From 4d3210044e22fc06140b435f79ed04c8302f832a Mon Sep 17 00:00:00 2001
From: 游雁 <zhifu.gzf@alibaba-inc.com>
Date: 星期一, 08 五月 2023 09:56:34 +0800
Subject: [PATCH] vad begin point bias
---
funasr/runtime/python/websocket/ws_server_offline.py | 9 ++++++---
1 files changed, 6 insertions(+), 3 deletions(-)
diff --git a/funasr/runtime/python/websocket/ws_server_offline.py b/funasr/runtime/python/websocket/ws_server_offline.py
index 7873918..c415703 100644
--- a/funasr/runtime/python/websocket/ws_server_offline.py
+++ b/funasr/runtime/python/websocket/ws_server_offline.py
@@ -78,6 +78,7 @@
is_speaking = message["is_speaking"]
websocket.param_dict_vad["is_final"] = not is_speaking
+ websocket.wav_name = message.get("wav_name", "demo")
if speech_start:
frames_asr.append(audio)
speech_start_i, speech_end_i = await async_vad(websocket, audio)
@@ -95,8 +96,9 @@
if not is_speaking:
websocket.vad_pre_idx = 0
frames = []
+ websocket.param_dict_vad = {'in_cache': dict()}
else:
- frames = frames[-10:]
+ frames = frames[-20:]
except websockets.ConnectionClosed:
@@ -136,8 +138,9 @@
rec_result = inference_pipeline_punc(text_in=rec_result['text'],
param_dict=websocket.param_dict_punc)
# print(rec_result)
- message = json.dumps({"mode": "offline", "text": [rec_result["text"]]})
- await websocket.send(message)
+ message = json.dumps({"mode": "offline", "text": [rec_result["text"]], "wav_name": websocket.wav_name})
+ await websocket.send(message)
+
--
Gitblit v1.9.1