Handbook for Sound Engineers

Designing for Speech Intelligibility 1393

It must not be forgotten that any noise occurring at
the microphone itself will reduce the perceived
SNR—indeed this is directly additive to the SNR at the
listener’s position. At least 20 dBA should be aimed for
and preferably >25 dBA. A number of techniques can
be employed to achieve this, including:

• Close talking/noise canceling microphones.


• Use of highly directional microphones (e.g., gun
  microphones or adaptive arrays).


• Providing a noise hood or preferably by locating the
  microphone in a suitable quiet room or enclosure.


• Digital, noise canceling and processing can also be
  used in extreme conditions to improve the SNR.

36.7 Reverberation Time and Direct-to-

Reverberant Ratios

Just as noise can mask speech signals so too can exces-

sive reverberation. However, unlike the simpler case of

SNR, the way in which the direct-to-reverberant (D/R)

ratio affects speech intelligibility is not constant but

depends on the room reverberation time, the level of the

reverberant sound field and on the nature of the speech

itself.

The effect is illustrated in Fig. 36-17A–C. The upper

trace is the speech waveform of the word back. The

word starts suddenly with the relatively loud “ba”

sound. This is followed some 300 ms later by the conso-

nant “ck” sound. Typically the “ck” sound will be 20 dB

• 25 dB lower in amplitude than the “ba” sound.

With short reverberation times—e.g., 0.6 s—the “ba”

sound has time to die away before the start of the “ck”

sound. Assuming a 300 ms gap, the “ba” will have

decayed by around 30 dB and will not mask the later

“ck.” However, if the reverberation time increases to 1 s

and if the reverberant level in the room is sufficiently

Figure 36-15. Noise-time history profile of underground

trains entering and leaving station.

Figure 36-16. Noise-time history analysis of soccer game

crowd noise—note short-term variability and peak of

111 dB as compared to average of 82 dB.

A. Broadband linear recording.

B. ”A” weighted time history.

Figure 36-17. Waveform of the word back

1 octave 2000 Hz

C. Temporal amplitude response with effects

if reverberation time is superimposed.

A. Basic waveform.

200 400 600 800Time–ms

Data input–V

0.10

0.08

0.06

0.04

0.02

0.00

0.02

0.04

0.06

0.08

0.10

auto

12.0

0.0

-12.0

-24.0

-36.0

-48.0

-60.0

Magnitude - dB

12.0

0.0

12.0

24.0

36.0

48.0

60.0

Magnitude - dB

B. Temporal amplitude response–dB.

0.00 0.10 0.20 0.30 0.40 0.50 0.60 0.70 0.80 0.90

Time–s

0.00 0.10 0.20 0.30 0.40 0.50 0.60 0.70 0.80 0.90

Time–s

1 octave 2000 Hz

RT=0.6 s

RT=1.0 s