diff --git a/backend/app/admin/schema/audit_log.py b/backend/app/admin/schema/audit_log.py
index 60bff19..5d12e29 100755
--- a/backend/app/admin/schema/audit_log.py
+++ b/backend/app/admin/schema/audit_log.py
@@ -18,7 +18,7 @@ class AuditLogSchemaBase(SchemaBase):
     status_code: int = Field(description="HTTP状态码")
     error_message: Optional[str] = Field("", description="错误信息")
     called_at: Optional[datetime] = Field(None, description="调用时间")
-    image_id: int = Field(description="关联的图片ID")
+    image_id: Optional[int] = Field(None, description="关联的图片ID")
     user_id: int = Field(description="调用用户ID")
     api_version: str = Field(description="API版本")
     dict_level: Optional[str] = Field(None, description="词典等级")
diff --git a/backend/app/ai/api/qa.py b/backend/app/ai/api/qa.py
index b027975..943b9f5 100644
--- a/backend/app/ai/api/qa.py
+++ b/backend/app/ai/api/qa.py
@@ -77,7 +77,7 @@ async def reply_conversation(request: Request, session_id: int, obj: Conversatio
 
 
 @router.post('/conversations/{session_id}/recognize_audio', summary='识别音频内容', dependencies=[DependsJwtAuth])
-async def recognize_audio(request: Request, obj: ConversationRecognitionRequest) -> ResponseSchemaModel[ConversationRecognitionResponse]:
+async def recognize_audio(request: Request, session_id: int, obj: ConversationRecognitionRequest) -> ResponseSchemaModel[ConversationRecognitionResponse]:
     res = await qa_service.recognize_audio(file_id=int(obj.file_id), user_id=request.user.id, session_id=session_id)
     return response_base.success(data=ConversationRecognitionResponse(**res))
 
diff --git a/backend/app/ai/service/qa_service.py b/backend/app/ai/service/qa_service.py
index 12f5790..21731a9 100644
--- a/backend/app/ai/service/qa_service.py
+++ b/backend/app/ai/service/qa_service.py
@@ -897,6 +897,7 @@ class QaService:
         
         # 1. 验证会话
         exercise_id = None
+        image_id = None
         async with async_db_session.begin() as db:
             session = await qa_session_dao.get(db, session_id)
             if not session:
@@ -904,6 +905,10 @@ class QaService:
             if session.starter_user_id != user_id:
                 raise errors.ForbiddenError(msg="Forbidden")
             exercise_id = session.exercise_id
+            
+            exercise = await qa_exercise_dao.get(db, exercise_id)
+            if exercise:
+                image_id = exercise.image_id
 
         # 2. 获取文件信息
         file_obj = await file_service.get_file(file_id)
@@ -935,7 +940,7 @@ class QaService:
                 audio_url = temp_file_path
             
             # 调用Qwen ASR
-            res = await Qwen.recognize_speech(audio_url, user_id=user_id)
+            res = await Qwen.recognize_speech(audio_url, user_id=user_id, image_id=image_id)
             
             if not res.get("success"):
                 raise errors.ServerError(msg=res.get("error") or "ASR failed")
diff --git a/backend/middleware/qwen.py b/backend/middleware/qwen.py
index 3688c8e..0f14e4d 100755
--- a/backend/middleware/qwen.py
+++ b/backend/middleware/qwen.py
@@ -317,11 +317,12 @@ class Qwen:
         )
 
     @staticmethod
-    async def recognize_speech(file_path: str, user_id: int = 0) -> Dict[str, Any]:
+    async def recognize_speech(file_path: str, user_id: int = 0, image_id: int | None = None) -> Dict[str, Any]:
         """
         调用通义千问API识别语音内容
         :param file_path: 音频文件路径 (本地路径)
         :param user_id: 用户ID
+        :param image_id: 关联的图片ID
         :return: 识别结果
         """
         api_key = _get_primary_qwen_api_key()
@@ -377,7 +378,7 @@ class Qwen:
                 status_code=status_code,
                 error_message=None,
                 called_at=start_at,
-                image_id=0,
+                image_id=image_id,
                 user_id=user_id,
                 cost=0,
                 api_version=settings.FASTAPI_API_V1_PATH,
@@ -428,7 +429,7 @@ class Qwen:
                     duration=time.time() - start_time,
                     status_code=status_code,
                     error_message=error_message,
-                    image_id=0,
+                    image_id=image_id,
                     user_id=user_id,
                     called_at=start_at
                 )