Comparative Evaluation Of A New Method For Helium Speech Unscrambling

Current helium speech unscramblers use a time-domain process which, while fairly effective, produces poor quality speech, is vulnerable to noise-induced errors, and suffers fundamental limitations which reduce its usefulness at extreme depths. A new frequency-domain algorithm has been developed which overcomes all of these problems. This paper presents a comparison of the new algorithm to various time-domain devices. The results of three intelligibility tests are reported, one for the first time. Speech quality is compared via a demonstration tape. The results show that the new unscrambler offers intelligibility equivalent to current processors at moderate depth and clearly superior to them at extreme depths, while giving better speech quality and noise robustness at all depths.