作　　者： ; ;

机构地区： 罗定职业技术学院

出　　处： 《广东交通职业技术学院学报》 2014年第2期42-46,共5页

摘　　要： 目前基于GMM的说话人辨认系统主要在微机上由软件实现,难以胜任大语音流的多路实时处理任务。鉴于FPGA强大的流水和并行处理能力,提出了一种以FPGA为应用平台,基于GMM的与文本无关说话人实时辨认系统的硬件实现方法。抽取NIST2003语音库的语料进行试验,结果表明,与在PC上的软件实现相比,识别率几无差别,但实时处理速度提高了约90倍。 Nowadays GMM-based speaker identification system is mainly realized by software on PC. It is difficult to deal with large number of voice streams in real time. In view of powerful pipelining and parallesim of FPGA, the article describes a method for realizing a GMM-based text-independent speaker identification system, which based on FPGA as the application platform. By drawing the voice material from NIST2003 database, an experiment is conducted to compare the effect of GMM-based speaker identification system with the one PC software implementation.The result shows that there is no significant difference in terms of identification accuracy, but the identification speed increased by 90 times.

关 键 词： 高斯混合模型 说话人辨认 文本无关 识别率

领　　域： [自动化与计算机技术] [自动化与计算机技术]