基于卢氏特征安全模板的语音生物哈希检索算法

Speech biological Hash retrieval algorithm based on Lu's feature security template

为提高对语音特征的保护及生物哈希构造过程的多样性,解决现有语音生物哈希检索算法哈希构造过程单一、不能对语音特征形成有效保护等问题,提出一种基于卢氏特征安全模板的语音生物哈希检索算法.首先,通过伽马通倒谱系数(GTCC)算法提取原始语音的音频特征,并利用Toeplitz和循环测量矩阵对音频特征进行降维处理;然后,将降维后特征向量差分处理并通过支持向量机(SVM)分类,根据分类结果产生一一对应的密钥,并将密钥作为卢氏混沌映射初始值进行多轮迭代构建对应密钥的卢氏特征安全模板;最后,降维后的特征向量通过对应卢氏特征安全模板量化得到生物哈希.实验结果表明:所提出的卢氏特征安全模板提高了生物哈希构造过程的安全性和多样性;同时,卢氏特征安全模板产生的生物哈希具有较好的区分性和鲁棒性,对音量调节、滤波、重采样及格式压缩等内容保持操作语音具有较好的检索性能.

To improve the protection of speech features and the diversity of biological Hash construction process,and address the problem that the existing speech biological Hash retrieval algorithm could not form effective protection of speech features due to the single Hash construction process,a speech biological Hash retrieval algorithm based on Lu's feature security template was proposed.First,gammatone cepstral coefficient(GTCC)algorithm was used to extract the audio features of the original speech,and Toeplitz and cyclic measurement matrix were used to reduce the dimension of the audio features.Then,the dimensionally reduced feature vectors were processed differentially and classified by support vector machine(SVM).According to the classification results,one to one corresponding key was generated,and the key was used as the initial value of Lu's chaos mapping to construct the Lu's feature security template of corresponding key.Finally,the feature vectors after dimensionality reduction were quantized by the corresponding Lu's feature security template to obtain the biological Hash.Experimental results show that the proposed security template can improve the security and diversity of the biological Hash construction process.Meanwhile,the biological Hash generated by the security template has good differentiation and robustness,and can maintain the retrieval of operational speech for volume adjustment,filtering,resample and format compression.