158 Chapter 7
The results of the subjective intelligibility test are
greatly influenced by speech velocity which includes the
number of spoken syllables or words within the articula-
tion time (articulation velocity) and the break time.
Therefore so-called predictor sentences are often used to
precede the words or logatoms that are not part of the
test. These sentences consist of three to four syllables
each, for example: “Mark the word...”, “Please write
down...,” “We’re going to write....” Additionally to
providing a continuous flow of speech, this also serves
for guaranteeing that the evaluation takes place in an
approximately steady-state condition of the room.
There is a close correlation between the subjectively
ascertained syllable intelligibility and room-acoustical
criteria. For example, a long reverberation time reduces
the syllable intelligibility^20 Fig. 7-12 owing to the
occurrence of masking effects, despite an increase in
loudness, see Eq. 7-8.
Quite recently, comprehensive examinations
concerning the frequency dependence of speech-
weighting room-acoustical criteria were conducted in
order to find the influence of spatial sound coloration.^12
It was ascertained that with broadband frequency
weighting between 20 Hz and 20 kHz the definition
measure C 50 (see Section 7.2.2.6) correlates very insuf-
ficiently with the syllable intelligibility. Through a
frequency evaluation across three to four octaves
around a center frequency of 1000 Hz, however, the
influence of the sound coloration can sufficiently be
taken into account. Even better results regarding the
subjective weightings are provided by the frequency
analysis, if the following frequency responses occur,
Fig. 7-13.
As the definition declines with rising frequency due
to sound coloration, the intelligibility of speech is also
low (bad intelligibility o 3). This includes also the
definition responses versus frequency with a maximum
value at 1000 Hz, poor intelligibility o 4 in Fig. 7-13.
The definition responses versus rising frequency tobe aimed at for room-acoustical planning should either
be constant (good intelligibility o 1) or increasing
(very good intelligibility o 2). With regard to auditory
psychology, this result is supported by the importance
for speech intelligibility of the consonants situated in
this higher-frequency range.
The determination of speech intelligibility through
the definition measure C 50 can easily lead to faulty
results as the mathematical integration limit of 50 ms is
not a jump function with regards to intelligibility
without knowledge of the surrounding sound reflection
distribution.
The best correlation with the influence of the spatial
sound coloration exists between the subjective speech
intelligibility and the center time tS (see Section 7.2.2.4)
with a frequency weighting between the octave of
500 Hz to the octave of 4000 Hz. According to Hoff-
meier,^12 the syllable intelligibility V measured at the
point of detection is then calculated as(7-34)
where,
Vsp is the influence factor of the sound source (trained
speaker Vsp= 1, untrained speaker Vsp|0.9),Table 7-4. Correlation between the Intelligibility
Values and the Ratings
Rating Syllable
Intelligibility
VL in %Sentence
Intelligibility
VS in %Word
Intelligibility
VW in %Excellent 90–96 96 94 –96
Good 67–90 95–96 87–94
Satisfactory 48–67 92–95 78–87
Poor 34–48 89–92 67 –78
Unsatisfactory 0–34 0–89 0–67
Figure 7-12. Syllable intelligibility factor as a function of
reverberation time.Figure 7-13. Correlation between attainable intelligibility
and frequency dependence of the definition measure C 50.V=0.96uVspuVSNRuVR