Voice activity detection for best signal selection in air traffic management and control systems

The Best Signal Selection (BSS) in air traffic management and control systems has to decide among several signal instances of the same source which one offers the highest speech intelligibility. In these systems, the source signal is not available, thus, objective speech quality tests could not be used. However, information with regards to the speech quality could be obtained from the score of voice activity detection (VAD) algorithms. In this paper the correlation between speech quality and the score of VAD algorithms is analyzed. The results showed that the VAD score-based methods do not saturate for higher SNR, as the Perceptual Evaluation of Speech Quality (PESQ) does. A new VAD algorithm as a solution for the best signal selection problem is also proposed.