
Project objectives: To identify electrical activity within noisy ECG/EMG/EEG sensor data. To recognize rhythms and patterns of orderly progression of heart depolarization, skin motion, and brain functioning changes. To separate various noise from the measurement data. To apply the mutual information-based approach in heart rate variability measurement for a real time stress management system. To form a prognosis framework for health-status-prediction using multiple sensors.
Broader Societal and Commercial Impact: Stress manifests itself and can affect us at the office, at home and at public places. It can infiltrate our society in countless ways. Stress also contributes to chronic diseases, such as high blood pressure, heart disease and diabetes. In addition to pharmaceutical effort in new medication development, industrial companies have developed relaxation applications and sensors to detect heart rate variability to estimate stress level.
Breakthrough potential, innovative concept: Wearable sensor system designs have several inter-correlated issues that require trade-offs and new design strategies. One of the biggest challenges is noise. It is well understood that wearable sensors such as electrocardiogram (ECG), electroencephalogram (EEG), electromyography (EMG), etc., are designed to process biopotential signals taken from various locations on human bodies. Typically, these signals have very low amplitude (e.g., tens to hundreds of μV) with less than 1kHz frequency range. Instead of fitting the measurement data to one particular model using least mean square, the proposed project identifies the dynamics of the human heart/brain/skin movements and functions through data processing. Then, the extracted data patterns/templates can help reduce noise. This approach is different from the abovementioned approaches which focus on noise reduction. The aim of this approach is to identify heart electrical activity within noisy sensor data. Use ECG as an example. By recognizing the rhythms and patterns of the orderly progression of heart depolarization, we are able to separate noise contributions from the measurement data. Therefore, the proposed new approach is applicable not only to ECG-based sensor data, but also to other sensors where rhythms and patterns of data may be recognized.
Value propositions: The proposed mutual information-based approach extracts real time dynamics of stress associated physiological changes, and thus reduces wearable sensor noises by 30%. Industrial relevance includes wearable sensor companies, sport training and utility companies, analog front-end design companies, and other data service providers that may use data patterns/templates to identify children/seniors and their health status.