Slope of the performance-intensity function and reaction time for speech in different noise types

Speech recognition was measured in four different noise types for two groups in a project for developing a screening test for the hard of hearing organization in Norway (HLF – Hørselshemmedes Landsforbund). Group 1 consisted of 28 audiology students with normal hearing. Group 2 consisted of 70 persons with varying degrees of hearing loss. Audiometry was also performed for the frequencies of 250, 500, 1000, 2000, 3000, 4000, 6000 and 8000 Hz.

The evaluated noise types were:
1. Stationary speech-shaped noise from the Norwegian Matrix test with high frequencies (HF > 1400 Hz) attenuated by 15 dB.
2. Amplitude-modulated (AM) speech-shaped noise with HF attenuated. The noise was 100% amplitude modulated by a 16-Hz sine tone.
3. Reverse speech babble – four matrix sentences (with silent periods > 200 ms removed) played backwards – with HF attenuated.
4. Stationary speech-shaped noise from the Norwegian Matrix test.

A pilot screening test was developed using Matlab software to measure speech recognition thresholds for words in these four noise types. For each noise type (1-4), 40 test words selected among 10 nouns were used in an adaptive method, where the signal-to-noise ratio was adjusted in ±2 dB steps according to the result of the preceding word. The user responded on a touch screen with symbols and text for each of the nouns. For each participant, the following was measured for each word and noise type: level, selected response, success, time used, estimated threshold.

From these results, a rough estimate of the slope of the performance-intensity function for the different noise types and different hearing thresholds can be obtained. Variations in response time will also be evaluated and presented.