Ubiquitous Gis, Data Acquisition and Speech Recognition

The research asked whether Mobile GIS that incorporated speech recognition was a viable tool for locating defects in pavement, curbs and footpaths. A Xybernaut MA IV wearable computer and Dragon NaturallySpeaking’s speech recognition software were tested, and the Geography Markup Language, GML 2.0, was used to implement an application schema for street condition surveys. Both technical and overall accuracy exceeded 95% over six speech recognition tests in environments that were quiet or constantly loud. However, for three tests while walking along a busy road during which the noise level varied, the accuracy of the speech recognition plummeted to 58%. A “standing” test for capturing the position of defects (N=30) gave an error of 0.41m at the 95% confidence interval. Finally, a web-based questionnaire was answered by 80 GIS project managers, who indicated that they are unhappy with the quality of their data, although they do not require the data in real-time.