Abstract: The audio algorithm of Multi-Band-Excitation（MBE） is widely used in secret communication.Recently，secret-embedded audios based on MBE appear along with diversification of carrier.In order to ensure the safety of secret communication，the analysis and optimization of MBE algorithm are proposed.By the methods of optimizing the surd-sonant judge module，translating floating-point to fix-point，using iteration to realize evolution，inquiring table to realizing exceeding function and so on，experiments on the FPGA of SOC/Leon3 show that 45% of hardware resource is reduced and 29% of the real-time capability of algorithm is raised.These methods supplies a platform for steganalysis of MBE.

Key words: Multi-Band-Excitation（MBE）, iteration, fix-point, leon3, Field-Programmable Gate Array（FPGA）

摘要： 多带激励语音压缩（MBE）算法广泛应用于保密通信中，近来，随着语音压缩编解码技术的进步，信息隐藏载体的多样化发展，出现了以MBE为载体的密写文件。为了保障保密通信安全，提出了一种MBE算法的分析与优化技术。通过对算法的清浊音判决模块优化，浮点转定点，牛顿迭代法实现开平方，对超越函数查表实现等方法，并采用基于SOC/Leon3微处理器架构的现场可编程门阵列（FPGA）实现，减少了近45%硬件资源的消耗，并且运算实时性能提升了约29%，为针对MBE的隐写分析提供了一个很好的平台。

关键词: 多带激励语音压缩, 牛顿迭代法, 定点, Leon3, 现场可编程门阵列（FPGA）