Refactor SpeechMonitor to use active audio frame duration instead of count

- Replace `active_audio_frame_count` with `active_audio_frame_duration` for more precise threshold control.
- Adjust `ACTIVE_FRAME_THRESHOLD` and associated calculations to use duration-based logic.

src/voice_dialogue/services/speech/monitor.py

@@ -26,10 +26,10 @@ from voice_dialogue.utils.logger import logger
 class SpeechMonitorConfig:
     """语音监控配置类"""
     MIN_AUDIO_AMPLITUDE = 0.01  # 最小音频振幅阈值
-    ACTIVE_FRAME_THRESHOLD = 10  # 连续活跃帧数阈值
     QUEUE_TIMEOUT = 0.1  # 队列获取超时时间（秒）
 
     # 时间阈值（毫秒）
+    ACTIVE_FRAME_THRESHOLD = 0.32 * 1000  # 连续活跃帧数阈值
     USER_SILENCE_THRESHOLD = 1 * 1000  # 用户静音阈值
     SILENCE_THRESHOLD = 0.3 * 1000  # 静音检测阈值
     AUDIO_FRAMES_THRESHOLD = 5 * 1000  # 音频帧时长阈值
@@ -80,7 +80,7 @@ class SpeechStateMonitor(BaseThread):
     def _reset_monitoring_state(self):
         """重置监控状态"""
         self.silence_audio_frame_count = 0
-        self.active_audio_frame_count = 0
+        self.active_audio_frame_duration = 0
         self.user_silence_duration = 0
         self.task_id = None
 
@@ -148,10 +148,11 @@ class SpeechStateMonitor(BaseThread):
 
         # 重置静音计时
         self.user_silence_duration = 0
-        self.active_audio_frame_count += 1
+        duration = self._calculate_frame_duration_ms(audio_frame)
+        self.active_audio_frame_duration += duration
 
         # 检查是否需要中断当前任务
-        if self.active_audio_frame_count > self.config.ACTIVE_FRAME_THRESHOLD:
+        if self.active_audio_frame_duration > self.config.ACTIVE_FRAME_THRESHOLD:
             voice_state_manager.interrupt_task_id = self.task_id
 
         return True
@@ -169,7 +170,7 @@ class SpeechStateMonitor(BaseThread):
         Returns:
             tuple: (更新后的音频帧缓存, 是否需要继续处理)
         """
-        self.active_audio_frame_count = 0
+        self.active_audio_frame_duration = 0
         duration = self._calculate_frame_duration_ms(audio_frame)
 
         if is_audio_frames_empty: