Comparison of spectral analysis models for feature extraction in word recognition

Authors

  • Albert James M. Licup ⋅ PH Department of Physics, University of San Carlos
  • Marlo A. Flores ⋅ PH Department of Physics, University of San Carlos
  • Edcel John L. Salumbides ⋅ PH Department of Physics, University of San Carlos
  • Kees Karremans ⋅ NL Vrije Universiteit Amsterdam, The Netherlands

Abstract

In this paper several speech features that could best represent a set of isolated words are compared. Features are derived from the power spectrum, the spectrogram, and the linear predictive code (LPC). Using a k-nearest neighbor classifier, we have obtained best results of 99.8% for the recognition rate when the features were obtained using LPC coefficients.

SPP 2002 Proceedings Cover

Downloads

Published

2002-10-23

Issue

2002: Proceedings of the 20th Samahang Pisika ng Pilipinas Physics Congress

Section

Image and Signal Processing

License

Copyright License Agreement for Full Articles

How to Cite

[1]
“Comparison of spectral analysis models for feature extraction in word recognition”, Proc. SPP, vol. 20, no. 1, pp. SPP–2002, Oct. 2002, Accessed: Apr. 05, 2026. [Online]. Available: https://proceedings.spp-online.org/article/view/SPP-2002-2H-03