Handbook for Sound Engineers

Designing for Speech Intelligibility 1409

announcement system may be required to meet category
“G” or above whereas a high-quality system for a the-
ater for a concert hall might be required to meet cate-
gory “D,” or an assistive hearing system might be
required to meet category “B” or above, etc. (Table
36-5). It is anticipated that the new scale will be adopted
by IEC 268-16 and form part of the fourth edition of the
standard.

Whereas STI does incorporate a degree of diagnostic
ability—e.g., it can readily be determined if the speech
intelligibility is primarily being reduced by noise or
reverberation and the presence of late reflections—a
visual display of the ETC or impulse response is invalu-

able when actually determining appropriate remedial

measures and identifying the underlying cause or

offending reflective surfaces. A combination of tech-

niques is therefore employed by the author when tack-

ling such problems including the use of directional and

polar ETC measurements.

36.14.2.5 SII Speech Intelligibility Index

This relatively new index (ANSI S3.5 1997) is closely

related to the Articulation Index (AI) but also makes use

of some STI concepts. SII calculates the effective

signal-to-noise ratio for a number of frequency bands

related to the speech communication bandwidth. Several

procedures with different frequency domain resolutions

are available. These include conventional^ -octave and

-octave as well as a twenty one band critical band-

width (ERB) analysis. An analysis based on seventeen

equally contributing bands is also incorporated. The

method would appear to be more suitable for direct

communication channels rather than for sound reinforce-

ment and speech announcement systems, but in situa-

tions where reverberation has little or no effect, the

method would be applicable. It should also be useful for

evaluating and quantifying the effectiveness of speech

masking systems.

36.14.3 The Future for Speech Intelligibility

Measurements

As can be seen from the foegoing discussions, we are

still a long way from truly measuring speech intelligi-

bility itself. Currently all we can do is to measure a

number of physical parameters than correlate under

certain conditions to intelligibility. An order of magni-

tude improvement is required for these to become less

anomalous and fallible. The power of the modern PC

should allow more perceptually based measurements to

be made, as is already happening with telephone

networks. However, it must not be forgotten that what is

needed in the field is a simple to operate system that

does not require highly trained staff to operate, as one

thing is for certain: the need to measure and prove that

intelligibility criteria have been met is going to rapidly

expand over the next few years—indeed the introduc-

tion of STIPA is already hastening the process. The

range of applications where such testing will need to be

performed is also going to rapidly expand and will

encompass almost all forms of public transport as well

as all forms of voice-based life safety systems. The

more traditional testing of churches, auditoria, class-

rooms, stadiums, and transportation terminals is also set

Table 36-5. Possible Rating Scheme for Sound
Systems

Cat-

egory

Typical Use Comment

A

B Theaters, speech auditoria,

HOH systems

High speech intelligibility,

C Theaters, speech auditoria,

teleconferencing

High speech intelligibility,

D Lecture theaters, class-

rooms, concert halls, mod-

ern churches

Good speech intelligibil-

ity,

E Concert halls, modern

churches

High-quality PA systems,

F Shopping malls, public

buildings offices, VA sys-

tems

Good quality PA systems

G Shopping malls, public

buildings offices, VA sys-

tems

Target requirement for

VA/PA

H VA & PA systems in diffi-

cult acoustic environments

Lower target for VA/ PA

I VA & PA systems in diffi-

cult spaces

J Not suitable for PA systems

U Not suitable for PA systems

Figure 36-36. New STI scale proposal.

STI

.38 .42 .46 .50 .54 .58 .62 .66 .70 .74

.36 .40 .44 .48 .52 .56 .60 .64 .68 .72 .76

U J IHGFE DCBA A+

(^1) » 3
(^1) » 1