EPA-600/1-77-025
May 1977
Environmental Health Effects Research Series
SPEECH LEVELS IN VARIOUS NOISE
ENVIRONMENTS
Office of Health and Ecological Effects
Office of Research and Development
U.S. Environmental Protection Agency
Washington, D.C. 20460
-------
RESEARCH REPORTING SERIES
Research reports of the Office of Research and Development, U.S. Environmental
Protection Agency, have been grouped into nine series. These nine broad cate-
gories were established to facilitate further development and application of en-
vironmental technology. Elimination of traditional grouping was consciously
planned to foster technology transfer and a maximum interface in related fields.
The nine series are:
1. Environmental Health Effects Research
2. Environmental Protection Technology
3. Ecological Research
4. Environmental Monitoring
5. Socioeconomic Environmental Studies
6. Scientific and Technical Assessment Reports (STAR)
7. Interagency Energy-Environment Research and Development
8. "Special" Reports
9. Miscellaneous Reports
This report has been assigned to the ENVIRONMENTAL HEALTH EFFECTS RE-
SEARCH series, This series describes projects and studies relating to the toler-
ances of man for unhealthful substances or conditions. This work is generally
assessed from a medical viewpoint, including physiological or psychological
studies. In addition to toxicology and other medical specialities, study areas in-
clude biomedical instrumentation and health research techniques utilizing ani-
mals — but always with intended application to human health measures.
This document is available to the public through the National Technical Informa-
tion Service, Springfield, Virginia 22161.
-------
TECHNICAL REPORT DATA
rratl tutiniciions t»i itir ?<•> < rsc »ct< A- c
1 REPORT NO
EPA-6QO/1-77-025
4. TiTL€ ANDSU8TITLE
Speech Levels in Various Noise Environments
3. RECIPIENT'S ACCES5I Of»NO.
5. REPORT OATi
May 1977 issuing date
S. PERFORMING ORGANISATION CODE
7. AUTHQFUSI
Karl S. Pearsons, Ricarda L. Bennett, Sanford Fidel!
8, PERFORMING ORGANIZATION REPORT NO
9. PERFORMING ORGANIZATION NAME AND ADDRESS
Bolt, Beranek and Newman. Inc.
Canoge Park, California 91305
10. PROGRAM ELEMENT NO.
1GA085
1 », CONTRACT/GRANT NO.
68-01-2466
12KSPONSQRING AGENCY NAME AND ADDRESS
Office of Health and Ecological Effects - wasn.,
Office of Research and Development
U.S.. Environmental Protection Agency
Washington, D.C. 20460
13. TYPi Of REPORT ANC PERIOD CCVERSQ
14, SPONSORING AGENCY CODE
EPA/600/18
15. SUPPLEMENTARY NOTES
16. ABSTRACT
the goal of this study was to determine average speech levels used by people when
conversing in different levels of background noise. The non-laboratory environ-
ments where speech was recorded were: high school classrooms,..homes,. hospitals"*
department stores, trains and commercial a1rcraft.
-------
-------
EPA-600/1-77-025
May 1977
SPEECH LEVELS IN VARIOUS
. 'NOISE ENVIRONMENTS .
by
Karl S. Pearsons
Ricarda L. Bennett
Sanford Pidell
Bolt, Beranek and Newman, Inc.
Canoge Park, California 91305
Contract No. 6S-01-2466
'Project Officer
George R. Simon
Health Effects Division
Office of Health and Ecological Effects
Washington, DC 20460
OFFICE OF HEALTH AND ECOLOGICAL EFFECTS
OFFICE OF RESEARCH AND DEVELOPMENT
U.S. ENVIRONMENTAL PROTECTION AGENCY
' WASHINGTON, DC 20^60
ib
-------
DISCLAIMER
This report has been reviewed by the Health Effects Division,
U.S. Environmental Protection Agency, and approved for publication.
Approval for publication does not signify that the contents necessarily
reflect the views and policies of the U.S. Environiriental Frotection Agency,
nor does mention of trade names or commercial products constitute
endorsement or recommendation for use.
-------
ABSTRACT
Research on speech level measurements was conducted under
laboratory and non-laboratory conditions. The 'goal of this
study was to determine average speech levels used by people
when conversing in different levels of background noise.
The non-laboratory or real-life environments where speech was-
recorded were: high school classrooms, homes, hospitals, de-
partment stores, trains and commercial aircraft. Briefly, the
results of speech measurements at schools confirmed that
teachers in typical classroom situations speak at a consistently
higher level (67-78 dB at one meter) than in face-to-face
conversation. Further, their vocal effort increased at the
rate of 1 dB/dB increase in background noise which ranged from
i<5 to 55 dB.
The speech levels recorded in face-tc-face conversation were
lower, averaging 55 dB at 1 meter for ambient levels less than
^8 dB, But, as the background level increased above *I8 dB to
70 dB, people correspondingly raised their voice levels up to
67 dB at the rate of 0.6 dB/dB as the ambient increased. It
was also noted that for background levels less than ^5 dB,
speech levels measured at the listener's ear - disregarding
distance between talkers - was also 55 dB.
The laboratory portion of the study was conducted in an anechoic
chamber. The analysis of approximately 100 observers for four
varied speech instructions ("Speak in a normal, raised, loud,
and shout voice") showed an orderly progression in level, and
iii
-------
shift in spectral emphasis as voice levels increased. A
comparison of male and female voice levels for the speech
categories normal and raised yielded minimal differences, .
thus negating conclusions by other researchers that background
levels should be lowered to accommodate female speech.
This report concludes with recommended background levels to
achieve speech, intelligibility for the various environments
investigated in this study, - - -.
iv
-------
TABLE OP CONTENTS
ABSTRACTS . lii
FIGURES ., • vi
ACKNOWIEDGIMIMT viil
I. INTRODUCTION -'.. . 1
II. CONCLUSIONS AND KECOMMENTATIONS*...,...,., 3
A. Conclusions 3
B. Recommendations.....,. • - 5
III.
BACKGROUND -...- .-. . 9
IV. RECORDING PROCEDURE ................................... 12
A. Classroom. ...... ................................... ' 12
B. Individual Face-to-Face - Conraunication ....... ........ 14
C. Anechoic Chamber Measurements . ...... . . ............. 15
V. ANALYSIS PROCEDURE ..... . ..................... ......... .16
VI. MEASUREMEM1 SHE DESCRIPTION ........... ., ____ ,.,.. ----- 17
A. Schools........ ....... . ........ . .......... ......... 17
B. Hones..... ................. . ......... . ....... .....-.•.. 17
C. Hospitals... ....... . ............................... • 17
D. Public Places. ....... .............................. 18
E. Transportation Vehicles . . ........ . ........... ....... 18
VII. RESULTS...... ........ . .............................. '... 19
A. Schools ... ....... . . .................... ..... ..... ... 19
B. Homes....................... ........ . ........ . ...... 23
C.' Hospitals. ....... . ..................... '....... ..... ..... 29
D. Public Places .............................. . ........ • 29
E. Transportation Vehicles ...... . ...................... 29
F. 'Anechoic Chamber ....... ' .......... .... ...... . ........ 29
.VIII. DISCUSSION
A. Lecturing in Schools . . ...... ......... ............. . . 43
B. Conversing in Various Environments, . . ............. . . 45
• C, Speech Measurements in an Anechoic Chamber. . ............ _ 54
REFERENCES - ,'60
APPENDICES 61
-------
LIST OF FIGURES
1. Microphone Locations in Classroom'
• 2. Distribution of Speech Levels in School Produced by
Teachers on Several Occasions
3, 'Speech Levels at 1 Meter Produced by Teachers During
Lectures
4. Speech Intelligibility In Classroom
5. • Distribution of Speech Levels at' Homes Produced by People
on .Several Occasions
6. Conversational Speech Levels in.Homes Normalized to 1 Meter
7. Distribution of Television Speech Levels at Various Distances.
8. Speech Levels Heard from TV as a Function of Background
Noise
9. Distribution of Speech' Levels in.Hospitals Produced by
Nurses and Patients on Several Occasions
10. Distribution of Speech Levels In Department. Stores Produced by
People on Several Occasions
11. Distribution of Speech Levels in Trains and Aircraft Produced
by People'on Several Occasions
12. Age Distribution of Subjects Employed for Anechoic Chamber
Speech Measurements
13- .Distribution of Speech Levels Produced by Males at Five
Vocal Efforts
14. Distribution of Speech Levels Produced by Females at Five'
Vocal Efforts
15* Distribution of Speech Levels Produced by Children at Five
Vocal Efforts
16. -Average Speech Spectra for Males at Five Vocal Efforts
vl
-------
LIST OF FIGURES (CONT'D.)
17. Average Speech Spectra for Females at Five Vocal Efforts
18. Average Speech Spectra for Children at Five Vocal Efforts
19- Differences Between Overall and A-Weighted Sound Pressure
Levels of Speech
20. Conversing Speech Level as a Function of Background Noise
in Several Environments
21, Conversing Speech Level Normalized to One Meter as a
Function of Background Noise in Several Environments
22. Interpersonal Distance Observed for Conversations as a
Function of Background Level in Various Environments
23. Articulation Indexes for Conversations in Various
Environments
24. Sentence Intelligibility for Conversations in.Various
Environments
25. Speech Levels for Various Vocal Efforts
A-l Equipment for Measuring Speech and Background Levels in
Classrooms
A-2 Microphone Locations in Classroom
A-3 Equipment Setup for Calibration of Miniature Microphone Used
in Conversational Speech Recording
A-<4 Speech and Background Analysis System '
vii
-------
ACKNOWLEDGEMENTS
The authors thank Suyeo Tonooka for his diligent assistance
in data collection- and analysis, Thanks also go to Brian
Curtis for his assistance in recording speech and background
levels and data analysis.
vlii
-------
I. INTRODUCTION
Speech communication, although an essential aspect of daily
life, is often degraded by the masking effects of background
noise. EPA has identified various noise levels Intended to
guarantee adequate speech communication. These noise levels
were identified on the basis of existing knowledge, rather
than specific research programs. The current research was
undertaken to provide a firmer basis for specifications
of noise levels that Insure adequate speech communication in
a variety of real world settings.
To provide information for specifying the noise level in
environments where speech communication may take place, one
needs to know most crucially the distance over which people
choose to communicate, and the speech levels at which people
normally converse. Secondary factors may influence speech
intelligibility as well, notably familiarity of the talker
and listener with the language, the hearing acuity of the
listener, visual cues, the amount of redundancy in the speech
material, and reverberant characteristics of the acoustical
environment. However these secondary factors remained fairly
constant for a given speech measurement situation.
The distance between the talker and listener is important
primarily when the conversation takes place in an outdoor envi-
ronment, in which speech levels are typically reduced 6 dB for
every doubling cf distance of separation between the talker
and listener. Indoors, particularly in home environments with
relatively small rooms, the distance between the talker and
listener is not as critical, since speech levels do not
-------
decrease In the same fashion as In the outdoor environment.
This is particularly true for distances greater than 1 meter,
since at these distances the listener is in the reverberant
field and speech levels usually remain nearly constant with
distance.
Thus, the principal factor that determines the adequacy of
speech communication is the level produced by the talker.
Most measurements of speech levels have been taken In very
quiet environments (such as anechoic chambers), with a talker
Instructed to read from prepared text or word lists. Brown
et al. (1976} have recently shown that even these data' can
be highly variable. Since it is Important to determine speech
Intelligibility in environments other than laboratories,
direct measurements of actual speech levels normally employed
in environments are needed.
The study reported here provides measures of typical speech
levels"in homes, schools, hospitals, public places, and trains,
and airplanes. To supplement this information and to make
available detailed information on speech spectra, measurements
were also made of speech levels In an anechoic chamber. Tabu-
lations of one-third octave band statistical distributions of
the speech levels for the anechoic measurements are provided
in the data supplement of this report.
-------
ii. CONCLUSIONS' AND RECOMMENDATIONS
A. Conclusions
Several conclusions may be drawn from, the data collected and
analyzed in this project.
1) Schools
In the classroom ( lecturing) environment the teachers' speech
levels increased at the rate of 1 dB/dp increase in background
noise For ambient levels of ^5 to 55 dB. The teachers' speech
levels at 1 meter ranged from 6? to 78. dB.
2) Face-to-Face Communication
a) FOF background levels less than 45 dB3 levels
measured at the listener^ ear averaged 55 dB.
b) For background levels less than to 8 dB people main-
tained an average voice level -of 55 dB when the effects of
distance were normalized to 1 meter.
c) For background levels above 48, up to 70 dB, people
began to raise their-voice levels, up to an average of 6? dE,
at the rate of 0.6 dB/dB increase in the ambient level. The speech
levels were normalized tp 1 meter,
d) Distances at which people communicate steadily .
decrease with increase-d background level. In ambient levels up
-------
to ^5 dB, such as those found In the homes, communication-
distances averaged 1 meter. For higher background levels
(above 70 dB) this communication distance decreased to an
average 0,4 meters.
e) High sentence Intelligibility of virtually
can be easily achieved when the speech to background ratio is
at least 10 dB. According to the results of this study, this
ratio or better can be maintained with a background level
below ^5 dB.
f) Sentence Intelligibility of $%%, according to
this study, Is possible with a zero speech to background ratio
in an ambient of 70 dB.
3) Anechoic Chamber'-'
The results of the laboratory study indicated that vocal
emphasis shifted from the low frequencies to the high fre-
quencies as the speech categories went from normal to shout.
This trend Is evidenced by a 1.6 octave shift in the maximum
one-third octave band from 500 Hz found in the normal voice
spectrum, to 1600 Hz in the shout spectrum.
-------
B. Recommendations
1 } Schools
The Environmental Protection Agency in the "Levels Document"
(19.^7) recommended an indoor level for classrooms not to
exceed L 0f?iiV°^ ^5 dB. This criterion was based on the
.consideration of providing an educational environment with a
minimum of speech interference activities. The results from
this study described in this report revealed that the average,
background level for occupied classrooms, with no talking during
a test, was an L of ^5 dB. However, the far more typical
classroom environment consisted of some student-teacher, or
interstudent communication. The ambient level during the normal
classroom activity was 50 dB, Therefore, it is .recommended that
for an occupied classroom, the background level could be 50 dB
which would provide 991 sentence intelligibility.
2) Face-to-Face Communication
a) Homes
The ^5 dB background level measured in this study for the indoor
residential environment agrees with the recommended criterion in
the EPA "Levels Document" (19?4). The EPA recommended an indoor
L-dn of 45 dB .for speech communication. This would permit vir-
tually 100$ sentence intelligibility. The recommended outdoor
L-, level was set at 55 dB which again corresponded to the
average ambient found in this study for both urban and suburban
environments. This level would permit an average sentence intelli-
gibility of 98% at 1 meter.
-------
b) Hospi tals
The hospital Interior level set by the "Levels. Document" (197*1)
.was L of--- ^5 dB. This criterion was based on minimizing acti-
vity interference and annoyance. The EPA .document, however,
failed to stipulate whether this level .was based, on measurements
taken in the patients' ro'oms, or the operating rooms, "or the
nurses' stations. The background measurements made for this ,
study yielded a background of 45 dB for the patients', rooms,
but 52 dB for noise measurements taken at the nurses' stations'.-
But, even with the higher background levels of 5'2 dB and a
resulting decrease in the speech to background ratio, the level
was such as to allow 99? sentence' intelligibility at 1 -meter.
c) Department Stores
Background levels in public places, such as department stores,
were higher than the indoor levels in homes or hospitals. How-
ever, people raised their voice levels to maintain an adequate
speech to noise ratio for communication. For such commercial
places as retail stores, restaurants and general office envi-
ronments,, a background of 55 dB (EPA 1974,, • Table D-1Q) is an
average level recommended by architects and noise control
engineers as ah acceptable noise background. The sample taken -
in department stores for this study agrees with the 55 dB level
and will provide- a communication environment to enable art
average 98$ sentence intelligibility at 1 meter.
d} Transportation Vehicles " ,
The noise exposure levels in the trains and airplanes .afforded
-------
less than desirable sentence Intelligibility (<95?)> The
EPA "Levels Document" (197*0 recommended background levels rio
higher than L = 70 dB over a 21! hour period in order to
protect against hearing loss. The ambient levels measured in
the current study, averaged 77 dB, presenting a danger of po-
tential hearing loss and most certainly impeding communication.
It is therefore suggested that a background level of 70 dB
be viewed as a goal for speech communication in both trains
and airplanes. This level would permit sentence intelligibility
of approximately 95^ at about 0.5 meter.
3) Future Research
It is further recommended that the Articulation Index calcu-
lation procedure (ANSI, 1969) should be reviewed to incorporate
the new'speech spectra information obtained in the anechoic
chamber laboratory study. Additional changes in the standard
would be the Inclusion of the new data on differences between
peak and long term rms speech levels.
The data collected in the present study was from participants
with normal hearing. The speech levels that other segments of
the populous use for communication in various environments might
bear some investigation. The elderly, or the hard of hearing at
all ages use public transportation, and in order to facilitate
proper usage of a transportation system such as a commuter train,
it is vital they be able to communicate adequately. Also it would
be important to determine their speech levels in residential
settings or public environments (such as hospitals or office
buildings) to aid in the development of speech privacy criteria.
-------
4} Speech Intelligibility and Annoyance
m
The results in this study have1been interpreted in 'terras-of
speech level (L )> with emphasis on the influence of back-
ground noise upon the"speech as translated by the Articulation
Index and Correlated with a percentage of sentence intelli-
gibility. However, no attempt was made to qualify sentence
intelligibility^ with a subjective evaluation of the background
level. Thus, a person might be able to communicate at 98$
sentence intelligibility but be very annoyed with the kind o'f
background noise or the.ambient level. Such a qualification of
the ambient level might be helpful in analyzing the difference
between 95% sentence,intelligibility and 99$ intelligibility.
Future research should concentrate on determining a relationship
between the Articulation Index and sentence Intelligibility and
the subjective evaluation of noise.
-------
III. BACKGROUND
At first glance, the literature on speech levels seems
reasonably complete. Early studies such as Dunn and White
(19^0) established nominal levels observed under controlled
conditions. Subsequent studies, such as those of French
and Steinberg (191!?) and Benson and Hlrsch (1953) replicated
the early findings with greater numbers of measurements taken
under somewhat wider conditions.
Prom these studies come much of the data still considered as
"standard" values of speech levels. For example, the widely
accepted approximation of 60-65 dB (long term rras overall
sound pressure level) at one meter for the level of a male
talker reading prepared text aloud with normal vocal effort
dates from these studies. Beranek's early (19^7) work on
speech communication, from which later measures such as SIL
and PSIL are derived, also is based on these studies.
By the mid-1950's, some of the limitations of the early work
had been recognized. The first deficiency of the data was
that it was taken under quiet conditions. Normal conversations
are not conducted exclusively in quiet background noise envi-
ronments; people converse in noisy places as well. Thus,
studies such as that of Korn (1954) were undertaken to quantify
the relationship between the background noise in which speech
is conducted with actual speech levels. Korn found that speech
levels varied by 17.5 dB over a range of 50 d3 in background
noise. He concluded that the best estimate of the rate of
increase of sDeech levels with background noise levels i«;as
-------
0,38 dB/dB, Kern's study had several methodological flaws,
however^ which encouraged further research in the area.
Subsequent studies, such as those of Pickett (1958), Webster'
and Klumpp (1963), and Gardner (1966) have produced o-ther
estimates of the so-called "Lombard effect" (the tendency
to raise the voice as the background noise increases) (Lombard,
1911).
A second deficiency was the absence of any real information
on speech levels that people produce when not in controlled
listening conditions. Under what conditions do people vary
their vocal 'effort from a whisper to. a shout? Are-certain
speech levels characteristic of certain social settings and
background noise environments? How does speech intelligibility
vary in these circumstances?
Thus, knowledge of speech levels was not wholly adequate
several years ago, when 'EPA sought to identify noise "levels
requisite to protect the public health and welfare with an
adequate•margin of safety". EPA based many of its recommenda-
tions for these health and welfare levels on speech'interference
effects, reasoning'tnat speech interference was the most
salient effect of nolst exposure less Intense than that asso-
ciated with hearing damage, yet more intense than that asso-
ciated with sleep interference. The basic phenomena of speech
interference seemed well understood: speech spectra were well
known; there was general if not detailed agreement on levels
observed in controlled conditions; and several measures of
speech intelligibility were well developed and in general use.
10
-------
Upon closer examination, however, the gaps -in the literature •'
noted above became apparent, The most important lack was
that of information"on the statistics of distributions'of
speech levels encountered in the real world} beyond laboratory'
walls. The present research project was -undertaken to prp^
vide more Information on speech levels and thus to create
a firmer bagis for environmental noise criteria necessary for
conversational speech,
11
-------
IV. RECORDING PROCEDURE
Two different procedures were used to make recordings of
speech levels. The first .procedure was used for school
classroom measurements, while the second was used in all
other situations for measuring personal (face-to-face)
communication. The main difference between the two proce-
dures was in the number and placement of the microphones.
The classroom situation used three microphones: two placed
at different distances from the teacher, and one (a lavalier
microphone) worn by the teacher. The second procedure
utilised a miniature tape recorder and a single microphone
located at the listener's ear while conversation was taking
place.
A. Classroom
Typical microphone placement used in the classroom situation
is indicated in Figure 1. In general, Position A measured
speech and background levels near the front of the class
approximately 2 meters from the teacher, while Position B
(approximately 7 meters from the teacher) was used to record
speech and background information at the rear of the.class.
All microphones including the one worn by the teacher were
connected to a multi-channel tape recorder by long cords.
This arrangement allowed the teacher normal freedom of move-
ment about the classroom. The speech levels recorded with
the microphone worn by the teacher were converted to equivalent
levels, i.e., those that would have been measured one meter
from the teacher's lips. Both teacher and students were
encouraged to carry on the normal classroom procedures which
12
-------
TEACHER'S MICROPHONE
(Worn by Teacher)
Teach
Desk
er's
Blackboard
Podium
POS.B
FIGURE 1. MICROPHONE LOCATIONS IN CLASSROOM
13
-------
Included,!) lecturing by the teacher^ 2) question and answer
interactions between teacher and class, aftd 3) study
situations with no.speech communication between teacher and
class. Data reduction concentrated on speech levels collected
during the teachers* lectures. Further'detail on the instru-
mentation employed-.for all measurements is included in
Appendix 'A*, . • ' . "
A speech intelligibility test was given to the students In
the classroom using phonetically balanced word lists. One
of seven 5Q~word lists was read by the teacher, who was'
asked to read them lh a customary classroom lecturing voice
to the students,. The students were asked to write, the word
heard oh test sheets, The word lists were read in a fixed
cadence, with no repeats. Complete Instructions and word
lists are included in Appendix B.
B. Individual FaCeito-kF'ace Coirtmunication
The procedure employed for all situations, other than schools,
was to record normal conversation at fixed, distances using a
single microphone mounted near the ear on an eyeglass frame '
worn by a listener. Background measurements were, made using
the same equipment, but Without conversation between the
participants. Several recordings were made to obtain at least 10
seconds of continuous conversation of• the-talker alone without
responses .from the listener.
Because of the. microphone Ib&atlon (very- near the head)-, it
was expected that•the speech levels recorded were somewhat higher
than would-have been observed if the microphone had been
-------
placed away from the head. However, the measured speech
levels were representative of those heard by the listener,
and therefore provided reasonable levels for estimating
the listener's intelligibility.
In all cases in the home measurement situations, there was
no difficulty in conversing. This appeared to be true for
most of the speech measurement environments except in the
transportation vehicles where there was some difficulty in
understanding speech. -Initially speech measurements were
made at the distance of one meter between the talker and
the listener. However, in later measurement sessions, this
restriction was relaxed, yet people seemed to voluntarily
select this one meter communication distance, at least in
the home environment. For transportation environments, this
distance diminished to about 0.5 meter.
C. Anechoic Chamber Measurements
Measurements were made in an anechoic chamber one meter from
the talker to determine speech spectra for men, women and
children. The subjects were asked to repeat from memory
the phrase "Joe took father's shoe bench out; she was waiting
at my lawn;" for approximately 10 seconds at different vocal
efforts. The stipulated vocal efforts were labeled Normal,
Raised* Loudf and Shout, Complete instructions are reproduced
in Appendix B. In addition, a brief conversation between the
experimenter and the subject was carried on before the formal
test.began; this speech was labeled casual conversation.
During the casual conversation phase, the experimenter stood
near the microphone at the one meter distance.
15
-------
V. ANALYSIS PROCEDURE
The analysis was conducted with a one-third octave band real
time analyzer interfaced to a digital computer, as discussed
in Appendix A. All speech samples were at least 10 seconds
in length, which allowed at least 100 samples to be taken
at 0.1 sec. intervals. , The spectrum analyzer's integration
time was equivalent to "fast" on a sound level meter. Back-
ground noise analyses, were completed in a similar fashion.
All speech level and background data are reported,in A-
weighted sound pressure levels unless otherwise noted.. All
^speech levels were corrected as ne.cessary to account for
possible background noise influence. To provide levels of
vocal output at a constant distance, the speech levels were
normalized to equivalent levels.at one meter. -
16
-------
VI. MEASUREMENT SITE DESCRIPTION
A. Schools
Measurements were made in two schools. One was located on
a moderately busy street while the other was situated on a
quiet street under the landing path for Los Angeles Inter-
national Airport. Since the noise from aircraft- was lower
than expected and since -no background noise measurements
Included aircraft noise,.the schools are referred to as I
and II to avoid misinterpretation. Measurements were made
in a total of 20 classrooms. Classrooms typically were
occupied by 23 students. Windows in the classrooms were
usually closed during the normal classroom activities.
Homes
Speech background measurements were made both inside and
outside 25 homes. Some of the homes were located on quiet
suburban streets and others were situated in areas of high
traffic noise exposure. None of the homes were located under
an airport landing path. Outdoor measurements were made in
the backyard or patio areas not directly facing the street.
C. Hospitals
Measurements were made at 23 hospital locations in four medium
sized hospitals. Speech and background measurements were made
while conversing with patients in their rooms, and also while
talking with on-duty nurses at nurses' stations.
17
-------
D. Public Places
Speech background noise measurements were made at 19 loca-
tions in 7 large department stores while talking with on-
duty sales personnel.
E. Transportation Vehicles
Recordings of speech and background levels were made while
conversing with 11 passengers on the Bay Area Rapid Transit
System (BART) in San Francisco. Speech level recordings of
12 passengers in 5 different commercial aircraft were also
made. The measurement of speech and background levels for
each passenger was made while the plane was cruising at
its normal speed and altitude. Aircraft included Boeing
70?s, Boeing 72?s, Douglas DC-9s, a Lockheed L-1011, and a
Lockheed Electra,
18
-------
VH. RESULTS
A. Schools
A summary of the speech levels measured in the schools Is
shown in histogram form in Figure 2. The'speech and back-
ground levels are given in A-weighted sound pressure level
which was used exclusively in this report unless otherwise
noted. The figure summarizes levels measured in the twenty
classrooms at the two different schools, as well as at the
three different microphone locations in the classroom repre-
sented by Positions A and B, and'the teacherrs microphone.
The histograms indicate considerable variation in speech
levels measured in different classrooms. The speech levels
at school II for all microphone locations were higher on the
average by 5 dB than those found in school I. Higher back-
ground levels (average 3 dB) were also noted for school II
over school I, An analysis of the speech to background noise
ratio for all microphone locations revealed that the teachers
at both schools maintained about the same ratio. The average .
speech level was 15 dB higher than the background for school I
and 16 dB for school II.
Figure 3 summarizes all of the teachers' speech levels measured
with the teacher's microphone and normalized to one meter from
the teacher's lips. The results indicate that the teachers'
speech level In the range of 67 to 78 dB In the classroom in-
creased at the same rate (1 dB/dB) as the background noise, over
a range of 45 to 55 dB.
Figure 4 display's the results of the speech intelligibility tests
administered in the classrooms. Articulation Indices (AI) based
upon samples of the teacher's speech during class lectures were
19
-------
IV)
o
Schooll
One Meter From Teacher
10
0)
Q_
E
0
o 5
_Q
E
* 0
—
1 Me
n — <
55 60 65
;a n
ji
i i
70 75
10
5
_l 0
-
-
Position A
n
i
i 1 Mean
80 50
, ,1
55 60
,
Position B
10
5
I 0
—
Mean
. ii i i i
65 70 75 45 50 55 60 65 70
Leq of Speech in dB
One Meter From Teacher
10
a.
E
D
"- 5
ID
-Q
E
D
Z
0
—
-
il 1
1
Mean
T '
10
5
J 0
—
-
School
n
Position A
n
i
r~
Mean
,|
Position B
10
^5
. . .
-
-
r~i
i i Mean 1
P-i ,1 , h
55 60 65 70 75 80
50 55 60 65 70 75
Leq of Speech in dB
45 50 55 60 65 70 75
FIGURE 2. DISTRIBUTION OF SPEECH LEVELS IN SCHOOL PRODUCED BY TEACHERS
ON SEVERAL OCCASIONS
-------
80
75
CO
T3
70
V
Q_
o
CD
65
60
55
35
r = .76
• School I
• School II
I
40 45 50 55
Leq of Background noise in dB
60
FIGURE 3. SPEECH LEVELS AT 1 METER PRODUCED BY
TEACHERS DURING LECTURES
21
-------
y Understood
s 00 C
> o c
*• OU
0)
i.
o
U
in
"O
1 40
u-
0
*->
4)
U
°~ 20
n
X
/
/
/
/
/
A
Note: Th
an
/
/
• 1000
ese re
ey de
d skil
/
/
_*f
X
x^
o School I
A School JJ
PB W«
lotions
send u
of to
Drds
are a
pon ty
Ikers <
pprox?
pe of
^nd lis
<"
4
a
mate.
mater
teners
a
*
a
A
fcl _
0.2
0.4 0.6
Articulation Index
0.8
1.0
FIGURE 4. SPEECH INTELLIGIBILITY IN CLASSROOM
22
-------
calculated and compared In Figure ^ to the percentage of
correctly understood phonetically balanced words recited to
the students during the intelligibility test. AT scores
represent the percentage of speech material available to the
listener; i.e., that which is not masked by -background noise.
The AI calculation uses the differences in one-third octave
band levels between the speech and the background noise. This
result is then weighted according to a procedure specified by
ANSI (1969).
he results for school I are in good agreement with the
relationship of percent correct versus Articulation Index given
in the Articulation Index Calculation Standard (ANSI, 1969),
represented by the curve in Figure 4, The average percent 'of
words correct for school I is 91$. However, the results for
school II were in minimal agreement with this curve and the
average percent correct was only 11%.
B, Homes
Figure 5 shows the results of speech level measurements made
in the homes. Speech samples were recorded both inside and
outside homes which were located in suburban and urban areas.
As indicated by the histograms, the average difference between
the speech levels recorded inside the homes in the suburban
or urban areas was 2 dB; whereas the difference in the observers'
speech levels recorded outside the homes for the same areas was
10 dB. The higher speech levels were associated with the mea-
surements in the urban areas .
As anticipated, higher background levels were' found both inside
and outside the homes in the urban areas . The average noise
exposure level in the urban areas was 55 dB. This was 10 dB
23
-------
INSIDE
OUTSIDE
ro
-Cr
CL
E
D
Q.
E
o
0
4
20
15
10
c
0
,
5 50
1
5
^^^B
M
Me
J
5
eon
>a n
I.
60
L
6
4
c
"III 0
5 70 75 4
Leq of Speech
10
-n , ,
URBAN
FT ei r~i
5 50 55 60 65 '. 70 75
in dB
1 SUBURBAN
Mean 1
I 1 1 4 1 1 I 1 1
45
50 55 60 65
70 75 45 50
Leq of Speech in dB
55 60 65 70 75
FIGURE 5. DISTRIBUTION OF SPEECH LEVELS AT HOMES PRODUCED
BY PE,OPLE ON SbVERAL OCCASIONS
-------
higher than the average ambient in the suburban areas
with 45 dB. A comparison of speech levels to background noise
suggested that people maintain about a 5-8 dB speech to noise
ratio when conversing outside their homes and a 9-1^ dB speech
to noise ratio when talking inside their homes. Thus, the
intelligibility was maintained at a higher level inside rather
than outside the homes.
Figure 6 illustrates the effects of background noise on
speech level measured in the home. As the background noise
level increased above a certain level (approximately- 45 dB),
in the homes, speech levels for the most part increased also.
The lines connecting the points indicate that the same observer
was recorded both inside and outside the home. The actual levels
were then normalized to reflect what the speech level would have
been if measured at 1 meter. As indicated by the horizontal
lines in Figure 6 for background noise levels below 45 dB,
speech levels measured either inside or outside the home remained
the same. In some cases they remained the same up to a, background
level of 50 dB. However, in general, above a 45 dB background
level the observers tended to raise their voice levels. Speech
level tended to increase with background level above 45 dB,
by about 0.5 dB for every 1 dB increase in background level.
Measurements were also made of television speech levels. The
recordings were made'with the microphone located at the observer's
ear. Figure 7 shows a histogram of thos'e levels wdth an average
of 61 dB. The observers were told to adjust the television volume
to their preferred listening level depending upon the distance
they chose to sit from the television set. The average distance
of observers from a television was 3 meters. A plot of television
speech levels as a function of background noise is shown in Figure
8. This figure indicates that people increase the volume on the
television 0.7 dB for every 1 dB increase in background level.
25
-------
cr\
75.
70
65
OQ
-a
60
a_
IS)
55
50
45
HOMES
INSIDE
OUTSIDE
I
I
I
I
35
40
45 50 55 60
Leq of Background Noise in dB
65
70
FIGURE 6.
CONVERSATIONAL SPEECH LEVELS IN HOMES NORMALIZED
TO 1 METER
-------
TELEVISION
10
D-
E
D
0)
_D
E
Mean
45 50 55 60 . 65 70
Leq of Speech in dB
FIGURE 7. DISTRIBUTION OF TELEVISION SPEECH
LEVELS AT VARIO US 1>1 b FAN CES
27
-------
CORRELATION
COEFFICIENT
40
40 45 50 55 60
Leq of Background Noise in d B
65
FIGURE 8.
SPEECH LEVELS HEARD FROM TV AS A FUNCTION
OF BACKGROUND NOISE
28
-------
C. Hospitals
Conversational speech measurements were made for both nurses
and patients in the hospital environment. Figure 9 shows these
results in histogram form. The speech level for the patients
was only 2 dB lower than the speech level for the nurses. At
the nurses' stations there was only a 5 dB speech to noise ratio,
as compared to the 10 dB speech to noise ratio found when
measuring patients' speech in their hospital rooms.
D. Public Places
Speech measurements were also made In department stores. The
histogram for speech level distributions in this'environment
is shown in Figure 10. The average speech level, measured at
various distances from the listeners' ear, was 6l dB. The
background level had an average of 5^ dB,^thus there was a 7 dB
speech to noise ratio.
E. Transportation Vehicles
Speech levels were obtained for two types of transportation:
trains (as represented by the San Francisco BART system)3 and
conventional aircraft. Histograms of these speech levels are
shown in Figure 11. The average speech-level inside aircraft
and trains averaged 75 dB, the average ambient level at 77 dB.
The average distance between speaker and listener for both mea-
surement situations was 0,4 mefcer.
F. Anechoic Chamber
Speech measurements were also made in a quiet laboratory setting
in an anechoic chamber. Male and female talkers of all ages
-------
NURSES
PATIENTS
I0r
Q.
£
D
_D
£
3
Mean
J>_
45 50 55 60
I
65 70 45
Leq of Speech in dB
Mean
50 55 60 65
FIGURE 9. DISTRIBUTION OF SPEECH LEVELS IN HOSPITALS
PRODUCED BY NURSES AND PATIENTS ON SEVEKAL
OCCASIONS
30
-------
DEPARTMENT STORE
IX
E
D
to
_D
£
D
z
15,-
10
0
Mean
50 55 60 65 70
Leq of Speech In dB
FIGURE 10.
DISTRIBUTION OF SPEECH LEVELS IN DEPARTMENT
STORE'S PRODUCED BY PEOPLE ON SEVERAL
OCCASIONS
31
-------
15 i-
-------
participated in this phase of the study. The range of ages
for approximately 100 talkers was from 6 to 60 years, as shown
in Figure 12. The average age was 2^1 years. The observers
i^ere grouped as males, females and children (talkers under
age 13).
Histograms for the three groups and for the five different
vocal efforts which were designated casual, normal, raised,
loud, & shout are shown in Figures 13, 14 and 15. A summary
of the means and standard deviations is found in Table I, For
the categories casual, normal & raised* there was a small
difference in measured voice level between the males, females
and children. Larger differences in voice levels betxreen male
and female observer groups were found for the loud and shout
categories. As expected, the males produced the highest average
vocal output in the shouting and loud voice categories regis-
tering approximately 5 dB higher than the female group or the
children.
The variability in voice level between talkers increased
with vocal effort. For example, the voice level variability
between male speakers for the normal vocal effort was JJ dB.
But the difference between male speakers voice levels was more
pronounced (7 dB) when instructed to recite Joe's Passage at
a shouting voice level. A similar increase in speech level
variability between speakers was also noted for both the female
and childrens groups.
Figures 16 through 18 show a further analysis of this data in
the plots of the voice spectra for males, females and children.
The complete tabulation of all one-third octave band speech data
recorded in the ane-choic chamber can be found in the Data
33
-------
Q
D
1/1
3
z
10 -
0
6-10 11-15 16-20 21-25 26-30 31-35 36-40 41-45 46-50 51-55 56-60
Age in Years
FIGURE 12. AGE DISTRIBUTION OF SUBJECTS EMPLOYED FOR
ANECHOIC CHAMBER SPEECH MEASUREMENTS
-------
U)
15r
CASUAL
a
o 10
1/1
01
_c
15 i-
n_
10
Mean
J 0
NORMAL
10
Mean
, i,
RAISED
40 45 50 55 60 65 45
50 55 60 65 70
Leq of Speech in dB
50 55 60 65 70 75 80
10
D.
E
o
1/1
t-»-
o 5
-------
CASUAL
NORMAL
RAISED
10
0)
CL
E
o
uo
tt)
_Q
E
ior
Mean
i
I ill 0
ior
r
Mean
I Mean
H . i.
40 45 50 55 60 65
45 50 55 60 65
Leq of Speech In dB
50 55 60 65 70 75
LOUD
SHOUT
10
a.
E
o
k.
o 5
_Q
E
j i rr i
Mean
1
55 60 65 70 75 80
I
85
Mean
1
J I I I W I I I I
60 65 70 75 80 85 90 95
Leq of Speech in dB
FIGURE 14. DISTRIBUTION OF SPEECH LEVELS PRODUCED BY FEMALES AT FIVE VOCAL EFFORTS
-------
CASUAL
NORMAL
RAISED
•n 10
0)
Q.
E
o
1/1
*o 5
L-
E
Z
0
-
n ,
n
f
Mean
10
i 1 i n
40 45 50 55 60 65
ior
01
o_
E
o
on
o 5
L.
(V
_D
E
0
n ,
LOUD
i
A. ,
10
rn
I M ea n 1
1 , , i , 71
5
. n
-
-
1 L t \
1 i
1 i i w i i i
45 50 55 60 65 70 55 60 65 70 75 80
Leq of Speech in dB
]0r SHOUT
5
ll ll 0
-i , r
n
\n
J J i J
till II
55 60 65 70 75 RO 85
90 95 55
Leq of Speech in dB
60 65 70 75 80 85 90 95
FIGURE 15. DISTRIBUTION OF SPEECH LEVELS PRODUCED BY CHILDREN AT FIVE VOCAL EFFORTS
-------
Supplement for this report. The speech spectra across observers
were relatively uniform in shape for increased vocal efforts;
with a trend towards greater high frequency content at the
higher voice levels. The main difference between male and
female speech spectra can be noted in Figures 16 and 17, in the
frequencies below 200 Ez. The spectra summarising the results
for male speakers (Figure 16) show a greater concentration of
vocal energy in the one-third octave bands below 200 Hz. The
speech levels at low frequencies, however, increased only
slightly relative to the increased vocal effort. Furthermore,
for all three groups, the levels at the low frequencies remained
fairly constant.
The relationship between the overall level of speech and A-
weighted sound pressure level was studied because the Articulation
Index calculation procedure utilizes an overall measure of
speech, The difference between the two measures was plotted as a
function of A-weighted sound pressure level for all of the data
collected in the anechoic chamber. A plot of these results is
shown in Figure 19. A best fitting second order equation is
provided for these data, as shown in the figure. Note that at
high levels of speech, the average difference between A-level and
overall level of speech is near 0, whereas at the lower levels
(such as those associated with casual conversation) typical
differences of 5 dB occur.
38
-------
MD
ONE-THIRD OCTAVE BAND SOUND PRESSURE LEVEL IN dB
RE 20 MICROPASCALS
r-o oo .>. ui CN vj oo co oo
O o o o ° ° O^JOD
3 .,-.-,
-|
MH
I
|
1
-
:0
1—
— -»l
h—
— 1
i^_
— (
A-A
• «
_
••^M
^
—
1
1
i
-/
/
/
J\
•\!
\
\
— *
/*!
i
1
(
i
i/
X
1
.')
«/
' 1
fy
^
^
1
I
/
f
J /
I/
t
/
I
1
1
f
X
/•
r
^
^
\
>
^4
f
/
/
/
/
/
J
f
1
/
1
i
1
/*
/*"
x-
X
f
f
s
>
\
\
^
%
\
\
\
\
\
>
^
\r
— \
i
v
*
*
"^
^^^
^or
^™
Cos
L
^^
^
^
/
r
^4
x
mal
_^4
ual
x
^
^.
x.
>
\
s
VR
\
>
\
\
\
•
^^«
. Loud.
\
\
\
\
| — ^
aise
i^
\
^\
V
\
»
1
*^
>H-
v
V
^* •
••!•
V
ll
""
^
•\
\
L ~~\
V
s
/
r
Sho
X
\
V
\
\
\
\
1
w
V
\
\
V
ul
^"»»
^ ,
— -*i
.^^
^^^
>
s
\
s
k
s
\
\
V
•^^^
V
\
A
v
\
^
w
\
\
V
\
125 200 315 500 800 1250 2OOO 3150 5000 8000 12,500 20,000
160 250 400 63O 1000 1600 2500 4000 6300 10,000 16,000
ONE-THIRD OCTAVE BAND CENTER FREQUENCIES IN Hz (cps)
FIGURE 16. AVERAGE SPEECH SPECTRA FOR MALES AT FIVE VOCAL EFFORTS
-------
-t=-
o
m
-o
UJ
>
UJ
-1 co
Ul -I
cc <
3 O
co tn
(0 <
UJ Q.
IT O
O. K
O
Z Ul
4 (E
m
UJ
I
o
o
o
(K
UJ
Z
o
125 200 315 500 80O 1250 20OO 3150 500O 8000 12,500 20,000
160 250 400 630 IOOO I60O 250O 40OO C3OO 10,000 I6.OOO
ONE-THIRD OCTAVE BAND CENTER FREQUENCIES IN Hz (cps)
FIGURE 17. AVERAGE SPEECH SPECTRA FOR FEMALES AT FIVE VOCAL EFFORTS
-------
CD
•V
z
Ul
>
gc
_*
to
UJ
UJ
Ul
o
o
o
DC
X
t-
I
Ul
z
o
30
20
125
800 315 500 800 I35O 2OOO 3I5O 50OO 800O I2.50O 20,000
160 250 400 630 1000 1600 2500 4000 6300 10,000 16,000
ONE-THIRD OCTAVE BAND CENTER FREQUENCIES IN Hz (cps)
FIGURE 18. AVERAGE SPEECH SPECTRA FOR CHILDREN AT FIVE VOCAL EFFORTS
-------
IVJ
CO
TJ
4)
-------
VIII DISCUSSION
A. Lecturing in Schools
The average speech level computed For all teachers (normalized
to 1 meter) was 71 dB, This was compared to the other speech
measurement situations. This level was 13 (IB greater than
the average voice level (normalized to 1 meter) employed either
inside or outside the home. The teachers' voice levels were
also compared with the laboratory study of speech measured in
the anechoic chamber. As rioted in Table I, the 71 d3 average
voice level for the teachers would fall between the raised and
loud vocal effort.
The increase in background level had a noticeable effect on
the teachers' vocal efforts. The speech level increased with
background level at 1 dB/dB whereas for all other speech measure-
ment situations the rate was 0.6 dB/dB (Figure 20). Over one-
quarter (28%) of the teachers sampled spoke at an average voice
level of 75 dB or more. All of these were teachers measured at
school II. The average lecturing level at the back of the
classroom (Position B-estimated at 7 neters from the teacher) was
also measured for all teachers in both schools and was 60 dB.
The increase in voice level did not seem to help performance in
the word 'intelligibility test given to the students. Thus, even
though the teachers' speech levels at school II were OH the
average 5 dB higher than that used at school I, the students in
school I achieved 1U% better scores on their word Intelligibility
test. A possible explanation is that the students in school II
lacked the motivation to adequately perform on this test.
It is important to mention also that although the sites for
the school were selected as being representative of a traffic
-------
TABLE I
SPEECH LEVELS (dB) AT VARIOUS VOCAL EFFORTS MEASURED IN
AN ANECHOIC CHAMBER* (BACKGROUND LEVEL L = 16' dB)
SG
VOICE LEVEL
Casual
Normal
Raised
Loud
Shout
MALE
eq o
52.0 4,0
58.0 4.0
65.0 5.0
76.0 6.0
89.0 7.0
'
FEMALE
eq a
50.0 4.0
55-0 4.0
63.0 4.o
71.0 6.0
82.0 7.0
CHILDREN
eq a
53-0 5-0
58.0 5-0
65.0 7.0
74.0 9.0
82.0 9.0
AVERAGE
L
eq
52.0
57.0
64.0
73-0
85.0
^Results were rounded off to the nearest decibel.
-------
noise environment and an aircraft•noise environment, the main
source of background noise, particularly in the aircraft
exposed school, was produced by the students themselves in
the classroom.
B, Conversing in Various Environments
The means and standard deviation of the speech levels measured.
under non-laboratory conditions are summarized in Table II.
These data reflect speech and background levels measured under
conditions, as judged by the observers, of adequate speech
intelligibility. These conditions take into consideration, among
other variables, the distance between the talker and the listener,
the visual cues and the length of the conversation. Thus, for
conversations recorded in high background noise environments (above
70 dB), such as trains or airplanes, the distance of 0.4 meters
between the participants was shorter than between participants
recorded in quieter environments such as the home where the dis-
tance was approximately 1.0 meter. Communication in the high
ambient environment also necessitated careful attention to the
speakers' phraseology in addition to visual cues to achieve
adequate intelligibility.
The spread in speech levels between talkers and between speech
samples was considered. The average variability of speech levels
computed from the speech samples collected from each talker in
the hone was approximately 2 dB, as compared to an average
standard deviation of 5 dB between talkers for speech measured
in the home. The variability between talkers for the laboratory
condition designated as casual, normal & raised ranged from ^ dB
to 7 dB (Table I). The intertalker variability in speech levels
increased even more when the talkers spoke at a shouting level
(ranging from 7 to 9 dB).
-------
TABLE II
AVERAGE SPEECH LEVELS IN VARIOUS ENVIRONMENTS''
Schools - I
II
Homes -
_ , . , Urban
Outside
Suburban
Inside Urban
Suburban
Hospitals - Nurses
Patients
Department Stores
Trains
Aircraft
Background
Levels (dB)
Leq
48. O2
51. O2
61.0
48.0
18.0
ill. 0
52. 0
45.0
54.0
74.0
79-0
a
2.0
3.0
5.0
4.0
2.0
3.0
5-0
2.0
4.0
1.0
3-0
Speech Levels, dB
1 Meter
eq a
69.0 4.0
73.0 4,0
Corrected
To 1 Meter
65-0 4.0
55-0 5-0
57.0 6.0
55>. 0 5-0
56.0 3.0
56,0 2.0
58.0 3.0
66.0 2.0
68'. 0 4.0
Pos . A Pos. B
2 Meters 7 Meters
eq c
62,0 5.0
66,0 5.0
Conversation
Distance
66.0 4,o
56.0 5.0
57-0 6.0
55-0 5.0
57-0 4.0-
55.0 1.0
61.0 3.0
73-0 3.0
Lea
57.0
62.0
77-0 4.0
a
It.O
6.0
1 Results were rounded off to the nearest decibel.
Measurements were made with typical student activity. Background values
of classrooms during the phonetically balanced word test and other "quiet
periods" were 4? for School I and 43 dB for School II.
146
-------
The measurement of speech levels in face-to-faee communication
revealed a corresponding increase in speech level as background
levels rose above ^5 dB. Figure 20 summarizes this effect for
all of the speech levels measured in a two-way communication
situation. The points are coded for the type of environment.
While all the speech measurements were not taken at the same
distance for all speakers, the criteria of adequate observer
assessed intelligibility was maintained.
Regression lines were computed for the data plotted in Figure
20. An approximation of the regression line was also drawn.
This approximation falls well within the confidence interval
of the regression lines. The results for this study suggested
that for background levels below M5 d3, the level at the listener's
ear remained constant at 55 dB. Thereafter, the speech level
increased up to approximately 80 dB at a rate of 0,6 dB/dB in-
crease in background level which ranged from 45 to 8l dB.
By utilizing Table I & II, it was possible to compare the results
from the face-to-face speech measurements to the data collected
under laboratory conditions in the anecholc chamber. The casual
conversation measure was conducted in a manner similar to the
face-to-face communication situation, only in a controlled back-
ground level of 16 dB in the anechoic chamber. The average speech
level measured for all observers under this laboratory condition
was only 3 dB below the speech levels obtained at 1.0 meter in
the suburban home environment. The average speech level obtained
for observers speaking in a normal voice in the chamber was 57 dB,
only 2 d3 higher than the quiet home situation.
-------
-f
en
• HOMES
• HOSPITALS
A DEPT STORE
• TRAIN
• AIRCRAFT
REGRESSION LINE
APPROXIMATION
50 55 60 65
Leq of Bockground Noiie Level in d B
FIGURE 20. CONVERSING SPEECH LEVEL AS A FUNCTION OF BACKGROUND NOISE
IN SEVERAL ENVIRONMENTS
-------
In Figure 21, a more precise comparison was made between speech
results obtained in the anechoic chamber and those collected
under non-laboratory conditions. The speech levels--measured in
the different noise exposure situations were originally recorded
at varying distances, however for this analysis, all results
were adjus-ted to approximate speech levels measured at 1.0
meter. It was noted from this plot that speech levels used
in the homes and hospital -and department stores could be
characterized as casual to normal voice levels, as determined
from the anechoic chamber measurements. People in transportation
Interior environments such as trains or aircraft appeared to
speak at .what could be compared to raised or loud voice levels
in the laboratory situation.
Three regression lines were calculated 'for the data plotted
in Figure 21. An approximation of the three regression lines
was also drawn In Figure 21. This simplified relationship
Indicated that speech level stayed at about 55 dB when background
levels were below 48 dB. This is only a 3 dB increased dif-
ference In background-level from results in Figure 20, where
speech levels were plotted disregarding distance between speakers,
It is noted in Figure 21 that people started raising their voice
level after 48 dB and continued to do so at the rate of 0.6 d'B/dB
increase In the background level. At an ambient of 70 dB, the
speech data appeared to level out at 67 d3 which indicated that
most people stopped raising their voice above a 70 dB ambient.
An explanation for the 'dramatic difference between Figures 20 and
21 In the speech measurements above 70 dB background level, was
that the high levels measured at the listener's ear and plotted
in Figure 20 were due to the short distance between the speaker
and the-listener and not necessarily increased- voice level due to
49'
-------
o
0)
0)
a
75
70
I)
s: 65
60
55
50
45
• HOMES
• HOSPITALS
A DEPT STORE
• TRAIN
• AIRCRAFT
REGRESSION LINE
APPROXIMATION
LOUD
RAISED
CASUAL
35
40
FIGURE 21
45
50
55
60
65'
70
75
80
85
Leq of Background Noise Level in dB
CONVERSING SPEECH LEVEL NORMALIZED TO ONE METER AS A FUNCTION OF
BACKGROUND NOISE IN SEVERAL ENVIRONMENTS
E
o
j:
U
u
o
NORMAL
-------
increased background level. Therefore in Figure ?1 when the
levels were adjusted for the same 1.0 meter distance, 86? of
the speech measurements taken in an ambient environment above
70 dB dropped below the 70 dB speech level,
A plot of the Interpersonal communication distance between
speaker and listener as a function of the background level is
seen in Figure 22. As the slope of the line indicates, the
distance between the participants in the quieter environments
such as homes or hospitals was approximately 1.0 meter. The
average background level corresponding to this distance was -43 dB.
When the background level increased to JO dB such as in the trans-
portation environments, there the distance between the partici-
pants decreased to 0.5 meters.
A subjective determination of speech intelligibility was not the
sole criterion. Speech intelligibility was also defined objec-
tively by the Articulation Index (ANSI, 1969). The relationship
between AI and background level is seen in Figure 23 • The corre-
lation coefficient for the regression line drawn through this data
was r =-0.82. Reading from the slope of the regression line, at
the background level of AO dB, the AI = 0.82. For an Increased
ambient level of 70 dB, the AI = 0.44. Thus, it was observed
that as the background level increased, the Articulation index
decreased.
The curve plotted in Figure 2*1 is a translation of the regression
line in Figure 23- This was achieved by converting the Articulation
Indices into speech intelligibility scores taken from the psycho-
metric function which describes the percentage of sentences
51
-------
2.0
vn
1.5
1.4
1.3
1.7
1.1
1.0
0.9
0.6
o.:
0.4
0.3
35
• HOMES
• HOSPI1ALS
* DEPT STORE
• TRAIN
• AIRCRAH
A A
J_
I
I
40
45
50 55 60 65 70
Leq of Bnckgrounci Noi^e Level in dB
75
=50
90
' FIGURE 2?. INTERPERSONAL DISTANCE OBSERVED FOR CONVERSATIONS AS A
FUNCTION OF BACKGROUND LEVEL IN VARIOUS ENVIRONMENTS
-------
U)
1.0
•X
A*
• HOMES
• HOSPITALS
A DEPT STORE
• TRAIN
• AIRCRAFT
.2
35
40 45 SO 55 . 60 65 70 75 80
Leq of Background Noise Level in dB
FIGURE 23. ARTICULATION INDEXES FOR CONVERSATIONS IN VARIOUS ENVIRONMENTS
85
-------
correctly understood (ANSI, 1969). Thus, for an AI = 0.50
the sentence intelligibility score is 971 and this occurs at -a
background of 65 dB. This curve can now be utilized'to predict
sentence intelligibility given an ambient level.. For example3
at an .ambient of 80 dB, the sentence intelligibility score was
81%.
A detailed comparison was made of the Articulation Indices and
sentence intelligibility scores for the speech•levels utilized in
the five environmental situations with decidedly different back-
ground levels. As representative of the quiet environments, the
average AI for homes was 0.71 with virtually 1001 sentence in-
telligibility; and for hospitals the AI = 0.63 with 99? intelli-
gibility. As the background level increased above ^5 dB3 the
Articulation Index decreased. Thus3 for department stores the
AI - 0.61 with 98% Intelligibility; for trains the AI - 0.^4
with 9^% sentence intelligibility; and for airplanes the AI =
0.38 with 90$ intelligibility. Thus, it was apparent that intelli-
gibility was inversely correlated to background levels. However,
the decrease did not impede communication until the background level
was above 70 dB5 then the AI dropped two-tenths to AI = O.i4k and
sentence intelligibility was calculated at less than 95$.
C. Speech Measurements in an Anechoic Chamber
Table I summarizes the results of the anechoic chamber mea-
surements. There was approximately a 30 dB difference between
the average voice levels designated casual and shout. But the
progressive increase in level for the five speech categories
-------
ui
VJI
100
90
o
II
o
U
80
70
c
o
« 60
c
a>
o
o 50
Q.
40
30
35
I
40
I
45
EQUIVALENT REGRESSION LINE'
OF FIGURE 23 (SEE TEXT)
I
I
J_
I
I
I
75
80
50 55 60 65 70
L°q of Background Noise Level in d B
FIGURE 24. SENTENCE INTELLIGIBILITY FOR CONVERSATIONS IN VARIOUS ENVIRONMENTS
85
-------
(casual, normalj raised^ loud & shout] ranged from 5 to 12 dB.
The smallest Increase was between casual speech and normal
-speech; the largest increase was betiveen loud speech and shout,
A comparison of the speech levels In the categories of casual,
normal & raised for male and female speakers showed approxi-
mately .a 2 dB difference. These results would not support
Beranek's (195*0 recommendation that background levels be
lowered by 5 dB to accommodate the voice levels used by female
speakers. The real effect of vocal effort on speech level is
more evident in measurements made for the loud and shout cate-
gories where the difference between male and female speech
levels was 5 and ? dB respectively.
Figure 25 shows the results of this phase of the speech study and
compares them with an earlier study by Beranek (195^) In which
the criteria for the Speech Interference Level (STL) were developed,
A comparison of the voice range between normal and shout revealed
that in the current study the- difference was 28 dB, but In
Beranek's results the difference was only 21 dB. Both studies '
agreed '(within 1 d3) on approximately 73 dB for the loud speech
level. However, for the other speech categories (normal, raised
& shout} the results from the two studies differed by 3 to 4 dB.
The normal (57 dB) and raised (64 dB) voice levels In this study
were lower than those suggested by Beranek with 61 dB and 67 dB
respectively; while shout was higher by 4 dB.
As Figure 25 Indicates the standard deviation between speakers
Increased with vocal effort from approximately 4 dB for casual
speaking to 9 dB at the shouting level. This increase in
56
-------
-o
c
fll
>
V
u
4)
(1)
CL
cr
4)
90
80
70
60
50
:$:jiil Present StuHy
iS Bern nek
Shaded Area
Indicafes Standard
Deviation
g
Casual Normal
Raised
Loud
Shout
Voca! Efforts
FIGURE 25. SPEECH LEVELS FOR VARIOUS VOCAL EFFORTS
57
-------
variability between individuals may be attributable to several
factors. One explanation is that while all subjects were given
the same instructions for measuring their speech in the anechoic
chamber, the individuals may have differed in their personal
interpretation -of the five vocal effort descriptors. For example,
it might be more difficult (especially for the children as evi-
denced by the 9 dB standard deviation) to understand how much
vocal effort the experimenter meant when the instructions were to
shout. The subjects seemed better able to uniformly relate to
the instruction's to speak in a normal voice, with a resulting de-
crease in inter'subj ect variability.. Another factor contributing
to the variance between subjects may be related to the difference
in individual capabilities to speak at the various voice levels.
Thus, most subjects were able to maintain levels within the
speech range of the first three speech categories. However, for
the vocal effort in the loud and shout categories, the capacities
of the individuals to maintain these levels differed greatly.
Finally, most individuals speak everyday at a speech level which
would be characterized as either aasual, normal* or vaised.
Therefore, when asked to speak at a loud or shout level, they
would be less familiar with what level to maintain and they would
be far less accustomed to exercising' this level of speech.
The shape of the speech spectra also changed in an orderly fashion,
providing higher level components at high frequencies for increased
voice level. An indication of this trend is the shifting of the
maximum one-third octave band from 500 Hz to 1600 Hz (which is
approximately 1.6 octaves) as the voc.al effort progressed from
normal to shout. The comparison between speech spectra among
males, females and children also indicated similarity, except at
58
-------
the higher speech levels. In all cases, however, the speech
spectrum presented in the Articulation Index standard (ANSI,
1969), contains less irregularities than in the spectrum ob-
tained for the present data.
59
-------
REFERENCES
ANSI, American National Standards Institute, "Methods for the
Calculation of the Articulation Index", ANSI 83.5-1969.
Benson, R. W. , I. J. Hirsh, "Some Variables in Audio Spectro-
metry", J. Acoust. Soc. Am. H,'499-505, 1953.
Beranek, Leo L., "Acoustics", McGraw-Hill Electrical and Electronic
Engineering Series, 1954.
Brown, W. S., Jr., Thomas Murry & David Hughes, "Comfortable
Effort Level: An Experimental Variable", Journal of the Acoustical
Society of America, Vol. 60, No. 3, September, 1976,
•Dunn, H. K., S. D. White, "Statistical Measurements on Conversa-
tional Speech", J. Acoust. Soc. Am. 11, 27-8-288, 1940.
Environmental Protection Agency, "Information on Levels of Envi-
ronmental Noise Requisite to Protect Public Health and Welfare with
an Adequate'Margin of Safety", 550/9-74-004, Environmental Protection
Agency, Washington, D. G., March 1974.
French, N. R. and J. C. Steinberg, (1947), "Factors Governing
'the Intelligibility of Speech Sounds", J. Acoust. Soc. Am. 19, 90-119
Gardner, Mark B., "Effect of Noise, System Gain, and Assigned Task '
on Talking Levels in Loudspeaker Communication", Journal of Acous-
tical Society of1 America, 4_0_, No. 5, 955-9653 1966.
Klumpp, Ri G. and J. C. Webster (1963), "Physical Measurements of
Equally Speech-interfering Navy 'Noises", J. Aeoust. Soc. Amer., 35,
1328-1338. _ . -
Korn, T. S. (1954), "Effect of Psychological Feedback on Conversa-
tional Noise Reduction in Rooms", J. Acoust. Soc. Am., 26, 793-794.
Lombard, E., "Le Signe de L'elevation de la voix", Ann. Mai.-
Orel 1.- Larynx, 37, 101-119, (1911).
Pickett, J; M. (1958), ."Limits of Direct Speech Communication in
Noise", J. Acoust. Soc. Amer., 30, 278-281.
60
-------
APPENDIX A
INSTRUMENTATION EMPLOYED IN SPEECH
MEASUREMENT AND ANALYSIS
-------
INSTRUMENTATION APPENDIX A
Block diagrams are presented•In this section of instru-
mentation used to acquire speech level data, 'calibrate
equipment, and reduce data,
1 . Data Acquisition
Equipment used to measure speech levels In classrooms is
shown in Figure A-l, Data were recorded on three independent
tracks of a standard tape recorder, Figure A-? shows the.
typical microphone placement within a classroom.
2. Equipment Calibration
Calibration of the teacher's microphone was achieved in
an anechoio chamber under eo'nditipns outlined in Figure A-3.
The basic procedure was to place the miniature microphone
immediately adjacent to a standard instrumentation microphone
(a 1" B & K condenser microphone). Output levels produced by
the two microphones 1 meter from a loudspeaker were then com-r
pared at a variety of frequencies and levels, A correction
spectrum so developed was incorporated into all subsequent
processing involving data recorded by the miniature microphone,
Calibration of the miniature microphone in the field was
accomplished via a B1 & K type ^230 (94 £B') calibrator, for
which an adaptor was specially prepared.
3. Data Analysis
All data reduction was accomplished by 33K's real time
one-third octave band analysis system, shown in Figure A~i|. The
process involved playing magnetic tape recordings-into a spectrum
analyzer, processing the frequency analyzed-'data digitally, using
a specially designed computer program, punching paper tape for
long term storage, a,nd listing the paper tapes on a line printer.
-------
Teacher's
Microphone
Microphone
Calibrator
Microphone
Pre-Amplifier
PoslHon A
Microphone
Pre-Amplifier
Position B
Sound
Level
Merer
Sound
Level
Meter
Sound
Level
Meter
Annotation I I
Microphone I - J
nnnn
Magnetic Tape
Recorder
A-l EQUIPMENT FOR MEASURING 'SPEECH AND BACKGROUND
LEVELS IN CLASSROOMS
Teacher's Microphone
Microphone
Random Incidence Corrector
Pre Amplifier
Power Supply
Sound Level Meter
Magnetic Tape Recorder
Piston Phone Calibrator
Random Incidence Corrector Adaptor
Transound, Model 74-A (Minimic)
B & K, Condenser, 1
B & K, Type UA0055
HPr Type 15108B
HP, Type 15114A
B & K, Type 2205
Sony, Model 854-4S
B&K, Type 4220
B& K, Model 152
0 Inch
63
-------
TEACHER'S MICROPHONE
(Worn by Teacher)
POS.B
Blackboard
• Podium
•POS.A
FIGURE A-2. MICROPHONE LOCATIONS IN CLASSROOM
-------
Reference Microphone
Miniature Microphone
• ANECHO1C CHAM BER ,
A-3 EQUIPMENT SETUP FOR CALIBRATION OF MINIATURE
MICROPHONE USED IN CONVERSATIONAL. SPEECH
RECORDING
Reference Microphone
Pre-Amplifier
Sub-Miniature Microphone
Sound Level Merer (SIM)
Graphic Level Recorder
Sine Random Generator
Power Amplifier
Speaker
Piston Phone Calibrator
9&K, Type 4133, 1/2 Inch
G-R, Type P42
Transound, Minimic, Model 74-A
or BBN, Eleclret Microphone
B& K, Type 2205
8 & K, Type 2305
B & K, Type 1024
JBL, SE400S
JBL, C40
B & K, Type 4220
-------
Magnetic Tope
Recorder
External
Filter
Sound
Level
Meter
ReaS-Time
One-Third
Octave
Band
Analyzer
Digital
Computer
A-4 SPEECH AND BACKGROUND ANALYSIS SYSTEM
Magnetic Tape Recorder
Sound Level Meter (SLM)
External Filter
Real Time Third Octave Analyzer
Sony, Model 854-4S or Nagra, Model SN
B & K, Type 2203
BBN
HP, Type 8054A
66
-------
APPENDIX B
INSTRUCTIONS AND WORD LISTS USED IN CLASSROOMS
AND ANECHO1C CHAMBER
67
-------
CLASSROOM INSTRUCTIONS
To the Teacher
Please read the following word list to
your students. Read the words one at a
time and do not repeat the word even If
asked by a student to do so. Read the
words at a normal pace and maintain the
same classroom lecturing voice level
throughout the presentation.
To the Student
Listen carefully to the words the teacher
will read. They will be read only once.
Do not ask the teacher to repeat a word
that you have missed. This is not a
spelling test, nor does it count on your
grade. Pay close attention and do the
best you can.
68
-------
PHONETICALLY BALANCED WORDS READ IN CLASSROOM
LIST 1
CANE
THERE
DISH
HID
HEAP
PANTS
HUNT
NO
BAR
PAN
FUSS
CREED
BOX
STRIPE
DIKE
NOT
FORD
END
THEN
BASK
FRAUD
SMILE
DEATH
ARE
BAD
PEST
SLIP
RUB
FEAST
DEED
CLEANSE
FOLK
NOOK
MANGE
SUCH
USE
CRASH
RIDE
PILE
RAT
RAG
IS
WHEAT
RISE
HIVE
GROVE
TOE
PLUSH
CLOVE
FERN
LIST 2
TANQ
FATE
SUCH
ELSE
PIT
GILL
CHARGE
BOUGHT
CLOUD
MUTE
BEAN
SCYTHE
VAST
RIB
PICK
HOCK
OUR
HIT
JOB
WISH
NUT
DAB
FROG
LOG
SNUFF
BLUSH
NAB
BAIT
BUD
RAP
MOOSE
TRASH
GLOSS
PERK
VAMP
START
EARL
CORPSE
SLUDGE
TAN
WAYS
BOUNCE
NIECE
AWE
THEM
NEED
QUART
FIVE
HIRE
SHOE
LIST 3
WHY
TURP
GNAW
DROP
JAM
FLUSH
ROUSE
NECK
SOB
TRIP
DILL
THRASH
DIG
RATE
FAR
CHECK
AIR
BEAD
SPED
CAST
CLASS
LUSH
SHOUT
BALD
CAPE
SIZE •
WEDGE
DECK
HURL
WHARF
LEAVE
CRAVE
VOW
LAW
STAG
OAK
NEST
SIT
CRIME
MUCK
FAME
TAKE
WHO
TOIL
PATH
PULSE
FIG
BARB
PLEASE
ACHE
LIST H
FLOAT
SAGE
CLOAK
RACE
TICK
TOUCH
HOT
POD
FROWN
RACK
BUS
BLONDE
PERT
SHED
KITE
RAW
HISS
PIN
SCAB
'HOW
STRAP
SLAP
PINCH
OR
STARVE
NEW
RUT
NEAT
DODGE
SKETCH
MERGE
BATH
COURT
OILS
SHIN
PECK
BEAST
HEED
EEL
MOVE
EARN
BUDGE
SOUR
RAVE
BEE
BUSH
TEST
HATCH
COURSE
DUPE
69
LIST 5
FEED
GAPE
SICK
GREEK
ROE
CHOOSE
TRUE
PASS
BROWSE
PUNT
SHOVE
HILL
BLACK
HIGH
RIND
VASE
RODE
PUFF
INCH
BRONZE
SOLVE
BATHE
ADD
REAR
SHINE
SLY
WRATH
LOVE
BECK
THICK
FLAP
CHEAT
WINK
ZONE
ODDS
KID
TRADE
SCARE
MAST
PIPE
GOOD
LEND
YAWN
WATCK
THUD
TUG
CURSE
OWLS
NOSE
GRUDGE
LIST 6
AS
PUN
ROUGH
NIGH
BEST
JAG-
TONGUE
EITCH
BOG
ROOMS
FOWL
REAP
WRITE
WIFE
CLOTHES
GAGE
FORGE
PRIME
SCAN
GROPE
SUP
SLOUCH
THUS
PRIG
FLICK
BADGE
CLOTH
KEPT
PLOP
FALL
WASP
ODE
HULL
PEE
LAG
THIGH
CHART
WAIT
COB
MASH
EYES
RAISE
DEEP
SHANK
RAY
GAP
CRIB
PUS
EAT
DAD
LIST 7
GASP
WOO
THOUGH
ACT
DWARF
SCOUT
SLEDGE
SNIFF
FLING
COOK
DOPE
GUN
JUG
MUD
PLOD
FAKE
PHASE
RASH
RICH
BUT
POUNCE
WHIFF
PIG -
ROAR
SAG
BY
AM
NINE
WIRE
AIM
SHAFT
SOUTH
WOE
CHOP
KNIT
RAID
SIN-
CUT
HIM
DOSE
QUIZ
SIEGE
COAST
GRADE
FORT
COMES
OFF
PENT
RANGE
MOTE
-------
INSTRUCTIONS FOR SPEECH MEASUREMENTS IN ANECHOIC CHAMBER
PLEASE MEMORISE THE FOLLOWING SENTENCE:
"JOE TOOK FATHER'S SHOE BENCH OUT; SHE WAS
WAITING AT MY LAWN."
1. Speak in a NORMAL VOICE - that which you would use in
everyday conversation.
2, Then speak In' a RAISED VOICE.
3. Then speak' as LOUDLY AS POSSIBLE without straining your
. vocal cords.
k. Then speak at a SHOUTING VOICE level.
FOR EACH VOICE LEYEL, REPEAT THE ABOVE SENTENCE UNTIL NOTIFIED
TO STOP.
70
------- |