A more flexible method for recognizing signals using back propagation: Piecewise linear regression vectors

  • Greg Makowski
Track 2: Artificial Intelligence
Part of the Lecture Notes in Computer Science book series (LNCS, volume 507)


Using the neural network architecture of back propagation applied to speech recognition, a new input data structure is developed which improves shift invariance when recognizing signal data in the form such as a formant. The preliminary development of the data structure, piecewise linear regression vectors, is reported. The new input data structure reduces the amount of data presented to the network by as much as an order of magnitude, giving a computational advantage in execution speed.


Speech Recognition Back Propagation Time Slice Test Pattern Dynamic Time Warping 
These keywords were added by machine and not by the authors. This process is experimental and the keywords may be updated as the learning algorithm improves.


Unable to display preview. Download preview PDF.

Unable to display preview. Download preview PDF.


  1. [1]
    Y. Bengio and R. De Mori. Speaker normalization and automatic speech recognition using spectral lines and neural networks. In Proceedings of the 1988 Connectionist Models Summer School, pages 388–397. Carnegie Mellon University, 1988.Google Scholar
  2. [2]
    Michael A. Franzini. Learning to recognize spoken words: A study in connectionist speech recognition. In Proceedings of the 1988 Connectionist Models Summer School, pages 407–416. Carnegie Mellon University, 1988.Google Scholar
  3. [3]
    Richard P. Lippmann. An introduction to computing with neural nets. IEEE ASSP, Vol 3, No 4:4–22, 1987.CrossRefGoogle Scholar
  4. [4]
    Richard P. Lippmann. Neural network classifiers for speech recognition. The Lincoln Laboratory Journal, Vol 1, No 1:107–124, 1988.Google Scholar
  5. [5]
    Richard P. Lippmann. Review of neural networks for speech recognition. Neural Computation, Vol 1, No 1:1–38, 1989.Google Scholar
  6. [6]
    Thomas W. Parsons. Voice and Speech Processing. McGraw-Hill, 1987.Google Scholar
  7. [7]
    Lawrence R. Rabiner, A.E. Rosenberg, and S. Levinson. Considerations in dynamic time warping algorithms for discrete word recognition. IEEE Transactions on Acoustics, Speech, and Signal Processing, VOL ASSP-26, NO. 6:575–582, 1978.CrossRefGoogle Scholar
  8. [8]
    M. L. Rossen and J. A. Anderson. Representational issues in a neural network model of syllable recognition. In International Joint Conference on Neural Networks, pages I.19–I.21. IEEE, 1989.Google Scholar
  9. [9]
    David E. Rumelhart and James L. McClelland. Explorations in Parallel Distributed Processing. MIT Press, 1988.Google Scholar
  10. [10]
    David E. Rumelhart and James L. McClelland. Parallel Distributed Processing, Vol I,II. MIT Press, 1988.Google Scholar
  11. [11]
    Terrence J. Sejnowski and Charles R. Rosenberg. Nettalk: a parallel network that learns to read aloud. Technical Report JHU/EECS-86/01, 32pp., The Johns Hopkins University, Electrical Engineering and Computer Science, Johns Hopkins, 1986.Google Scholar
  12. [12]
    E. Vidal, H. Rulot, C. Casacuberta, and J. Benedi. On the use of a metric-space search algorithm (aesa) for fast dtw-base recognition of isolated words. IEEE Transactions on Acoustics, Speech, and Signal Processing, VOL ASSP-36, NO. 5:651–660, 1988.CrossRefGoogle Scholar
  13. [13]
    Alex Waibel, H. Sawai, and K. Shikano. Modularity and scaling in large phonemic neural networks. Technical Report TR-I-0034, 25pp, ATR Interpreting Telephony Research Laboratories, Japan, 1988.Google Scholar

Copyright information

© Springer-Verlag Berlin Heidelberg 1991

Authors and Affiliations

  • Greg Makowski
    • 1
  1. 1.Computer Science DepartmentWestern Michigan UniversityKalamazoo

Personalised recommendations