(甚)低速率语音编码关键问题研究

项目名称： (甚)低速率语音编码关键问题研究

项目编号： No.61271248

项目类型： 面上项目

立项/批准年度： 2013

项目学科： 无线电电子学、电信技术

项目作者： 赵晓群

作者单位： 同济大学

项目金额： 60万元

中文摘要： 本项目针对频带资源有限的水声通信，研究(甚)低速率水声语音编码中的关键问题，主要包括：1)声道参数：利用自适应帧内、帧间联合预测编码的方法降低声道参数的编码比特，或将节省下来的比特数用于提升编码器的鲁棒性，使其能抵抗水声信道的强干扰环境；2)声门激励：利用线性预测分析方法和希尔伯特-黄等非线性信号分析方法对不同帧模式下的声门激励信号进行分析和重构，尽可能地逼近语音短时平稳模型的理想情况，提高编码器的合成语音自然度。本项目所设计的实验将在我们已经构建的"浅海水声信道综合实验软件平台"的基础上开展研究，并最终通过FPGA硬件平台对算法进行实现和验证。通过对上述问题的研究，预期可以提升(甚)低速率语音编码方案的鲁棒性和合成语音的自然度，促进其在水声通信中的应用。另外，依托本项目进一步加强国际合作与交流，同时为我国在语音编码及水声通信领域的人才培养做出贡献。

中文关键词： 低速率语音编码；语音特征分析；压缩编码；语音建模；水声通信

英文摘要： This project studies key problems in (very)low rate speech coding for bandwidth-limited underwater acoustic communication, which include: 1)vocal tract parameters: adaptive intra-frame & inter-frame joint prediction coding can reduce the coding bits of the vocal tract parameters. The saved bits can be used to improve the robustness of the encoder, so the system can resist the strong turbulent environment of underwater acoustic communication channel; 2)glottal excitation: the linear predictive analysis method and Hilbert-Huang nonlinear signal analysis method are used to do analysis and reconstruction of the glottal excitation signal in a different frame mode to approach the ideal situation of the short-time stationary model speech, so the naturalness of synthetic speech by the encoder can be improved. The proposed methods can be implemented and tested on "shallow water acoustic channel experimental software platform" which we have built, and ultimately the algorithm is performed by FPGA hardware platform. Based on the above mentioned problems, it can be expected to enhance the robustness of the (very) low bit rate speech coding schemes and the naturalness of the synthesized speech, and promote its application in underwater acoustic communication. It is supposed to strengthen international cooperation and communi

英文关键词： Low bit rate speech coding；Speech characteristics analysis；Compress coding；Speech modeling；Underwater acoustic communication