A Wearable Platform utilizing off-the-shelf Components and performing Quality Analysis of Physiological Data

In this paper we present our efforts towards establishing a wearable platform that utilizes off-the-shelf Bluetooth-enabled sensors in order to preprocess and analyze streaming physiological recordings. A smart-phone running multi-threaded J2ME software is utilized for handling multiple simultaneous Bluetooth connections and a network socket connection with a remote workstation. Received measurements of signals such as the ECG and PPG are decomposed through appropriately selected Wavelet Transforms with the purpose of identifying unusable segments that have been severely corrupted by noise and de-noising the remaining usable data portions. We study the use of the undecimated wavelet packet transform for ECG noise removal. Provided results illustrate the advantages of the proposed decomposition for wavelet denoising compared to conventional approaches, at the cost however of performing more computations. The described wearable platform along with the documented data preprocessing steps is employed as the front end of a closed-loop intelligent and interactive system termed Prognosis. This system is capable of facilitating ubiquitous and unsupervised round-the-clock health monitoring of people at risk, as it is able to a) address the issue of unsupervised data collection and b) to interact with the patient via an automated speech-dialogue system.