Delay estimation using sinusoidal modeling: application to epileptic seizures

David L. Sherman, Yien Che Tsai, Marek A. Mirski, Nitish V. Thakor

Research output: Contribution to journalArticle

Abstract

The ESPRIT analysis method reveals information including direction-of-arrival (DOA) and frequency about constituent sinusoids. The PRO-ESPRIT algorithm extends its use as a time delay method for temporal delay. After calculating appropriate model order and frequency estimates, phase delays are calculated from signal auto- and crosscorrelation matrices. A low-rank approximation of signal matrix pencil known as the core rotations matrix is made to calculate delays. A key advantage of PRO-ESPRIT is that both phase shifts and frequency are estimated by a data-association method. The simulations show estimation accuracy in scenarios with correlated sinusoids as well as low coherence channels. Tracing propagation delays occurring in the spread of the brain's epileptic seizure-related electrical signals is an important issue in understanding the functional anatomy of epilepsy. Experimentally, it has been observed that the development of a clonic seizure is accompanied by electroencephalogram (EEC) periodic signals. Exploiting this periodicity, the signal is modeled as a collection of coherent sinusoids. The PROESPRIT method is then used to analyze the EEC phase delays during seizure episodes in an experimental epilepsy model. In a model of epileptic seizure propagation, this method is useful for calculating small phase delays between the cerebral cortex and mid-brain centers.

Original languageEnglish (US)
Number of pages1
JournalIEEE Transactions on Signal Processing
Volume45
Issue number5
StatePublished - Dec 1 1997

    Fingerprint

ASJC Scopus subject areas

  • Signal Processing
  • Electrical and Electronic Engineering

Cite this