基于FPGA的高性能MFCC特征参数提取  

High-Performance MFCC Features Extraction Based on FPGA

作　　者：高谦[1] 张国杰[1] 张树才[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：《通信技术》2008年第6期153-154,157,共3页Communications Technology

摘　　要：为了实现高速语音特征参数的提取,在分析了美尔频率倒谱特征参数提取算法的基础上,提出了算法的硬件设计方案,介绍了各模块的设计原理。该方案增加了语音激活检测功能,可对语音信号中的噪音帧进行检测,提高了特征参数的可靠性。最后将设计实现于Stratix II系列FPGA上,仿真结果表明设计可以实现高速、高精度的MFCC特征参数提取。In order to realize extraction of the high-speed speech features, the hardware design for algorithm is proposed and the design principle of each module is described based on the analysis of Mel frequency cepstral features extraction algorithm. The voice activiation detection function is added in hardware implementation, and thus the noisy frame in voice signal can be detected and the characteristic parameters reliability is improved. Finally, the design is implemented successfully on to the Stratix II FPGA, and the simulation results indicate that the design can realize high-speed and high-precision extraction of MFCC characteristic parameters.

关 键 词：美尔频率倒谱 特征提取 FPGA 

分 类 号：TP391.42[自动化与计算机技术—计算机应用技术]