Refactor audio processing pipeline to normalize data in `SpeechMonitor` and streamline queuing in `AudioCapture`

src/voice_dialogue/services/audio/capture.py

@@ -11,7 +11,6 @@ import threading
 import time
 from multiprocessing import Queue
 
-import numpy as np
 import pyaudio
 
 from voice_dialogue.config.paths import LIBRARIES_PATH
@@ -96,12 +95,9 @@ class AudioCapture(BaseThread):
                 data = stream.read(chunk)
 
                 if self.is_paused:
-                    time.sleep(0.01)
                     continue
 
-                # 将音频数据转换为 [-1.0, 1.0] 范围内的浮点数
-                audio_frame = np.frombuffer(data, dtype=np.int16).astype(np.float32) / np.iinfo(np.int16).max
-                self.audio_frames_queue.put(audio_frame)
+                self.audio_frames_queue.put(data)
 
         except Exception as e:
             logger.error(f'PyAudio 音频捕获器运行时发生错误: {e}')
@@ -134,13 +130,10 @@ class AudioCapture(BaseThread):
 
                 if data_ptr and size.value > 0:
                     audio_data = bytes(data_ptr[: size.value])
-                    # 将音频数据转换为 [-1.0, 1.0] 范围内的浮点数
-                    audio_frame = np.frombuffer(audio_data, dtype=np.int16).astype(np.float32) / np.iinfo(
-                        np.int16).max
 
                     if not self.is_paused:
                         # 将音频帧和语音活动状态一同放入队列
-                        self.audio_frames_queue.put((audio_frame, is_voice_active.value))
+                        self.audio_frames_queue.put((audio_data, is_voice_active.value))
 
                     # 释放原生库分配的内存
                     audio_recorder.freeAudioData(data_ptr)

src/voice_dialogue/services/speech/monitor.py

@@ -45,24 +45,23 @@ class SpeechStateMonitor(BaseThread):
     - 音频帧的缓存和处理
     """
 
-    def __init__(self, group=None, target=None, name=None, args=(), kwargs=None, *, daemon=None,
-                 audio_frame_queue: Queue,
-                 user_voice_queue: Queue,
-                 device_sample_rate: int = 16000
-                 ):
+    def __init__(
+            self, group=None, target=None, name=None, args=(), kwargs=None, *, daemon=None,
+            audio_frame_queue: Queue,
+            user_voice_queue: Queue,
+    ):
         """
         初始化语音状态监控器
         
         Args:
             audio_frame_queue: 音频帧队列
             user_voice_queue: 用户语音队列
-            device_sample_rate: 设备采样率，默认16000Hz
         """
         super().__init__(group, target, name, args, kwargs, daemon=daemon)
 
         self.audio_frame_queue = audio_frame_queue
         self.user_voice_queue = user_voice_queue
-        self.sample_rate = device_sample_rate
+        self.sample_rate = 16000
 
         # 配置参数
         self.config = SpeechMonitorConfig()
@@ -101,10 +100,16 @@ class SpeechStateMonitor(BaseThread):
         if self.user_silence_duration >= self.config.USER_SILENCE_THRESHOLD:
             silence_over_threshold_event.set()
 
+    def _normalize_audio_frame(self, data: bytes) -> np.ndarray:
+        """将 int16 格式的音频字节数据转换为 [-1.0, 1.0] 范围的 numpy 浮点数组。"""
+        return np.frombuffer(data, dtype=np.int16).astype(np.float32) / np.iinfo(np.int16).max
+
     def _get_audio_frame_from_queue(self):
         """从队列获取音频帧"""
         try:
-            return self.audio_frame_queue.get(block=False, timeout=self.config.QUEUE_TIMEOUT)
+            data, is_voice_active = self.audio_frame_queue.get(block=False, timeout=self.config.QUEUE_TIMEOUT)
+            audio_frame = self._normalize_audio_frame(data)
+            return audio_frame, is_voice_active
         except Empty:
             return None, None