統計学輪講（第9回）

日時      2011年07月05日(火)    15時50分〜16時40分
場所      経済学部新棟３階第３教室
講演者    Pattara Rujeerapaiboon  (情報理工M2)
演題      Statistical Model for Electrocardiogram

概要
Electrocardiogram (ECG) is a series of electrical signals produced by heart
pulses recorded by skin electrodes over a period of time. It captures
important information of the cardiac health and condition of patients. ECG 
is usually
analyzed and interpreted by medical practitioners to gain information on the
functioning of patients' hearts, which are important in cardiovascular
diseases or cardiac condition diagnosis. ECG exhibits periodicity.
However, the peridocity of ECG is variational.
We developed a new model for ECG that relies on this property. We
proposed a statistical generative model for ECG signals based on an 
autoregressive
model with variational time lag. Speci cally speaking, we model ECG data 
based on
a 2-level hierarchical autoregressive model. The rst level of the model is
an autoregressive model that generates variational time lags (periods), 
while
the second level is an autoregressive model that generates ECG data based on 
the
times lags generated by the rst level. The performance comparison shows that
our proposed model yields in better performance than the existing model. Our
model is hoped to be further developed to be able to enhance effectiveness
and efficiency of ECG data analysis.