feat(exercise): 优化死虫式训练姿态检测算法

- 调整视频处理频率从每帧处理改为每2帧处理
- 添加膝角趋势平滑算法减少单帧抖动误判
- 改进对角伸展检测逻辑支持准备位手臂上举
- 优化状态机确保严格回到准备姿态才计数
- 添加姿态丢失时的候选帧清理机制
- 更新音频文件生成路径至resources目录
- 改进macOS音频生成使用AIFF格式提高质量
- 添加详细的帧处理日志输出间隔配置

app/webrtc/video_receiver.py

@@ -33,7 +33,12 @@ class VideoReceiver:
 
     async def run(self) -> None:
         """持续接收视频帧并进行姿态检测、渲染和语音播报"""
-        logger.info("Start receiving video frames, process_every_n={}", config.video.process_every_n_frames)
+        log_every_n_frames = max(1, config.video.log_every_n_frames)
+        logger.info(
+            "Start receiving video frames, process_every_n={}, log_every_n={}",
+            config.video.process_every_n_frames,
+            log_every_n_frames,
+        )
 
         frame_count = 0
         processed_count = 0
@@ -70,7 +75,7 @@ class VideoReceiver:
                 display_img = last_annotated if last_annotated is not None else raw_img
                 show_frame(display_img)
 
-                if frame_count % 100 == 0:
+                if frame_count % log_every_n_frames == 0:
                     logger.info(
                         "Received {} frames, processed={}, raw_shape={}, reps={}, phase={}, feedback={}, {}",
                         frame_count,