"This content is being updated."

In our study, we used smartphones, smartwatches, sleep sensors, and self-recording apps to collect daily life logs and sleep health records of study participants in 2024.

The data collection procedures and methods followed a similar approach to those used in previous studies [1, 2, 3].

Here, we publicly provide the following 12 data items, which comprise a total of 700 days' worth of lifelog data, strictly for non-commercial and academic research purposes only.

- mACStatus: Indicates whether the smartphone is currently being charged.

- mActivity: Value calculated by the Google Activity Recognition API.

- mAmbience: Ambient sound identification labels and their respective probabilities.

- mBle: Bluetooth devices around individual subject.

- mGps: Multiple GPS coordinates measured within a single minute using the smartphone.

- mLight: Ambient light measured by the smartphone.

- mScreenStatus: Indicates whether the smartphone screen is in use.

- mUsageStats: Indicates which apps were used on the smartphone and for how long.

- mWifi: Wifi devices around individual subject.

- wHr: Heart rate readings recorded by the smartwatch.

- wLight: Ambient light measured by the smartwatch.

- wPedo: Step data recorded by the smartwatch.

For the purpose of training a learning model to predict sleep health, fatigue, and stress, the following six metrics were derived from sleep sensor data and self-reported survey records. Each metric consists of values categorized into either two levels (0, 1) or three levels (0, 1, 2), depending on the specific metric. The detailed classification criteria for each metric's levels will be provided in a separate document.

- Q1: Overall sleep quality as perceived by a subject immediately after waking up.

- Q2: Physical fatigue of a subject just before sleep.

- Q3: Stress level experienced by a subject just before sleep.

- S1: Adherence to sleep guidelines for total sleep time (TST).

- S2: Adherence to sleep guidelines for sleep efficiency (SE).

- S3: Adherence to sleep guidelines for sleep onset latency (SOL, or SL).

Acknowledgments

This work was supported by Electronics and Telecommunications Research Institute (ETRI) grant funded by the Korean government. 

([24ZB1100, Core Technology Research for Self-Improving Integrated Artificial Intelligence System], [25ZB1100, Core Technology Research for Self-Improving Integrated Artificial Intelligence System])

References 

[1] ETRI_Lifelog_Dataset_2020. (2021) (https://nanum.etri.re.kr/share/schung/ETRILifelogDataset2020).

[2] Chung, Seungeun, et al. "Real‐world multimodal lifelog dataset for human behavior study." ETRI Journal 44.3 (2022): 426-437.

[3] Oh, Se Won, et al. "Human Understanding AI Paper Challenge 2024--Dataset Design." arXiv preprint arXiv:2403.16509 (2024).