Articulation Index Tutorial

These pages were created to give the reader a general introduction to AI methods created specifically for use in a clinical setting. If the reader would like to read futher on this topic, they are encouraged to see the list of Works Cited on page three.

The Articulation Index (AI) was first described by French and Steinberg (1947) as a way to express the amount of average speech information that is available to patients with various amount of hearing loss. It is usually described as a number between 0 and 1.0 or as a percentage, 0% to 100%. The AI can be calculated by dividing the average speech signal into several bands and obtaining an importance weighting for each band. Based on the amount of information that is audible to a patient in each band and the importance of that band for speech intelligibility, the AI can be computed.

According to Mueller & Killion (1990), one of the primary reasons the AI was not widely used in the clinic was there was not an easy-to-use-and-understand dB HL calculation for the AI. Methods to compute the AI for research purposes involved up to 20 divisions of the speech signal, the dB SPL scale, and the need of a calculator or computer. Obviously, these methods would be of limited use in the clinic. This is because hearing is usually measured at 6-10 frequencies in dB HL and, extra time to perform extensive computations (even on a calculator) are usually not available.

Pavlovic (1988) addressed this problem by creating a simplified method of computing the AI from audiometric information. A speech spectrum was plotted on the dB HL scale with a dynamic range from 20-50 dB HL at 500, 1000, 2000, and 4000 Hz and each band was given an equal importance weighting for speech.
The audiogram on the left shows the speech spectrum with m representing the minimum intensity level of speech and p representing the maximum intensity level of conversational speech. (adapted with permission from Pavlovic, 1988).

Pavlovic (1998) used this speech spectrum to compute the AI using his proposed Ao method. First, the number of dB that are audible at 500, 1000, 2000, & 4000 Hz are added (grey area). This number is then divided by 120 (total dB available) to give the AI. The number of "audible decibles" in the example to the right are 30, 30, 10 and 0, for a total of 70. 70/120 gives an AI of .58 for this hearing loss. An aided AI is computed by adding the insertion gain to the minima and maxima for the speech spectrum at 500, 1000, 2000, and 4000 Hz and recalculating the AI.

Next Page of The Tutorial

Disclaimer
The material on these pages has been composed for viewing audiologists who are solely responsible for the audiological care of their own patients and are strongly urged to maintain close liaison with the ENT and primary care physicians who provide medical care to these individuals in their own community. All queries of a medical nature must be directed to these physicians. The authors and coordinators accepts no direct or implied responsibility for the application of information contained herein.

Audiologyinfo.com Home