Merge pull request #2190 from xinnan-tech/py_fix_voiceprint

update:声纹识别添加相似度阈值检查机制

main/manager-api/src/main/java/xiaozhi/modules/config/service/impl/ConfigServiceImpl.java

@@ -298,6 +298,20 @@ public class ConfigServiceImpl implements ConfigService {
             Map<String, Object> voiceprintConfig = new HashMap<>();
             voiceprintConfig.put("url", voiceprintUrl);
             voiceprintConfig.put("speakers", speakers);
+            
+            // 获取声纹识别相似度阈值，默认0.4
+            String thresholdStr = sysParamsService.getValue("server.voiceprint_similarity_threshold", true);
+            if (StringUtils.isNotBlank(thresholdStr) && !"null".equals(thresholdStr)) {
+                try {
+                    double threshold = Double.parseDouble(thresholdStr);
+                    voiceprintConfig.put("similarity_threshold", threshold);
+                } catch (NumberFormatException e) {
+                    // 如果解析失败，使用默认值0.4
+                    voiceprintConfig.put("similarity_threshold", 0.4);
+                }
+            } else {
+                voiceprintConfig.put("similarity_threshold", 0.4);
+            }
 
             result.put("voiceprint", voiceprintConfig);
         } catch (Exception e) {

main/manager-api/src/main/resources/db/changelog/202509081140.sql

@@ -0,0 +1,4 @@
+-- 添加声纹识别相似度阈值参数配置
+delete from `sys_params` where id = 115;
+INSERT INTO `sys_params` (id, param_code, param_value, value_type, param_type, remark)
+VALUES (115, 'server.voiceprint_similarity_threshold', '0.4', 'string', 1, '声纹识别相似度阈值，范围0.0-1.0，默认0.4，数值越高越严格');

main/manager-api/src/main/resources/db/changelog/db.changelog-master.yaml

@@ -303,10 +303,17 @@ databaseChangeLog:
         - sqlFile:
             encoding: utf8
             path: classpath:db/changelog/202508131557.sql
+  - changeSet:
+      id: 202509081140
+      author: cgd
+      changes:
+        - sqlFile:
+            encoding: utf8
+            path: classpath:db/changelog/202509081140.sql
   - changeSet:
       id: 202508271113
       author: cgd
       changes:
         - sqlFile:
             encoding: utf8
-            path: classpath:db/changelog/202508271113.sql
+            path: classpath:db/changelog/202508271113.sql

main/xiaozhi-server/config.yaml

@@ -151,6 +151,9 @@ voiceprint:
     - "test1,张三,张三是一个程序员"
     - "test2,李四,李四是一个产品经理"
     - "test3,王五,王五是一个设计师"
+  # 声纹识别相似度阈值，范围0.0-1.0，默认0.4
+  # 数值越高越严格，减少误识别但可能增加拒识率
+  similarity_threshold: 0.4
 
 # #####################################################################################
 # ################################以下是角色模型配置######################################

main/xiaozhi-server/core/utils/voiceprint_provider.py

@@ -19,6 +19,8 @@ class VoiceprintProvider:
         self.original_url = config.get("url", "")
         self.speakers = config.get("speakers", [])
         self.speaker_map = self._parse_speakers()
+        # 声纹识别相似度阈值，默认0.4
+        self.similarity_threshold = float(config.get("similarity_threshold", 0.4))
         
         # 解析API地址和密钥
         self.api_url = None
@@ -62,7 +64,7 @@ class VoiceprintProvider:
                     # 进行健康检查，验证服务器是否可用
                     if self._check_server_health():
                         self.enabled = True
-                        logger.bind(tag=TAG).info(f"声纹识别已启用: API={self.api_url}, 说话人={len(self.speaker_ids)}个")
+                        logger.bind(tag=TAG).info(f"声纹识别已启用: API={self.api_url}, 说话人={len(self.speaker_ids)}个, 相似度阈值={self.similarity_threshold}")
                     else:
                         self.enabled = False
                         logger.bind(tag=TAG).warning(f"声纹识别服务器不可用，声纹识别已禁用: {self.api_url}")
@@ -169,12 +171,14 @@ class VoiceprintProvider:
                         
                         logger.bind(tag=TAG).info(f"声纹识别耗时: {total_elapsed_time:.3f}s")
                         
-                        # 置信度检查
-                        if score < 0.5:
-                            logger.bind(tag=TAG).warning(f"声纹识别置信度较低: {score:.3f}")
+                        # 相似度阈值检查
+                        if score < self.similarity_threshold:
+                            logger.bind(tag=TAG).warning(f"声纹识别相似度{score:.3f}低于阈值{self.similarity_threshold}")
+                            return "未知说话人"
                         
                         if speaker_id and speaker_id in self.speaker_map:
                             result_name = self.speaker_map[speaker_id]["name"]
+                            logger.bind(tag=TAG).info(f"声纹识别成功: {result_name} (相似度: {score:.3f})")
                             return result_name
                         else:
                             logger.bind(tag=TAG).warning(f"未识别的说话人ID: {speaker_id}")