vad预处理

app/controllers/chat.py

@@ -69,6 +69,11 @@ def split_string_with_punctuation(current_sentence,text,is_first):
             result.append(current_sentence)
             current_sentence = ''
     return result, current_sentence, is_first
+
+def vad_preprocess(audio):
+    if len(audio)<1280:
+        return ('A'*1280)
+    return audio[:1280],audio[1280:]
 #--------------------------------------------------------
 
 # 创建新聊天
@@ -461,6 +466,8 @@ async def streaming_chat_lasting_handler(ws,db,redis):
 async def voice_call_audio_producer(ws,audio_queue,future,input_finished_event):
     logger.debug("音频数据生产函数启动")
     is_future_done = False
+    audio_data = ""
+    try:
         while not input_finished_event.is_set():
             voice_call_data_json = json.loads(await ws.receive_text())
             if not is_future_done: #在第一次循环中读取session_id
@@ -470,7 +477,12 @@ async def voice_call_audio_producer(ws,audio_queue,future,input_finished_event):
                 input_finished_event.set()
                 break
             else:
-            await audio_queue.put(voice_call_data_json["audio"]) #将音频数据存入audio_q
+                audio_data += voice_call_data_json["audio"]
+                while len(audio_data) > 1280:
+                    vad_frame,audio_data = vad_preprocess(audio_data)
+                    await audio_queue.put(vad_frame) #将音频数据存入audio_q
+    except KeyError as ke:
+        logger.info(f"收到心跳包")
 
 #音频数据消费函数
 async def voice_call_audio_consumer(audio_q,asr_result_q,input_finished_event,asr_finished_event):