Applying learning algorithms to extract anxiety levels using the heart rate variability measure

Abstract

The classification problems in biological measures have been studied since mathematical methods and statistical tools were created to determine difference between two distinct samples. In this paper we present a mathematical methodology capable of differing 29 non-clinical volunteers with distinct degrees of trait anxiety (high or low) according to the State and Trait Anxiety Inventory (STAI-T) using an electrocardiogram (ECG) data as starting point. Specifically, the wavelet transforms and its statistical measures were used to extract simple patterns from the resting ECG and classify the group as low or high trait anxiety. The Daubechies, Haar and Symlet mother function were used to filter the original ECG data. Then, by means of the Weka Learning Algorithm and using only 5 attributes (Pearson Coefficient from Haar and Symlet, Median from Haar and Mode of Haar and Daubechies) we achieved a higher level of reliability, 96.90% (p < .05), with low training percentages. The results showed the efficiency of this methodology to classify volunteers according to their anxiety levels through an ECG data collection.