Time-frequency signal decompositions for audio and speech processing


  • Karthikeyan Umapathy Dept. of Elec. and Comp. Engg., University of Western Ontario, London, Ont. N6A 5B8
  • Sridhar Krishnan Dept. of Elec. and Comp. Engg., Ryerson University, Toronto, Ont. M5B 2K3


Audio acoustics, Frequencies, Functions, Pattern recognition, Speech processing, Frequency band mapping (TWFB), Gaussian TF function, Matching pursuit (MP), TF decomposition


A unified adaptive TF decomposition based methodology for processing audio and speech signals was presented. The proposed methodology uses an adaptive TF transform (ATFT) based on the matching pursuit (MP) algorithm with Gaussian TF function. MP when used with a dictionary of TF functions yields an adaptive time-frequency transformation. A novel time-width vs frequency band mapping (TWFB) is also developed which demonstrates high potential to form as a versatile parametric visualization/pattern recognition tool.

Additional Files



How to Cite

Umapathy K, Krishnan S. Time-frequency signal decompositions for audio and speech processing. Canadian Acoustics [Internet]. 2005 Sep. 1 [cited 2024 Jun. 15];33(3):58-9. Available from: https://jcaa.caa-aca.ca/index.php/jcaa/article/view/1744



Proceedings of the Acoustics Week in Canada