 |
|
The intelligibility of air traffic control (ATC) radio speech is subject to degradations owing to the additive noise, narrow bandwidth and limited dynamic range of the legacy analog transmission channel. Noise reduction at the receiver is a well explored research area, but it is difficult to achieve results that effectively are more intelligible. In this thesis a different approach is taken: speech is processed before transmission to make it more intelligible at the far end. Five enhancement algorithms are compared using a simulated ATC channel and subjective and objective measures.
Subjective rhyming test results show that significant intelligibility gains can be obtained at all noise levels. The best system, a multiband dynamic range compressor, offers an overall gain in monosyllabic recognition ratio of 12.6%-points. Improvements are larger for strong noise conditions (with gains up to 17.7%-points); while in low noise there is no or little improvement. The results show that in the tested conditions the technically simplest systems are those that showed the highest intelligibility gains. It is also seen that objective measures are not capable of predicting subjective intelligibility, and listening tests are necessary to evaluate and optimize this type of enhancement systems.
|
|
| |
 |
The full report |
|
(13.5 MB) |
|
|
|
|
|
|
|
|
|
|
|
|
air traffic control (ATC), voice communication, channel enhancement, amplitude compression, analog voice radio, intelligibility, modified rhyming test, speech enhancement, speech intelligibility index, speech processing
|
|
|
|
|
EEC Technical/Scientific Report No. 2009-013
|
|
|
|
