speech_recognition/FunASR
1
0
Fork 0
mirror of https://github.com/modelscope/FunASR synced 2025-09-15 14:48:36 +08:00
Code Issues Actions Packages Projects Releases Wiki Activity

websocket

Browse Source
This commit is contained in:
游雁 2023-03-23 17:12:02 +08:00
parent d781aa27e1
commit 7a8bedace2
2 changed files with 17 additions and 30 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

7
funasr/bin/vad_inference_online.py
View File

@ -30,14 +30,7 @@ from funasr.models.frontend.wav_frontend import WavFrontendOnline
from funasr.models.frontend.wav_frontend import WavFrontend
from funasr.bin.vad_inference import Speech2VadSegment
header_colors = '\033[95m'
end_colors = '\033[0m'
global_asr_language: str = 'zh-cn'
global_sample_rate: Union[int, Dict[Any, int]] = {
'audio_fs': 16000,
'model_fs': 16000
}
class Speech2VadSegmentOnline(Speech2VadSegment):

40
funasr/runtime/python/websocket/ASR_server.py
View File

@ -56,7 +56,9 @@ vad_pipline = pipeline(
model_revision="v1.2.0",
output_dir=None,
batch_size=1,
mode='online'
)
param_dict_vad = {'in_cache': dict(), "is_final": False}
# 创建一个ASR对象
param_dict = dict()
@ -85,17 +87,20 @@ start_server = websockets.serve(ws_serve, args.host, args.port, subprotocols=["b
def vad(data): # 推理
global vad_pipline
global vad_pipline, param_dict_vad
#print(type(data))
segments_result = vad_pipline(audio_in=data)
#print(segments_result)
# print(param_dict_vad)
segments_result = vad_pipline(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 False
elif segments_result["text"][0][0] != -1:
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
elif segments_result["text"][0][1] != -1:
if segments_result["text"][0][1] != -1:
speech_end = True
return speech_start, speech_end
@ -135,20 +140,21 @@ def main(): # 推理
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("检测到人声...")
# print("检测到人声...")
# speech_detected = True # 标记为检测到语音
frames = []
frames.extend(buffer) # 把之前2个语音数据快加入
# silence_count = 0 # 重置静音次数
elif speech_end_i or RECORD_NUM > 300:
if speech_end_i or RECORD_NUM > 300:
# silence_count += 1 # 增加静音次数
# speech_end = speech_end_i
speech_start = False
# if RECORD_NUM > 300: #这里 50 可根据需求改为合适的数据快数量
print("说话结束或者超过设置最长时间...")
# print("说话结束或者超过设置最长时间...")
audio_in = b"".join(frames)
#asrt = threading.Thread(target=asr,args=(audio_in,))
#asrt.start()
@ -170,16 +176,4 @@ s = threading.Thread(target=asr)
s.start()
asyncio.get_event_loop().run_until_complete(start_server)
asyncio.get_event_loop().run_forever()
asyncio.get_event_loop().run_forever()