节点文献
语音密码机中的语音压缩改进算法研究
The Improvement of Speech Compression Coding on Voice Cipher
【作者】 王国文；
【导师】 赵耿；
【作者基本信息】 西安电子科技大学 ， 密码学， 2011， 硕士
【摘要】 现阶段,为了满足高速发展的数字通信和其他商业应用需求,语音压缩编码技术得到了良好发展。语音密码机的实现的核心技术是语音压缩编解码技术,在本论文中选择了MELP语音压缩算法来实现。在语音编码算法中,MELP算法由于能更好的模拟自然语音的特征,能在低速率上合成高质量的语音,因此成为现在低速率语音编码中很有潜力的算法。它在原有的LPC(线性预测)编码的基础上,采用了混合激励、非周期脉冲、自适应谱增强、脉冲整形滤波和傅氏级数幅度值等五项新技术,使得合成语音能更好地拟合自然语音。从而较好的实现了低码率的语音编码。本论文深入分析了MELP的语音编解码算法的原理,对它的编解码过程中的关键技术进行了归纳总结,其中的一些公式进行了理论推导,并在这基础上对MELP算法在LSF量化方面提出了改进,并做畸变比较分析。数据结果表明,改进后的MELP在2.4kb/s上得到更好的语音合成,更好的拟合成自然语音。最后基于MATLAB进行了语音信号频谱分析,同时本文提出了DSP上语音数据的获取和存储的实现方法,以及使用CCS软件进行优化代码的方法。最终实现了语音密码机中的MELP算法的改进和优化。
【Abstract】 In order to satisfy demands of the digital communication and other commercial applications,the speech compression technology has been developed rapidly. The realization of the voice cipher core technology is speech compression encoding &decoding technology, in this thesis chose MELP speech compress algorithm to realize.In speech coding algorithm, because MELP algorithm can better simulating natural speech characteristics and can synthesize a high quality speech in the low bite rate, thus became now low speed speech coding has great potential algorithm. It in the original LPC rules (linear forecast) code, and on the basis of using hybrid excitation, the periodic impulsive, the adaptive enhancement, pulse shaping filter and Fourier series amplitude value five other new techniques, making synthesized speech can better fitting natural voice. Thus better realize low bit rate speech coding.This paper deeply analyzed the MELP speech decoding algorithm of principle, to its key technology in the decoding process has summarized, and some of these formulas for the theoretical derivation, and on this basis to MELP algorithm in LSF quantification and puts forward the improvement and distortion comparison analysis. Experimental results show that the modified MELP in 2.4 kb/s on get better speech synthesis, better fitting natural voice synthesis.Finally, based on MATLAB for the speech signal spectrum analysis, the paper proposed DSP on speech data acquisition and storage, and the realization method of using CCS optimized method of software code. Finally achieved the voice cipher MELP algorithm is improved and optimized.