An investigation of methods of improving the intelligibility of audio-frequency speech in noise

Abstract

A discussion of the nature of speech is presented, followed by a review of speech processing to date, with emphasis on the characteristics of speech which must be retained for intelligibility. Methods of measuring speech intelligibility are described. The relative merits of abrupt and gradual audio clipping of speech are investigated, and two tone and articulation test results are presented showing that there is no significant difference in these methods of clipping with respect to speech intelligibility. Processing of speech to radio frequencies, filtering and retranslat ion to audio to improve the peak to average value ratio of the audio frequency prior to transmitting it through a noisy channel is investigated. Two tone and articulation test results are presented showings, that this processing results in a 20% improvement in speech intelligibility over audio clipping and filtering.