Sleep is increasingly recognized as a critical component of healthy development. To better understand the impact of daily physical activity on sleep behaviour, new tools are needed. This project aims to develop an algorithm capable of detecting sleep and wake parts using a wearable device. To this end, exploration of state-of-the-art techniques was needed as well as data collection to verify the techniques and develop new algorithms. The research phase consisted of the identification of existing techniques and algorithm performances that had a direct impact on the developing phase. The development phase started obtaining the data for the study. Because of that, a protocol was defined and launched in IMEC. For three weeks, 6 subjects wore a Chill+ watch for 7 days in a row, day and night. The dataset was evaluated with several acceleration and gyroscope data processing techniques to develop an algorithm to accomplish the objective of the project. To check whether these signals can be used to assess the sleep of the subject, self-assessment methods were used. Sensitivity (84.62  19.85 %) and specificity (90.78  14.73 %) of the results proved that our new technique can strongly detect the sleep and wake phases. Actigraphy and gyroscope represent a useful diagnostic tool for the sleep medicine practitioner. The performance of the algorithm is high and makes it suitable to measure sleep. In conclusion, a new method is presented for sleep analysis which is accurate and follows the design criteria guidelines. This algorithm proves its validity by being faster and accurate while improving the detection of sleep phases. Sleep analysis has a wide range of opportunities to discover due to the health importance which it involves. New signals, tools and devices need to be explored to better understand all the aspects which influence the sleep and thus improve its quality.

Curs 2018-2019