Design of Low-Cost Noise Measurement Sensor Network: Sensor Function Design

Ismo Hakala, Ilkka Kivela, Jukka Ihalainen, Jari Luomala, Chao Gao
University Of Jyvaskyla
Kokkola University Consortium Chydenius
P.O.Box 567, FI-67701, Kokkola, Finland
{ismo.hakala, ilkka.kivela, jukka.ihalainen, jari.luomala, chao.gao}@chydenius.fi
AbstractIn this paper, we report the sensor function design
and implementation of a wireless sensor network application
for measuring environmental acoustic noise. The system is built
on ATmega128 and CC2420 platform. The protocol stack is
based on CiNet stack with a global synchronization scheme and
supports multi-hop communications. Strict filtering function
specified by ITU-R 468 (namely A-weighting) is followed.
Both the indoor and outdoor test results were compared with
standard sound level meters (CESVA SC-20c and Pulsar94) and
showed a less than 2dB error in both short-term and longterm measurement. Power consumption has been measured
that a single AA-type battery can sustain the application.
Comparing to the traditional noise measurement method, our
wireless sensor network solution is much lower in cost, able
to offer real-time data with sensed data timely coherent, and
requests least attention after deployment.
Keywords-sensor node design; noise measurement; environment monitoring

I. I NTRODUCTION AND M OTIVATION

In WSN (Wireless Sensor Network) development, sensor
node design is never a trivial task due to the small size,
low-cost, resource-limit nature of WSN applications. A requirement of measuring acoustic noise in both industrial and
residential environments is present in Ostrobothnia area of
Western Finland. In such a measurement system, a number
of wireless sensor nodes must be scattered into a concerned
area. Each node measures the noise level at its location and
the data is collected by a sink node, which forwards the
data to a web-based database. The network should be able to
cover a large area such as a university campus, an industrial
park, or a residential block.
In many countries the environmental noise has been
regarded as a critical metric of working and living comfort.
However, traditional way of conducting noise measurement
exhibits great inconvenience: A test technician needs to carry
a sound level meter to a measuring location and set the meter
up for a specific measurement duration, which is usually
several hours, and repeat this procedure for all the measuring
points. The disadvantages of this method are obvious: 1)
commercial sound meters are expensive, making large scale
measurement very costly, 2) point-by-point measurement
makes the result incoherent in time, 3) due to the lack of
communication facility, the measured result is not available
in real-time, and 4) the sound meter has to be fully attended,

which increases the work load. These give the motivation

of our WSN design. The goal of sensor node design is to
minimize the necessity of peripheral components of a given
microcontroller chip so that the cost of sensor is minimized.
The target WSN consists of a sink and a set of noise
sensor nodes. The network topology is therefore a multihop tree as the sink node the root of the tree. This network
can be deployed in both indoor and outdoor environments.
The sensor nodes should keep measuring the noise level for
e.g., a whole day [1]. Comparing to the traditional method,
our system has the following significant advantages: 1) cost
reduction in both sensing devices and workload; 2) real-time,
multi-point, coherent measurement; and 3) least attention.
This WSN application has very special features because
it demands excessive CPU power for data acquisition: the
sensor nodes have to continuously sample the noise, thus
leaving very short time for communication. The aim of
this paper is to present the solution approaches to achieve
accurate sensing and efficient communication for excessive
data acquisition WSN applications.
In order to accomplish multi-hop communications, we
built our application on CiNet platform [2], which is a
IEEE802.15.4-based platform using ATmega128 microcontroller with a cross-layer protocol stack containing efficient
routing, topology management, and synchronization functions [3]. The block diagram of sensor node architecture can
be seen in Figure 1. This document mainly covers the sensor
function design. The networking and data communication
issues are briefly illustrated in Section VI. However, they
are thoroughly presented in our technical report, which will
be published later.
The rest of this document is organized as follows: A brief
survey of related work is given in Section II; Background of
sound level and its calculation is illustrated in Section III;
Section IV covers the hardware design; Section V covers
the software design and device calibration; Section VII
shows the results of measurement and the comparison to
a referencing device; Finally, Section VIII summarizes the
design.
II. R ELATED W ORK
The work carried in [4] is probably the closest one
comparing to our project. In their project, Tmote Sky

LAN/WAN

Cinet

Database/
HTTP server

Sensor node

Crosslayer

Application
SYNC2SINK
Protocol
Stack

Power
Saving

Management

Topology Synchro
Control
nization

MAC(802.15.4)
Radio (CC2420)

Figure 1.

System and node architecture

platform [5] were used, and the sensors were deployed to

measure road traffic noise. Furthermore, from the measured
data it is able to count the number and type of vehicles.
The authors asserted that large-scale noise measurement
using WSN solution is possible. However, the precision of
measurement was not mentioned, the calibration of nodes
was an open issue, and the sampling rate of sensor was
set at 8kHz due to the CPU/ADC limit. In [6], the authors
claimed that an environment monitoring system must have
autonomy, reliability, robustness, and flexibility characteristics. They implemented a testbed network based on 8MHz
TinyNode platform with 868MHz XE1025 transceiver to
monitor the hydrological model of Alps area in Swissland.
The authors in [7] proposed a WSN framework for environmental monitoring. The authors introduced a Sensorbased, Distributed Signal Processing (SbDSP) infrastructure
to handle the acoustic data in a distributed way. However,
detailed implementation was not included in the paper. In
[8], a WSN was developed for coal mine monitoring to
(1) rapidly detect the collapse area and report to the sink
node, (2) to maintain the system integrity when the sensor
network structure is altered, and (3) provide a sound and
robust mechanism for efficiently handling queries over the
sensor network under unstable circumstances. The system
was built on Mica2 motes platform and a prototype has been
deployed to D.L. coal mine in China.
There is a bluetooth-based solution for noise measurement
available in market [9]. In this solution, a bluetooth piconet,
which supports maximum 5 noise nodes, can be deployed.
It does not support multi-hop communication and therefore
the application is limited within a small scale.

III. BACKGROUND : S OUND L EVEL

Sound is a traveling wave, which is an oscillation of pressure transmitted through a solid, liquid, or gas, composed
of frequencies within the range of hearing and of a level
sufficiently strong to be heard, or the sensation stimulated in
organs of hearing by such vibrations [10]. This wave energy
can be picked up (i.e., heard) by the human ears. The human
ears are very sensitive devices, which can hear sound power
in a wide range of 12 to 13 decimal magnitudes [11]. Sound
power level is thus defined as logarithm scale, denoted as
decibels, and can be calculated by
!


p2rms
prms
Lp = 10 log10
dB,
(1)
=
20
log
10
p2ref
pref
where prms and pref are RMS (Root Mean Square) sound
pressure and reference sound pressure, respectively, and
pref = 20Pa (RMS).
The human ear is able to hear frequencies from 20 Hz to
20 kHz. However, it does not have a flat response to all the
frequencies in the range, therefore in sound level measurement the sound signal is often frequency-weighted so that
the measured level will match the ear-perceived level. Such
a frequency weighting devices is also called weighting filter
due to its frequency response feature. The IEC (International
Electrotechnical Commission) has defined several weighting
schemes, namely A, B, C, and D weighting. A-weighting is
valid for relatively quiet sounds, which fits the application
of our project.
A-weighting filter is a frequency-select filter picking up
the frequency range around 3-6 kHz, to which the human ear
is most sensitive, while attenuating very high and very low
frequencies to which the ear is insensitive. The corresponding transfer function of A-weighting filter is given according
to [12] as
7.39705 109 s4
.
(s + 129.4)2 (s + 676.7)(s + 4636)(s + 76655)2
Figure 2 shows the frequency response of A-weighting filter,
in comparison with B- and C-weighting filter curves. Sound
level which is measured after A-weighting filter is also
denoted as LA .
The effective sound pressure is the RMS value of the
instantaneous sound pressure over a given interval of time
(a.k.a., time-averaging sound level). In general, it is called
equivalent sound level over time T and denoted as LeqT .
Three of these time-weightings have been standardized in
IEC-61672(2003): S (T = 1 second) originally called
Slow, F (T = 125 milliseconds) originally called Fast
and I (T = 35 milliseconds) originally called Impulse.
A general equation of LeqT for discrete system is denoted
as
!
N 1
1 X p2i
,
(2)
LeqT = 10 log
N i=0 p2ref
HA (s) =

A. Signal Amplification and Dual Channel for Dynamic

Range

10

Frequency Response (dB)

10

10

10

10

Bweighting
Cweighting
Design
Aweighting

10

10

10

Figure 2.

10
Frequency (Hz)

10

A-weighting filter frequency response

PD4
GH
MAX4465

AWeighting

ADC

Mic
MAX4524

GL
MAX4465

Figure 3.

ATmega128L

Overall block diagram of noise measurement circuits

where N is the number of samples taken in time T .

In our design, a set of digital and analog amplifications are
introduced for a suitable ADC (Analog-to-Digital Converter)
range. All these amplifications, together with pref , can be
treated as a single constant and able to be retrieved by
comparing the ADC readings with a standard calibrator.

IV. D ESIGN A PPROACH : H ARDWARE

Research has shown that a long-term exposure to sound
level over 85dB will cause hearing damage [13]. The average
sound level of motorway traffic is 70dB and definitely
regarded as an uncomfortable noise. Typical environmental
or background noise level in residential areas ranges from
30dB to 80dB. Considering some measurement margins, this
requires a dynamic range of 60dB more of measurement
range. The focus of this chapter is to describe our hardware
design that covers a dynamic range of 60+dB.
The overall block diagram of sensor circuits is given in
Figure 3. The sound pickup device is Monacor MCE-400
high-quality electret omnidirectional microphone cartridge
[14]. The frequency range of MCE-400 fits well into Aweighting frequency curve.

As mentioned previously, one of the design objectives is

to measure the noise level in a range of [30,90+]dB, which
requires the data sampling and conversion system to cover a
dynamic range of 60+dB. ATmega128 comes with a 10-bit
build-in successive approximation ADC. 10-bit resolution
can compensate a dynamic range of 20 log10 210 60dB
only, which does not meet the requirement.
In order to solve the problem, we used two different preamplification gains to measure the sound levels in different
ranges. As shown in Figure 3, two MAX4465 pre-amplifiers
are in parallel arranged after the microphone. When the input
level is low, the high-gain channel is selected; otherwise
the low-gain channel is selected. A MAX4524 4-channel
multiplexer is applied to select proper channel with the
selection controlled by PD4 output pin of ATmega128. We
here denote the two gains, GL , GH , respectively. In our
design, GH = 390, GL = 10. Therefore, the dynamic range
G2
is increased by 10 log10 GH2 = 31.82dB correspondently.
L
The sampling program starts from measuring the highgain channel input and keeps tracing the ADC input level.
When a threshold is reached, PD4 is pulled low and the ADC
input is switched to the low gain channel. Figure 4(a) is the
oscilloscope snapshot of channel selection moment. It can
be seen, that when the high-gain input gets saturated, PD4 is
pulled down immediately and the low-gain input is selected.
When the low-gain channel is selected, and the sound level
goes below a threshold, the system will switch back to the
high-gain, as shown in Figure 4(b).
B. A-Weighting Filter
According to the transfer function HA (s), the Aweighting filter can be approximated by a cascading of 3
high-pass plus 2 low-pass filters. The frequency response of
such a filter is also plotted in Figure 2. Need to mention that
theoretically such a filter can be implemented by software
as a digital filter. However, a digital filter involves excessive
floating point calculation which definitely surpasses the limit
of 8-bit ATmega128.
V. D ESIGN A PPROACH : S OFTWARE
The software is designed to be able to perform both slow
and fast measurements: the RMS power calculation is done
every 125ms and 1s corresponding to F/S time-weightings
in IEC 61672, respectively.
A. 10-bit ADC setting
The 10-bit ADC is set in differential channel mode at
33kHz sampling rate. Such a sampling rate has frequency
response from 0 to 16.5kHz, which covers the range of
A-weighting filter well, therefore most of sound power is

3.5

lected. Thus we can accept an instantaneous peak voltage

to the reference of Vp|peak =1.5-Vn =Vn -1.0=0.25V. If we
assume that the input sound wave is a sinusoidal for the
first order approximation, theRMS voltage of microphone
input Vrms|mic =Vp|peak /GL / 2 = 17.7mV. The MCE-400
has sensitivity of 7.9mV/Pa, so 17.7mV corresponds to
prms =2.24Pa sound pressure. Applying equation (1), it indicates that our system is able to measure sound level up to
100.98dB.

3
2.5

Channels + Offsets

2
1.5
1
0.5

B. Signal sampling and averaging

0
0.5
High Gain input
Low Gain input
Aweigh Filter output
MUX select

1
1.5

0.8

0.6

0.4

0.2

0
0.2
Time(s)

0.4

0.6

0.8

(a) Oscilloscope record of channel selection

3.5
High Gain input
Low Gain input
Aweigh Filter output
MUX select

2.5

Channels + Offsets

1.5

0.5

The basic algorithm is to take samples and square each of

them. At the end of each 1ms, if GH is being applied, in case
the signal is too strong and clip has been observed, the MUX
is turned into GL channel; if GL is being applied, in case the
signal is weaker than a threshold , the MUX is turned into
GH . For every 125ms, an RMS average is calculated using
equation (2). Algorithm 1 shows the algorithm in details,
and Table I is the variable description for the algorithm.
N1
Nm
N1ms
xi
S1
Sm
H
c
C

RMS()

number of repeats in 1 second

number of 1ms repeats in different modes (i.e., I/F/S)
number of samples can be converted in 1ms
instantaneous sample
accumulative sum in 1 second
accumulative sum in different modes
threshold that a HIGHGAIN signal will clip the amplifier
counter for the HIGHGAIN channel saturated
threshold count that LOWGAIN channel should be selected
threshold that LOWGAIN channel gives too small input
RMS operation
Table I
VARIABLE D ESCRIPTION FOR A LGORITHM 1

0.5

1.5
0.02

C. Calibration
0

0.02

0.04

0.06

0.08
Time(s)

0.1

0.12

0.14

0.16

0.18

(b) Multiplexer selection pin activities

Figure 4.

High/low gain channel selection

picked up by ADC. The conversion formula is given as [15]:

ADC =

(Vp Vn ) GAIN 512

,
Vr

(3)

where Vn =1.25V is the negative input pin of ADC0,

Vr =2.56V is the selected voltage reference, and GAIN =10
is the ADCs internal gain. These three are constant values.
Vp is the instantaneous voltage of positive input pin ADC1
fed by the A-weighting filter. According to (3), Vp must be
in range of [1.0,1.5V] so that ADC output is in range of
[-512, 511].
When Vp is at boundaries of range [1.0,1.5]V, the
strongest sound pressure level is seen at the microphone
input, and in this case the low-gain channel GL is se-

Calibration was done with CESVA SC-20c, which is a

high performance Type I integrating-averaging sound level
meter. SC-20c is able to send measurement data to a host
PC via RS232 interface and the data can be stored in file
for further analysis.
The calibration acoustic source must be 1kHz sine wave.
The ADCrms values are compared with CESVA measurement in range of [30,90] decibels and an empirical curve
fitting equation is drawn as
LpA = 8.1522 ln(ADCrms ) + 32.7148(dB)

(4)

The calibration curve can be seen in Figure 5.

VI. N ETWORKING & S YNCHRONIZATION
The application requires multi-hop communication to send
sensed data back to a sink, and a global synchronization
is necessary to offer time-coherent noise information. We
proposed a SYNC2SINK application-oriented protocol to
accomplish the task. In this section we briefly describe
SYNC2SINK mechanism. The details of networking and

Algorithm 1 Sound Level Calculation Algorithm

Tsyn

for k=0 to N1 -1 do
S1 = 0
for i=0 to Nm 1 do
Sm = 0, c = 0
for j=0 to N1ms 1 do
readADC to xi
if xi > H then
c++
end if
Sm = Sm + x2i
end for
if HIGHGAIN then
if c C then
set LOWGAIN
end if
else
Sm = Sm (GH /GL )2
if Sm < then
set HIGHGAIN
end if
end if
q
1
RMS(Sm )= N N
Sm
m 1ms
end for
record max(RMS(Sm )) and min(RMS(Sm ))
S1 = S1 + RMS(Sm )
end for
S1
mean(S1 ) = N
1
record mean(S1 ),max,min

Tcom
Time

Sink

Network
nodes

Sampling ADC, constructing DATA frame

Figure 6.

Time

Timeline of the network operations

receiving a copy of it, and sets the sender of that SYNC

as its predecessor to sink, thus a passive route to Sink is
established for every node throughout the network. Then
the node goes into noise sampling phase which has been
described in previous section. During this phase, the node
turns off its radio transceiver so that the RF electromagnetic
interference to the sampling circuits is eliminated.
Each period is 5 seconds. The SYNC-broadcasting phase
is no longer than 20ms and negligible. Sampling phase is
4 seconds and totally 32 125ms Fast RMS values are
produced, leaving the last second for data communication.
Each data frame contains 64 bytes RMS values (2 bytes for
each 125ms) and some application layer overhead, resulting
in a total of 96 bytes of payload are pushed to the IEEE
802.15.4 MAC layer.

100
LpA= (log(ADC) + 4.0135)/0.1227 (dB)

90

Sound level LpA (dB)

Tsen
Send to database server via GPRS

VII. M EASUREMENT R ESULT & C OMPARISON

80

The final finish of our sensor node can be seen in

Figure 7. The nodes are sealed by metal cover and waterproof material so that they can be deployed outside.

70

60

A. Power consumption
50
CESVA
Fitting Curve

40

30

200

Figure 5.

400

600
ADC readings

800

1000

1200

Calibration and Curve Fitting

synchronization issues such as throughput control over

multi-hop and clock skew control are presented in another
document.
SYNC2SINK is built on CiNet protocol stack [3] (Fig. 1).
SYNC2SINK works periodically and each period consists of
three phases: SYNC broadcasting, noise sampling, and data
communication with durations denoted as Tsyn , Tsen , and
Tcom respectively (Figure 6). Each period is started by the
Sink node broadcasting a SYNC frame, which contains a
monotonically increasing sequence number and the current
time of the sink. Every node re-broadcasts the SYNC after

Instantaneous power consumption was tested by adding a

5.25 resistor next to the power supply. We measured the
voltage over the resistor and calculated the current using
Ohms rule. Similar method has been used in [16], [17]. In
order to indicate the phases mentioned above, an output pin
is triggered when there is a phase transition.
In each period T the C keeps reading ADC value and
calculates the sound level. At the end of each period the
radio module is turned on and a data frame is sent to the
sink. When a sensor node is not scheduled to do radio
communication, the radio module CC2420 is turned off to
save energy. In this state the node consumes a constant,
small amount of energy as shown in Figure 8(a). In general,
the sensor nodes have sleep schedules. However, in this
application the sensor nodes must keep sensing the noise
thus no sleep mode is applied to C. Nevertheless, we have
observed less than 1mA current drain during the sleeping
mode. We are interested in the real-time power consumption
when the node is active and its radio is turned on. Totally
five phases are needed to send a frame [18]:

Average = 26.90mA
60

50

Current (mA)

40

30

20

10
Power Consumption(mA)
Timeindication pin
0
200

200

400

600

800
1000
Time (ms*5)

1200

1400

1600

1800

(a) Power consumption in a sensing cycle (5sec)

Average = 36.74mA
180

160

Figure 7.

t t

Final finish of the sensor node

t

140

Current (mA)

120

1) Initialize radio module: the radio transceiver is turned

on and the oscillator starts to oscillate. This phase
takes a few milliseconds.
2) Frame formatting: a frame is copied to the radio buffer
from C. This phase duration varies, depending on the
frame length.
3) Waiting RSSI: the radio module waits for a clearance
of channel. This phase duration varies, depending on
the channel situation.
4) Random backoff: IEEE 802.15.4 uses CSMA/CA media access control (MAC), so a random backoff will
reduce the chance of collision.
5) Sending the frame: the frame is sent out. The duration
is proportional to the frame length.
The above phases can be seen in Figure 8(b).
As the time transitions given by the time-indication-pin
in Figure 8(b), the duration t2 t1 is the radio initialization
phase, which takes approximately 5ms. There is an impulse
power consumption at the beginning of this time when the
radio circuits oscillator starts to run.
The second duration t3 t2 indicates the local communication between ATmega128 and CC2420. In our test, a
total of 37 bytes (including MAC header 9 bytes, IP header
20 bytes and payload 8 bytes) were copied to CC2420.
SPI (Serial Peripheral Interface) was set at a bus clock rate
7.7MHz/16 so that 37 bytes will take 0.615ms to transmit.
Together with SPI bus commands, we have observed approximately 1ms for this phase. The power consumption raises
5mA from a floor of 30mA during this time.

100

80

60

40

20
Power Consumption(mA)
Timeindication pin
0

5
Time (ms)

(b) Power consumption during transmission

Figure 8.

Power consumption

The third time is short. This is the time that CC2420

acknowledges to the C that frame buffer is ready.
The next duration t5 t4 is the channel access time.
CC2420 must find a valid RSSI (Received Signal Strength
Indicator) before the attempting of transmission. During this
time the radio module keeps listening to the channel so a
few power peaks are observed.
The duration t6 t5 is the random back-off time of
CSMA/CA, and t7 t6 the frame transmitting time. During
the transmission, the power consumption stays on a floor of
60mA.
At last, during t8 t7 the radio module is turned off.
In average, sending a frame consumes 36.74mA in
approximately 10ms, and the entire cycle average is

1,5 m

84

3 sensor nodes
and Pulsar

Pulsar
Node Avg
Node Avg+
Node Avg

82

80

0,73 m

0,73 m

Sink node

Sound level (dB)

78

76

74

72

Figure 9.

70

Indoor test configuration

68

26.90mA. Consider a typical AA-type battery with capacity

of 2500mAh [19], it can support the node for 2500
3600/26.90 = 334570 seconds, which is nearly 93 hours,
which satisfies the demand. Our field test experience has also
proved this conclusion. Actually by network control/synchronization, the sensor nodes can be put into sleep mode,
and wake up on demand. The network life time can be
significantly increased by doing so.

66

C. Indoor accuracy test

All the nodes have been tested with Pulsar94, a type II
sound meter, in a noise-immune room. The test configuration can be seen in Figure 9. Nodes are tested and their
microphone pick-up points are very close to the sound meter
1.5m away from the sound source. The sound source is
traffic noise recorded at Colin McRae Gathering down the
M6, Sandbach [20]. In order to reduce the error, we put
Pulsar94 at different places in the array. The test duration is
30 minutes with each configuration 5 minutes. The result
is shown in Figure 10. The result shows that our nodes
give very close data to that of Pulsar94. The mean standard
deviation among our nodes is 0.91dBA throughout the all
test phases.
D. Outdoor accuracy test
The field test was conducted on an open ground close to a
motorway. We arranged 5 noise nodes in a line perpendicular
to the motorway with 10 meters separation one each other,
and Node 1 being the closest to the road. We put Cesva SC20c together with each node and measured for 5 minutes.
Because there is no synchronization between our network
and SC-20c, we gave an impulse sound to mark the start of
measurement. Figure 11 shows the results. We can see that
our nodes have less than 2dB error for all the nodes.

50

100

150
Time (sec)

200

250

300

Figure 10. Indoor test results, presented as 3 nodes average together with
Pulsar94. Deviations of nodes measurement are also given
100
Cesva
Node 5 ( 1.18)
Node 4 (0.34)
Node 3 (0.83)
Node 2 (0.26)
Node 1 (0.57)

90

B. Node cost

80

Sound level (dB)

The cost of one sensor node is about 100 Euros. Comparing to the price of commercial sound meters in market,
with the typical price for a class I sound meter 2000 Euros
or more, this WSN application is very competitive.

70

60

50

40

200

400

600

800

1000

1200

1400

1600

Time (s)

Figure 11.
Outdoor test motorway noise (The measurement errors
comparing to Cesva are given in the legend after each node, respectively)

VIII. C ONCLUSION & F UTURE W ORK

In this paper, the design of wireless sensor node for
environmental noise measurement is described. Technical
challenges during the design are reported and corresponding
solutions are explained. The comparison with standard Type
1 sound meter is made and the objective of design is
achieved. The power consumption is small enough so that
a typical AA-type battery can support the nodes for several
days of measuring without sleep. ATmega128 build-in A/D
converter is enough for environmental noise measurement
thus the cost of hardware platform is minimized.
The sensor node platform is built on ATmega128L with
4kB RAM, and its internal 10-bit ADC can operate a peak
sampling rate of 33kHz. Both of these two resources have
been pushed to their limits in our design. A better alternative
should definitely have wider ADC range and more RAM.
For this reason, we have decided our next development will

be continued on Jennic JN5139 platform, which has a 32-bit

architecture, 16MHz clock, 12-bit ADC, and 96kB RAM.
R EFERENCES
[1] European Commission, Directive 2002/49/ec of the european
parliament and of the council of 25 june 2002 relating to the
assessment and management of environmental noise, Official
Journal of the European Communities, 07 2002.
[2] I. Hakala, M.Tikkakoski, and I. Kivela, Wireless sensor
network in environmental monitoring - case foxhouse, in
Proceedings of the Second International Conference on Sensor Technologies and Applications (SENSORCOMM 2008),
Cap Esterel, France, August 25-31 2008.
[3] I. Hakala and M. Tikkakoski, From vertical to horizontal
architecture: a cross-layer implementation in a sensor network
node, in InterSense 06: Proceedings of the first international
conference on Integrated internet ad hoc and sensor networks.
New York, NY, USA: ACM, 2006, p. 6.
[4] S. Santini, B. Ostermaier, and A. Vitaletti, First experiences
using wireless sensor networks for noise pollution monitoring, in REALWSN 08: Proceedings of the workshop on Realworld wireless sensor networks. New York, NY, USA: ACM,
2008, pp. 6165.
[5] Moteiv Corporation. Now Sentilla (www.sentilla.com),
TMote Sky Datasheet, Date retrieved: 20th of April
2010. [Online]. Available: http://sentilla.com/files/pdf/eol/
tmote-sky-datasheet.pdf
[6] G. Barrenetxea, F. Ingelrest, G. Schaefer, and M. Vetterli,
Wireless Sensor Networks for Environmental Monitoring:
The SensorScope Experience, in The 20th IEEE International Zurich Seminar on Communications (IZS 2008), 2008,
invited paper.
[7] K. Lu, Y. Qian, D. Rodriguez, W. Rivera, and M. Rodriguez,
Wireless sensor networks for environmental monitoring applications: A design framework, in Global Telecommunications Conference, 2007. GLOBECOM 07. IEEE, Nov. 2007,
pp. 11081112.
[8] M. Li and Y. Liu, Underground coal mine monitoring with
wireless sensor networks, ACM Trans. Sen. Netw., vol. 5,
no. 2, pp. 129, 2009.
[9] Metravib.com. Wed007 noise dosimeter exposimeter. Web
page. Date retrieved: 20th of April 2010. [Online].
Available: http://www.01db-metravib.com/environment.13/
products.16/wed007.460/?L=1
[10] W. Morris, Ed., The American Heritage Dictionary of the
English Language. Houghton Mifflin, 2000.
[11] W. Boyes, Instrumentation Reference Book, 3rd ed.
Butterworth-Heinemann, 2002.
[12] IEC 61672 Ed.1.0, Electroacoustics - Sound level meters,
Electroacoustics Std., 2003.
[13] World Health Organization, Occupational and community
noise, Web page. Date retrieved 20th of April 2010, 02
2001. [Online]. Available: http://www.who.int/mediacentre/
factsheets/fs258/en/

[14] Monacor International. Mce-400 high-quality electret

microphone cartridge. Web page. Date retrieved: 20th
of April 2010. [Online]. Available: http://www.monacor.de/
typo3/index.php?id=84&L=1&artid=5171&spr=EN&typ=full
[15] ATmega128 8-bit AVR microcontroller datasheet, ATMEL
Co.Ltd.
[16] E. Erdogan, S. Ozev, and L. Collins, Online snr detection for
dynamic power management in wireless ad-hoc networks, in
Research in Microelectronics and Electronics, 22 2008-April
25 2008, pp. 225228.
[17] B. Hohlt, L. Doherty, and E. Brewer, Flexible power scheduling for sensor networks, in IPSN 04: Proceedings of the 3rd
international symposium on Information processing in sensor
networks. New York, NY, USA: ACM, 2004, pp. 205214.
[18] TI CC2420 2.4GHz IEEE 802.15.4/ZigBee-ready
Transceiver Datasheet, Texas Instruments.

RF

[19] A. L. P. Aroul, A. Manohar, D. Bhatia, and L. Estevez,

Power efficient multi-band contextual activity monitoring for
assistive environments, in PETRA 08: Proceedings of the 1st
international conference on PErvasive Technologies Related
to Assistive Environments. New York, NY, USA: ACM, 2008,
pp. 17.
[20] Youtube video, Colin mcrae gathering down the m6,
sandbach, Date retrieved: 20th of April 2010. [Online].
Available: http://www.youtube.com/watch?v=QXQEP37fxU