|Atienza Alonso David |
An integral part of advanced health monitoring systems today is continuous bio-signal monitoring through embedded devices in combination with signal processing and machine learning techniques. With the number of users scaling up rapidly, continuous health surveillance through existing paradigms requires a high sampling rate of up to hundreds or even thousands of times per second. To deal with resulting problems such as the management of large amounts of generated data and strive for energy efficiency, some modern wearables have adopted a non-uniform sampling scheme to reduce the amount of data to be captured, processed, transmitted and stored. However, this scheme prevents the use of most of the existing biosignal processing algorithms.
The aim of this project is to develop novel biosignal sampling strategies that reduce the amount of acquired data by orders of magnitude. For this, we adopt the so-called event-based paradigm, in which the signal is only acquired when something relevant is observed. Using this system, we propose to create event-based heart-rate analysis devices, including a novel algorithm for QRS detection that is able to process electrocardiogram signals acquired irregularly and much below the theoretically-required Nyquist rate. In the process, we can drastically reduce the average sampling frequency of the signal and, hence, the energy needed to process it and extract the relevant information.
We implemented both the proposed event-based algorithm and a state-of-the-art version of the regular Nyquist rate-based sampling on an ultra-low power hardware platform. The experimental results show that the event-based version reduces the energy consumption in runtime up to 15.6 times, while the detection performance is maintained at an average F1 score of 99.5%.