基于多特征I-Vector的说话人识别算法

兰州理工大学学报 ›› 2021, Vol. 47 ›› Issue (5): 93-98.

• 自动化技术与计算机技术 • 上一篇下一篇

基于多特征I-Vector的说话人识别算法

赵宏^*, 岳鲁鹏, 常兆斌, 王伟杰

兰州理工大学计算机与通信学院, 甘肃兰州 730050

收稿日期:2019-12-19 出版日期:2021-10-28 发布日期:2021-11-18
作者简介:赵宏(1971-),男,甘肃西和人,博士,教授,博导.Email:594286500@qq.com
基金资助:
国家自然科学基金(51668043),赛尔网络下一代互联网技术创新项目(NGII20160311, NGII20160112)

Speaker recognition algorithm based on multi-featured I-Vector

ZHAO Hong, YUE Lu-peng, CHANG Zhao-bin, WANG Wei-jie

College of Computer and Communication, Lanzhou Univ. of Tech., Lanzhou 730050, China

Received:2019-12-19 Online:2021-10-28 Published:2021-11-18

摘要/Abstract

摘要： 针对单一声学特征无法精准高效地辨识说话人身份的问题,提出了一种基于多特征I-Vector的说话人识别算法.该算法首先采集不同的声学特征并将其构成一个高维特征向量,然后通过主成分分析法有效地剔除高维特征向量的关联,确保各种特征之间正交化,最后采用概率线性判别分析进行建模和打分,并在一定程度上降低空间维度.在TIMIT语料库上利用Kaldi进行实验,算法运行结果表明,该算法较当前流行的基于I-Vector的单一梅尔频率倒谱系数和感知线性预测系数的特征系统在等错误率上分别提高了8.18%和1.71%,在模型训练时间上分别减少了60.4%和47.5%,具有更好的识别效果和效率.

关键词: 说话人识别算法, 多特征I-Vector, 主成分分析, 概率线性判别分析, Kaldi

Abstract: Aiming at the problem of inaccurate and inefficient speaker recognition presented by single acoustic feature, a speaker recognition algorithm was proposed based on multi-featured I-Vector. Firstly,different acoustic feature vectors were extracted and combined into a high-dimensional feature vector.Then principal components analysis (PCA) was used to effectively remove the correlation of these feature vectors, so that the features became orthogonalized. Finally, probabilistic linear discriminant analysis(PLDA) was used for modeling and scoring, which led to reduce the spatial dimension to a certain degree.Experiments were carried out on TIMIT corpus in combination with Kaldi speech recognition toolkit, and the results compared with the single-featured systems including Mel-frequency cepstral coefficients (MFCC) and perceptual linear predictive (PLP) coefficients based on I-Vector, the equal error rate (EER) of the purposed algorithm were increased by 8.18%and 1.71%, respectively;the model training time were decreased respectively by 60.4% and 47.5%,respectively.Therefore, the purposed algorithm has betterspeaker recognition performance and efficiency.

Key words: speaker recognition algorithm, multi-featured I-Vector, principal components analysis, probabilistic linear discriminant analysis, Kaldi speech recognition toolkit

中图分类号:

TN391

赵宏, 岳鲁鹏, 常兆斌, 王伟杰. 基于多特征I-Vector的说话人识别算法[J]. 兰州理工大学学报, 2021, 47(5): 93-98.

ZHAO Hong, YUE Lu-peng, CHANG Zhao-bin, WANG Wei-jie. Speaker recognition algorithm based on multi-featured I-Vector[J]. Journal of Lanzhou University of Technology, 2021, 47(5): 93-98.

参考文献

[1] 仲伟峰,方祥,范存航,等.深浅层特征及模型融合的说话人识别 [J].声学学报,2018,43(2):263-272.
[2] 艾斯卡尔·肉孜,王东,李蓝天,等.说话人识别中的分数域语速归一化 [J].清华大学学报(自然科学版),2018,58(4):337-341.
[3] SUN E L,GU T,XIE K,et al.Text-independent speaker identification based on deep Gaussian correlation supervector [J].International Journal of Speech Technology,2019,22(2):449-457.
[4] AGRAWAL P,GANAPATHY S.Unsupervised modulation filter learning for noise-robust speech recognition [J].The Journal of the Acoustical Society of America,2017,142(3):1686-1692.
[5] 李作强,高勇.基于CFCC和相位信息的鲁棒性说话人辨识 [J].计算机工程与应用,2015,51(17):228-232.
[6] 张二华,王明合,唐振民.加性噪声条件下鲁棒说话人确认 [J].电子学报,2019,47(6):1244-1250.
[7] REYNOLDS D A,QUATIERI T F,DUNN R B.Speaker verification using adapted Gaussian mixture models [J].Digital Signal Processing,2000,10(1/2/3):19-41.
[8] HAN L,LI T,ZHENG W,et al.Recognition of voiceprint using deep neural network combined with supportvector machine [C]//International Conference on Intelligent and Interactive Systems and Applications.Bangkok Thailand:Springer,2019:3-11.
[9] DEHAK N,KENNY P J,DEHAK R,et al.Front-end factor analysis for speaker verification [J].IEEE Transactions on Audio Speech and Language Processing,2010,19(4):788-798.
[10] 栗志意,张卫强,何亮,等.基于总体变化子空间自适应的i-vector说话人识别系统研究 [J].自动化学报,2014,40(8):1836-1840.
[11] KANAGASUNDARAM A,DEAN D,VOGT R,et al.Weighted LDA techniques for i-vector based speaker verification [C]//IEEE International Conference on Acoustics,Speech and Signal Processing.Kyoto,Japan:IEEE,2012:4781-4784.
[12] LUO D,XIA L,ZHANG C,et al.Automatic pronunciation evaluation in high-states English speaking tests based on deep neural network models [C]//International Conference on Artificial Intelligence and Big Data.Chengdu:IEEE,2019:124-128.
[13] TANG H,LEI Z,HUANG Z,et al.The GMM and I-Vector systems based on spoofing algorithms for speakerspoofing detection [C]//Chinese Conference on Biometric Recognition.Zhuzhou:Springer,2019:502-510.
[14] MACHLICA L,ZAJIC Z.An efficient implementation of probabilistic linear discriminant analysis [C]//IEEE International Conference on Acoustics,Speech and Signal Processing.Vancouver,Canada:IEEE,2013:7678-7682.
[15] POVEY D,GHOSHAL A,BOULIANNE G,et al.The Kaldi speech recognition toolkit [C]//IEEE 2011 workshop on automatic speech recognition and understanding.Hilton Waikoloa Village,Big Island,Hawaii,US:IEEE,2011:1-4.
[16] 吴文昭.基于多约简α-GMM和SVM的说话人确认 [J].兰州理工大学学报,2015,41(4):113-116.
[17] 周萍,沈昊,郑凯鹏.基于MFCC与GFCC混合特征参数的说话人识别 [J].应用科学学报,2019,37(1):24-32.
[18] DHANUSH B K,SUPARNA S,AARTHY R,et al.Factor analysis methods for joint speaker verification and spoof detection [C]//IEEE International Conference on Acoustics,Speech and Signal Processing.New Orleans,LA,USA:IEEE,2017:5385-5389.
[19] 王昕,张洪冉.基于DNN处理的鲁棒性I-Vector说话人识别算法 [J].计算机工程与应用,2018,54(22):167-172.
[20] ARCIA-ROMERO D,ESPY-WILSON C Y.Analysis of i-vector length normalization in speaker recognition systems [C]//Twelfth Annual Conference of the International Speech Communication Association.Florence,Italy:[s.n.],2011:249-252.
[21] PRINCE S J D,ELDER J H.Probabilistic linear discriminant analysis for inferences about identity [C]//IEEE 11th International Conference on Computer Vision.Florence,Italy:IEEE,2007:1-8.
[22] NIST speaker recognition evaluation [EB/OL].[2019-09-05].http://www.itl.nist.gov/iad/mig/tests/spk/2010/index.html.

基于多特征I-Vector的说话人识别算法

Speaker recognition algorithm based on multi-featured I-Vector

PDF

可视化

摘要/Abstract

引用本文

使用本文

参考文献

相关文章 1

编辑推荐

Metrics

本文评价