH] Merge branch 'dev_gzf' of github.com:alibaba-damo-academy/FunASR into dev_gzf
---
funasr/runtime/python/websocket/ASR_server.py | 191 +++++++++++++++++++++++++++++++++++++++++++++++
1 files changed, 191 insertions(+), 0 deletions(-)
diff --git a/funasr/runtime/python/websocket/ASR_server.py b/funasr/runtime/python/websocket/ASR_server.py
new file mode 100644
index 0000000..20de20a
--- /dev/null
+++ b/funasr/runtime/python/websocket/ASR_server.py
@@ -0,0 +1,191 @@
+import asyncio
+import websockets
+import time
+from queue import Queue
+import threading
+import argparse
+
+from modelscope.pipelines import pipeline
+from modelscope.utils.constant import Tasks
+from modelscope.utils.logger import get_logger
+import logging
+
+logger = get_logger(log_level=logging.CRITICAL)
+logger.setLevel(logging.CRITICAL)
+
+
+parser = argparse.ArgumentParser()
+parser.add_argument("--host",
+ type=str,
+ default="0.0.0.0",
+ required=False,
+ help="host ip, localhost, 0.0.0.0")
+parser.add_argument("--port",
+ type=int,
+ default=10095,
+ required=False,
+ help="grpc server port")
+parser.add_argument("--asr_model",
+ type=str,
+ default="damo/speech_paraformer-large_asr_nat-zh-cn-16k-common-vocab8404-pytorch",
+ help="model from modelscope")
+parser.add_argument("--vad_model",
+ type=str,
+ default="damo/speech_fsmn_vad_zh-cn-16k-common-pytorch",
+ help="model from modelscope")
+
+parser.add_argument("--punc_model",
+ type=str,
+ default="damo/punc_ct-transformer_zh-cn-common-vad_realtime-vocab272727",
+ help="model from modelscope")
+parser.add_argument("--ngpu",
+ type=int,
+ default=1,
+ help="0 for cpu, 1 for gpu")
+
+args = parser.parse_args()
+
+print("model loading")
+voices = Queue()
+speek = Queue()
+
+# vad
+inference_pipeline_vad = pipeline(
+ task=Tasks.voice_activity_detection,
+ model=args.vad_model,
+ model_revision=None,
+ output_dir=None,
+ batch_size=1,
+ mode='online',
+ ngpu=args.ngpu,
+)
+param_dict_vad = {'in_cache': dict(), "is_final": False}
+
+# asr
+param_dict_asr = {}
+# param_dict["hotword"] = "灏忎簲 灏忎簲鏈�" # 璁剧疆鐑瘝锛岀敤绌烘牸闅斿紑
+inference_pipeline_asr = pipeline(
+ task=Tasks.auto_speech_recognition,
+ model=args.asr_model,
+ param_dict=param_dict_asr,
+ ngpu=args.ngpu,
+)
+
+param_dict_punc = {'cache': list()}
+inference_pipeline_punc = pipeline(
+ task=Tasks.punctuation,
+ model=args.punc_model,
+ model_revision=None,
+ ngpu=args.ngpu,
+)
+
+print("model loaded")
+
+
+
+async def ws_serve(websocket, path):
+ global voices
+ try:
+ async for message in websocket:
+ voices.put(message)
+ #print("put")
+ except websockets.exceptions.ConnectionClosedError as e:
+ print('Connection closed with exception:', e)
+ except Exception as e:
+ print('Exception occurred:', e)
+
+start_server = websockets.serve(ws_serve, args.host, args.port, subprotocols=["binary"], ping_interval=None)
+
+
+def vad(data): # 鎺ㄧ悊
+ global vad_pipline, param_dict_vad
+ #print(type(data))
+ # print(param_dict_vad)
+ segments_result = inference_pipeline_vad(audio_in=data, param_dict=param_dict_vad)
+ # print(segments_result)
+ # print(param_dict_vad)
+ speech_start = False
+ speech_end = False
+
+ if len(segments_result) == 0 or len(segments_result["text"]) > 1:
+ return speech_start, speech_end
+ if segments_result["text"][0][0] != -1:
+ speech_start = True
+ if segments_result["text"][0][1] != -1:
+ speech_end = True
+ return speech_start, speech_end
+
+def asr(): # 鎺ㄧ悊
+ global inference_pipeline2
+ global speek
+ while True:
+ while not speek.empty():
+ audio_in = speek.get()
+ speek.task_done()
+ rec_result = inference_pipeline_asr(audio_in=audio_in)
+ rec_result_punc = inference_pipeline_punc(text_in=rec_result['text'], param_dict=param_dict_punc)
+ print(rec_result_punc)
+ time.sleep(0.1)
+ time.sleep(0.1)
+
+
+def main(): # 鎺ㄧ悊
+ frames = [] # 瀛樺偍鎵�鏈夌殑甯ф暟鎹�
+ buffer = [] # 瀛樺偍缂撳瓨涓殑甯ф暟鎹紙鏈�澶氫袱涓墖娈碉級
+ # silence_count = 0 # 缁熻杩炵画闈欓煶鐨勬鏁�
+ # speech_detected = False # 鏍囪鏄惁妫�娴嬪埌璇煶
+ RECORD_NUM = 0
+ global voices
+ global speek
+ speech_start, speech_end = False, False
+ while True:
+ while not voices.empty():
+
+ data = voices.get()
+ #print("闃熷垪鎺掗槦鏁�",voices.qsize())
+ voices.task_done()
+ buffer.append(data)
+ if len(buffer) > 2:
+ buffer.pop(0) # 濡傛灉缂撳瓨瓒呰繃涓や釜鐗囨锛屽垯鍒犻櫎鏈�鏃╃殑涓�涓�
+
+ if speech_start:
+ frames.append(data)
+ RECORD_NUM += 1
+ speech_start_i, speech_end_i = vad(data)
+ # print(speech_start_i, speech_end_i)
+ if speech_start_i:
+ speech_start = speech_start_i
+ # if not speech_detected:
+ # print("妫�娴嬪埌浜哄０...")
+ # speech_detected = True # 鏍囪涓烘娴嬪埌璇煶
+ frames = []
+ frames.extend(buffer) # 鎶婁箣鍓�2涓闊虫暟鎹揩鍔犲叆
+ # silence_count = 0 # 閲嶇疆闈欓煶娆℃暟
+ if speech_end_i or RECORD_NUM > 300:
+ # silence_count += 1 # 澧炲姞闈欓煶娆℃暟
+ # speech_end = speech_end_i
+ speech_start = False
+ # if RECORD_NUM > 300: #杩欓噷 50 鍙牴鎹渶姹傛敼涓哄悎閫傜殑鏁版嵁蹇暟閲�
+ # print("璇磋瘽缁撴潫鎴栬�呰秴杩囪缃渶闀挎椂闂�...")
+ audio_in = b"".join(frames)
+ #asrt = threading.Thread(target=asr,args=(audio_in,))
+ #asrt.start()
+ speek.put(audio_in)
+ #rec_result = inference_pipeline2(audio_in=audio_in) # ASR 妯″瀷閲岃窇涓�璺�
+ frames = [] # 娓呯┖鎵�鏈夌殑甯ф暟鎹�
+ buffer = [] # 娓呯┖缂撳瓨涓殑甯ф暟鎹紙鏈�澶氫袱涓墖娈碉級
+ # silence_count = 0 # 缁熻杩炵画闈欓煶鐨勬鏁版竻闆�
+ # speech_detected = False # 鏍囪鏄惁妫�娴嬪埌璇煶
+ RECORD_NUM = 0
+ time.sleep(0.01)
+ time.sleep(0.01)
+
+
+
+s = threading.Thread(target=main)
+s.start()
+s = threading.Thread(target=asr)
+s.start()
+
+asyncio.get_event_loop().run_until_complete(start_server)
+asyncio.get_event_loop().run_forever()
\ No newline at end of file
--
Gitblit v1.9.1