Atrial fibrillation (AF) is one of the most diffused cardiac arrhythmias. Suffering from AF may lead to heart failure and to stroke, so an early detection and a continous monitoring are determining factors in the desease prevention. With the development of telemonitoring systems through wearable devices, the personalized medicine has reached a new level of improvement. An increasing number of telemonitoring systems base their functioning on the recording and the analysis of single-lead ECGs, with the purpose of detecting heart diseases using heart rhythm and rate features.

The purpose of this paper is to present an algorithm for the screening and monitoring of heart disease patients. While most of the wearable systems in this area are limited to the monitoring of ECG, this algorithm uses the combined analysis of different biosignals obtained with a sensorized t-shirt equipped with a single-lead ECG, a pulse-oximeter and a temperature sensor.

Since AF is known to alter heart rhythms’ dynamic and morfological characteristics of ECG signal, a time and frequency domain analysis is performed in order to extract the ECG features. Data collected from wearable devices are often exposed to different kind of artifacts, so morphological characteristic are not preferred because of their lack of robustness in noisy conditions. For this reason frequency and heart rate variability (HRV)-based analisys are used for features extraction. Through this ECG analysis and with the support of the other recorded biosignals, is therefore possibile to exctract features to perform an automatic detection of arrhythmias (specifically AF) and the classification of ECG signals.