Human Activity Recognition with Wrist-worn Sensor Device

Technological fields
Cutting-edge Technologies
  • Wearable sensors
  • Activity recognition
  • Elderly care support
Laboratory organization
NTT Communication Science Laboratories

Download PDF (341KB)


Activity recognition technology is one of the most important technologies for lifelogging and the care of elderly persons. The recognition method proposed in this study recognizes ‘what a user is doing’ by simply using a single wrist-worn sensor device. The device is equipped with a camera, a three-axis accelerometer, a microphone, a digital compass, and an illuminometer. The method recognizes the user's activities by employing objects the user is using, hand movements of the user, and sounds emitted in the activities.


  • We sense user's daily activities by focusing on his/her hand, which is the part of the body most often involved in daily activities.
  • We can recognize various activities with a single device.
  • A camera on the wrist captures images of objects held in the hand. This permits us to recognize activities by using visual information of objects being held.
  • We can preserve the user's privacy because our method recognizes activities by using abstracted image and sound information.

Application scenarios

  • Healthcare and care of elderly persons: Watch over elderly persons at a remote location
  • Lifelogging: Making a continuous record of activities in daily life
  • Context-aware services: Switch services automatically according to the user's activities
  • Context-aware advertising: Recommend ads according to trends in daily activities