Software design and optimization of ECG signal analysis and diagnosis for embedded IoT devices

The medical domain is one of the most rapidly expanding application areas of Internet of Things (IoT) technology. For chronic diseases, this technology can be highly useful for the patient, providing constant monitoring and ability for timely intervention of medical staff in case of an emergency. This intended system behavior imposes new requirements to the design and implementation of processing flows implemented on embedded IoT devices which are already constrained by limited computational capabilities and power budget. This work aims at designing and implementing such a bio-medical signal analysis flow based on the case study of arrhythmia detection using electrocardiogram signals and machine learning techniques. Different architectural decisions of the flow are explored at high level and the final optimized version is implemented on a state-of-the-art IoT node. The evaluation of the execution flow on this device provides information on the actual requirements of each sub-component of the flow combined with an analysis of its behavior as computational requirements of the machine learning algorithms scale up