Hass/xiaozhi-esp32-server
1
0
Fork 0
mirror of https://github.com/xinnan-tech/xiaozhi-esp32-server.git synced 2026-07-26 00:53:54 +08:00
Code Issues Packages Projects Releases Wiki Activity

update: 声纹识别添加相似度阈值检查机制

Browse Source
This commit is contained in:
3030332422
2025-09-07 15:30:04 +08:00
parent d04ec9d510
commit 25d9e6c8cf
2 changed files with 11 additions and 4 deletions
Download Patch File Download Diff File Expand all files Collapse all files
main/xiaozhi-server/config.yaml
+3
View File
@@ -148,6 +148,9 @@ voiceprint:
- "test1,张三,张三是一个程序员" - "test1,张三,张三是一个程序员"
- "test2,李四,李四是一个产品经理" - "test2,李四,李四是一个产品经理"
- "test3,王五,王五是一个设计师" - "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###################################### # ################################以下是角色模型配置######################################
main/xiaozhi-server/core/utils/voiceprint_provider.py
+8 -4
View File
@@ -19,6 +19,8 @@ class VoiceprintProvider:
self.original_url = config.get("url", "") self.original_url = config.get("url", "")
self.speakers = config.get("speakers", []) self.speakers = config.get("speakers", [])
self.speaker_map = self._parse_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 # 解析API地址和密钥
self.api_url = None 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 # 进行健康检查，验证服务器是否可用
if self._check_server_health(): if self._check_server_health():
self.enabled = True 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: else:
self.enabled = False 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}") 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") logger.bind(tag=TAG).info(f"声纹识别耗时: {total_elapsed_time:.3f}s")
# 置信度检查 # 相似度阈值检查
if score < 0.5: if score < self.similarity_threshold:
logger.bind(tag=TAG).warning(f"声纹识别置信度较低: {score:.3f}") logger.bind(tag=TAG).warning(f"声纹识别相似度{score:.3f}低于阈值{self.similarity_threshold}，拒绝识别")
return "未知说话人"
if speaker_id and speaker_id in self.speaker_map: if speaker_id and speaker_id in self.speaker_map:
result_name = self.speaker_map[speaker_id]["name"] 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 return result_name
else: else:
logger.bind(tag=TAG).warning(f"未识别的说话人ID: {speaker_id}") logger.bind(tag=TAG).warning(f"未识别的说话人ID: {speaker_id}")