回声待处理
This commit is contained in:
朱潮
parent
9523176d60
commit
e1d634af1f
@ -522,12 +522,14 @@ class OutputProcess:
|
||||
"""输出进程 - 借鉴 recorder.py 的优秀架构"""
|
||||
|
||||
def __init__(self, audio_queue: mp.Queue, config: Dict[str, Any] = None, event_queue: mp.Queue = None):
|
||||
print("🔊 OutputProcess __init__ 开始执行...")
|
||||
self.audio_queue = audio_queue # 主进程 → 输出进程（统一音频播放队列）
|
||||
self.event_queue = event_queue # 输出进程 → 主进程（事件通知）
|
||||
self.config = config or self._get_default_config()
|
||||
|
||||
# 初始化日志记录器
|
||||
self.logger = ProcessLogger("OutputProcess")
|
||||
print("🔊 OutputProcess 基本初始化完成")
|
||||
|
||||
# 音频播放参数 - 完全借鉴 recorder.py 的优化参数
|
||||
self.FORMAT = pyaudio.paInt16
|
||||
@ -542,7 +544,9 @@ class OutputProcess:
|
||||
self.total_chunks_played = 0
|
||||
self.total_audio_size = 0
|
||||
self.last_playback_time = 0 # 最后播放时间戳
|
||||
self.playback_cooldown_period = 1.5 # 播放冷却时间（秒）- 防止回声
|
||||
self.playback_cooldown_period = 0.1 # 播放冷却时间（秒）- 防止回声，减少到0.1秒
|
||||
self.playback_completed = False # 播放完成标志
|
||||
self.end_signal_received = False # 结束信号接收标志
|
||||
|
||||
# 智能缓冲系统 - 借鉴 recorder.py 的缓冲策略
|
||||
self.preload_buffer = [] # 预加载缓冲区
|
||||
@ -580,13 +584,24 @@ class OutputProcess:
|
||||
self.tts_speaker = "zh_female_wanqudashu_moon_bigtts"
|
||||
|
||||
# 启动工作线程 - 先启动播放线程，再启动TTS线程
|
||||
self._start_playback_worker()
|
||||
self._start_tts_worker()
|
||||
print("🔊 准备启动播放工作线程...")
|
||||
try:
|
||||
self._start_playback_worker()
|
||||
print("🔊 播放工作线程已启动，准备启动TTS工作线程...")
|
||||
self._start_tts_worker()
|
||||
print("🔊 TTS工作线程已启动")
|
||||
except Exception as e:
|
||||
print(f"❌ 工作线程启动失败: {e}")
|
||||
import traceback
|
||||
traceback.print_exc()
|
||||
|
||||
def _start_playback_worker(self):
|
||||
"""启动播放工作线程 - 借鉴 recorder.py 的播放线程模式"""
|
||||
print("🔊 创建播放工作线程...")
|
||||
self.playback_worker_thread = threading.Thread(target=self._playback_worker, daemon=True)
|
||||
print("🔊 启动播放工作线程...")
|
||||
self.playback_worker_thread.start()
|
||||
print("🔊 播放工作线程已启动")
|
||||
self.logger.info("播放工作线程已启动")
|
||||
|
||||
def _playback_worker(self):
|
||||
@ -597,14 +612,20 @@ class OutputProcess:
|
||||
max_wait_time = 10 # 最多等待10秒
|
||||
wait_start_time = time.time()
|
||||
|
||||
print(f"🔊 播放工作线程等待音频设备就绪... audio={self.audio is not None}, running={self.running}")
|
||||
|
||||
while (self.audio is None or not self.running) and (time.time() - wait_start_time) < max_wait_time:
|
||||
time.sleep(0.1)
|
||||
if (time.time() - wait_start_time) % 1 < 0.1: # 每秒打印一次状态
|
||||
print(f"🔊 仍在等待音频设备就绪... audio={self.audio is not None}, running={self.running}")
|
||||
|
||||
if self.audio is None:
|
||||
print("❌ 音频设备未就绪，播放工作线程退出")
|
||||
self.logger.error("音频设备未就绪，播放工作线程退出")
|
||||
return
|
||||
|
||||
print(f"✅ 音频设备已就绪，耗时: {time.time() - wait_start_time:.1f}秒")
|
||||
|
||||
print("🔊 音频设备已就绪，开始创建播放流")
|
||||
|
||||
# 创建音频播放流
|
||||
@ -643,27 +664,55 @@ class OutputProcess:
|
||||
|
||||
if in_cooldown:
|
||||
# 在冷却期内，跳过播放
|
||||
print(f"🔊 播放冷却中，跳过播放 ({time_since_last_play:.1f}s < {self.playback_cooldown_period}s)")
|
||||
if chunks_played == 0: # 只在第一次遇到冷却期时打印
|
||||
print(f"🔊 播放冷却中，跳过播放 ({time_since_last_play:.1f}s < {self.playback_cooldown_period}s)")
|
||||
self.logger.debug(f"播放冷却中，跳过播放 ({time_since_last_play:.1f}s < {self.playback_cooldown_period}s)")
|
||||
continue
|
||||
|
||||
# 播放音频块
|
||||
if chunks_played == 0: # 只在第一次播放时打印详细信息
|
||||
print(f"🔊 开始播放音频块 {chunks_played + 1}")
|
||||
print(f"🔍 播放工作线程检查: 音频块大小={len(audio_chunk)}字节, "
|
||||
f"冷却期检查={in_cooldown}, 距离上次播放={time_since_last_play:.2f}s, "
|
||||
f"冷却阈值={self.playback_cooldown_period}s")
|
||||
|
||||
# 标记正在播放
|
||||
self.currently_playing = True
|
||||
self.last_playback_time = current_time # 更新最后播放时间
|
||||
# 如果是第一次播放，不设置冷却期
|
||||
if chunks_played == 0:
|
||||
self.last_playback_time = 0 # 第一次播放不触发冷却期
|
||||
else:
|
||||
self.last_playback_time = current_time # 更新最后播放时间
|
||||
|
||||
# 播放音频（同步阻塞，直到播放完成）
|
||||
playback_stream.write(audio_chunk)
|
||||
chunks_played += 1
|
||||
total_size += len(audio_chunk)
|
||||
|
||||
# 减少进度显示频率
|
||||
if chunks_played % 10 == 0:
|
||||
if chunks_played % 10 == 0 or chunks_played <= 3:
|
||||
progress = f"🔊 播放工作: {chunks_played} 块 | {total_size / 1024:.1f} KB"
|
||||
print(f"\r{progress}", end='', flush=True)
|
||||
|
||||
# 播放完成后更新状态
|
||||
self.currently_playing = False
|
||||
|
||||
# 如果这是最后一个音频块，主动检查播放完成
|
||||
if (len(self.playback_buffer) == 0 and
|
||||
len(self.preload_buffer) == 0 and
|
||||
self.tts_task_queue.qsize() == 0 and
|
||||
not self.playback_completed): # 防止重复设置
|
||||
print(f"🔊 播放工作线程：播放完成，设置播放完成标志")
|
||||
print(f"🔊 播放工作线程：播放缓冲={len(self.playback_buffer)}, 预加载={len(self.preload_buffer)}, TTS队列={self.tts_task_queue.qsize()}")
|
||||
self.playback_completed = True
|
||||
print(f"🔊 播放工作线程：playback_completed标志已设置为{self.playback_completed}")
|
||||
# 不在这里直接调用_finish_playback，让主处理循环处理
|
||||
else:
|
||||
self.currently_playing = False
|
||||
else:
|
||||
# 缓冲区为空，检查是否还在接收数据
|
||||
# 缓冲区为空，短暂休眠，减少CPU占用
|
||||
self.currently_playing = False
|
||||
time.sleep(0.01) # 短暂休眠，减少CPU占用
|
||||
time.sleep(0.01)
|
||||
continue
|
||||
|
||||
except Exception as e:
|
||||
@ -712,7 +761,7 @@ class OutputProcess:
|
||||
tts_active = not self.tts_task_queue.empty()
|
||||
playback_active = self.currently_playing or len(self.playback_buffer) > 0
|
||||
|
||||
# 如果设备不健康、正在播放、TTS生成中、或在冷却期内，显示状态并跳过音频处理
|
||||
# 如果设备不健康、正在播放、TTS生成中、或在冷却期内，显示状态但继续处理播放完成检测
|
||||
if not self.audio_device_healthy or tts_active or playback_active or in_cooldown:
|
||||
# 显示播放状态
|
||||
tts_queue_size = self.tts_task_queue.qsize()
|
||||
@ -730,6 +779,11 @@ class OutputProcess:
|
||||
status = f"🔊 {playing_status}... TTS: {tts_queue_size} 播放: {queue_size}"
|
||||
|
||||
print(f"\r{status}", end='', flush=True)
|
||||
|
||||
# 关键修复：即使正在播放，也要检查播放完成
|
||||
if self.end_signal_received:
|
||||
self._check_playback_completion()
|
||||
|
||||
time.sleep(0.1) # 播放时增加延迟减少CPU使用
|
||||
continue
|
||||
|
||||
@ -747,7 +801,11 @@ class OutputProcess:
|
||||
# 4. 显示播放进度
|
||||
self._show_progress()
|
||||
|
||||
# 5. 借鉴 recorder.py: 根据播放状态调整休眠时间，优化性能
|
||||
# 5. 主动检查播放完成（无论什么状态都要检查）
|
||||
if self.end_signal_received:
|
||||
self._check_playback_completion()
|
||||
|
||||
# 6. 借鉴 recorder.py: 根据播放状态调整休眠时间，优化性能
|
||||
if self.is_playing and (self.playback_buffer or self.preload_buffer):
|
||||
time.sleep(0.005) # 播放时极短休眠，提高响应性
|
||||
else:
|
||||
@ -766,8 +824,9 @@ class OutputProcess:
|
||||
def _setup_audio(self):
|
||||
"""设置音频输出设备"""
|
||||
try:
|
||||
print("🔊 开始初始化音频设备...")
|
||||
self.audio = pyaudio.PyAudio()
|
||||
print("🔊 PyAudio实例已创建")
|
||||
print(f"🔊 PyAudio实例已创建: {self.audio}")
|
||||
|
||||
# 主进程不需要创建输出流，由播放工作线程负责
|
||||
# 这里只创建PyAudio实例供播放工作线程使用
|
||||
@ -798,22 +857,32 @@ class OutputProcess:
|
||||
|
||||
if audio_data is None:
|
||||
# 结束信号处理
|
||||
if end_signal_received:
|
||||
print(f"📥 输出进程已收到过结束信号，忽略重复信号")
|
||||
continue
|
||||
|
||||
print(f"📥 输出进程收到结束信号")
|
||||
end_signal_received = True
|
||||
self.end_signal_received = True
|
||||
|
||||
# 重置完成事件标记
|
||||
self.completion_sent = False
|
||||
# 重置播放完成标志
|
||||
self.playback_completed = False
|
||||
print(f"📥 已重置所有播放完成相关标志")
|
||||
|
||||
# 检查是否应该立即结束
|
||||
tts_queue_size = self.tts_task_queue.qsize()
|
||||
playback_queue_size = len(self.playback_buffer) + len(self.preload_buffer)
|
||||
|
||||
if tts_queue_size == 0 and playback_queue_size == 0 and not self.currently_playing:
|
||||
print(f"📥 所有队列已清空且未在播放，处理结束信号")
|
||||
# 双重确认机制：所有队列已清空且播放工作线程已停止
|
||||
print(f"📥 双重确认通过：TTS队列={tts_queue_size}, 播放缓冲={len(self.playback_buffer)}, 预加载={len(self.preload_buffer)}, 正在播放={self.currently_playing}")
|
||||
print(f"📥 所有音频已播放完成，处理结束信号")
|
||||
self._finish_playback()
|
||||
return
|
||||
else:
|
||||
print(f"📥 延迟处理结束信号 - TTS队列: {tts_queue_size}, 播放缓冲: {playback_queue_size}")
|
||||
print(f"📥 延迟处理结束信号 - TTS队列: {tts_queue_size}, 播放缓冲: {playback_queue_size}, 正在播放: {self.currently_playing}")
|
||||
# 重新放回队列，稍后重试
|
||||
self.audio_queue.put(None)
|
||||
time.sleep(0.05)
|
||||
@ -835,17 +904,37 @@ class OutputProcess:
|
||||
print(f"📥 输出进程收到音频数据: {len(audio_data)} 字节")
|
||||
|
||||
# 直接添加到预加载缓冲区
|
||||
print(f"🔍 添加音频到预加载缓冲区: 音频大小={len(audio_data)}字节, "
|
||||
f"添加前预加载缓冲区大小={len(self.preload_buffer)}, "
|
||||
f"添加前播放缓冲区大小={len(self.playback_buffer)}, "
|
||||
f"is_playing={self.is_playing}")
|
||||
self.preload_buffer.append(audio_data)
|
||||
print(f"🔍 添加后预加载缓冲区大小={len(self.preload_buffer)}")
|
||||
|
||||
# 检查是否应该开始播放
|
||||
if (not self.is_playing and
|
||||
len(self.preload_buffer) >= self.preload_size):
|
||||
# 将预加载的数据移到播放缓冲区
|
||||
# 检查是否应该开始播放或补充播放缓冲区
|
||||
if not self.is_playing and len(self.preload_buffer) >= self.preload_size:
|
||||
# 首次启动播放
|
||||
self.playback_buffer.extend(self.preload_buffer)
|
||||
self.preload_buffer.clear()
|
||||
self.is_playing = True
|
||||
self.last_playback_time = time.time()
|
||||
self.last_playback_time = 0 # 重置播放时间，避免立即触发冷却期
|
||||
print(f"🎵 开始播放音频（预加载完成），播放缓冲区大小: {len(self.playback_buffer)}")
|
||||
print(f"🔍 已重置last_playback_time，避免立即触发冷却期")
|
||||
elif self.is_playing and len(self.playback_buffer) < 3 and len(self.preload_buffer) > 0:
|
||||
# 正在播放时，保持播放缓冲区有足够的数据
|
||||
transfer_count = min(2, len(self.preload_buffer)) # 每次转移2个块
|
||||
for _ in range(transfer_count):
|
||||
if self.preload_buffer:
|
||||
self.playback_buffer.append(self.preload_buffer.pop(0))
|
||||
print(f"🔍 播放中补充数据: 转移{transfer_count}个块，播放缓冲区={len(self.playback_buffer)}, 预加载={len(self.preload_buffer)}")
|
||||
elif end_signal_received and not self.is_playing and len(self.playback_buffer) == 0 and len(self.preload_buffer) > 0:
|
||||
# 关键修复：收到结束信号后，如果播放缓冲区为空但预加载缓冲区有数据，强制转移
|
||||
print(f"🔍 结束信号模式下强制转移数据: 预加载缓冲区有 {len(self.preload_buffer)} 个数据块")
|
||||
self.playback_buffer.extend(self.preload_buffer)
|
||||
self.preload_buffer.clear()
|
||||
self.is_playing = True
|
||||
self.last_playback_time = 0
|
||||
print(f"🎵 强制开始播放音频，播放缓冲区大小: {len(self.playback_buffer)}")
|
||||
|
||||
else:
|
||||
print(f"📥 输出进程收到未知类型数据: {type(audio_data)}")
|
||||
@ -856,24 +945,67 @@ class OutputProcess:
|
||||
tts_queue_size = self.tts_task_queue.qsize()
|
||||
playback_queue_size = len(self.playback_buffer) + len(self.preload_buffer)
|
||||
|
||||
if tts_queue_size == 0 and playback_queue_size == 0 and not self.currently_playing:
|
||||
print(f"📥 播放完成，所有工作已完成")
|
||||
self._finish_playback()
|
||||
return
|
||||
print(f"📥 队列空时检查播放完成: TTS={tts_queue_size}, 播放缓冲={len(self.playback_buffer)}, 预加载={len(self.preload_buffer)}, 正在播放={self.currently_playing}")
|
||||
|
||||
if tts_queue_size == 0 and playback_queue_size == 0:
|
||||
# 三重确认机制：所有缓冲区都空了，检查播放状态
|
||||
if not self.currently_playing:
|
||||
print(f"📥 三重确认通过：TTS队列=0, 播放缓冲=0, 预加载=0, 播放状态=False")
|
||||
print(f"📥 调用 _finish_playback() 前，completion_sent={self.completion_sent}")
|
||||
self._finish_playback()
|
||||
print(f"📥 调用 _finish_playback() 后，completion_sent={self.completion_sent}")
|
||||
return
|
||||
else:
|
||||
# 播放工作线程可能还在播放最后一个块，等待一下
|
||||
print(f"📥 等待播放工作线程完成...")
|
||||
time.sleep(0.3) # 增加等待时间，确保播放完成
|
||||
if not self.currently_playing:
|
||||
print(f"📥 三重确认通过（等待后）：TTS队列=0, 播放缓冲=0, 预加载=0, 播放状态=False")
|
||||
print(f"📥 调用 _finish_playback() 前，completion_sent={self.completion_sent}")
|
||||
self._finish_playback()
|
||||
print(f"📥 调用 _finish_playback() 后，completion_sent={self.completion_sent}")
|
||||
return
|
||||
else:
|
||||
print(f"📥 播放工作线程仍在播放，继续等待...")
|
||||
elif tts_queue_size == 0 and playback_queue_size > 0:
|
||||
# 还有数据要播放，继续等待
|
||||
print(f"📥 还有 {playback_queue_size} 个音频块待播放，等待播放完成")
|
||||
time.sleep(0.2) # 增加等待时间
|
||||
elif tts_queue_size == 0 and len(self.playback_buffer) == 0 and len(self.preload_buffer) > 0:
|
||||
# 关键修复：播放缓冲区为空但预加载缓冲区还有数据，需要转移数据
|
||||
print(f"📥 播放缓冲区为空但预加载缓冲区有 {len(self.preload_buffer)} 个数据块，转移数据到播放缓冲区")
|
||||
transfer_count = min(len(self.preload_buffer), 3) # 一次转移最多3个块
|
||||
for _ in range(transfer_count):
|
||||
if self.preload_buffer:
|
||||
self.playback_buffer.append(self.preload_buffer.pop(0))
|
||||
print(f"📥 已转移 {transfer_count} 个数据块到播放缓冲区")
|
||||
time.sleep(0.2) # 增加等待时间
|
||||
|
||||
# 检查是否应该将预加载缓冲区的数据移到播放缓冲区
|
||||
if (not self.is_playing and
|
||||
len(self.preload_buffer) >= self.min_buffer_size):
|
||||
# 检查是否应该补充播放缓冲区的数据
|
||||
if not self.is_playing and len(self.preload_buffer) >= self.min_buffer_size:
|
||||
# 首次启动播放（最小缓冲区模式）
|
||||
self.playback_buffer.extend(self.preload_buffer)
|
||||
self.preload_buffer.clear()
|
||||
self.is_playing = True
|
||||
self.last_playback_time = time.time()
|
||||
self.last_playback_time = 0 # 重置播放时间，避免立即触发冷却期
|
||||
print(f"🎵 开始播放音频（最小缓冲区满足）")
|
||||
print(f"🔍 已重置last_playback_time，避免立即触发冷却期")
|
||||
elif self.is_playing and len(self.playback_buffer) < 2 and len(self.preload_buffer) > 0:
|
||||
# 正在播放时补充数据，避免播放缓冲区耗尽
|
||||
transfer_count = min(1, len(self.preload_buffer)) # 每次转移1个块
|
||||
for _ in range(transfer_count):
|
||||
if self.preload_buffer:
|
||||
self.playback_buffer.append(self.preload_buffer.pop(0))
|
||||
print(f"🔍 队列空时补充数据: 转移{transfer_count}个块，播放缓冲区={len(self.playback_buffer)}, 预加载={len(self.preload_buffer)}")
|
||||
|
||||
# 退出循环，避免过度占用CPU
|
||||
if processed_count > 0:
|
||||
break
|
||||
else:
|
||||
# 关键修复：即使没有处理数据，也要检查播放完成
|
||||
if self.end_signal_received:
|
||||
self._check_playback_completion()
|
||||
|
||||
time.sleep(0.01)
|
||||
|
||||
except Exception as e:
|
||||
@ -920,15 +1052,18 @@ class OutputProcess:
|
||||
"""完成播放 - 借鉴 recorder.py 的优雅完成机制"""
|
||||
# 防止重复发送完成事件
|
||||
if self.completion_sent:
|
||||
print("📡 输出进程：完成事件已发送过，跳过重复发送")
|
||||
return
|
||||
|
||||
print("📡 输出进程：开始执行播放完成逻辑")
|
||||
|
||||
self.is_playing = False
|
||||
self.playback_buffer.clear()
|
||||
self.preload_buffer.clear()
|
||||
self.last_playback_time = 0
|
||||
|
||||
# 通知主进程播放完成
|
||||
if self.event_queue and not self.completion_sent:
|
||||
if self.event_queue:
|
||||
try:
|
||||
# 发送播放完成事件到主进程
|
||||
completion_event = ProcessEvent(
|
||||
@ -941,13 +1076,65 @@ class OutputProcess:
|
||||
print("📡 输出进程：已发送播放完成事件到主进程")
|
||||
self.completion_sent = True
|
||||
except Exception as e:
|
||||
print(f"❌ 输出进程：发送播放完成事件失败: {e}")
|
||||
self.logger.error(f"发送播放完成事件失败: {e}")
|
||||
else:
|
||||
if not self.event_queue:
|
||||
print("⚠️ 输出进程：未设置事件队列，无法通知主进程播放完成")
|
||||
print("⚠️ 输出进程：未设置事件队列，无法通知主进程播放完成")
|
||||
|
||||
# 额外等待确保音频设备完全停止
|
||||
time.sleep(0.5)
|
||||
print("📡 输出进程：播放完成逻辑执行完毕")
|
||||
|
||||
def _check_playback_completion(self):
|
||||
"""检查播放完成状态 - 独立的播放完成检测方法"""
|
||||
if not self.end_signal_received:
|
||||
return
|
||||
|
||||
tts_queue_size = self.tts_task_queue.qsize()
|
||||
playback_queue_size = len(self.playback_buffer) + len(self.preload_buffer)
|
||||
|
||||
print(f"🔍 播放完成检查: TTS队列={tts_queue_size}, 播放缓冲={len(self.playback_buffer)}, 预加载={len(self.preload_buffer)}, 正在播放={self.currently_playing}, 播放完成标志={self.playback_completed}")
|
||||
|
||||
# 检查条件1: 播放完成标志被设置
|
||||
if self.playback_completed:
|
||||
print(f"✅ 检测到播放完成标志，触发播放完成")
|
||||
print(f"📥 调用 _finish_playback() 前，completion_sent={self.completion_sent}")
|
||||
self._finish_playback()
|
||||
print(f"📥 调用 _finish_playback() 后，completion_sent={self.completion_sent}")
|
||||
# 重要：重置播放完成标志，防止重复触发
|
||||
self.playback_completed = False
|
||||
print(f"📥 已重置播放完成标志，防止重复触发")
|
||||
return
|
||||
|
||||
# 检查条件2: 所有队列为空且没有在播放
|
||||
if tts_queue_size == 0 and playback_queue_size == 0 and not self.currently_playing:
|
||||
print(f"✅ 所有队列已清空且播放器空闲，触发播放完成")
|
||||
print(f"📥 调用 _finish_playback() 前，completion_sent={self.completion_sent}")
|
||||
self._finish_playback()
|
||||
print(f"📥 调用 _finish_playback() 后，completion_sent={self.completion_sent}")
|
||||
# 重要：重置结束信号标志，防止重复触发
|
||||
self.end_signal_received = False
|
||||
print(f"📥 已重置结束信号标志，防止重复触发")
|
||||
return
|
||||
|
||||
# 检查条件3: 所有队列为空但播放器还在播放（最后一个音频块）
|
||||
if tts_queue_size == 0 and playback_queue_size == 0 and self.currently_playing:
|
||||
print(f"⏳ 等待最后一个音频块播放完成...")
|
||||
time.sleep(0.3)
|
||||
if not self.currently_playing:
|
||||
print(f"✅ 最后一个音频块播放完成，触发播放完成")
|
||||
print(f"📥 调用 _finish_playback() 前，completion_sent={self.completion_sent}")
|
||||
self._finish_playback()
|
||||
print(f"📥 调用 _finish_playback() 后，completion_sent={self.completion_sent}")
|
||||
# 重要：重置结束信号标志，防止重复触发
|
||||
self.end_signal_received = False
|
||||
print(f"📥 已重置结束信号标志，防止重复触发")
|
||||
return
|
||||
else:
|
||||
print(f"⚠️ 最后一个音频块播放超时")
|
||||
# 即使超时也要重置标志，防止重复检测
|
||||
self.end_signal_received = False
|
||||
print(f"📥 已重置结束信号标志（超时情况）")
|
||||
|
||||
def _cleanup(self):
|
||||
"""清理资源"""
|
||||
|
||||
@ -28,6 +28,19 @@ from audio_processes import (
|
||||
RecordingState, ControlCommand, ProcessEvent
|
||||
)
|
||||
|
||||
def output_process_target(audio_queue, config, event_queue):
|
||||
"""输出进程的目标函数 - 在子进程中创建OutputProcess实例"""
|
||||
try:
|
||||
print("🔊 输出进程目标函数开始执行...")
|
||||
output_process = OutputProcess(audio_queue, config, event_queue)
|
||||
print("🔊 OutputProcess实例创建成功，开始运行...")
|
||||
output_process.run()
|
||||
print("🔊 输出进程运行完成")
|
||||
except Exception as e:
|
||||
print(f"❌ 输出进程出错: {e}")
|
||||
import traceback
|
||||
traceback.print_exc()
|
||||
|
||||
class ControlSystem:
|
||||
"""主控制系统"""
|
||||
|
||||
@ -213,11 +226,8 @@ class ControlSystem:
|
||||
}
|
||||
|
||||
self.output_process = mp.Process(
|
||||
target=OutputProcess(
|
||||
self.output_audio_queue,
|
||||
output_config,
|
||||
self.output_event_queue # 传递事件队列
|
||||
).run
|
||||
target=output_process_target,
|
||||
args=(self.output_audio_queue, output_config, self.output_event_queue)
|
||||
)
|
||||
|
||||
# 启动进程
|
||||
@ -312,6 +322,7 @@ class ControlSystem:
|
||||
|
||||
if event.event_type == 'playback_complete':
|
||||
print("📡 主控制：收到播放完成事件")
|
||||
print(f"📡 主控制：事件详情 - 类型: {event.event_type}, 元数据: {event.metadata}")
|
||||
self._handle_playback_complete(event)
|
||||
|
||||
except queue.Empty:
|
||||
@ -338,18 +349,31 @@ class ControlSystem:
|
||||
|
||||
def _handle_playback_complete(self, event: ProcessEvent):
|
||||
"""处理播放完成事件"""
|
||||
print(f"📡 主控制：开始处理播放完成事件")
|
||||
print(f"📡 主控制：当前状态 = {self.state.value}")
|
||||
print(f"📡 主控制：事件元数据 = {event.metadata}")
|
||||
|
||||
# 标记播放完成
|
||||
self.playback_complete = True
|
||||
print(f"📡 主控制：已设置 playback_complete = True")
|
||||
|
||||
# 更新统计
|
||||
self.stats['total_conversations'] += 1
|
||||
print(f"📡 主控制：已更新统计，对话数 = {self.stats['total_conversations']}")
|
||||
|
||||
# 切换到空闲状态
|
||||
old_state = self.state.value
|
||||
self.state = RecordingState.IDLE
|
||||
print(f"🎯 状态：PLAYING → IDLE")
|
||||
print(f"🎯 状态：{old_state} → IDLE")
|
||||
|
||||
# 重新启用输入进程录音功能
|
||||
self.input_command_queue.put(ControlCommand('enable_recording'))
|
||||
try:
|
||||
self.input_command_queue.put(ControlCommand('enable_recording'))
|
||||
print(f"📡 主控制：已发送 enable_recording 命令到输入进程")
|
||||
except Exception as e:
|
||||
print(f"❌ 主控制：发送 enable_recording 命令失败: {e}")
|
||||
|
||||
print(f"📡 主控制：播放完成事件处理完成")
|
||||
|
||||
def _process_audio_pipeline(self):
|
||||
"""处理音频流水线：STT + LLM + TTS"""
|
||||
@ -544,7 +568,6 @@ class ControlSystem:
|
||||
|
||||
if serialization_method == 0b0001: # JSON
|
||||
payload_msg = json.loads(str(payload_msg, "utf-8"))
|
||||
print(f"📋 解析后的JSON: {json.dumps(payload_msg, indent=2, ensure_ascii=False)}")
|
||||
|
||||
result['payload_msg'] = payload_msg
|
||||
result['payload_size'] = payload_size
|
||||
@ -721,7 +744,6 @@ class ControlSystem:
|
||||
return results
|
||||
else:
|
||||
print(f"❌ 未找到result字段，可用字段: {list(payload_msg.keys())}")
|
||||
print(f"完整payload: {json.dumps(payload_msg, indent=2, ensure_ascii=False)}")
|
||||
else:
|
||||
print(f"❌ Payload不是字典类型: {type(payload_msg)}")
|
||||
else:
|
||||
@ -854,11 +876,8 @@ class ControlSystem:
|
||||
if not chunk:
|
||||
continue
|
||||
|
||||
print(f"🔍 原始TTS响应块 {chunk_count + 1}: {chunk[:100]}...")
|
||||
|
||||
try:
|
||||
data = json.loads(chunk)
|
||||
print(f"🔍 解析后的TTS块 {chunk_count + 1}: {data}")
|
||||
|
||||
if data.get("code", 0) == 0 and "data" in data and data["data"]:
|
||||
chunk_audio = base64.b64decode(data["data"])
|
||||
@ -899,9 +918,41 @@ class ControlSystem:
|
||||
print(f"\n✅ TTS音频生成完成: {chunk_count} 块, {total_audio_size / 1024:.1f} KB")
|
||||
print(f"📊 队列大小变化: {queue_size_before} -> {self.output_audio_queue.qsize()}")
|
||||
|
||||
# 不再在这里发送结束信号，让输出进程自然播放完所有音频
|
||||
print(f"📦 TTS音频数据已全部发送，等待输出进程播放完成")
|
||||
# 等待音频数据被输出进程完全处理
|
||||
print(f"📦 TTS音频数据已全部发送，等待输出进程处理...")
|
||||
max_wait_time = 30 # 最多等待30秒
|
||||
wait_start_time = time.time()
|
||||
last_queue_size = self.output_audio_queue.qsize()
|
||||
|
||||
while time.time() - wait_start_time < max_wait_time:
|
||||
current_queue_size = self.output_audio_queue.qsize()
|
||||
|
||||
# 如果队列为空或队列大小不再变化，说明音频数据已被处理
|
||||
if current_queue_size == 0:
|
||||
print(f"✅ 音频队列已清空，可以发送结束信号")
|
||||
break
|
||||
elif current_queue_size == last_queue_size:
|
||||
# 队列大小不再变化，可能处理完成
|
||||
print(f"📦 队列大小稳定在 {current_queue_size}，等待确认...")
|
||||
time.sleep(1)
|
||||
if self.output_audio_queue.qsize() == current_queue_size:
|
||||
print(f"✅ 队列大小稳定，可以发送结束信号")
|
||||
break
|
||||
else:
|
||||
print(f"📦 等待队列处理: {current_queue_size} 个项目待处理")
|
||||
last_queue_size = current_queue_size
|
||||
|
||||
time.sleep(0.5)
|
||||
|
||||
# 发送结束信号，通知输出进程所有音频已发送完成
|
||||
print(f"📦 发送结束信号到输出进程")
|
||||
print(f"📊 音频队列当前大小: {self.output_audio_queue.qsize()}")
|
||||
print(f"📦 注意：结束信号不会立即触发播放完成，会等待所有音频播放完成")
|
||||
try:
|
||||
self.output_audio_queue.put(None)
|
||||
print(f"📡 已发送结束信号到输出进程")
|
||||
except Exception as e:
|
||||
print(f"❌ 发送结束信号失败: {e}")
|
||||
|
||||
return chunk_count > 0
|
||||
|
||||
|
||||
47
recorder.py
47
recorder.py
@ -401,12 +401,16 @@ class EnergyBasedRecorder:
|
||||
if time_since_last >= self.tts_accumulation_time and len(self.tts_buffer) >= self.tts_buffer_min_size:
|
||||
return True
|
||||
|
||||
# 检查句子特征 - 长句子优先（调整为100字符，防止回声）
|
||||
if len(sentence) > 100: # 超过100字符的长句子立即触发
|
||||
# 检查是否为完整句子（使用新的严格检测）
|
||||
if self._is_complete_sentence(sentence):
|
||||
return True
|
||||
|
||||
# 中等长度句子（80-100字符）如果有结束标点也触发
|
||||
if len(sentence) > 80:
|
||||
# 检查句子特征 - 长句子优先（50字符以上）
|
||||
if len(sentence) > 50: # 超过50字符的句子立即触发
|
||||
return True
|
||||
|
||||
# 中等长度句子（30-50字符）如果有结束标点也触发
|
||||
if len(sentence) > 30:
|
||||
end_punctuations = ['。', '！', '？', '.', '!', '?']
|
||||
if any(sentence.strip().endswith(p) for p in end_punctuations):
|
||||
return True
|
||||
@ -1808,12 +1812,18 @@ class EnergyBasedRecorder:
|
||||
if not text or len(text.strip()) == 0:
|
||||
return False
|
||||
|
||||
# 增加最小长度要求 - 至少10个字符才考虑作为完整句子
|
||||
if len(text.strip()) < 10:
|
||||
return False
|
||||
|
||||
# 句子结束标点符号
|
||||
sentence_endings = r'[。！？.!?]'
|
||||
|
||||
# 检查是否以句子结束符结尾
|
||||
if re.search(sentence_endings + r'\s*$', text):
|
||||
return True
|
||||
# 对于以结束符结尾的句子，要求至少15个字符
|
||||
if len(text.strip()) >= 15:
|
||||
return True
|
||||
|
||||
# 检查是否包含句子结束符（可能在句子中间）
|
||||
if re.search(sentence_endings, text):
|
||||
@ -1821,18 +1831,25 @@ class EnergyBasedRecorder:
|
||||
remaining_text = re.split(sentence_endings, text, 1)[-1]
|
||||
if len(remaining_text.strip()) > 0:
|
||||
return False
|
||||
# 对于包含结束符的句子，要求至少20个字符
|
||||
if len(text.strip()) >= 20:
|
||||
return True
|
||||
|
||||
# 对于较长的文本（超过50字符），即使没有结束符也可以考虑
|
||||
if len(text.strip()) >= 50:
|
||||
return True
|
||||
|
||||
# 对于较短的文本，如果包含常见完整句式模式
|
||||
common_patterns = [
|
||||
r'^[是的有没有来去在把被让叫请使].*[的得了吗呢吧啊呀]',
|
||||
r'^(你好|谢谢|再见|是的|不是|好的|没问题)',
|
||||
r'^[\u4e00-\u9fff]{2,4}[的得了]$' # 2-4个中文字+的/了/得
|
||||
]
|
||||
|
||||
for pattern in common_patterns:
|
||||
if re.match(pattern, text):
|
||||
return True
|
||||
# 对于中等长度的文本，如果包含常见完整句式模式
|
||||
if len(text.strip()) >= 20:
|
||||
common_patterns = [
|
||||
r'^[是的有没有来去在把被让叫请使].*[的得了吗呢吧啊呀]',
|
||||
r'^(你好|谢谢|再见|是的|不是|好的|没问题)',
|
||||
r'^[\u4e00-\u9fff]{4,8}[的得了]$' # 4-8个中文字+的/了/得
|
||||
]
|
||||
|
||||
for pattern in common_patterns:
|
||||
if re.match(pattern, text):
|
||||
return True
|
||||
|
||||
return False
|
||||
|
||||
|
||||
Loading…
Reference in New Issue
Block a user