A Wireless Network of Acoustic Sensors for Environmental Monitoring

Abstract

A distributed microelectronic system for the sound/acoustic monitoring of areas of environmental interest, based on a wireless network of acoustic sensors (microphones), and the automated generation of multi-level sound maps for environmental assessment has been recently proposed. This contribution focuses on the relation between the density of the wireless nodes and node operational parameters, like the required communication rate, data compression and power autonomy, in each one of the proposed area coverage schemes/ network architectures. The results provide the grounds both for the selection of the hardware architecture on network and on node level, but also for the decision on the distribution of the processing effort between the local and the central processing units.

Full text

A Wireless Network of Acoustic Sensors for Environmental Monitoring
Stelios M. Potirakis1,a, Bilel Nefzi2,b, N.-A. Tatlas1,c, Gurkan Tuna3,d, and M.
Rangoussi1,e
1Department of Electronics Engineering, Technological Education Institute (TEI) of Piraeus,
Aigaleo, Greece
2Independent Research, Paris, France
3Department of Computer Programming, Trakya University, Edirne, Turkey
aspoti@teipir.gr, bbilel.nefzi@live.com, cntatlas@teipir.gr, dgurkantuna@trakya.edu.tr,
emariar@teipir.gr
Keywords: wireless sensor networks, acoustic sensors, environmental monitoring.
Abstract. A distributed microelectronic system for the sound/acoustic monitoring of areas of
environmental interest, based on a wireless network of acoustic sensors (microphones), and the
automated generation of multi-level sound maps for environmental assessment has been recently
proposed. This contribution focuses on the relation between the density of the wireless nodes and
node operational parameters, like the required communication rate, data compression and power
autonomy, in each one of the proposed area coverage schemes/ network architectures. The results
provide the grounds both for the selection of the hardware architecture on network and on node
level, but also for the decision on the distribution of the processing effort between the local and the
central processing units.
Introduction
Wireless sensor networks (WSNs) have been one of most exciting and attractive topics in recent
years with variety of applications in the area of monitoring, control and sensing [1]. Electronic
monitoring of areas of environmental interest has recently attracted considerable research interest, at
a European and international level, e.g., [2], [3] and [4]. Although wireless sensors networks are
becoming “standard” technology, e.g., [5,6], the requirements of an open-air recording pose
practical limitations to the use of off-the-shelf networking solutions, thus leading to a custom
hardware, network architecture.
This paper presents an investigation concerning the main design parameters of a WSN that will
be the backbone of a sound mapping environmental assessment system [7,8], therefore employing
acoustic sensors. This kind of WSNs are often referred to as wireless acoustic sensor networks
(WASNs), e.g., [9]. They are challenging in terms of network topology, wireless protocols, used
bandwidth, and energy consumption. Some of these parameters are key hardware (HW) design
parameters. In this context, we performed a set of network simulations to investigate how are these
key HW design parameters compete with each other for specific network architecture and wireless
protocols.
WASN-Based Sound Maps for Environmentally Sensitive Areas
A WASN-based sound maps development for environmentally sensitive areas calls for the
effective coverage of a specified area of interest. A network like this could follow the scalable
architecture presented in Fig. 1. If the area of interest is narrow the one level solution depicted in
Fig. 1a can be employed, while in wider areas a scaling to the two level solution of Fig. 1b is
needed. According to the proposed approach, the same basic HW is upgradable and may be used as
any of the three involved units.

(a)
(b)
Fig. 1. WASN for environmental monitoring (a) Single-level setup (WLAN) of wireless acoustic
sensors. (b) Two-levels setup (WLAN-WAN), for wider area coverage.
WLAN setup. In its simplest form, the network shown in Fig.1a is a WLAN setup that provides
local area wireless network of acoustic sensors of relatively confined (narrow) areas. First, the
Integrated Peripheral Unit (IPU) collects the raw audio data through a digital microphone, and
temporarily stores them. In a following step, it transmits the data wirelessly to the Central Network
Unit (CNU). In order to reduce the bandwidth required for this transmission, data can be
compressed. The development of both the HW and embedded software (SW) of the system is
greatly simplified if the IEEE 802.11 protocol is selected for the IPU-NU communication.
WAN setup. In order to increase the area of coverage of the remote network, a different
architecture needs to be considered. This is shown in Fig.1b, where the IPU transmits all its data to
the Intermediate Network Unit (INU). The INU gets the data through the 802.11 network and then
repackages it for transmission on a different wireless network with much longer range. Although
there are more advanced solutions, the design decision was to use a 2G/3G network. Similarly to the
LAN setup, the CNU will receive the 2G/3G packets. Finally, the data will be formatted so that they
can be sent to the database computer (remote server) for handling and storage.
Competing Performance Factors: A Simulation-Based Investigation
In order to proceed to the selection of the HW architecture on network and on node level, but
also for the decision on the distribution of the processing effort between the local and the central
processing units, an elaboration is needed on the competing performance factors of the proposed
WASN. In this context, we performed a number of network simulations using the widely employed
OPNETTM simulation platform.
We considered two simulation scenarios composed of three WLAN networks and a WAN
network in a two-level setup following the architecture of Fig. 1.b, varying different parameters. It is
noted that each INU has a dual wireless card. The data rate was considered to be 11 Mbps for both
the WAN and the WLANs. This way, both scale cases presented in Fig. 1 are covered; either
collecting the data to CNU using 802.11, in the small scale example, or UMTS, in the large scale
example. We finally suppose that the transit delay in the WAN core network is negligible.
Therefore, we mainly study the access delays from the IPUs to their INU and from the INU to the
CNU in both directions. In the following, the term “duty-cycle” stands for the percentage of time for
which the node is generating data, also raw audio data were considered to be 16-bit/48kHz sampled.
In the first scenario, three WLANs consisted of five transmitting nodes (IPUs) each were
transmitting, while all possible combinations for the duty cycle among the set of values
{0.5,1,2,3,4,5,10,33,50,66,100} (%) and the load, resulting from different possible compression
rates, among the values {32,64,128,192,384,768} (kbps) were investigated. The network
performance was measured in terms of end-to-end delay and success rate. The objective of this
scenario was to determine the maximum duty-cycle for each audio quality so that the network

performance remains stable. The achievement of a success rate better than 90% was taken as an
indication of network stability. The obtained “stable” maximum duty cycles are given in Table 1.
Load (kbps)
768
384
192
128
64
32
Best duty cycle (%)
1
3
5
10
100
100
Table 1. Best duty cycle for different compression rates of 16-bit/48kHz audio.
From the obtained results we conclude that a raw (16-bit/48kHz) audio is a particularly heavy
load. In this case (768 kbps), the requirement for a stable network is that the overall audio events’
duration per transmitting node should be ≤ 1% of the total time of node operation. Nevertheless, the
recorded audio events can be more frequent if the acquired sound undergoes compressive coding.
Therefore, it is clear that the IPUs should perform some kind of encoding of adequate compression
rate. The higher the compression rate, the better the time coverage offered by the WASN.
In the second scenario, the number of IPUs per WLAN was varied among 5, 10 and 15 nodes per
WLAN, the load was considered to be constant at 64 kbps for all the investigated duty cycles, while
all other parameters were kept the same as in the first scenario. The objective of this scenario was to
determine the maximum duty-cycle for each network density so that the network performance
remains stable. The obtained “stable” maximum duty cycles are given in Table 2.
Number of nodes per WLAN
5
10
15
Best duty cycle
100
33
33
Table 2. Best duty cycle for different network densities and 64 kbps load.
From the results obtained for the second scenario, it is clear that the network density also affects
the network performance and therefore the maximum possible monitored time percentage. The
conclusion yielded from the up to this point presented analysis is that, further to the mandatory data
encoding before the wireless transmission, a careful network planning is necessary to achieve the
best possible environmental monitoring performance.
Furthermore, a WASN designed to be deployed in environmentally sensitive areas, should be
power autonomous to a high degree. Energy-wise, the best solution would be battery powered IPUs
and INUs that would be capable of reliably operating for as long as possible. Based on the above
presented simulation scenarios we investigated the energy consumption of such a WASN. Actually,
we focused on the energy consumption related to the wireless communication and the required
digital signal processing for the data compression prior to their transmission. We proceed to a case
study involving specific HW solutions for the main HW modules involved (the WiFi module, the
UMTS module and the microprocessor system) in order to present a realistic estimation of the
possible energy demands for the different simulated cases.
0
0.02
0.04
0.06
0.08
0.1
0.12
020 40 60 80 100
Power (W)
Duty Cycle
32
64
128
192
384
768
0
0.05
0.1
0.15
0.2
0.25
0.3
0.35
0.4
020 40 60 80 100
Power (W)
Duty Cycle
32
64
128
192
384
768
Fig. 2. Typical IPU (left) and INU (right) total power consumption of as a function of duty cycle
for different load values (compression rates), as calculated for the first simulation scenario.

Audio bitrate
768kbps (PCM)
384 (MP3)
192 (MP3)
128 (MP3)
Power Consumption
270 mW
328mW
325mW
320mW
Table 3. Microprocessor system (SoC) power consumption.
From Fig. 2 it seems that only a high degree of compression and for high duty cycles makes clear
difference on power consumption. Concerning the microprocessor system power consumption we
obtained the average power consumption shown in Table 3 for different MP3 cases.
Conclusions
We proposed a wireless acoustic sensors network for environmental monitoring that uses 802.11
at the WLAN level and UMTS, if necessary for wider area coverage, at the WAN level. By network
simulations, we concluded that the data encoding is mandatory before the wireless transmission,
while the higher the compression rate the higher the usable time for monitoring. Network density is
further competing with time-coverage; therefore a careful network planning is needed. The energy
consumption of course raises both with processing effort and transmission rate, therefore the
decisions regarding the HW and embedded SW should also take this into account.
Acknowledgments
Research co-funded by the EU (European Social Fund) and national funds, action “Archimedes III-
Funding of research groups in T.E.I.”, under the Operational Programme “Education and Lifelong
Learning 2007-2013”.
References
[1] N. Saputro, K. Akkaya, S. Uludag, A survey of routing protocols for smart grid
communications Computer Networks. 56 (2012) 2742-2771.
[2] G. Barrenetxea, F. Ingelrest, G. Schaefer and M. Vetterli, Wireless sensor networks for
environmental monitoring: the sensorscope experience, in Proc. 20th IEEE Intl. Zurich Seminar
on Communications (IZS 2008). (2008) 98-101.
[3] J. Tomić, M. B. Živanov, M. Kušljević, Đ. Obradović, J. Szatmari, Realization of measurement
station for remote environmental monitoring, Key Engineering Materials. 543 (2013) 105-108.
[4] F. Gui, X. Q. Liu, Design for multi-parameter wireless sensor network monitoring system based
on zigbee, Key Engineering Materials. 464 (2011) 90-94.
[5] I. F. Akyildiz, W. Su, Y. Sankarasubramaniam, E. Cayirci, Wireless sensor networks: a survey,
Computer Networks. 38 (2002) 393-422.
[6] J. Yick, B. Mukherjee, D. Ghosal, Wireless sensor network survey, Computer Networks. 52
(2008) 2292-2330.
[7] I. Paraskevas, S. M. Potirakis, I. Liaperdos, M. Rangoussi, Development of automatically
updated soundmaps for the preservation of natural environment. Journal of Environmental
Protection. 2 (2011) 1388-1391.
[8] M. Rangoussi, S. M. Potirakis, I. Paraskevas, On the development and use of sound maps for
environmental monitoring, in Proc. 128th AES Convention. (2010) paper no. 8113.
[9] A. Bertrand, Applications and trends in wireless acoustic sensor networks: A signal processing
perspective, in Proc. 2011 18th IEEE Symposium on Communications and Vehicular
Technology in the Benelux (SCVT). (2011), 1-6. doi: 10.1109/SCVT.2011.6101302.