Advertisement

Big Data, Deep Learning – At the Edge of X-Ray Speaker Analysis

  • Björn W. SchullerEmail author
Conference paper
Part of the Lecture Notes in Computer Science book series (LNCS, volume 10458)

Abstract

With two years, one has roughly heard a thousand hours of speech – with ten years, around ten thousand. Similarly, an automatic speech recogniser’s data hunger these days is often fed in these dimensions. In stark contrast, however, only few databases to train a speaker analysis system contain more than ten hours of speech. Yet, these systems are ideally expected to recognise the states and traits of speakers independent of the person, spoken content, language, cultural background, and acoustic disturbances at human parity or even super-human levels. While this is not reached at the time for many tasks such as speaker emotion recognition, deep learning – often described to lead to ‘dramatic improvements’ – in combination with sufficient learning data satisfying the ‘deep data cravings’ holds the promise to get us there. Luckily, every second, more than five hours of video are uploaded to the web and several hundreds of hours of audio and video communication in most languages of the world take place. If only a fraction of these data would be shared and labelled reliably, ‘x-ray’-alike automatic speaker analysis could be around the corner for next gen human-computer interaction, mobile health applications, and many further benefits to society. In this light, first, a solution towards utmost efficient exploitation of the ‘big’ (unlabelled) data available is presented. Small-world modelling in combination with unsupervised learning help to rapidly identify potential target data of interest. Then, gamified dynamic cooperative crowdsourcing turn its labelling into an entertaining experience, while reducing the amount of required labels to a minimum by learning alongside the target task also the labellers’ behaviour and reliability. Further, increasingly autonomous deep holistic end-to-end learning solutions are presented for the task at hand. Benchmarks are given from the nine research challenges co-organised by the author over the years at the annual Interspeech conference since 2009. The concluding discussion will contain some crystal ball gazing alongside practical hints not missing out on ethical aspects.

Keywords

Computational paralinguistics Automatic speaker analysis Big data Deep learning 

Notes

Acknowledgment

The author acknowledges funding from the European Research Council within the European Union’s 7th Framework Programme under grant agreement no. 338164 (Starting Grant Intelligent systems’ Holistic Evolving Analysis of Real-life Universal speaker characteristics (iHEARu)) and the European Union’s Horizon 2020 Framework Programme under grant agreement no. 645378 (Research Innovation Action Artificial Retrieval of Information Assistants - Virtual Agents with Linguistic Understanding, Social skills, and Personalised Aspects (ARIA-VALUSPA)). The responsobility lies with the author. The author would further like to thank his team colleague Anton Batliner at University of Passau/Germany as well as Stefan Steidl at FAU Erlangen/Germany and all other co-organisers and participants over the years for running the Interspeech Computational Paralinguistics related challenge events and turning them into a meaningful benchmark.

References

  1. 1.
    Adda, G., Besacier, L., Couillault, A., Fort, K., Mariani, J., De Mazancourt, H.: “Where the data are coming from?" ethics, crowdsourcing and traceability for big data in human language technology. In: Proceedings Crowdsourcing and Human Computation Multidisciplinary Workshop, Paris, France (2014)Google Scholar
  2. 2.
    Amiriparian, S., Gerczuk, M., Ottl, S., Cummins, N., Freitag, M., Pugachevskiy, S., Schuller, B.: Snore sound classification using image-based deep spectrum features. In: Proceedings INTERSPEECH, 5 p. ISCA, Stockholm (2017)Google Scholar
  3. 3.
    Arsikere, H., Lulich, S.M., Alwan, A.: Estimating speaker height and subglottal resonances using mfccs and gmms. IEEE Signal Process. Lett. 21(2), 159–162 (2014)CrossRefGoogle Scholar
  4. 4.
    Chang, J., Scherer, S.: Learning representations of emotional speech with deep convolutional generative adversarial networks. arXiv preprint (2017). arXiv:1705.02394
  5. 5.
    Chen, N., Qian, Y., Yu, K.: Multi-task learning for text-dependent speaker verification. In: Proceedings INTERSPEECH, 5 p. ISCA, Dresden, Germany (2015)Google Scholar
  6. 6.
    Chen, X.W., Lin, X.: Big data deep learning: challenges and perspectives. IEEE Access 2, 514–525 (2014)CrossRefGoogle Scholar
  7. 7.
    Covington, P., Adams, J., Sargin, E.: Deep neural networks for youtube recommendations. In: Proceedings 10th ACM Conference on Recommender Systems (RecSys), pp. 191–198. ACM, Boston (2016)Google Scholar
  8. 8.
    Davis, K.: Ethics of Big Data: Balancing risk and innovation. O’Reilly Media Inc., Newton (2012)Google Scholar
  9. 9.
    Deng, J., Schuller, B.: Confidence measures in speech emotion recognition based on semi-supervised learning. In: Proceedings of INTERSPEECH, 5 p. ISCA, Portland (2012)Google Scholar
  10. 10.
    Deng, L., Li, J., Huang, J.T., Yao, K., Yu, D., Seide, F., Seltzer, M., Zweig, G., He, X., Williams, J., et al.: Recent advances in deep learning for speech research at microsoft. In: Proceedings ICASSP, pp. 8604–8608. IEEE, Vancouver (2013)Google Scholar
  11. 11.
    Deng, X.N., Joshi, K.: Is crowdsourcing a source of worker empowerment or exploitation? understanding crowd workers perceptions of crowdsourcing career (2013)Google Scholar
  12. 12.
    Eyben, F., Wöllmer, M., Schuller, B.: A Multi-task approach to continuous five-dimensional affect sensing in natural speech. ACM Trans. Interact. Intell. Syst. Spec. Issue Affect. Interact. Nat. Environ. 2(1), 6 (2012)Google Scholar
  13. 13.
    Freitag, M., Amiriparian, S., Cummins, N., Gerczuk, M., Schuller, B.: An ‘end-to-evolution’ hybrid approach for snore sound classification. In: Proceedings INTERSPEECH, 5 p. ISCA, Stockholm (2017)Google Scholar
  14. 14.
    Goldberg, A.B., Zhu, X.: Seeing stars when there aren’t many stars: graph-based semi-supervised learning for sentiment categorization. In: Proceedings 1st Workshop on Graph Based Methods for Natural Language Processing, pp. 45–52. ACL, Stroudsburg (2006)Google Scholar
  15. 15.
    Guggilla, C.: Discrimination between similar languages, varieties and dialects using cnn-and lstm-based deep neural networks. VarDial 3, 185 (2016)Google Scholar
  16. 16.
    Hantke, S., Eyben, F., Appel, T., Schuller, B.: ihearu-play: Introducing a game for crowdsourced data collection for affective computing. In: Proceedings 6th biannual Conference on Affective Computing and Intelligent Interaction (ACII), pp. 891–897. aaac/IEEE, Xi’An (2015)Google Scholar
  17. 17.
    Hantke, S., Zhang, Z., Schuller, B.: Towards intelligent crowdsourcing for audio data annotation: integrating active learning in the real world. In: Proceedings INTERSPEECH, 5 p. ISCA, Stockholm, Sweden (2017)Google Scholar
  18. 18.
    Harris, C.G., Srinivasan, P.: Crowdsourcing and ethics. In: Altshuler, Y., Elovici, Y., Cremers, A.B., Aharony, N., Pentland, A. (eds.) Security and Privacy in Social Networks, pp. 67–83. Springer, Heidelberg (2013)CrossRefGoogle Scholar
  19. 19.
    Kranjec, J., Beguš, S., Geršak, G., Drnovšek, J.: Non-contact heart rate and heart rate variability measurements: a review. Biomed. Signal Process. Control 13, 102–112 (2014)CrossRefGoogle Scholar
  20. 20.
    Krizhevsky, A., Sutskever, I., Hinton, G.E.: Imagenet classification with deep convolutional neural networks. In: Advances in neural information processing systems, pp. 1097–1105 (2012)Google Scholar
  21. 21.
    Künzel, H.J.: How well does average fundamental frequency correlate with speaker height and weight? Phonetica 46(1–3), 117–125 (1989)CrossRefGoogle Scholar
  22. 22.
    Liu, P., Qiu, X., Huang, X.: Adversarial multi-task learning for text classification. arXiv preprint (2017). arXiv:1704.05742
  23. 23.
    Lu, J., Behbood, V., Hao, P., Zuo, H., Xue, S., Zhang, G.: Transfer learning using computational intelligence: a survey. Knowl. Based Syst. 80, 14–23 (2015)CrossRefGoogle Scholar
  24. 24.
    Lyakso, E., Frolova, O., Dmitrieva, E., Grigorev, A., Kaya, H., Salah, A.A., Karpov, A.: EmoChildRu: emotional child russian speech corpus. In: Ronzhin, A., Potapova, R., Fakotakis, N. (eds.) SPECOM 2015. LNCS, vol. 9319, pp. 144–152. Springer, Cham (2015). doi: 10.1007/978-3-319-23132-7_18 CrossRefGoogle Scholar
  25. 25.
    Majumder, N., Poria, S., Gelbukh, A., Cambria, E.: Deep learning-based document modeling for personality detection from text. IEEE Intell. Syst. 32(2), 74–79 (2017)CrossRefGoogle Scholar
  26. 26.
    Mao, Q., Dong, M., Huang, Z., Zhan, Y.: Learning salient features for speech emotion recognition using convolutional neural networks. IEEE Trans. Multimedia 16(8), 2203–2213 (2014)CrossRefGoogle Scholar
  27. 27.
    Mitchell, T.M., Cohen, W., Hruschka, E., Talukdar, P., Betteridge, J., Carlson, A., Mishra, B.D., Gardner, M., Kisiel, B., Krishnamurthy, J., et al.: Never-ending learning. In: Proceedings 29th AAAI Conference on Artificial Intelligence. AAAI, Austin (2015)Google Scholar
  28. 28.
    Miyato, T., Dai, A.M., Goodfellow, I.: Virtual adversarial training for semi-supervised text classification. Stat 1050, 25 (2016)Google Scholar
  29. 29.
    Moore, R.K.: A comparison of the data requirements of automatic speech recognition systems and human listeners. In: Proceedings INTERSPEECH, pp. 2582–2584, Geneva, Switzerland (2003)Google Scholar
  30. 30.
    Morschheuser, B., Hamari, J., Koivisto, J.: Gamification in crowdsourcing: A review. In: Proceedings 49th Hawaii International Conference on System Sciences (HICSS). pp. 4375–4384. IEEE (2016)Google Scholar
  31. 31.
    Nakov, P., Ritter, A., Rosenthal, S., Sebastiani, F., Stoyanov, V.: Semeval-2016 task 4: sentiment analysis in twitter. In: Proceedings International Workshop on Semantic Evaluations (SemEval), pp. 1–18 (2016)Google Scholar
  32. 32.
    Pokorny, F., Schuller, B., Marschik, P., Brückner, R., Nyström, P., Cummins, N., Bölte, S., Einspieler, C., Falck-Ytter, T.: Earlier identification of children with autism spectrum disorder: an automatic vocalisation-based approach. In: Proceedings INTERSPEECH, 5 p. ISCA, Stockholm (2017)Google Scholar
  33. 33.
    Poorjam, A.H., Bahari, M.H., Vasilakakis, V., et al.: Height estimation from speech signals using i-vectors and least-squares support vector regression. In: Proceedings 38th International Conference on Telecommunications and Signal Processing (TSP), pp. 1–5. IEEE, Prague (2015)Google Scholar
  34. 34.
    Poorjam, A.H., Bahari, M.H., et al.: Multitask speaker profiling for estimating age, height, weight and smoking habits from spontaneous telephone speech signals. In: Proceedings 4th International eConference on Computer and Knowledge Engineering (ICCKE). pp. 7–12. IEEE, Mashhad (2014)Google Scholar
  35. 35.
    Poria, S., Cambria, E., Hazarika, D., Vij, P.: A deeper look into sarcastic tweets using deep convolutional neural networks. arXiv preprint (2016). arXiv:1610.08815
  36. 36.
    Raina, R., Battle, A., Lee, H., Packer, B., Ng, A.Y.: Self-taught learning: transfer learning from unlabeled data. In: Proceedings 24th International Conference on Machine learning. pp. 759–766. ACM, Corvallis, OR (2007)Google Scholar
  37. 37.
    Rangel, F., Rosso, P., Verhoeven, B., Daelemans, W., Potthast, M., Stein, B.: Overview of the 4th author profiling task at pan 2016: cross-genre evaluations. Working Notes Papers of the CLEF (2016)Google Scholar
  38. 38.
    Schuller, B., Mousa, A.E.D., Vryniotis, V.: Sentiment analysis and opinion mining: on optimal parameters and performances. Wiley Interdisc. Rev. Data Min. Knowl. Disc. 5(5), 255–263 (2015)CrossRefGoogle Scholar
  39. 39.
    Schuller, B., Steidl, S., Batliner, A., Bergelson, E., Krajewski, J., Janott, C., Amatuni, A., Casillas, M., Seidl, A., Soderstrom, M., Warlaumont, A., Hidalgo, G., Schnieder, S., Heiser, C., Hohenhorst, W., Herzog, M., Schmitt, M., Qian, K., Zhang, Y., Trigeorgis, G., Tzirakis, P., Zafeiriou, S.: The INTERSPEECH 2017 computational paralinguistics challenge: addressee, Cold and Snoring.. In: Proceedings INTERSPEECH, 5 p. ISCA, Stockholm (2017)Google Scholar
  40. 40.
    Schuller, B., Vlasenko, B., Eyben, F., Wollmer, M., Stuhlsatz, A., Wendemuth, A., Rigoll, G.: Cross-corpus acoustic emotion recognition: variances and strategies. IEEE Trans. Affect. Comput. 1(2), 119–131 (2010)CrossRefGoogle Scholar
  41. 41.
    Schuller, B., Wöllmer, M., Eyben, F., Rigoll, G., Arsić, D.: Semantic speech tagging: towards combined analysis of speaker traits. In: Proceedings AES 42nd International Conference, pp. 89–97. AES, Ilmenau (2011)Google Scholar
  42. 42.
    Silver, D.L., Yang, Q., Li, L.: Lifelong machine learning systems: Beyond learning algorithms. In: Proceedings AAAI spring symposium series. AAAI, Palo Alto (2013)Google Scholar
  43. 43.
    Simonyan, K., Zisserman, A.: Very deep convolutional networks for large-scale image recognition. arXiv preprint (2014). arXiv:1409.1556
  44. 44.
    Strapparava, C., Mihalcea, R.: Semeval-2007 task 14: Affective text. In: Proceedings 4th International Workshop on Semantic Evaluations (SemEval), pp. 70–74. ACL, Swarthmore (2007)Google Scholar
  45. 45.
    Stuhlsatz, A., Meyer, C., Eyben, F., Zielke, T., Meier, G., Schuller, B.: Deep neural networks for acoustic emotion recognition: raising the benchmarks. In: Proceedings ICASSP, pp. 5688–5691. IEEE, Prague (2011)Google Scholar
  46. 46.
    Sun, X., Gao, F., Li, C., Ren, F.: Chinese microblog sentiment classification based on convolution neural network with content extension method. In: Proceedings 6th Biannual Conference on Affective Computing and Intelligent Interaction (ACII), pp. 408–414. aaac/IEEE, Xi’an (2015)Google Scholar
  47. 47.
    Tang, D., Qin, B., Liu, T.: Document modeling with gated recurrent neural network for sentiment classification. In: Proceedings Conference on Empirical Methods in Natural Language Processing (EMNLP). pp. 1422–1432. ACL, Lisbon, Portugal (2015)Google Scholar
  48. 48.
    Tarasov, A., Delany, S.J., Mac Namee, B.: Dynamic estimation of worker reliability in crowdsourcing for regression tasks: making it work. Expert Syst. Appl. 41(14), 6190–6210 (2014)CrossRefGoogle Scholar
  49. 49.
    Taylor, M.E., Stone, P.: Transfer learning for reinforcement learning domains: a survey. J. Mach. Learn. Res. 10, 1633–1685 (2009)MathSciNetzbMATHGoogle Scholar
  50. 50.
    Trigeorgis, G., Ringeval, F., Brückner, R., Marchi, E., Nicolaou, M., Schuller, B., Zafeiriou, S.: Adieu features? end-to-end speech emotion recognition using a deep convolutional recurrent network. In: Proceedings ICASSP, pp. 5200–5204. IEEE, Shanghai (2016)Google Scholar
  51. 51.
    Tzeng, E., Hoffman, J., Saenko, K., Darrell, T.: Adversarial discriminative domain adaptation (workshop extended abstract) (2017)Google Scholar
  52. 52.
    Van Dommelen, W.A., Moxness, B.H.: Acoustic parameters in speaker height and weight identification: sex-specific behaviour. Lang. Speech 38(3), 267–287 (1995)CrossRefGoogle Scholar
  53. 53.
    Walker, S., Pedersen, M., Orife, I., Flaks, J.: Semi-supervised model training for unbounded conversational speech recognition. arXiv preprint (2017). arXiv:1705.09724
  54. 54.
    Wöllmer, M., Eyben, F., Reiter, S., Schuller, B., Cox, C., Douglas-Cowie, E., Cowie, R.: Abandoning emotion classes - towards continuous emotion recognition with modelling of long-range dependencies. In: Proceedings INTERSPEECH, pp. 597–600. ISCA, Brisbane (2008)Google Scholar
  55. 55.
    Xia, R., Liu, Y.: Leveraging valence and activation information via multi-task learning for categorical emotion recognition. In: Proceedings ICASSP, pp. 5301–5305. IEEE, Brisbane (2015)Google Scholar
  56. 56.
    Zhang, B., Provost, E.M., Essi, G.: Cross-corpus acoustic emotion recognition from singing and speaking: a multi-task learning approach. In: Proceedings ICASSP, pp. 5805–5809. IEEE, Shanghai (2016)Google Scholar
  57. 57.
    Zhang, B., Provost, E.M., Essl, G.: Cross-corpus acoustic emotion recognition with multi-task learning: seeking common ground while preserving differences. IEEE Trans. Affect. Comput. (2017)Google Scholar
  58. 58.
    Zhang, Y., Coutinho, E., Zhang, Z., Adam, M., Schuller, B.: On rater reliability and agreement based dynamic active learning. In: Proceedings 6th Biannual Conference on Affective Computing and Intelligent Interaction (ACII), pp. 70–76. aaac/IEEE, Xi’an (2015)Google Scholar
  59. 59.
    Zhang, Y., Liu, Y., Weninger, F., Schuller, B.: Multi-task deep neural network with shared hidden layers: breaking down the wall between emotion representations. In: Proceedings ICASSP, pp. 4990–4994. IEEE, New Orleans (2017)Google Scholar
  60. 60.
    Zhang, Y., Weninger, F., Ren, Z., Schuller, B.: Sincerity and deception in speech: two sides of the same coin? a transfer- and multi-task learning perspective. In: Proceedings INTERSPEECH, pp. 2041–2045. ISCA, San Francisco (2016)Google Scholar
  61. 61.
    Zhang, Y., Weninger, F., Schuller, B.: Cross-domain classification of drowsiness in speech: the case of alcohol intoxication and sleep deprivation. In: Proceedings INTERSPEECH, 5 p. ISCA, Stockholm (2017)Google Scholar
  62. 62.
    Zhang, Y., Zhou, Y., Shen, J., Schuller, B.: Semi-autonomous data enrichment based on cross-task labelling of missing targets for holistic speech analysis. In: Proceedings ICASSP, pp. 6090–6094. IEEE, Shanghai (2016)Google Scholar
  63. 63.
    Zhang, Z., Coutinho, E., Deng, J., Schuller, B.: Cooperative learning and its application to emotion recognition from speech. IEEE/ACM Trans. Audio Speech Lang. Process. 23(1), 115–126 (2015)Google Scholar
  64. 64.
    Zhang, Z., Weninger, F., Wöllmer, M., Schuller, B.: Unsupervised learning in cross-corpus acoustic emotion recognition. In: Proceedings ASRU, pp. 523–528. IEEE, Big Island (2011)Google Scholar
  65. 65.
    Zhou, C., Sun, C., Liu, Z., Lau, F.: A c-lstm neural network for text classification. arXiv preprint (2015). arXiv:1511.08630
  66. 66.
    Zhu, X., Lafferty, J., Ghahramani, Z.: Combining active learning and semi-supervised learning using gaussian fields and harmonic functions. In: Proceedings ICML 2003 Workshop on the Continuum From Labeled to Unlabeled Data in Machine Learning and Data Mining, vol. 3, Washington, DC (2003)Google Scholar

Copyright information

© Springer International Publishing AG 2017

Authors and Affiliations

  1. 1.Department of ComputingImperial College LondonLondonUK
  2. 2.Chair of Complex and Intelligent SystemsUniversity of PassauPassauGermany
  3. 3.audEERING GmbHGilchingGermany

Personalised recommendations