Chapter 11 



VOICE COMMUNICATIONS: EFFECTS OF MASKING AND 



DISTORTION 



GEORGE A. MILLER 

 Harvard University 



Carefully selected and well trained per- 

 sonnel may use a highly audible vocabu- 

 lary and adhere to standard procedures, but 

 still be helpless victims of inadequate equip- 

 ment. If the signal arrives weak and dis- 

 torted while noise roars into the listener's 

 ears, vocal communications will be severely 

 crippled, perhaps impossible. 



The principal characteristics of the equip- 

 ment which determine its performance are: 

 (1) the signal level, (2) the noise level, (3) 

 the frequency-response characteristic, and 

 (4) the type and degree of non-linear am- 

 plification. Articulation tests (8, 11, 14) 

 have been used to evaluate the importance 

 of these characteristics for the intelligibility 

 of speech. We will review briefly some of 

 the results of this work, point out occasional 

 applications to problems of military com- 

 munication, and indicate how the subjec- 

 tive methods of evaluation might be pre- 

 dicted from a knowledge of the physical 

 characteristics of the system. 



/. Relation between Intelligibility and Intensity 

 of Speech 



If the listener and talker are connected by 

 an electrical communication system, it is a 

 simple matter to explore the relation of 

 intelhgibility to intensity by varying the 

 amount of amplification or attenuation in 

 the system. Over a wide range of intensi- 

 ties listeners will respond quite uniformly 

 by writing down words or sentences spoken 

 by an announcer. At very low intensities 

 some of the speech sounds become inaudible 

 and some of the test items are missed. At 

 very high intensities the sound becomes pain- 



ful and the listener refuses to cooperate. 

 The determination of these limits tells us the 

 range of speech intensities that is available 

 for vocal communications. 



The most intense vowel sound, the vowel 

 in ought, is just detectable half the time 

 when the average level of the speech is zero 

 db re 0.0002 dj'ne per cm.^, but the faintest 

 sound, the th in thin, does not become au- 

 dible until this average level is raised 25 or 

 30 db (10). The presence of about half the 

 words can be detected when the average 

 level of speech in a free field is about 5 db 

 (17). When the speech is heard via ear- 

 phones, the threshold of detectability for 

 half the words may be about 5 db higher 

 (16). 



The intensity necessary to identify haK 

 the test items correctly depends upon the 

 type of speech material. Nonsense syllables 

 and monosyllabic words give 50 percent ar- 

 ticulation scores at levels between 30 and 

 35 db (6, 12, 17). Polysyllabic words, sen- 

 tences, and continuous discourse give thresh- 

 olds in the range from 20 to 30 db (6, 16, 25). 

 Binaural listening gives thresholds of intel- 

 ligibility about 3 db lower than monaural 

 listening (25). 



At the other end of the scale, listening 

 becomes uncomfortable at about 117 db, 

 begins to tickle the ear at 128 db, and is 

 painful at 138 db. After some experience 

 with these intense levels, however, thresh- 

 olds of discomfort, tickle and pain are raised 

 about 10 db (6). 



The usable range of speech intensities is, 

 therefore, well over 100 db, a power ratio 

 of 101" to 1. 



243 



