Assessment of speech intelligibility in background noise and reverberation

Reliable methods for assessing speech intelligibility are essential within hearing research, audiology, and related areas. Such methods can be used for obtaining a better understanding of how speech intelligibility is affected by, e.g., various environmental factors or different types of hearing impairment.

In this thesis, two sentence-based tests for speech intelligibility in Danish were developed. The first test is the Conversational Language Understanding Evaluation (CLUE), which is based on the principles of the original American-English Hearing in Noise Test (HINT). The second test is a modified version of CLUE where the speech material and the scoring rules have been reconsidered.

An extensive validation of the modified test was conducted with both normal-hearing and hearing-impaired listeners. The validation showed that the test produces reliable results for both groups of listeners. An important deviation between the two new tests and the original HINT is a new procedure used for equalizing the intelligibility of the speech material during the development process.

This procedure produces more accurately equalized sentences than achieved with the original HINT procedure. This study also investigates a fundamentally different method for assessing speech intelligibility. This method is based on the identification of the stop-consonant [t] in a short test-word. The method was originally developed in order to measure the impact of reverberation on speech intelligibility and, in particular, to measure whether the intelligibility of the test-word depends on the reverberation added to a surrounding speech carrier.

It has been shown that the intelligibility of a reverberant test-word increases when the same amount of reverberation is also added to the carrier. In the literature, this observation has been interpreted as evidence of an extrinsic compensation mechanism for reverberation in the human auditory system.

However, in the present study, it is shown that the listener's perception of the test-word is not only related to the carrier reverberation but also to other of the carrier's acoustic-phonetic properties. The evidence of the extrinsic compensation mechanism is therefore questionable. Overall, the results from the present study may contribute to the development of future speech intelligibility tests in Danish and other languages.

The two developed sentence tests are expected to be useful for assessing speech intelligibility with Danish NH and HI listeners.