A Smart Health (sHealth)-Centric Method toward Estimation of Sleep Deficiency Severity from Wearable Sensor Data Fusion

Sleep deficiency impacts the quality of life and may have serious health consequences in long run. Questionnaire-based subjective assessment sleep has many limitations. On other hand, objective is challenging. In this study, we propose a polysomnography-based mathematical model for computing baseline severity score then investigated estimation using features available only from wearable sensor data including heart rate variability single-channel electroencephalography dataset 500 subjects. We used Monte-Carlo feature selection (MCFS) inter-dependency discovery selecting best removing multi-collinearity. For developing Regression both frequentist Bayesian approaches. An artificial neural network achieved performance RMSE = 5.47 an R-squared value 0.67 estimation. The developed method comparable to conventional methods Functional Outcome Questionnaire Epworth Sleepiness Scale assessing impact apnea on deficiency. Moreover, results pave way reliable interpretable EEG.