Development of a Real-time Chronic Stress Visualization System from Long-term Physiological Data.

We have developed a real-time system which can estimate and display chronic stress levels determined from a long-term physiological data. It consists of wearable sensors that measure physiological data, a smartphone application that receives data from the sensors and displays chronic stress levels, and a cloud system that estimates them on the basis of received data. To operate it, we have to treat irregularly uploaded user-physiological-data of varying sizes, calculate chronic stress levels from long-term features without delay on a daily basis, and display them in real-time on the smartphone application. For this purpose, we have developed a system that requires relatively little memory and processing time with one six-hundredth of maximum memory usage and one twentieth of processing time as compared to conventional method by subdividing uploaded physiological data, calculating features from them, and creating long-term features by combining the subdivided features.

Accurate Chronic Stress Estimation with Personalized Models Based on Correlation Maximization.

We propose an accurate chronic stress estimation system that utilizes personalized models based on correlation maximization between physiological features and ground truth, which helps determine physiological features effective for the estimation. The personalized models are trained using features respectively found for each individual classes among which the relationships between features and ground truth differ. Which class a new user belongs to can be estimated from the results of a personality questionnaire, as well as by means of conventional methods. W.r.t. evaluation data, with the cooperation of 168 subjects, 599 sets of 1-month wearable-sensor data and ground-truth Perceived Stress Scale (PSS) data were collected, along with the Big Five Personality Traits for each subject. In chronic stress estimation evaluations using this above data, we have confirmed that the proposed classification system achieved 69.1% estimation accuracy in terms of increase/decrease in PSS, as compared to 59.3% and 56.8% achieved, respectively, with two conventional methods, one employing no classification and the other employing k -means clustering.