Using Airport Layout Information to Improve Automatic Speech Recognition for Digitizing Taxi Instructions

Abstract

During taxiing, an Electronic Flight Bag (EFB) showing airport and route-data can support pilots. In the present study, the possibility of using a Voice Recognition System (VRS) to enable pilots to enter route instructions, for usage in situations where digital data link is not available, was investigated. The potential of techniques that utilize information on airport layout to detect and correct voice recognition errors was explored. Experiment I investigated the accuracy rates that can be achieved with the basic VRS. Although high, the resulting accuracy was insufficient, for safety reasons as well as user-acceptance of the system. A configurable system that interprets user utterances based on possible taxi routes on the airport, was designed and implemented. Experiment II explored the potential improvement of the different configuration options on error detection and correction performance. With one of the configurations, an increase of 42% in overall recognition rates was achieved.

Manual versus speech input for entering a taxiroute into an Electronic Flight Bag

“…With manual input, the user uses the touchsensitive display to select each next leg, with speech input a Voice Recognition System (VRS) recognizes the spoken entries. In an earlier study (Roefs, Theunissen, Jinkins & Moncur, 2006), a concept was developed and implemented to improve the recognition rates for taxiroutes entered by voice. The concept uses information about the airport layout to constrain the amount of options in the set that the VRS uses to compare the utterances against, and proved capable of significantly improving route recognition.…”