This project aims to create a wearable device which could help the user feel the gaze from nearby people. Once there is a gaze on the user, the device gives directional feedback indicating from where the gaze is originating. After continuous wear, the user’s develop an intuition related to the device feedback.


The idea of the project come from some unpleasant stories. One of my friend was robbed on his way back home. Two men came from his back and rapped his wallet. During Winter holiday, two of my friends come to visit me from Italy. They were also robbed when they stay here.

When it happened once and again, I realized anyone could be potential target of a crime, But if people can feel the criminals’ observation or gaze and realize there is some danger. Some trouble could be avoided.


The function I want to make at first is a crime detection based on computer vision, but how to identify crime arouse a lot debate. So according to Neil Harbisson’s comment  I do not set any function at beginning and focus on stared detection sense itself.


The sensor is camera, which capture the image of surrounding. The raspberry analysis the data, once there is a gaze on the user, the vibrating motor which corresponding the direction where gaze come from will be activated to warn the user.


The device is mostly used during working, and the should is a part can keep steady while working, and neck skin is exposed outside and sensitive to feel the vibration. So the design will focus on this part of body.


Each shoulder is equipped by on camera, I need to make some special design for the camera holders. The development has three iterations, the first iteration has weak curve branches, the second iteration is to short and could drop down when the user really wear it, so I get the final iteration.

Camera feature

  • 5MP sensor, 2592×1944 pixels
  • 1080p video at 30 FPS (or 60 FPS at 720p, 90 FPS at 480p)
  • f2.9 lens, 3.60 mm focal length
  • 120 degrees horizontal, 41.41 degrees vertical field of view (standard and no IR filter versions)
  • Approx. dimensions of circuit and camera: 60 x 11.4 x 5.1mm


Among three method, eye tracking need the camera close to eye and high performance microcomputer. Head pose estimation can detect far place but has low accuracy. But if I let the camera can detect front face it can do the same thing and require a low performance. So I chose face detection as my final method.


The first iteration was like a concept combination, shoulder piece, circuity and camera. The second iteration I begin to think about really garment design.


For the final iteration I thought about fashionable design combine with circuity. The whole device was divided into two parts, heavy part contain raspberry board and power bank. This part was collected in three pockets of the belt. Light part is camera and vibrator. Vibration was sewed in the collar. Between the two parts I used two string connect them the conductive wires and camera cables a in it.


Limited by performance and resolution, the prototype can only detect near place.


Imagine using the device in different environments. Collecting more information and making wiser choice, this is the meaning the project title, keen.


  • A list of vibration code represent different information
  • Using Bluetooth transport data instead of cable
  • Enable the camera detecting surrounding in different environment
  • Several separate ornaments looks like just fashionable ornaments
  • Search for different application


KEEN-Stared Detected Device is a project of IaaC, Institute for Advanced Architecture of Catalonia developed at the Master in Advanced Architecture in 2019 by:
Student: Hongyu Wang
Faculty: Luis Fraguada and Elizabeth Bigger
Physical Computing Expert: Cristian Rizzuti