Advertisement

Feature fusion methods research based on deep belief networks for speech emotion recognition under noise condition

  • Yongming HuangEmail author
  • Kexin Tian
  • Ao Wu
  • Guobao Zhang
Original Research

Abstract

The speech emotion recognition accuracy of prosody feature and voice quality feature declines with the decrease of signal to noise ratio (SNR) of speech signals. In this paper, we propose novel sub-band spectral centroid weighted wavelet packet Cepstral coefficients (W-WPCC) for robust speech emotion recognition. The W-WPCC feature is computed by combining the sub-band energies with sub-band spectral centroids via a weighting scheme to generate noise-robust acoustic features. And deep belief networks (DBNs) are artificial neural networks having more than one hidden layer, which are first pre-trained layer by layer and then fine-tuned using back propagation algorithm. The well-trained deep neural networks are capable of modeling complex and non-linear features of input training data and can better predict the probability distribution over classification labels. We extracted prosody feature, voice quality features and wavelet packet Cepstral coefficients (WPCC) from the speech signals to combine with W-WPCC and fused them by DBNs. Experimental results on Berlin emotional speech database show that the proposed fused feature with W-WPCC is more suitable in speech emotion recognition under noisy conditions than other acoustics features and proposed DBNs feature learning structure combined with W-WPCC improve emotion recognition performance over the conventional emotion recognition method.

Keywords

Speech emotion recognition Weighted wavelet packets Cepstral coefficients (W-WPCC) Feature fusion Deep belief networks (DBNs) 

Notes

Acknowledgements

An earlier version of this paper was presented at the International Conference on Network, Communication and Computing (ICNCC 2016).This work was supported by open Fund of Jiangsu Province Natural Science Foundation (No. BK20140649) and National Natural Science Foundation (No. 61503081, No. 61473079).

References

  1. Ali Hassan R, Damper, Niranjan M (2013) On acoustic emotion recognition: compensating for covariate shift. IEEE Trans Audio Speech Lang Process 21(7):1458–1468CrossRefGoogle Scholar
  2. Atal BS (1974) Effectiveness of linear prediction characteristics of the speech wave for automatic speaker identification and verification. J Acoust Soc Am 55(6):1304–1312CrossRefGoogle Scholar
  3. Bahreini K, Nadolski R, Westera W (2016) Towards multimodal emotion recognition in e-learning environments. Inter Learning Environ 24(3):590–605CrossRefGoogle Scholar
  4. Bengio Y (2009) Learning deep architectures for AI. Now Publ Inc 2(1):67–76MathSciNetzbMATHGoogle Scholar
  5. Bengio Y (2012) Deep learning of representations for unsupervised and transfer learning. J Mach Learning Res Proc Track 27(2), 17–36MathSciNetGoogle Scholar
  6. Brisson J, Martel K, Serres J, Sirois S, Adrien JL (2014) Acoustic analysis of oral productions of infants later diagnosed with autism and their mother. Inf Ment Health J 35(3):285–295CrossRefGoogle Scholar
  7. Burkhardt F, Paeschke A, Rolfes M, Sendlmeier W, Weiss B (2005) A database of german emotional speech. In: proceeding interspeech 2005, ISCA, pp 1517–1520Google Scholar
  8. Caponetti L, Buscicchio CA, Castellano G (2011) Biologically inspired emotion recognition from speech. Eurasip J Adv Signal Process 2011(1):1–10CrossRefGoogle Scholar
  9. Chang CC, Lin CJ (2011) LIBSVM: a library for support vector machines. ACM Trans Intell Syst Technol (TIST) 2(3), pp 1–27CrossRefGoogle Scholar
  10. Crumpton J, Bethel CL (2015) A survey of using vocal prosody to convey emotion in robot speech. Int J Social Robot 8(2):271–285CrossRefGoogle Scholar
  11. Deng J, Xia R, Zhang Z, Liu Y (2014) Introducing shared-hidden-layer autoencoders for transfer learning and their application in acoustic emotion recognition. Icassp IEEE international conference on acoustics, pp 4818–4822Google Scholar
  12. Farooq O, Datta S (2001) Mel filter-like admissible wavelet packet structure for speech recognition. Signal Process Lett IEEE 8(7):196–198CrossRefGoogle Scholar
  13. Fastl H, Zwicer E (1999) Psychoacoustics Facts and Models[M], 2nd edn. Springer, New YorkGoogle Scholar
  14. Feng Z, Zheng WX (2015) On extended dissipativity of discrete-time neural networks with time delay. IEEE Trans Neural Netw Learning Syst 26(12):3293–3300MathSciNetCrossRefGoogle Scholar
  15. France DJ, Shiavi RG, Silverman S et al (2000) Acoustical properties of speech as indicators of depression and suicidal risk. IEEE Trans Biomed Eng 47(7):829–837CrossRefGoogle Scholar
  16. Guzman M, Correa S, Munoz D et al (2013) Influence on spectral energy distribution of emotional expression. J Voice 27(1):129.e1–129.e10CrossRefGoogle Scholar
  17. Hinton GE, Salakhutdinov RR (2006) Reducing the dimensionality of data with neural networks. Science 313(5786):504–507MathSciNetCrossRefzbMATHGoogle Scholar
  18. Idris I, Salam MS (2015) Voice quality features for speech emotion recognition. J Info Assur Secur 10(4):183–191Google Scholar
  19. Iliev AI, Scordilis MS (2011) Spoken emotion recognition using glottal symmetry. Eurasip J Adv Sig Process 2011(1):1–11CrossRefGoogle Scholar
  20. Kandali AB, Routray A, Basu TK (2009) Vocal emotion recognition in five native languages of Assam using new wavelet features. Int J Speech Technol 12(1):1–13CrossRefGoogle Scholar
  21. Karmakar A, Kumar A, Patney RK (2007) Design of optimal wavelet packet trees based on auditory perception criterion. Ieee Signal Process Lett 14(4):240–243CrossRefGoogle Scholar
  22. Lee CM, Narayanan SS (2005) Toward detecting emotions in spoken dialogs. IEEE Trans Speech Audio Process 13(2):293–303CrossRefGoogle Scholar
  23. Lee TH, Park MJ, Park JH, Kwon OM, Lee SM (2014) Extended dissipative analysis for neural networks with time-varying delays. IEEE Trans Neural Netw Learning Syst 25(10):1936–1941CrossRefGoogle Scholar
  24. Mallat SA (2009) Wavelet tour of signal processing, 3rd edn. Academic Press, BurlingtonzbMATHGoogle Scholar
  25. Malta L, Miyajima C, Kitaoka N et al. (2009) Multimodal estimation of a driver’s spontaneous irritation. Intelligent vehicles symposium, 2009 IEEE, pp 573–577Google Scholar
  26. Mingyu Y, Chun C, Jiajun B et al. (2006) Emotion recognition from noisy speech. In: multimedia and expo, IEEE international conference on 2006, pp 1653–1656Google Scholar
  27. Morrison D, Wang RL, De Silva LC (2007) Ensemble methods for spoken emotion recognition in call-centres. Speech Commun 49(2):98–112CrossRefGoogle Scholar
  28. Paliwal KK (1998) Spectral subband centroid features for speech recognition. Acoustics, speech and processings. Proceedings of the IEEE international conference on 1998, pp 617–620Google Scholar
  29. Petrushin V (2000) Emotion recognition in speech signal experimental study, development, and application. ICSLP 2000, Beijing, pp 222–225Google Scholar
  30. Sarikaya R, Gowdy JN (1997) Wavelet based analysis of speech under stress[C]. Southeastcon ‘97. engineering new century., proceedings IEEE, pp 92–96Google Scholar
  31. Schuller B, Rigoll G, Lang M (2004) Speech emotion recognition combining acoustic features and linguistic information in a hybrid support vector machine-belief network architecture[C]. Acoustics, speech, and signal processing, proceedings (ICASSP ‘04). IEEE international conference on 2004, pp I-577–580Google Scholar
  32. Shah M, Chakrabarti C, Spanias A (2015) Within and cross-corpus speech emotion recognition using latent topic model-based features. Eurasip J Audio Speech Music Process 2015(1):1–17CrossRefGoogle Scholar
  33. Shamiand M, Verhelst W (2007) Anevaluation of the robustness of existing supervised machine learning approaches to the classification of emotions in speech. Speech Commun 49(3):201–212CrossRefGoogle Scholar
  34. Stephane M (2009) A wavelet tour of signal processing, 3rd edn. Academic Press, BurlingtonzbMATHGoogle Scholar
  35. Tahon M, Devillers L (2016) Towards a small set of robust acoustic features for emotion recognition: challenges. IEEE ACM Trans Audio Speech Lang Process 24(1):16–28CrossRefGoogle Scholar
  36. Tahon M, Sehili MA, Devillers L (2015) Cross-corpus experiments on laughter and emotion detection in HRI with elderly people. In: International Conference on Social Robotics, vol 31. Springer, pp 633–642Google Scholar
  37. Vlasenko B, Schuller B, Wendemuth A et al (2007) Frame vs. turn-level: emotion recognition from speech considering static and dynamic processing[C]. Affect Comp Intell Interact Proc 781:139–147CrossRefGoogle Scholar
  38. Wang X, He Q (2004) Enhancing generalization capability of svm classifiers with feature weight adjustment. International Conference on Knowledge-Based and Intelligent Information and Engineering Systems, vol 3213. Springer, Heidelberg, pp 1037–1043Google Scholar
  39. Yongming H, Ao W, Guobao Z, Yue L (2014a) Speech emotion recognition based on coiflet wavelet packet Cepstral coefficients. Chinese conference on pattern recognition, pp 436–443Google Scholar
  40. Yongming H, Guobao Z, Yue L, Ao W (2014b) Improved emotion recognition with novel task-oriented wavelet packet features, vol 8588. In: 10th international conference, ICIC 2014, Taiyuan, China, August 3–6, pp 706–714Google Scholar
  41. Zeng ZH, Tu JL, Pianfetti BM et al (2008) Audio-visual affective expression recognition through multistream fused HMM[J]. IEEE Trans Multimed 10(4):570–577CrossRefGoogle Scholar
  42. Zeng ZH, Pantic M, Roisman GI et al (2009) A survey of affect recognition methods: audio, visual, and spontaneous expressions. IEEE Trans Pattern Anal Mach Intell 31(1):39–58CrossRefGoogle Scholar
  43. Zhang WS, Zhao DH, Chai Z, Yang LT, Liu X, Gong FM, Yang S (2017) Deep learning and SVM-based emotion recognition from Chinese speech for smart affective services. Softw Pract Exp 47(8):1127–1138Google Scholar
  44. Zhou GJ, Hansen JHL, Kaiser JF (2001) Nonlinear feature based classification of speech under stress. IEEE Trans Speech Audio Process 9(3):201–216CrossRefGoogle Scholar
  45. Zhu LZ, Chen LM, Zhao DH, Zhou JH, Zhang WS (2017) Emotion recognition from chinese speech for smart affective services using a combination of SVM and DBN. Sensors 17(7):1694CrossRefGoogle Scholar

Copyright information

© Springer-Verlag GmbH Germany, part of Springer Nature 2017

Authors and Affiliations

  • Yongming Huang
    • 1
    • 2
    Email author
  • Kexin Tian
    • 1
    • 2
  • Ao Wu
    • 1
    • 2
  • Guobao Zhang
    • 1
    • 2
  1. 1.Laboratory of Measurement and Control of Complex Systems of EngineeringSoutheast UniversityNanjingChina
  2. 2.Ministry of Education, School of AutomationSoutheast UniversityNanjingChina

Personalised recommendations