Evaluation of Speech Engines for ATC Simulator

Abstract

Many general speech engines are available that includes Automatic Speech Recognition (ASR) technologies and have been embraced for general applications. The present work is to examine the adoption of ASR technology in Air Traffic Control (ATC) simulation environment. The paper presents a survey and evaluation of these speech engines that can be integrated into ATC simulators. The ATC system is heavily dependent on verbal communication between the Air Traffic Controller and pilots which happens in varied phases between pre-flight to landing. The challenges that are normally faced to bring in the ASR technology to ATC domain are call sign detection, poor input signal quality, problems with ambiguity and use of non-standard phraseology. In this paper, we closely examine the performance of current ASR technologies based on different test scenarios and thus an assessment of current ASR technologies to apply to ATC is presented with quantitative figures and recommendations.