基于GMM的移动终端声纹持续认证  

Continuous Voiceprint Authentication System for Mobile Terminal Devices based on GMM

作　　者：石元兵[1] 何雪海[1] 

机构地区：[1]卫士通信息产业股份有限公司,四川成都610041

出　　处：《通信技术》2017年第8期1800-1804,共5页Communications Technology

摘　　要：移动终端安全是安全领域中一个重要的研究问题,目前通常应用在开机密码、锁屏登陆阶段,具体包括密码口令、锁屏图形、指纹等常见安全措施认证。此外,还可增强安全措施、添加加密卡提升安全,但无法解决当设备密码被人破译后的使用权限管理问题。为此,设计一种利用声纹进行持续认证的安全系统方案。采用主流的声学识别特征和文本无关的说话人确认算法高斯混合模型进行说话人确认,并依据模型分数进行相应的安全响应策略。在Android移动设备上进行验证实验,结果表明该系统具有较高的正确接收率和正确拒绝率,能识别冒认者,提高了移动终端设备的安全功能。Mobile terminal security, as an important research issue in the field of security, is usually applied to in the boot password, screen-lock log-on phase, involving certifications of passwords, screen-lock graphics, fingerprints and other common security measures. In addition, it can also strengthen the security measures, and by adding encryption cards, improve the security, but cannot solve the password management problem when the device password is deciphered. For this, the security system scheme based on voiceprint for continuous authentication is designed. And the Gauss mixture model with principal acoustic identification feature and the text-independent speaker verification algorithm also established to confirm the speaker, while the corresponding security response strategy is implemented in accordance with the model score. Validation experiments on Android mobile terminal devices indicate that this system has fairly high correct-reception rate and correct-rejection rate, and could identify the impostor and improve the security function of mobile terminal devices.

关 键 词：声纹认证 高斯混合模型 端点检测 移动终端安全 SM4 

分 类 号：TP309[自动化与计算机技术—计算机系统结构]