说话人识别系统的FPGA硬件电路实现  

FGPA Implementation for Speaker Identification

作　　者：范羽 陈金塔 FAN Yu;CHEN Chinta(College of Electronics and Electrical Engineering,Zhaoqing University,Zhaoqing,Guangdong 526061,China)

机构地区：[1]肇庆学院电子与电气工程学院,广东肇庆526061

出　　处：《肇庆学院学报》2021年第5期22-27,共6页Journal of Zhaoqing University

摘　　要：针对当前基于DSP等嵌入式说话人识别系统存在训练和识别时间较长的问题,本文提出了一种以Cyclone IVFPGA为硬件平台的说话人识别系统设计方法.该方法用快速傅里叶变换方法从一组普通话说话人中提取语音特征,取长时平均频谱,通过测试语音与训练语音之间的欧氏距离的最小值来识别说话人.FPGA硬件实现平台的实验结果验证了仿真的效率和准确性,并且在保证识别率的前提下,给出了一种能实时进行说话人识别的硬件系统设计思路.In view of the problem of long training and recognition time existing in embedded speaker identifi-cation based on DSP,this paper proposes a speaker identification system design method with Cyclone IV FPGA as the hardware platform.This method uses Fast Fourier Transform to extract speech features from a group of mandarin speakers and takes the long time average spectrum.The speaker is identified by testing the minimum Euclide distance between the speech and the trained speech.The efficiency and accuracy of simulation are veri-fied by numerical results,and under the premise of ensuring recognition rate,the FPGA hardware circuit imple-mented in this design can effectively improve the speed of training and recognition.

关 键 词：说话人识别 FFT FPGA 

分 类 号：G642.0[文化科学—高等教育学]