首页> 外文期刊>Journal of Discrete Mathematical Sciences and Cryptography >Performance evaluation of speaker recognition system
【24h】

Performance evaluation of speaker recognition system

机译:扬声器识别系统的性能评估

获取原文
获取原文并翻译 | 示例
           

摘要

Significance of accuracy in Speaker Recognition System (SRS) is still a challenging area of research. In this paper, the accuracy of a SRS based on Mel Frequency Cepstral Coefficient (MFCC) using Correlation is analysed on a limited amount of speech corpus. The experiments have been performed on 2 Bilingual Speech Corpus. A set of two different experiments has been carried out to study the impact of recording text, recording languages, duration of speech sample and recording devices on the accuracy of SRS. On the basis of the performed experiments, it has been concluded that recognition rate degrades due to acoustic mismatch in training and testing phase. The SRS achieved accuracy of 96.67% for text-dependent and device dependent while it degrades by 30% for Language independent and device independent.
机译:扬声器识别系统(SRS)中准确性的意义仍然是一个具有挑战性的研究领域。 在本文中,在有限量的语音语料库上分析了基于MEL频率谱系码(MFCC)的SRS的精度。 该实验已经在2种双语语料库中进行。 已经进行了一组两个不同的实验,以研究录制文本,记录语言,语音样本持续时间和记录设备对SRS精度的影响。 在进行所进行的实验的基础上,已经得出结论,由于训练和测试阶段的声学不匹配,识别率降低。 对于文本相关的和设备,SRS的准确性为96.67%,而无论是无关的语言和设备无关的依赖性依赖于依赖于设备。

著录项

相似文献

  • 外文文献
  • 中文文献
  • 专利
获取原文

客服邮箱:kefu@zhangqiaokeyan.com

京公网安备:11010802029741号 ICP备案号:京ICP备15016152号-6 六维联合信息科技 (北京) 有限公司©版权所有
  • 客服微信

  • 服务号