一种基于感知特性的鲁棒性语音认证算法

A robust speech authentication algorithm based on perceptual characteristics

提出了一种基于感知域的鲁棒性语音认证算法,将语音的感知特性与签名算法相结合,在满足内容认证和身份认证的同时,能够有效地抵抗通信噪声微扰.算法基于语音的掩蔽效应和非线性效应等人耳感知特性,着重去除其时频域掩蔽阈值下的冗余信息,进行非线性滤波后提取感知参数,并运用改进的Rainbow算法对这些语音参数进行签名.实验证明,该算法的唯一性和针对通信噪声的鲁棒性都很好,兼有Rainbow签名的安全性保证,可以满足语音通信中的鲁棒性认证要求.

At present, there is a growing need for multimedia information authentication, but studies on speech authentication are rare. Based on perceptual characteristics, a robust speech authentication algorithm is proposed to combine the psychoacoustic properties with signature algorithm. It meets the requirements of entity authentication and content authentication, and resists the channel noise as well. Analyzing the perceptual properties, such as masking effect and non-linear effect, speech redundancy in both temporal field and frequency field is eliminated, and perceptual parameters are extracted. Then an improved Rainbow algorithm is used to sign the extracted data. Experiment results demonstrate good robustness and uniqueness of the algorithm when applied to the robust authentication of audio communication.