Abstract: Extracting discriminative speaker-specific representations from speech signals and transforming them into fixed length vectors are key steps in speaker identification and verification systems. In this study, we propose a latent discriminative representation learning method for speaker recognition. We mean that the learned representations in this study are not only discriminative but also relevant. Specifically, we introduce an additional speaker embedded lookup table to explore the relevance between different utterances from the same speaker. Moreover, a reconstruction constraint intended to learn a linear mapping matrix is introduced to make representation discriminative. Experimental results demonstrate that the proposed method outperforms state-of-the-art methods based on the Apollo dataset used in the Fearless Steps Challenge in INTERSPEECH2019 and the TIMIT dataset.

用于说话人识别的潜在可区分性表征学习

[1]Al-Kaltakchi MTS, Woo WL, Dlay SS, et al., 2016. Study of statistical robust closed set speaker identification with feature and score-based fusion. IEEE Statistical Signal Processing Workshop, p.1-5.

[2]Al-Kaltakchi MTS, Woo WL, Dlay SS, et al., 2017. Speaker identification evaluation based on the speech biometric and i-vector model using the TIMIT and NTIMIT databases. Proc 5^th Int Workshop on Biometrics and Forensics, p.1-6.

[3]Chen NX, Qian YM, Yu K, 2015. Multi-task learning for text-dependent speaker verification. Proc 16^th Annual Conf of the Int Speech Communication Association, p.185-189.

[4]Chen XB, Cai YF, Chen L, et al., 2015. Discriminant feature extraction for image recognition using complete robust maximum margin criterion. Mach Vis Appl, 26(7-8):857-870.

[5]Cumani S, Plchot O, Laface P, 2013. Probabilistic linear discriminant analysis of i-vector posterior distributions. IEEE Int Conf on Acoustics, Speech and Signal Processing, p.7644-7648.

[6]Davis S, Mermelstein P, 1980. Comparison of parametric representations for monosyllabic word recognition in continuously spoken sentences. IEEE Trans Acoust Speech Signal Process, 28(4):357-366.

[7]Dehak N, Kenny PJ, Dehak R, et al., 2011. Front-end factor analysis for speaker verification. IEEE Trans Audio Speech Lang Process, 19(4):788-798.

[8]Desai D, Joshi M, 2013. Speaker recognition using MFCC and hybrid model of VQ and GMM. Proc 2^nd Int Symp on Intelligent Informatics, p.53-63.

[9]Dey S, Motlicek P, Madikeri S, et al., 2017. Template-matching for text-dependent speaker verification. Speech Commun, 88:96-105.

[10]Fisusi A, Yesufu T, 2007. Speaker recognition systems: a tutorial. Afr J Inform Commun Technol, 3(2):42-52.

[11]Garofolo JS, Lamel LF, Fisher WM, et al., 1993. DARPA TIMIT acoustic-phonetic continous speech corpus CD-ROM. NIST Speech Disc 1-1.1. NASA STI/Recon Technical Report N, 93:27403, NASA, USA.

[12]Hansen JHL, Sangwan A, Joglekar A, et al., 2018. Fearless steps: Apollo-11 corpus advancements for speech technologies from Earth to the Moon. Proc 19^th Annual Conf of the Int Speech Communication Association, p.2758-2762.

[13]Heigold G, Moreno I, Bengio S, et al., 2016. End-to-end text-dependent speaker verification. IEEE Int Conf on Acoustics, Speech and Signal Processing, p.5115-5119.

[14]Hermansky H, 1990. Perceptual linear predictive (PLP) analysis of speech. J Acoust Soc Am, 87(4):1738-1752.

[15]Huang XD, Acero A, Hon HW, 2001. Spoken Language Processing: a Guide to Theory, Algorithm and System Development. Upper Saddle River, Prentice Hall PTR, USA.

[16]Jiang HJ, Wang RP, Shan SG, et al., 2017. Learning discriminative latent attributes for zero-shot classification. IEEE Int Conf on Computer Vision, p.4233-4242.

[17]Kenny P, Boulianne G, Ouellet P, et al., 2007. Speaker and session variability in GMM-based speaker verification. IEEE Trans Audio Speech Lang Process, 15(4):1448-1460.

[18]Kim MJ, Yang IH, Kim MS, et al., 2017. Histogram equalization using a reduced feature set of background speakers’ utterances for speaker recognition. Front Inform Technol Electron Eng, 18(5):738-750.

[19]Kumar R, Yeruva V, Ganapathy S, 2018. On convolutional LSTM modeling for joint wake-word detection and text dependent speaker verification. Proc 19^th Annual Conf of the Int Speech Communication Association, p.1121-1125.

[20]Lei Y, Scheffer N, Ferrer L, et al., 2014. A novel scheme for speaker recognition using a phonetically-aware deep neural network. IEEE Int Conf on Acoustics, Speech and Signal Processing, p.1695-1699.

[21]Li C, Ma XK, Jiang B, et al., 2017. Deep speaker: an end-to-end neural speaker embedding system. https://arxiv.org/abs/1705.02304

[22]Luo Y, Liu Y, Zhang Y, et al., 2018. Speech bottleneck feature extraction method based on overlapping group lasso sparse deep neural network. Speech Commun, 99:56-61.

[23]Mao QR, Dong M, Huang ZW, et al., 2014. Learning salient features for speech emotion recognition using convolutional neural networks. IEEE Trans Multim, 16(8):2203-2213.

[24]Peri R, Pal M, Jati A, et al., 2019. Robust speaker recognition using unsupervised adversarial invariance. https://arxiv.org/abs/1911.00940

[25]Rabiner LR, 1989. A tutorial on hidden Markov models and selected applications in speech recognition. Proc IEEE, 77(2):257-286.

[26]Reynolds DA, Rose RC, 1995. Robust text-independent speaker identification using Gaussian mixture speaker models. IEEE Trans Speech Audio Process, 3(1):72-83.

[27]Reynolds DA, Quatieri TF, Dunn RB, 2000. Speaker verification using adapted Gaussian mixture models. Dig Signal Process, 10(1-3):19-41.

[28]Sadjadi SO, Slaney M, Heck L, et al., 2013. MSR Identity Toolbox v1.0: a MATLAB Toolbox for Speaker Recognition Research. Microsoft Research Technical Report, Piscataway, NJ, USA.

[29]Schroff F, Kalenichenko D, Philbin J, 2015. FaceNet: a unified embedding for face recognition and clustering. IEEE Conf on Computer Vision and Pattern Recognition, p.815-823.

[30]Singh S, Rajan EG, 2011. Vector quantization approach for speaker recognition using MFCC and inverted MFCC. Int J Comput Appl, 17(1):1-7.

[31]Snyder D, Ghahremani P, Povey D, et al., 2016. Deep neural network-based speaker embeddings for end-to-end speaker verification. IEEE Spoken Language Technology Workshop, p.165-170.

[32]Snyder D, Garcia-Romero D, Povey D, et al., 2017. Deep neural network embeddings for text-independent speaker verification. Proc 18^th Annual Conf of the Int Speech Communication Association, p.999-1003.

[33]Snyder D, Garcia-Romero D, Sell G, et al., 2018. X-vectors: robust DNN embeddings for speaker recognition. IEEE Int Conf on Acoustics, Speech and Signal Processing, p.5329-5333.

[34]Togneri R, Pullella D, 2011. An overview of speaker identification: accuracy and robustness issues. IEEE Circ Syst Mag, 11(2):23-61.

[35]van Leeuwen DA, Saeidi R, 2013. Knowing the non-target speakers: the effect of the i-vector population for PLDA training in speaker recognition. IEEE Int Conf on Acoustics, Speech and Signal Processing, p.6778-6782.

[36]Variani E, Lei X, McDermott E, et al., 2014. Deep neural networks for small footprint text-dependent speaker verification. IEEE Int Conf on Acoustics, Speech and Signal Processing, p.4052-4056.

[37]Wan V, Campbell WM, 2000. Support vector machines for speaker verification and identification. Neural Networks for Signal Processing X. Proc IEEE Signal Processing Society Workshop, p.775-784.

[38]Wen YD, Zhang KP, Li ZF, et al., 2016. A discriminative feature learning approach for deep face recognition. Proc 14^th European Conf on Computer Vision, p.499-515.

[39]Yadav S, Rai A, 2018. Learning discriminative features for speaker identification and verification. Proc 19^th Annual Conf of the Int Speech Communication Association, p.2237-2241.

[40]Yoshimura T, Koike N, Hashimoto K, et al., 2018. Discriminative feature extraction based on sequential variational autoencoder for speaker recognition. Asia-Pacific Signal and Information Processing Association Annual Summit and Conf, p.1742-1746.

[41]Young S, 1993. The HTK Hidden Markov Model Toolkit: Design and Philosophy. Department of Engineering, Cambridge University, Cambridge.

[42]Yu K, Mason J, Oglesby J, 1995. Speaker recognition using hidden Markov models, dynamic time warping and vector quantisation. IEE Proc Vis Image Signal Process, 142(5):313-318.

[43]Zhang C, Koishida K, 2017. End-to-end text-independent speaker verification with triplet loss on short utterances. Proc 18^th Annual Conf of the Int Speech Communication Association, p.1487-1491.

[44]Zhang FF, Zhang TZ, Mao QR, et al., 2018. Joint pose and expression modeling for facial expression recognition. IEEE Conf on Computer Vision and Pattern Recognition, p.3359-3368.