说话人识别系统中VQ判决模块的FPGA实现  

FPGA implementation of VQ-based decision module in speaker recognition system

作　　者：张树才[1] 张国杰[1] 王晓雷[1] 高谦[1] 

机构地区：[1]国家数字交换系统工程技术研究中心,河南郑州450002

出　　处：《电子技术应用》2008年第7期41-44,共4页Application of Electronic Technique

摘　　要：目前,基于PC或DSP的系统其处理能力无法满足海量语音信号高速处理需求的增长。本文分析了VQ(矢量量化)搜索算法的硬件实现复杂度,针对说话人识别过程中运算量最大、耗时最长的判决过程,提出了一种基于标签的说话人判决模型实现方案。该设计用FPGA实现,可对多路电话信道说话人进行实时判决识别。On the basis of analyzing the computational complexity of vector quantization ,a tag-based scheme of speaker recognition is presented aiming at the ever increasing demand of high-speed processing for massive voice signals which exceed the ability of systems based on PC or DSP. The design is implemented with FPGA which can carry on real-time recognition of the speakers on multiple telephone channel.

关 键 词：说话人识别 VQ 标签 多路 FPGA 

分 类 号：TN912.34[电子电信—通信与信息系统] TP312[电子电信—信息与通信工程]