Publication result detail

The Number of States of the Hidden Markov Model in ECG Signal Processing

BARDOŇOVÁ, J., PROVAZNÍK, I.

Original Title

The Number of States of the Hidden Markov Model in ECG Signal Processing

English Title

The Number of States of the Hidden Markov Model in ECG Signal Processing

Type

Paper in proceedings (conference paper)

Original Abstract

In the study, an influence of the number of states of a hidden Markov model on recognition efficiency was examined. The model was used for detection of acute myocardial ischemia from ECG signals. The time-frequency wavelet transform was applied to preprocess the recorded data. Thus, HMMs were able to recognize subtle electrophysiological changes caused by ischemia. The models with selected number of states were tested on signals from 10 experiments. An optimal number of states K=3 was found in the study.

English abstract

In the study, an influence of the number of states of a hidden Markov model on recognition efficiency was examined. The model was used for detection of acute myocardial ischemia from ECG signals. The time-frequency wavelet transform was applied to preprocess the recorded data. Thus, HMMs were able to recognize subtle electrophysiological changes caused by ischemia. The models with selected number of states were tested on signals from 10 experiments. An optimal number of states K=3 was found in the study.

Key words in English

Electrocardiographic signal, wavelet transform, vector quantization, hidden Markov model

Authors

BARDOŇOVÁ, J., PROVAZNÍK, I.

Released

23.10.2002

Publisher

Západočeská Univerzita

Location

Plzeň

ISBN

80-7082-904-4

Book

Elektrotechnika a informatika 2002

Pages from

95

Pages count

4

BibTex

@inproceedings{BUT5447,
  author="Jana {Kolářová} and Valentýna {Provazník}",
  title="The Number of States of the Hidden Markov Model in ECG Signal Processing",
  booktitle="Elektrotechnika a informatika 2002",
  year="2002",
  pages="4",
  publisher="Západočeská Univerzita",
  address="Plzeň",
  isbn="80-7082-904-4"
}