A LINGUISTIC MODEL FOR WAVEFORM ANALYSIS.

Abstract

Many important pattern recognition problems deal with waveform analysis, e.g., electrocardiograms, engine noise signatures, etc. Spectral analysis is usually not helpful in such cases. A method is described by which a time function can be represented as a concatenation of simpler functions instead of a sum as is the case with spectral analysis. This representation emphasizes features of short duration which tend to be obscured in the spectral analysis. Furthermore, one can use the results of the analysis to represent the waveforms as strings of symbols. A procedure for imperfect matching of such strings is given and the problem of detecting common substrings which characterize the various classes is discussed. (Author)

Document Details

Document Type
Technical Report
Publication Date
Mar 01, 1969
Accession Number
AD0684360

Entities

People

  • Theodosios Pavlidis

Organizations

  • Princeton University

Tags

DTIC Thesaurus Topics

  • Electrocardiography
  • Engine Noise
  • Health Services
  • Identification
  • Noise
  • Pattern Recognition
  • Recognition
  • Waveforms

Readers

  • Computational Linguistics
  • Speech Processing/Speech Recognition.
  • Statistical inference.

Technology Areas

  • AI & ML