Advertisement

Journal of the Korean Physical Society

, Volume 74, Issue 1, pp 12–18 | Cite as

Origin of the Higher Difficulty in the Recognition of Vowels Compared to Handwritten Digits in Deep Neural Networks

  • Hyunjae Kim
  • Woo Seok Lee
  • Jaeyun Yoo
  • Maruchan Park
  • Kang Hun AhnEmail author
Article
  • 3 Downloads

Abstract

We investigate the origin of the significantly different error rates between handwritten digit machine recognition and vowel sound machine recognition. While the error rate for five Korean vowel sounds, [ɑ], [ʊ], [ɪ], [ο], and [ɛ], is about 10 percent, that of handwritten digit recognition is less than 1 percent for convolutional neural networks (CNNs) with raw data. We first dilute the information of the sound by subtracting its temporal fine structure, with the assumption that sorting out extraneous sound information will improve the accuracy of vowel recognition. Simulation results show no improvement though, indicating that the recognition rate difference does not arise from unnecessary sound information. Rather, conserving subtle information with no information reduction can be helpful to improve recognition rates; however, even the model with the highest accuracy does not reach the accuracy for handwritten digit recognition we desired. Finally, we find that the main difficulty of Korean vowel sound recognition comes from the similarity of [ο] and [ɛ]; without [ɛ], recognition of the remaining vowels is up to 99 percent. The similarity can be seen through their formant structure. Humans overcome the similarity to adeptly differentiate the two, and human vowel recognition remains far superior to the best performing CNNs. This indicates room to develop deep neural networks beyond the CNN still exists.

Keywords

Vowel recognition Deep neural network Mel-frequency cepstral coefficients Formant analysis 

Preview

Unable to display preview. Download preview PDF.

Unable to display preview. Download preview PDF.

References

  1. [1]
    Y. LeCun, Y. Bengio and G. E. Hinton, Nature 521, 436 (2015).ADSCrossRefGoogle Scholar
  2. [2]
    Y. LeCun, C. Cortes and C. J. C. Burges, http://yann.lecun.com/exdb/mnist/.
  3. [3]
    C. Lopes and F. Perdigao, Speech Technologies (InTech, Rijeka, 2011), p. 285.Google Scholar
  4. [4]
    D. H. Hubel and T. N. Wiesel, Brain and visual perception: the story of a 25-year collaboration (Oxford University Press, New York, 2004), p. 59.CrossRefGoogle Scholar
  5. [5]
    Y. LeCun et al., Neural Comput. 1, 541 (1989).CrossRefGoogle Scholar
  6. [6]
    I. I. Groen, E. H. Silson and C. I. Baker, Phil. Trans. R. Soc. B 372, 20160102 (2017).CrossRefGoogle Scholar
  7. [7]
    A. W. Harley, in International Symposium on Visual Computing (Las Vegas, NV, USA, December 14–16, 2015), p. 867.Google Scholar
  8. [8]
    S. B. Davis and P. Mermelstein, Readings in speech recognition (Morgan Kaufmann Publishers, Inc., San Mateo, CA, 1990), p. 65.CrossRefGoogle Scholar
  9. [9]
    L. Wan, M. Zeiler, S. Zhang, Y. Le Cun and R. Fergus, International Conference on Machine Learning (Atlanta, Georgia, USA, June 16–21, 2013), p. 1058.Google Scholar
  10. [10]
    B. C. Moore, JARO 9, 399 (2008).CrossRefGoogle Scholar
  11. [11]
    S. Rosen, Phil. Trans. R. Soc. Lond. B 336, 367 (1992).ADSCrossRefGoogle Scholar
  12. [12]
    R. V. Shannon, F. G. Zeng, V. Kamath, J. Wygonski and M. Ekelid, Science 270, 303 (1995).ADSCrossRefGoogle Scholar
  13. [13]
    L. Varnet, M. C. Ortiz-Barajas, R. G. Erra, J. Gervain and C. Lorenzi, JASA 142, 1976 (2017).CrossRefGoogle Scholar
  14. [14]
    V. Nair and G. E. Hinton, in International Conference on Machine Learning (Haifa, Israel, June 21–24, 2010), p. 807.Google Scholar
  15. [15]
    D. E. Rumelhart and J. L. McClelland, Parallel distributed processing: explorations in the microstructure of cognition,. volume 1. foundations (MIT Press, Cambridge, Massachusetts, 1986).Google Scholar
  16. [16]
    J. Bouvrie, Notes on convolutional neural networks (2006).Google Scholar
  17. [17]
    S. Ioffe and C. Szegedy, arXiv:1502.03167, 2015.Google Scholar
  18. [18]
    D. P. Kingma and J. Ba, arXiv:1412.6980, 2014.Google Scholar
  19. [19]
    J. Sundberg, Sci. Am 236, 82 (1977).CrossRefGoogle Scholar
  20. [20]
    C. J. Seong, J. Acoust. Soc. Korea 23, 454 (2004).Google Scholar
  21. [21]
    H. Lee, W. Shin and J. Shin, Phon. Speech Sci. 9, 39 (2017).Google Scholar
  22. [22]
    S. M. Cho, Korean Lang. Cult. 24, 427 (2003).Google Scholar
  23. [23]
    P. Boersma, http://www.praat.org/ (2006).
  24. [24]
    H. Kang and J. I. Han, Phon. Speech Sci. 5, 33 (2013).CrossRefGoogle Scholar
  25. [25]
    J. Kang and E. Kong, Phon. Speech Sci. 8, 39 (2016).Google Scholar

Copyright information

© The Korean Physical Society 2019

Authors and Affiliations

  • Hyunjae Kim
    • 1
  • Woo Seok Lee
    • 1
  • Jaeyun Yoo
    • 1
  • Maruchan Park
    • 1
  • Kang Hun Ahn
    • 1
    Email author
  1. 1.Department of PhysicsChungnam National UniversityDaejeonKorea

Personalised recommendations