UIST 09- Always On Muscle Input

T. Scott Saponas, Daniel Kelly, and Babak A. Parviz (University of Washington, Seattle) and Desney S. Tan (Microsoft Research)

Summary

The focus of this paper is combining interface design principals with prior work on muscle sensing and gesture recognition to create a always available muscle input for real world use. To do this effectively the researchers had to answer an important question about detecting gestures versus relaxation. That is to say, registering a correct gesture at the correct time and not registering a gesture when user is in a transition state. One of the major complicatios of this problem is the EMG technology for sensing small muscle finger movements is not near as fast or accurate as large muscle movements. Prior work has show that these large muscle movements lack the wide gesture set needed for everyday use. The researchers came up with a unique bi-manual solution, building on past discoveries. Their bi-manual approach was to have the user form gestures in his dominant hand and then use his non dominant hand to signal when a gesture is formed. This method allows for the input to be always available without registering incorrect gestures. They used a set of four gestures for their experiment, pinching thumb and index, thumb and middle, etc. The non-dominant hand used only one gesture, squeezing a fist, which is a large muscle movement. In their experiment they tested gesture recognition accuracy over three hand states, free hand, holding a cup and holding a heavy bag. The systems gesture accuracy for each case: 79% free hand, 85% hold a mug, and 88% carrying a weighted bag.

Thoughts

I was very impressed with the user study results. I thought they were very thorough in their testing. The fact that they tested the gestures for the conditions of holding a cup and carrying a bag answered a lot of questions about possibly functionality. I also liked the fact that they included a real world scenario in their experiment. The gesture set is not yet complete enough to be truly useful but it is encouraging that they are focusing on questions of actual usability. I was disappointed by the fact that they don't actually have an EMG arm band like the picture. Instead they used a series of electrodes and the participants had to remain seated next to a large EMG device. Although the technology they used is not at a real-world level, this paper answered a significant amount of important questions with a unique solution.