Skinput

Background

This research article from Microsoft and Carnegie Mellon discusses the use of the body as a type of input surface, mainly by identifying the location of finger taps on the arm. The goal of this research is to use the surface area in the environment for interaction so there is no need to enlarge the device to create more interaction space. Video can be seen here. Related video from TED

Methods

Skinput makes use of the natural acoustic conduction of the human body, and this particular method had inspirations from non-body acoustic input surfaces as well. When a finger taps the skin there are several distinct forms of acoustic energy. For this expirement they decided to use an armband device that has two arrays of 5 sensing elements.

5 males and 7 females were put into three possible groups for testing. Each group had a different input location set; fingers, whole arm, and forearm. After collecting training data the subjects were given text instructions to tap different locations. Results showed that the average accuracy was 87.6%. This can be increased though by collapsing input locations into groups.

  • Subjects with the highest BMI's had the lowest average accuracies.
  • System never produced a false input when tested while jogging and running.
  • Single handed gestures had an overall accuracy of 89.6%
  • differing types of taps can be identified (tip, tap, knuckle)

Rating

The article was very good at going beyond some of the previous research that has been done in this area. It detailed its results very well and went in depth to explain different ways to go about getting better results in the future. Having different groups for testing gave more range to the results and the data was shown in multiple ways to avoid misrepresentation. Overall the research was conducted extremely well and now they need to expand upon it.