DOO-RE: A dataset of ambient sensors in a meeting room for activity recognition

TL;DR

DOO-RE is the first dataset to support the recognition of both single and group activities in a real meeting room with reliable annotations and is the first dataset to support the recognition of both single and group activities in a real meeting room with reliable annotations.

Abstract

With the advancement of IoT technology, recognizing user activities with machine learning methods is a promising way to provide various smart services to users. High-quality data with privacy protection is essential for deploying such services in the real world. Data streams from surrounding ambient sensors are well suited to the requirement. Existing ambient sensor datasets only support constrained private spaces and those for public spaces have yet to be explored despite growing interest in research on them. To meet this need, we build a dataset collected from a meeting room equipped with ambient sensors. The dataset, DOO-RE, includes data streams from various ambient sensor types such as Sound and Projector. Each sensor data stream is segmented into activity units and multiple annotators provide activity labels through a cross-validation annotation process to improve annotation quality. We finally obtain 9 types of activities. To our best knowledge, DOO-RE is the first dataset to support the recognition of both single and group activities in a real meeting room with reliable annotations.

Figures (7)

• Figure 1: The meeting room testbed layout and installed location of ambient sensors. Stuff related to meetings, such as tables and a projector, are basically arranged. Various types of ambient sensors are installed to track user actions. We attach suitable types of sensors for each area after observing user activities in the meeting room.
• Figure 2: Sensors located in the meeting room testbed. Digi XBee Sensor /L/T/H and Phidget are used for environmental sensors. Phidget and Monnit are leveraged for user-driven sensors. Phidget and API are utilized for actuator-driven sensors.
• Figure 3: An illustration of a value change graph for environment-driven sensors (i.e. Brightness, Temperature, and Humidity sensors) and the 'Presenter Detection' sensor in the file 'Seminar_0.csv'. The figure on the left shows how each environment-driven sensor state can change in one episode. The figure on the right shows that the value of the 'Presenter Detection' sensor increases when there is a presenter in front of the podium.
• Figure 4: Visualization of whether each sensor occurs or not in an activity episode. The figure helps us roughly understand the characteristics of each activity. The darker the color, the more likely a certain sensor appears in an activity episode. For example, on average, 'Projector' appear in all episodes of Lab meeting, but rarely in all episodes of Eating together.
• Figure 5: Value distributions of 'Environmental-driven' sensors and the 'Presenter Detection' sensor. Those sensors produce numerical values as states, thus, they are represented by numerical distributions. Each box plot represents the distribution of values across all episodes of each activity.