Self-optimizing voice control user interface

A voice control system converts spoken commands into control actions, a process which is always imperfect due to errors of the speech recognizer. Most speech recognition research is focused on decreasing the recognizers' error rates. In this paper, we present an alternative approach: the development of a strategy that minimizes the number of errors visible on the user interface, given a fixed speech recognizer performance. We introduce a unified view on the recognition act, i.e. the classification of an incoming sound pattern, and derive a self-optimization algorithm for the user interface. With this algorithm, each recognition act is optimized with respect to its own criterion, independently from all other recognition acts. As a result, the voice control system adapts flexibly to the demands of the respective situation.