A HYBRID SENSOR NETWORK FOR
The project investigates a wireless, acoustic sensor network application—monitoring
cane toads in Kakadu National
Park, Northern Territory, Australia. Our goal is to use
automatic recognition of animal vocalizations to detect the existence of cane
toads. This is a challenging application as it requires high frequency acoustic
sampling, complex signal processing and wide area sensing coverage.
We set up two prototypes of wireless sensor networks that recognize
vocalizations of up to 9 frog species in northern Australia. The first prototype is
simple and consists of only resource-rich Stargate devices. The second prototype
is more complex and consists of a hybrid mixture of Stargates and
inexpensive, resource-poor Mica2 devices operating in concert. In the hybrid
system, the MICA2s are used to collect acoustic samples, and expand the sensor
network coverage. The Stargates are used for resource-intensive tasks such as
Fast Fourier Transform (FFT) and machine learning. To enable the hybrid system,
we design and incorporate three algorithms to account for the sampling, processing
and communication bottlenecks of the MICA2s:
high frequency sampling
compression and noise reduction, to reduce data
transmission by up to 90%
sampling scheduling, which exploits the sensor
network redundancy to increase effective sample processing rate
We evaluate the performance of both systems over a range of scenarios, and
demonstrate that the feasibility and benefits of a hybrid systems approach
justify the additional systems complexity.
Cane-toad Monitoring Issue
Cane toads (Bufo marinus) were introduced to control sugar pests in Australia about
seventy years ago. Their continuous spreading, due to lack of natural predators
through-out northern Australia
have caused major concerns by biologists and ecologists. The dark region in
following figure shows the current distribution of cane toads in northern Australia.
Kakadu National Park is a World Heritage area,
and its status is threatened by the possibility of colonization by cane toads.
Cane toads have recently been spotted at approximately 15 to 30 km to the east
of the park, which makes it important to monitor them in this area.
MICA mote family: MICA2 is the
third generation of Berkeley
mote manufactured commercially by Crossbow. It
has a 7.7MHz Atmega processor and 512KB on-board flash memory. It can
transmit at a maximum data rate of about 19 kbps and is powered by two AA
size batteries. Its recent cousin MICAz has a ZigBee compliant RF transceiver
and can support up to 250 kbps transmission rate. We use the MICA2 sensors as
our resource-poor sensors.
X-Scale Single Board Computer - Stargate,
manufactured by Crossbow, is a high
performance processing platform that offers much more resources than MICA
motes in terms of computation power, memory, energy, and transmission
capability. It operates on a 400MHz Intel PXA 255 processor and has 96MB memory
in total (64MB SDRAM and 32MB flash memory). It can be powered by a li-Ion
battery and can support Wi-Fi (11mbps when using IEEE 802.11b) transmission.
We use Stargates as our resource-rich sensors.
High-level System Structures
Figure 1: Stargate-only System Structure
The system samples acoustic data using a desktop microphone.
The sound spectrogram is then generated to get signals in frequency domain from
input in time domain. The sound attributes, including local peaks and other
necessary variables, are extracted from the spectrogram and used as the input of
machine learning classifiers, one for each frog species. To increase correctness
and reliability of the recognition, a hierarchical recognition structure is
employed, termed as voting process in above figure.
Hybrid System Structure
Figure 2: Hybrid System Structure
In this system, instead of sampling acoustic data using desktop
microphone, we use 2 MICA2s to sample acoustic data and compress it before
sending to the Stargate via radio channel. Upon receiving data from the satellite
motes, the Stargate will decompress received data before processing it. As more
than 1 MICA2 is used, a scheduling algorithm is designed and employed to have
them sense acoustic data in alternate periods.
Result Screen Shots
In this experiment, two frogs (Bufo marinus/cane toad, Cyclorana cryptotis)
are detected successfully by our system. More screen shots:
Design and Evaluation of a Hybrid Sensor Network for Cane-toad Monitoring".
Wen Hu, Van Nghia Tran, Nirupama Bulusu, Chun-tung Chou, Sanjay Jha, Andrew
Taylor. In Proceedings of Information Processing in Sensor Networks (IPSN
2005/SPOTS 2005), Los Angeles, CA, April 2005. To appear.
- "Cane-toad Monitoring
in Kakadu National Park Using Wireless Sensor Networks". S. Shukla,
N. Bulusu and S. Jha. Network Research Workshop 2004, Cairns, Australia
in 2-7 July 2004 as part of 18th APAN Meetings. This paper addresses the
complementary problem of finding deployment hotspots for cane-toad monitoring.
Presentation and News
We are porting the system from Mica2 to MicaZ to get higher bandwidth between Mica and Stargate. (Jan. 2005)