Home Page

About Us

Induction Loop Systems

Induction Loop Receivers

Induction Loop Testing Equipment

Ambient Noise Compensation

Sound Field System Accessory

Multisensory Sound Lab

Airplane ALS

Loop Systems & Accessories Pricing & Ordering

Contact Us

Using Digital Signal Processing (DSP) to Improve the Performance of Classroom Sound Field Systems 

by Norman Lederman, M.S., Director of Research & Development, Oval Window Audio
and Cheryl DeConde Johnson, EdD., Colorado Department of Education

Abstract
Manually adjusted classroom sound field systems generally improve listening
conditions but do not address the fact that background noise and teacher
microphone input levels often change throughout the day.

This article reports on the development and field validation of adaptive digital
signal processing (DSP) technology that, when interfaced with classroom
sound field systems, "listens" to changing background noise and automatically
adjusts the system's output in order to maximize signal-to-noise ratio, improve
speech intelligibility and reduce teacher vocal stress.  Field test data
demonstrated the positive effects of this adaptive signal processing technology
on speech perception in the presence of competing background noise. 

The unfavorable acoustical environments often found in classrooms have led
to the proliferation of small sound reinforcement systems generally referred to
as "sound field systems".  Previous studies detail how speech perception is
negatively affected by variable background noise conditions created by poor
acoustics, room heating/cooling systems, noisy audiovisual and computer
equipment, and other noise generated inside, as well as external to the
classroom (Crandell, Smaldino, & Flexer, 1995).

The majority of classrooms equipped with conventional sound field systems
are set up for one particular room condition, that is, one background noise level
and an optimum teacher microphone placement.  Unfortunately, this condition
is not a constant.  Unless the volume control is re-adjusted throughout the day,
the systems operate at only one level of output/volume everyday.  This also
assumes that the teacher microphone has not moved from its original position.
As a result, the effectiveness of these systems may vary throughout each day
as the background noise, vocal effort, and microphone-to-mouth distance change.
Based on this observation, a classroom sound field system was conceptualized that
"listens" to the environment in which it is operating in and automatically makes
adjustments to its output in order to maintain a favorable signal-to-noise ratio. This
digital signal processing  (DSP) technology, generally referred to as "ambient noise compensation" (ANC), has been in use for several years in the professional sound reinforcement industry with applications ranging from sports arenas to train stations.
ANC has never been applied to commercially available classroom sound field systems. 

The project’s initial objective was to develop and evaluate a classroom sound field
system that would automatically self-adjust to changing microphone input signals
and varying background noise conditions.  The project team later decided to focus
on the development of an "add-on" signal processor that would easily connect with,
and improve, the performance of existing sound field systems.  This approach
makes the technology economical and accessible to more people who can
benefit by an improved "retrofitted" sound field system.  This decision ultimately
resulted in a new sound field system accessory named SmartSpeaker
“Intelligence” ™ that is easily connected to any sound field system that utilizes
an external wireless microphone receiver.

ANC DSP circuitry depends on a high quality signal in order to make the
appropriate "decisions" about the signal-to-noise status of the system.  After a
brief comparative study of microphones was performed, it was concluded that
only head-worn microphones could consistently deliver the required signal quality
in the typical dynamic classroom setting (Lederman & Hendricks, 2000).

The system’s ANC monitors the signal-to-noise conditions in the room through a
"sense microphone" that is located 6' away from one of the system loudspeakers,
and automatically makes adjustments of up to 12 dB to compensate for changing
background noise.  The ANC employs DSP that continuously measures the
system’s true signal-to-noise ratio, comparing the loudspeakers’ acoustic output
signal with the original electrical input signal while taking into account signal level,
spectrum and signal statistics.  A variable adaptive response time ensures that
changes to the signal level occur unobtrusively.  The ANC system requires an
initial calibration that may be performed with a sound level meter or by ear. Once
established, calibration does not have to be repeated unless the sound field system's
control settings or loudspeaker to sense microphone distances are altered. 

Phase I field testing of an ANC equipped sound field system was conducted in first
and third grade elementary school classrooms where sound field systems were
already being utilized. Thirty-two students comprised the first grade class, the
majority of whom were bilingual.  Student desks were grouped in clusters of four
facing one another. While all students participated in the listening tasks, seven
students from the first grade class were eliminated from the data analysis due to
their very limited English skills.  The third grade class was composed of twenty-three
students.  Sound field amplification was utilized in this classroom because one student
was identified with a unilateral hearing loss. The classroom was rectangular with
student desks oriented in a traditional row configuration. Occupied-quiet background
noise levels of 53 dBA  and 54 dBA were measured respectively in each room. The
sound field systems were adjusted to deliver a +15dB signal-to-noise ratio, resulting
in average sound levels of 68dBA and 69dBA in each room.

The listening task for Phase I employed Ross and Lerman’s  Word Intelligibility by
Picture Identification Test (WIPI). The test was presented twice; first, with the
classroom's existing sound field system and then repeated using a different word list
and an ANC equipped sound field system.  The 25 word list was divided so that there
were 5 presentations of 5 words each at the following signal-to-noise ratios: +10db,
+5B, 0dB, -5dB, -10dB.  A separate sound system was used for presenting the
competing background noise.  Recordings of the WIPI and consonant-vowel
discrimination stimuli were recorded on the left track of a compact disc. The right
track contained a recording of a 50% mix of pink noise and 12 voice babble at the
varying intensity levels required for the listening tasks. This noise mix was created
in order to represent industrial (e.g. heating, ventilation, air conditioning) noise and
human made (e.g., speech) background noise.   A type II sound level meter was
used to measure background noise conditions and to set the playback levels of the
noise track and recorded speech materials. 

Students marked their responses on individual worksheets while proctors facilitated
the proceedings.   In addition to the scored listening task, the ANC equipped sound
field system system was used in each classroom for the following week. At the end
of the trial period, the teachers and students were asked to subjectively report on the performance of the system.

The results of the Phase I field test are reported in Figure 1 (below). The bar graph
depicts average word identification performance for each age group as a function of signal-to-noise ratio. Scores are compared for conventional sound field amplification
and the ANC equipped sound field system.  The graphed results illustrate a substantial
increase in word identification performance with ANC at the -10 signal-to-noise condition…33.9% to 96.5% for third grade students and 11.2% to 81.6% for first
graders. After using the system for a week, subjective comments from the teacher
and student evaluations that referenced experience with conventional sound field
systems were also positive. Teachers reported:  “…it was easier for everyone to concentrate…” and, “I could hear my own voice more clearly". Third grade student
evaluations indicated that 16/20 could differentiate between the two systems, 19/20
preferred the ANC system to the conventional system, 19/20 reported the ANC system
helped them understand the teacher, and 15/20 said that the ANC helped them pay
attention better. 


Figure 1. Phase I: Average WIPI word identification performance across subject categories
as a function of signal-to-noise ratio for conventional sound field (CSF) and ANC equipped SmartSpeaker (SS) sound field amplification.

Participants for Phase II of the project were third grade students at the same
elementary school as Phase I.  This school was again chosen for its population
of students who were considered to be at greater risk for learning problems. 
Phase II had a longer project period that allowed more time for gathering student
profile information and creating a more in depth field test.  Learning risk for these
students may be further characterized by their performance on the third grade
Colorado Student Assessment Program (CSAP) reading test in which only 21%
achieved a proficient or advanced performance level as compared to the state average
of 67%. In addition, a group of elementary age students who were hard of hearing and
attended a center-based deaf/hard of hearing program also participated in order to
evaluate the effects of the ANC technology on students with hearing loss. The
breakdown of the 84 participating students were: 40 regular education 3rd graders,
20 3rd graders identified as English Language Learners (ELL), 14 3rd graders who
received special education services, 1 3rd grader who was both ELL and in special
education, and 9 students with hearing loss (2 3rd grade, 3 4th grade, and 4 5th grade). 

Eight of the nine students with hearing loss attended a center-based school district
deaf/hard of hearing program and received their education in regular classrooms
which were co-taught by a regular education and a deaf education teacher. The
other student was a third grader from the target school. Only students who were
hard of hearing participated. The better ear mean pure tone average was 53.3dB
and the poorer ear mean pure tone average was 58.3dB (range was 37-75dB for
both). These students did not have any other significant disabilities, and all were
consistent hearing instrument users. To verify performance, a listening check of
all instruments worn by the students was conducted prior to the beginning of the testing.

In order to eliminate the variables that existed between the two sound field
systems in Phase I, one system with integrated SmartSpeaker “Intelligence”
was used with a switch that could bypass the ANC circuitry.  The two classrooms
that were used had respective occupied-quiet background noise levels of 52dBA and
55dBA.  As in Phase I, the sound field systems were adjusted to deliver an average
+15dB signal-to-noise ratio resulting in respective average sound levels of 67dBA
and 70dBA.

Three listening tasks were completed by each student under two listening conditions
-- ANC ambient noise compensation “on” and ANC ambient noise compensation
“off”. First, the WIPI listening task was administered using a worksheet with the
picture and word representation of each of the 45 stimulus items and foils in a 6
forced choice response format. Test items were broken into groups of 15 items with
the first 15 presented at +6dB SNR, the second 15 items at a 0dB SNR, and the final
15 items at a -6dB SNR. Students were asked to mark the picture that represented the
word they heard. The second task was a 54 item discrimination test using the consonant-
vowel combinations of BA, DA, and GA. Students were required to identify whether the
two sounds that were presented were the same or different (e.g., BA/BA or BA/GA) by circling the word "same" or "different" on their response form. The test items were divided
into three groups, this time with 18 items at each of the three signal-to-noise ratios. Practice items were given for each of these listening tasks. The third measure was a magnitude of estimation of quality (MEQ) scale in which the students were to place their listening
experience on a sound quality continuum. This 100 point scale was used for the students
to judge their perception of the sound quality during the listening tasks (on a continuum
of 1= “very poor” to 100 = “very good”). 

An analysis of variance (ANOVA) with an alpha level set at .05 was used for the data analysis. The mean scores for each of the listening tasks for the total population and for
each of the groups are presented in Figure 2. When considering the scores of the entire
group of subjects, performance improved in all conditions with the ANC "on". Statistically significant increases were evident in 5/8 conditions (WIPI word recognition at 0 SNR &
-6 dB signal-to-noise ratios, same/different at -6 dB SNR, MEQ for word and same/
different tests). When considered by subgroups, the ELL students had statistically
significant better performance during the WIPI test at all SNR levels as well as on
both MEQs, the special education students had better performance on 7/8 tests
although none were at the .05 level, regular education students demonstrated
statistically significant improvement in 5/8 areas (WIPI at each signal-to-noise ratio, same/different at -6 dB signal-to-noise ratio, and MEQ words) and the students with
hearing impairment performed statistically better on the same/different test at -6 dB signal-to-noise ratio. 

Collectively performance was improved in all but 2/40 conditions with 16/40 conditions
reaching significance at the .05 level. These are identified in bold italicized print in 
Figure 2 (see below). 

Figure 2.   Mean performance for word recognition, same/different, and MEQ tasks across subject categories. Significance at the >05 level is noted in bold italics. For Trial 1, ANC is off; for Trial 2,
ANC is on.

The most noteworthy aspect of the field test results was the difference in performance
on the speech perception tasks between Phase I and II. While performance was
improved with ANC in nearly all conditions in both phases of the field tests, the degree
of difference was less dramatic in Phase II. This finding is likely attributable to several experimental design decisions. First, the poorest signal-to-noise ratio used in Phase II
was -6 dB as compared to -10dB in Phase I. The effects of the poorer signal-to-noise
ratio were evident in the Phase I performance changes on the WIPI (3rd grade, 33.9% to 96.5% and 1st grade, 11.2% to 81.6%) as compared to -6dB in Phase II (77.6% to 93%). Performances of the Phase I  –5dB condition and the Phase II  –6dB condition were
similar however with Phase I 3rd grade scores improving from 88% to 94.7% as
compared to Phase II improvement of 77.6% to 93%. The decision to use an easier signal-to-noise ratio was based on the difficulty the students exhibited on task in Phase
I and because this background noise condition is more likely to be encountered by students
in most classrooms.   Also, the consonant-vowel same/different task and the WIPI may
not have been sufficiently sensitive to identify speech perception difficulties under the
adverse listening conditions utilized.

Another design change between the two phases that may have attributed to the overall
higher performance in Phase II were the sound field systems used and the inherent
technology differences. In Phase I, the performance of a prototype ANC sound field
system was compared to the conventional sound field system currently utilized in the classrooms. During Phase II, the prototype ANC sound field system was its own
control with performance comparisons made with the system’s ANC technology
either on or off.  The high quality loudspeakers, electronics and analog and digital
signal processing components of the Phase II ANC sound field system may actually
have accounted for a superior signal and therefore better performance across test
conditions. The decision to use the prototype ANC sound field system as its own
control in Phase II was a design decision to reduce the number of variables that might
affect the outcome.

The use of sound field amplification with children who have diagnosed hearing losses
and who wear hearing aids is not well documented. The 9 hard of hearing listeners in
Phase II of this project exhibited statistically significant improvement in only one
experimental condition out of eight (see Figure 2). Why might this high-risk group not
show benefit with sound field amplification?  First, based on their favorable performance
on the listening tasks, it was evident that they were hearing quite well with their personal hearing aids. While the addition of sound field amplification did improve their relative performance in 5/8 conditions, their performance was poorer in 2/8 conditions and un-
changed in 1/8 conditions. Perhaps the “mix” of personal hearing instrumentation and
sound field amplification produces an effect that may not always be advantageous.
Further, it is possible that the materials used were not adequately sensitive to detect the
changes in performance when the ANC was on. While the results of this group of 9
students is too small to make any generalizations, it does support the need for further
study of the interaction between hearing aids and sound field FM amplification. 

The results of other special education and at-risk populations of students shown in 
Figure 2 were also somewhat inconsistent. The ELL group demonstrated relatively
better performance in all conditions with ANC turned on. The improvement at three
of the conditions was statistically significant. While the special education group's
performance improved in 7/8 conditions, none were statistically significant. In spite of
this, there is a definite trend in the data to suggest that these populations benefited from
the ANC.  The small numbers of subjects and heterogeneity typical of these populations
may have conspired to prevent statistically significant differences in performance.

In conclusion, this study suggests that DSP-based ambient noise compensation increases
the performance of classroom sound field amplification technology and is worthy of further research and development for the educational market.  Since the success of sound field amplification depends on a number of technical variables such as system integrity, sound
quality, room set-up and management variables including acoustical environment,
instructional style and teacher and student abilities, it is important to have a variety of
options from which audiologists can choose when designing a sound field amplification plan
for a classroom. Accordingly, DSP equipped sound field systems may provide an option for classrooms that suffer from background noise problems that are intermittent or variable in nature. 
 

Adapted from Lederman, N., DeConde Johnson, C., Crandell, C., Smaldino, J., The Development
and Validation of an “Intelligent” Classroom Sound Field Frequency Modulation (FM) System.
Journal of the Educational Audiology Association (in press).

Norman Lederman is  Director of Research & Development at Oval Window Audio, Nederland, CO.  Cheryl DeConde Johnson is a consultant for Hard of Hearing Disabilities
and Audiology Services at the Colorado Department of Education.

ACKNOWLEDGEMENTS:  The authors gratefully acknowledge the support of the
Small Business Innovation Research Division of the U.S. Department of Education.

A special thank you to: Paula Hendricks, M.A., Educational Director at Oval Window
Audio for her invaluable contributions to the project's design and her editing of related
proposals and documentation, and Jefferson Elementary School and the Weld County
School District #6 Deaf and Hard of Hearing Program, in Greeley CO, for their
enthusiastic involvement as field test sites.
 

For more information, please contact:
Norman Lederman, Director of Research & Development
Oval Window Audio
33 Wildflower Court
Nederland, CO 80466
Telephone/fax; 303-447-3607
E-mail:  info@ovalwindowaudio.com

References

Crandell, C., Smaldino, J.,  & Flexer, C. (1995).  Sound-field FM Amplification:
Theory and Practical Applications. Singular Publishing Group, Inc., San Diego, CA.

Lederman, N., Hendricks, P. (2000).  A comparison of assistive listening system microphones.  Educational Audiology Review, 17(1), 16-17.

Ross, M. & Lerman, J. (1979). A picture identification test for hearing impaired children. Journal of Speech and Hearing Research, 13, 44-53.

For more information please click on the underlined product references above or on the menu below.

PLEASE CONTACT US BY E-MAIL, OR PHONE/FAX 303-447-3607
FOR ADDITIONAL INFORMATION ON ANY OF OUR
SOUND FIELD SYSTEMS.


Home Page // About Us // Induction Loop Assistive Listening Systems // Induction Loop Receivers // Induction Loop Testing Equipment // Ambient Noise Compensation // Sound Field System Accessory // Vibrotactile/Multisensory Sound Lab // Airplane Assistive Listening System //  Loop Systems & Accessories Pricing & Ordering // Contact Us

This site created with the help of The Write Direction, Kauai, HI.