Classification of time-varying signals using time-frequency atoms

Author

Wellig, Peter ; Moschytz, George S.

Author_Institution

Signal & Inf. Process. Lab., Swiss Fed. Inst. of Technol., Zurich, Switzerland

Volume

2

fYear

1999

fDate

36434

Abstract

Extracting relevant features from signals is a key element in classification of signals, e.g., for the decomposition of electromyograms (EMG signals). We present an algorithm which uses time-frequency dictionaries and adaptively selects a small number of discriminant time-frequency atoms. Using our method, simulations show reduced misclassification rates compared to commonly-used linear classifiers

Keywords

digital simulation; discrete wavelet transforms; electromyography; medical signal processing; signal classification; time-frequency analysis; EMG decomposition algorithms; EMG segments; class separability; classification; discriminant quality function; discriminant time-frequency atoms; discriminant wavelet packet atoms; electromyogram signals; local discriminant basis algorithm; misclassification rates; simulations; time-frequency dictionaries; time-varying signals; wavelet packet atoms; Data mining; Dictionaries; Electromyography; Electronic mail; Feature extraction; Information processing; Linear discriminant analysis; Signal processing; Time frequency analysis; Wavelet packets;

fLanguage

English

Publisher

ieee

Conference_Titel

[Engineering in Medicine and Biology, 1999. 21st Annual Conference and the 1999 Annual Fall Meetring of the Biomedical Engineering Society] BMES/EMBS Conference, 1999. Proceedings of the First Joint

Conference_Location

Atlanta, GA

ISSN

1094-687X

Print_ISBN

0-7803-5674-8

Type

conf

DOI

10.1109/IEMBS.1999.804107

Filename

804107