Articulation Index and Speech Intelligibility Index

The articulation index (AI) is a measure of the intelligibility of speech in a continuous noise. The AI was first proposed by French and Steinberg [16] and was extended later by Beranek and Ver [17]. Speech has a dynamic range of about 30 dB in each one‐third octave band from 200 to 6000 Hz, and the long‐term root mean square (rms) overall SPL at the speaker’s lips is about 65 dB. In speech, vowels and consonants are joined together to produce not only words but sounds that have a distinctive personal nature as well. The vowels usually have greater energy than consonants and give the speech its distinctive characteristics. This is because vowels have definite frequency spectra with superimposed short‐duration peaks. The AI ranges from AI = 0 to 1.0 corresponding to 0 and 100% intelligibility, respectively. If the AI is less than about 0.3, speech communication is unsatisfactory (only about 30% of monosyllabic words are understood); while if the AI is greater than about 0.6 or 0.7, speech communication is generally satisfactory (with more than 80% of monosyllabic words understood). Methods to calculate the AI are somewhat complicated and are given in American National Standard (ANSI S3.5–1969) [18] and explained in several books [10, 13]. Since the calculation of AI is complicated, it will not be explained in detail here. In 1997 the ANSI S3.5–1969 standard was updated and then further revised in 2002, 2007, and 2012 [19]. In this later standard, the AI has been renamed as the speech intelligibility index (SII). As before, the SII is calculated from acoustical measurements of speech and noise. The reader is referred to the new standard for the changes from AI to SII and complete details of the calculation of this index [19]. Because of the complication in the calculation of AI and SII, many favor the use of the speech interference level (SIL), which is easier to calculate. SIL is described in the next section of this chapter.