 |
People keep asking me about what I did in my diploma thesis. So I decided to write a short post about the story behind "Motion Gestures: false positive prediction and prevention". A lot of you will know motion gestures from the Nintendo Wii or the "shake to shuffle" gesture on Apple's iPhone. By shaking the phone user's can switch in the music player's shuffle mode. Those of you that have the "shake to shuffle" functionality on and quickly descended a stair while reading an email once will know this problem: The functionality triggers accidentally and the player will switch to shuffle mode. This kind of error is called a false positive. When building a motion gesture system for everyday use one will always face the problem of "accidentally triggering functionality".
 |
False positives: left - the intended gesture (moving phone up / down)
right - gesture triggered from up and down movement while walking
|
Finding these errors is not an easy task. Neither for interaction designers nor for trained pattern recognition experts. For both it is hard to tell gestures triggering falsely in everyday life from gestures that don't. In most cases these errors will get noticed very late in development which will increase expenses or in the worst case in the final product where it will cause user dissatisfaction. Long story short. In my thesis I describe how to build a fast data base of movements from everyday life. Designers can check their gestures against this database. The result of this search is a number indicating whether to keep or to discard a gesture because it will false trigger in everyday life. Thats basically it! For a more technical description I will cite the abstract to my thesis:
"False positives are a common problem for interfaces that rely on gesture recognition. Often a gesture can seem fine in development but is found to trigger accidentally during an initial deployment of the interface, restarting development and increasing expense. This works discusses fast methods to predict and prevent the false positive rate of a gesture recognition system. Furthermore it introduces MAGIC 2.0, a technique for false positive prediction and prevention that can be used interactively during the interface design process. To ground my research, I implemented MAGIC 2.0 as a web application and developed a gesture interface using sensors on common Android mobile phone platforms. I use iSAX to enable interactive searching (2 sec/example) of a large database (1,500,000 sec) of everyday user movements on a standard workstation to determine if a candidate gesture will trigger accidentally during use of an interface. Furthermore I introduce a novel algorithm called template SAX to speed up Nearest Neighbour search. In order to evaluate those methods I perform a user-independent study that suggests that the number of matches to this Everyday Gesture Library (EGL) database is indeed predictive of a candidate gesture's suitability. I compare iSAX and tSAX to Hidden Markov Models (HMMs) and Nearest Neighbor with respect to accuracy and speed for the EGL search. Using iSAX on the EGL, I also develop a ``garbage'' class and show that including this class in recognition reduces errors. Last but not least a parameter sensibility study for the iSAX search method is performed. From the results I developed an automatic parameter tuning method for SAX based searches ( iSAX, tSAX)."