Sound Synthesis by Flexible Activation Function Recurrent Neural Networks

  • Aurelio Uncini
Conference paper
Part of the Lecture Notes in Computer Science book series (LNCS, volume 2486)


In this paper we investigate on the use of adaptive spline neural networks, to define a new general class of physical-like sound synthesis models, based on a learning from examples strategy (in particular in this paper we study single-reed woodwind instruments). It is well known that one of the main problems in physical modeling concerns the difficulty of parameter identification and the definition of the exciter nonlinear function shape (which plays a key rule in the instrument timbre). In the proposed neural model we make use of FIR-IIR synapses followed by a Catmul-Rom spline based flexible nonlinear function whose shape can be modified by adapting its control points. This general structure can imitate an entire class of instruments by learning all the parameters (synaptic weights and spline control points) from recorded sounds. In order to obtain an efficient hardware/software implementation, the synaptic weights are constrained to be power-of-two terms while the nonlinear function can be implemented as a simple spline interpolation scheme or through a small lookup table. In order to demonstrate the effectiveness of the proposed model, experiments on a single-reed woodwind instruments have been carried out.


sound synthesis physical model flexible activation function spline neural networks power-of-two neural networks 


Unable to display preview. Download preview PDF.

Unable to display preview. Download preview PDF.


  1. [1]
    S. Y. Kung, J. N. Hwang, “Neural Networks for Intelligent Multimedia Processing”, Proceedings of IEEE, Vol. 86, No. 6, pp. 1244–1272, June 1998.CrossRefGoogle Scholar
  2. [2]
    B. L. Vercoe, W. G. Gardner, E. D. Scheirer, “Structured Audio: Creation, Transmission, and Rendering of Parametric Sound Representations”, Proceedings of IEEE, Vol. 86, No. 5, pp. 922–940, May 1998.CrossRefGoogle Scholar
  3. [3]
    J. O. Smith, “Physical Modeling Using Digital Wave-guides”, Computer Music Journal, vol 16, n. 4, pp 74–91, 1992.CrossRefGoogle Scholar
  4. [4]
    G. Borin, G De Poli, and A. Sarti, “Sound Synthesis by Dynamic Systems Interaction”, in Readings in Computer-Generated Music, pp. 139–160, IEEE Comp. Soc. Press, D. Baggi ed., 1992.Google Scholar
  5. [5]
    Gary Paul Scavone, “An acoustical analysis of single-reed woodwind instruments with an emphasis on design and performance issues and digital waveguide modeling techniques”, Ph.D. thesis, Music Dept., Stanford University, March 1997.Google Scholar
  6. [6]
    N.H. Fletcher, T.D. Rossing, “The Physics of Musical Instruments”, Springer-Verlang, New-York, 1981.Google Scholar
  7. [7]
    Gary P. Scavone, Perry Cook, “Real Time Computer Modeling of woodwind instruments”, International Symposium on Musical Acoustics, Leavenworth, WA 1998.Google Scholar
  8. [8]
    J. O. Smith, “Technique for Digital Filter Design and System Identification with Application to the Violin”, PhD Thesis, CCRM, Standford University, Report n. STAN-M-14, 1983.Google Scholar
  9. [9]
    J. Vuori and V. Vlimki, “Parameter Estimation of Non-linear Physical Models by Simulated Evolution: Application to the Flute Model”, Proc. Int. Comp. Music Conf., pp. 402–404, Tokyo 1993.Google Scholar
  10. [10]
    Carlo Drioli, Davide Rocchesso, “Learning Pseudo-Physical Models for Sound Synthesis and Trasformation”, IEEE International Conference on Systems, Man, and Cybernetics, vol. 2, pp 1085–1090, 1998.Google Scholar
  11. [11]
    M. A. Casey, “Understanding musical sound with forward models and physical models”, Connection Sci., vol. 6, pp. 355–371, 1994.CrossRefGoogle Scholar
  12. [12]
    A. Horner, B. Beauchamp, and L. Haken, “FM matching synthesis with genetic algorithms”, Comput. Music J., vol. 17, no. 4, pp. 17–29, Winter 1993.CrossRefGoogle Scholar
  13. [13]
    Stefano Guarnieri, Francesco Piazza and Aurelio Uncini, “Multilayer Feedforward Networks with AdaptiveSpline Activation Function”, IEEE Trans. On Neural Network, Vol. 10, No. 3, pp. 672–683, May 1999.CrossRefGoogle Scholar
  14. [14]
    Lorenzo Vecci, Francesco Piazza and Aurelio Uncini, “Learning and Approximation Capabilities of Adaptive Spline Activation Function Neural Networks”, Neural Networks, Vol. 11, No. 2, pp 259–270, March 1998.CrossRefGoogle Scholar
  15. [15]
    Aurelio Uncini, Lorenzo Vecci, Paolo Campolucci and Francesco Piazza, “Complex-valued Neural Networks with Adaptive Spline Activation Function for Digital Radio Links Nonlinear Equalization”, IEEE Trans. on Signal Processing, Vol. 47, No. 2, February 1999.Google Scholar
  16. [16]
    Paolo Campolucci, Aurelio Uncini, Francesco Piazza and Bhaskar D. Rao, “OnLine Learning Algorithms for Locally Recurrent Neural Networks”, IEEE Trans, on Neural Network, Vol. 10, No. 2, pp. 253–271 March 1999.CrossRefGoogle Scholar
  17. [17]
    F. Glover, M. Laguna, “Tabu search”, Kluver Academic Publisher, 1997.Google Scholar
  18. [18]
    Stefano Traferro and Aurelio Uncini, “Power-of-Two Adaptive Filters Using Tabu Search”, IEEE Transactions on Circuits and Systems-II: Analog and Digital Signal Processing, Vol. 47, No. 6, June 2000.Google Scholar

Copyright information

© Springer-Verlag Berlin Heidelberg 2002

Authors and Affiliations

  • Aurelio Uncini
    • 1
  1. 1.INFOCOM dept.University of Rome “La Sapienza”RomeItaly

Personalised recommendations