Handbook for Sound Engineers

654 Chapter 18

loudspeakers for sound reinforcement in highly rever-
berant or reflective environments will be shown.
We will discuss effective communications and define
intelligibility and how to measure it both subjectively
and objectively. We will look at architecture and acous-
tics and at reverberation and its effect on intelligibility
in large public spaces. Finally we’ll look at digitally
steerable column arrays, their design considerations,
and their performance and benefits when used in large
reverberant spaces.
Some of the basic principles involved in voice
communications are:

• In voice communications intelligibility is the capa-
  bility of being understood.


• It assumes the existence of a communication process
  between a talker and a listener, or between a source
  and a listener.


• For the conveyance of meaning, the English language
  is highly dependant upon the effective receipt and
  comprehension of consonants. This is how we differ-
  entiate words based on similar vowels. For example,
  Zoo, Two, New.


• In terms of frequency response, speech ranges
  between 100 Hz and 8 kHz, with maximum energy
  around 250 Hz.


• In speech, the frequency range that conveys the most
  consonant information is the octave around 2 KHz.

18.7.1 What Affects Intelligibility

Major Influences that affect intelligibility are:

• Elocution and pronunciation of the talker. It’s hard to
  understand someone who mumbles under any condi-
  tion.


• Hearing acuity of listener. An often overlooked influ-
  ence, those with a hearing loss have trouble under-
  standing what’s being said.


• SNR. We’ve all been places where it was so noisy we
  couldn’t understand what was being said.


• Direct to reverberant ratio. The higher the reverbera-
  tion level, the more difficult it is to understand what’s
  being said.


• Directivity of the loudspeaker or loudspeakers.
  Highly directional loudspeakers direct more of the
  sound onto the audience and less onto the reflective
  walls and ceilings.


• The number of loudspeakers. Larger numbers of
  loudspeakers translate into more acoustic energy
  being transmitted into the room and higher reverbera-
  tion levels.
  
  
  • Reverberation time. The longer the reverberation time,
    the more likely it will interfere with intelligibility.
  
  
  • Distance of source to listener, The closer the listener
    is to the loudspeaker, the less likely reverberation
    will interfere.
  
  
  
  

Secondary Influences are:

• Gender of talker.


• Microphone technique.


• Vocabulary and context of speech information.


• Direction of main sound to listener and/or direction
  of reflections and echoes.


• System fidelity, equalization, and distortion.


• Uniformity of coverage.

18.7.2 Measuring Intelligibility

18.7.2.1 Subjectively

Statistical tests with trained talkers and listeners can be

the most reliable metric for determining the intelligi-

bility of a system. To ensure that all speech sounds are

represented in a test, Phonemically Balanced (PB) word

lists are commonly used. These word list can be a long

as 1000 words. Tests using nonsense syllables or loga-

toms, and Modified Rhyme Tests are also used. These

tests are very time consuming and are difficult to set up.

18.7.2.2 Objectively

Articulation Index. Articulation Index or AI was one

of the first attempts to quantify intelligibility with

measurements. AI is primarily concerned with the affect

of noise on speech. The index ranges from 0 to 1 with 0

representing no intelligibility.

%ALcons. %ALcons or the articulation loss of conso-

nants was developed by Peutz in Holland during the

1970’s. %ALcons takes both noise and reverberation

into account and is based the importance of the octave

around 2000 Hz in conveying consonant information.

%ALcons uses a scale running downwards from 0

where 0 is perfect intelligibility, or 0% articulation loss.

Although Peutz used 2000 Hz as the center

frequency and 2000 Hz is still the European standard,

many acousticians in the USA prefer using 1000 Hz. As

a general rule, %ALcons calculated at 1000 Hz show a

higher articulation loss than ones calculated at 2000 Hz.

STI. STI or Speech Transmission Index considers the

source/room/listener as a transmission channel and