From 8b05b03a84044434b574e122d21a950e2bf13494 Mon Sep 17 00:00:00 2001
From: zhifu gao <zhifu.gzf@alibaba-inc.com>
Date: 星期四, 23 三月 2023 10:10:45 +0800
Subject: [PATCH] Merge pull request #284 from cgisky1980/main
---
funasr/runtime/python/websocket/ASR_server.py | 143 +++++++++++++++++++++++++++++++++++
funasr/runtime/python/websocket/ASR_client.py | 73 ++++++++++++++++++
2 files changed, 216 insertions(+), 0 deletions(-)
diff --git a/funasr/runtime/python/websocket/ASR_client.py b/funasr/runtime/python/websocket/ASR_client.py
new file mode 100644
index 0000000..7dce880
--- /dev/null
+++ b/funasr/runtime/python/websocket/ASR_client.py
@@ -0,0 +1,73 @@
+import pyaudio
+import websocket #鍖哄埆鏈嶅姟绔繖閲屾槸 websocket-client搴�
+import time
+import websockets
+import asyncio
+from queue import Queue
+import threading
+voices = Queue()
+async def hello():
+ global ws # 瀹氫箟涓�涓叏灞�鍙橀噺ws锛岀敤浜庝繚瀛榳ebsocket杩炴帴瀵硅薄
+ uri = "ws://localhost:8899"
+ ws = await websockets.connect(uri, subprotocols=["binary"]) # 鍒涘缓涓�涓暱杩炴帴
+ ws.max_size = 1024 * 1024 * 20
+ print("connected ws server")
+async def send(data):
+ global ws # 寮曠敤鍏ㄥ眬鍙橀噺ws
+ try:
+ await ws.send(data) # 閫氳繃ws瀵硅薄鍙戦�佹暟鎹�
+ except Exception as e:
+ print('Exception occurred:', e)
+
+
+
+asyncio.get_event_loop().run_until_complete(hello()) # 鍚姩鍗忕▼
+
+
+# 鍏朵粬鍑芥暟鍙互閫氳繃璋冪敤send(data)鏉ュ彂閫佹暟鎹紝渚嬪锛�
+async def test():
+ #print("2")
+ global voices
+ FORMAT = pyaudio.paInt16
+ CHANNELS = 1
+ RATE = 16000
+ CHUNK = int(RATE / 1000 * 300)
+
+ p = pyaudio.PyAudio()
+
+ stream = p.open(format=FORMAT,
+ channels=CHANNELS,
+ rate=RATE,
+ input=True,
+ frames_per_buffer=CHUNK)
+
+ while True:
+
+ data = stream.read(CHUNK)
+
+ voices.put(data)
+ #print(voices.qsize())
+ await asyncio.sleep(0.01)
+
+
+
+
+
+async def ws_send():
+ global voices
+ print("started to sending data!")
+ while True:
+ while not voices.empty():
+ data = voices.get()
+ voices.task_done()
+ await send(data)
+ await asyncio.sleep(0.01)
+ await asyncio.sleep(0.01)
+
+async def main():
+ task = asyncio.create_task(test()) # 鍒涘缓涓�涓悗鍙颁换鍔�
+ task2 = asyncio.create_task(ws_send()) # 鍒涘缓涓�涓悗鍙颁换鍔�
+
+ await asyncio.gather(task, task2)
+
+asyncio.run(main())
\ No newline at end of file
diff --git a/funasr/runtime/python/websocket/ASR_server.py b/funasr/runtime/python/websocket/ASR_server.py
new file mode 100644
index 0000000..3627d3a
--- /dev/null
+++ b/funasr/runtime/python/websocket/ASR_server.py
@@ -0,0 +1,143 @@
+# server.py 娉ㄦ剰鏈緥浠呭鐞嗗崟涓猚lent鍙戦�佺殑璇煶鏁版嵁锛屽苟鏈澶歝lient杩炴帴杩涜鍒ゆ柇鍜屽鐞�
+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)
+import asyncio
+import websockets #鍖哄埆瀹㈡埛绔繖閲屾槸 websockets搴�
+import time
+from queue import Queue
+import threading
+
+print("model loading")
+voices = Queue()
+speek = Queue()
+# 鍒涘缓涓�涓猇AD瀵硅薄
+vad_pipline = pipeline(
+ task=Tasks.voice_activity_detection,
+ model="damo/speech_fsmn_vad_zh-cn-16k-common-pytorch",
+ model_revision="v1.2.0",
+ output_dir=None,
+ batch_size=1,
+)
+
+# 鍒涘缓涓�涓狝SR瀵硅薄
+param_dict = dict()
+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_pipeline2 = pipeline(
+ task=Tasks.auto_speech_recognition,
+ model="damo/speech_paraformer-large-contextual_asr_nat-zh-cn-16k-common-vocab8404",
+ param_dict=param_dict,
+)
+print("model loaded")
+
+
+
+async def echo(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(echo, "localhost", 8899, subprotocols=["binary"],ping_interval=None)
+
+
+def vad(data): # 鎺ㄧ悊
+ global vad_pipline
+ #print(type(data))
+ segments_result = vad_pipline(audio_in=data)
+ #print(segments_result)
+ if len(segments_result) == 0:
+ return False
+ else:
+ return True
+
+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_pipeline2(audio_in=audio_in)
+ print(rec_result)
+ 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
+ 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_detected:
+ frames.append(data)
+ RECORD_NUM += 1
+
+ if vad(data):
+ 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
+ else:
+ silence_count += 1 # 澧炲姞闈欓煶娆℃暟
+
+ if speech_detected and (silence_count > 4 or RECORD_NUM > 50): #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()
+
+
+
+
+
+
+
+
+
+
+
+
--
Gitblit v1.9.1