Machine learning and data fusion for an intelligent wearable device

Diseases related to the cardiac and respiratory systems are the single largest causes of death worldwide. Noninvasive signals provide doctors a powerful tool for diagnosis. Due to the recent development in wearable technology, large amount of sensor data is generated. The goal of this project is to develop signal processing, feature extraction, machine learning and data fusion method for such wearable devices. A novel R peak detection scheme is proposed for processing of the noisy ECG Signals. This method is successfully applied to analyze ECG data arising from cardiac rehabilitation patients. Additionally, an adaptive filtering technique for noise cancellation in ECG is also explored. The instantaneous respiration frequency is successfully derived from the ECG signal and its performance was found to be superior to conventional methods. Machine learning and data fusion for classification of patient age group based on ECG and PCG signal is implemented using a combination of SVMs and KNN. Combination of ensembles of neural networks is implemented for automatic heart sound anomaly detection. Data fusion of the output from ECG and PCG based learners was implemented to improve classification accuracy. The methods discussed in this project will be useful for future studies for wearable patient monitoring and home based tele-health applications. Parts of this work have been accepted for publication in IEEE International Symposium on Circuits and Systems (ISCAS) 2018 conference.Bachelor of Engineerin