Compressive Data Gathering using NACS in Wireless Sensor Network

International Research Journal of Engineering and Technology (IRJET) e-ISSN: 2395 -0056
Volume: 04 Issue: 03 | Mar -2017 www.irjet.net p-ISSN: 2395-0072
© 2017, IRJET | Impact Factor value: 5.181 | ISO 9001:2008 Certified Journal | Page 909
Compressive Data Gathering using NACS in Wireless Sensor Network
Sabari Girison P.S.1, Vinoth S.2, Ranjeeth Kumar S.O.3, Veeralakshmi P.4
1,2,3Student, Department of IT, Prince Shri Venkateshwara Padmavathy Engineering College, Chennai, Tamilnadu,
India
4Associate Professor, Department of IT, Prince Shri Venkateshwara Padmavathy Engineering College, Chennai,
Tamilnadu, India
---------------------------------------------------------------------***---------------------------------------------------------------------
Abstract - Wireless sensor networks are useful in such area
where the human being is unable to go and monitor.
Compressive sensing (CS) has been used widely for the
gathering of data in wireless sensor networks for the purpose
of the communication overhead recent years. Structured
Random Matrix (SRM) offers a practical method to sampling.
It has high sparsity, low complexity, fast computation
properties and has good sensing performance comparable to
that of completely random sensing matrices. In this paper,
Neighbor-Aided Compressive Sensing(NACS) scheme is
proposed for efficient gathering of data without any data loss
in spatial and temporalcorrelatedWSNs. Duringeverysensing
period, the sensor node sends the raw readings within the
sensing period to a uniquely selected shortest neighbor.
Simulation results demonstrate that compared with the
conventional KCS (Kronecker Compressive Sensing) models
and SRM, the proposed NACS model can achieve efficient data
gathering and recovery performance with much fewer
transmissions.
Key Words: Compressive sensing, WSNs, Data gathering,
Structured Random Matrix,KroneckerCompressiveSensing,
Neighbor-Aided Compressive Sensing.
I. INTRODUCTION
Wireless Sensor Networks [1] (WSNs) is a collection of
sensors that are spatially connected together with the
network to monitor the environmental and physical
condition such as pressure, sound, temperature, humidity
etc. and transfer the data to the server location. Fig.1 shows
the connection of sensor nodes with gateway node. Wireless
sensor network is used in some of the applications like
precision agriculture, medicine and health care, machine
surveillance and preventive measures and so on.
Compressed Sensing or Compressive Sensing [2] is about
acquiring and recovering a sparse signal inthemost efficient
way possible (subsampling) with the help of an incoherent
projecting basis. Unlike conventional sampling
methods, Compressive Sensing provides a new framework
for acquiring sparse signals in a mutiplexed manner.
Compressive Sensing (CS) provides a new approach to
simultaneous sensing and compression that promises a
potentially large reduction in sampling costs and the
required number of measurements to recover the original
signal. The main requirement for CS is that the signal to
compress has to be sparse in some basis to be able to
recover the original signal from the shorter compressed
version.
Fig -1: Wireless sensor network
Structured Random Matrix [3] (SRM)isa sensing matrixthat
offers a practical method of sampling. It has high sparsity,
low complexity, fast computationpropertiesandhassensing
performance comparable to that of completely random
sensing matrices.
Kronecker compressive sensing [4] (KCS) is recently
introduced compressive sensing method to exploit general
correlation patterns by combining the possibly distinct
sparsifying bases from each signal dimension into a single
basis matrix. In terms of improving compression
performanceanddecreasing sensor energyexpenditurewith
signals featuring typical WSN data characteristics, KCS has
been empirically shown to outperform.
Neighbor-Aided CompressiveSensing[5](NACS)scheme isa
proposed system for data gathering throughwirelesssensor
network for transfer of data to sink node with efficient
performance.
Notations: We use boldface letters to denote vectors
(lowercase) and matrices (capital), andcalligraphyletters to
denote sets. An entry of matrix A at its i-th row and j-th
column is denoted as aij . The matrix A of size N × N is
denoted as AN and when A is the i-th matrix of a matrices
set, it is denoted as A(i) . (·)T denotes the matrix transpose,
⊗ denotes the Kronecker product, vec (A) stacks the
columns of A into a column vector, and ℜs,t (A) reshapes
matrix A of size s × t to a matrix of size p × q (st = pq).

II. EXISTING SYSTEM
SRM is a conventional technique used for a practical method
of sampling data. It has low complexity, high sparsity and
also fast computation properties. Sensing performance of
SRM is comparable to that of randomly sensing matrices.
And, KCS compressive technique is used to combine distinct
sparsity bases from each signal dimension into single basis
matrix. In terms of improvingcompressionperformance and
decreasing sensor energy expenditure withsignalsfeaturing
typical WSN data characteristics, KCS has been empirically
shown to outperform single-dimensional Compressive
sensing approaches. However, the SRM is not applicable for
WSNs and the KCS model suffers increased data dimension
which could result in degraded recovery performance.
III. PROPOSED SYSTEM
To improve the sensing performance and the energy
efficiency of compressive data gathering there was several
different contributions were made. Fig 2 describes the full
flow of data from source node to sink node. Firstly, a
neighbor aided compressive data gathering framework is
proposed to exploit both spatial and temporal correlations
with less data transmissions. In this technique, the sensor
node just send the raw data with consecutive time slots to
shortest neighbor node. The neighbor node applies the
compressive sensing measurementandsendsthedata tothe
sink node. The NACS requires only one neighbor node for
each sampling node, resulting in lower communication cost
and higher energy efficiency. Secondly, we generalized the
conception of KCS and proved that the equivalent sensing
matrix can be constructed by the Kronecker product of
temporal and spatial sensing matrices. Last but not least, as
the main contribution, the relationship between KCS and
SRM is studied, and the sensingperformanceofconventional
KCS is further improved by introducing the idea of SRM to
KCS to form equivalent sensing matrices.
Fig - 2: System architecture
IV. MODULES
1) Configuring the nodes
Firstly, M of N nodes are randomly and uniformly selected
for gathering. And gathering commands are sent to these
nodes by the sink node. As shown in Fig 3 Each node of the
sensor is represented by different colors. Note these nodes
can also be activated by predefined periodic scheduling
scheme in practice.
Fig - 3: creation of nodes
2) Forwarding data to neighbor node
Once a node received the gathering commands, it randomly
selects a neighbor. Then, it uses its original sensor readings
to form a transmission packet. After that, the nodetransmits
the packet to the selected neighbor. Each node is settoact as
a unique role between generator and neighbor, namely a
node can only be sampling node or the unique neighbor of a
specified sampling node. As shown in fig. 4 data is
transferred from node 3 to node 5 with less data loss. If a
sampling node receiveda transmissionpacket,thenthe node
selects another neighbor of itself randomly and uniformly.
The node updates the packet and transmits the packettothe
updated neighbor. In case of the received second packet and
the later packets, the neighbor node relays the packet to its
neighbor as above. After this procedure, the updated
neighbor ID list b′ satisfies.
Fig - 4:Forward data to neighbor node
3) Compressing data packet
When a neighbor node received a sensing packet, firstly,
it mixes the received data and its own data by Secondly, it
permutes the vector using a fast permutation algorithm
(Algorithm 1), whose computational complexity is o(N).The
fast permutation operation is equivalent to left product a

permutation matrix P(c) Thirdly, the mixed data isreshaped
to a data block of size and fast transform of the data block is
computed with an orthogonal transformmatrix W(c).Atlast,
the transformed data block is vectored and uniformly down
sampled at random by a random down samplingmatrix D(c)
as measurements. Then the packet is updated.
Algorithm 1: Fast Permutation Algorithm
Input: Raw data sequence: λc = (λ1, λ2, . . . λn )
Maximum number of iteration:n
Output: Permuted data sequence: λc = (λ1, λ2, . . . λn )
for i from 1 to n do
Generate a random number r ∈ [1, i]
Exchange λi and λr
end for
4) Gathering neighbor node data
The proposed NACS scheme considers a realistic scenario,
where the sensor readings across all the nodes exhibit both
spatial and temporal correlations. Denote that the routing
protocol has been initialized and the unique ID and the
random seed for each node have been set. During every
sensing period t, the sensor network of n nodes produces a
compressible sensor reading block X ∈ R n×t . Let X = (x1, x2,
. . ., xn), where xi = ( x1,i , x2,i , . . ., xt,i )T is the i-th column of
the reading block. The node with ID l, l ∈ χ is denoted as Θl
and its generated (excluding the modified)packetisdenoted
as θl . The data structures of θl is given by
(1)
Where the .src field is the ID of the node who generates it,
the .nbr field is the ID of the node who processes it, and the
.dat field stores the data or payload of the packet. 1)
Initialization: Firstly, M of N nodes are randomly and
uniformly selected for gathering. The IDs of the selected
nodes are denoted by l = (l1, l2, . . ., lm) ∈ χ. And gathering
commands are sent to these nodes by the sink node. Note
these nodes can also be activated by predefined periodic
scheduling scheme in practice. 2) Forwarding: Once a node
Θl , l ∈ l received the gathering commands, it randomly
selects a neighbor Θb b ∈ χ. Then, it uses its original sensor
readings to form a transmission packet by
(2)
After that, the node transmits the packet to the selected
neighbor. Each node is set to act as a unique role between
generator and neighbor, namely a nodecanonlybesampling
node or the unique neighbor of a specified sampling node. If
a sampling node received a transmission packet θl , then the
node selects another neighbor of itself randomly and
uniformly, whose ID is denoted by b ′ . The node updates the
packet by θl.nbr = b ′ and transmits the packet to the
updated neighbor. In case of the received second packet and
the later packets, the neighbor node relays the packet to its
neighbor as above. After this procedure, the updated
neighbor ID list b ′ satisfies b ′ ∩ l = ∅ and for any i , j, b ′ i , b ′
j .
V. NETWORK MODEL
We consider a single-sink multi-hop WSN for data
gathering, which consists of N sensors, with identification
numbers (ID) of χ = {1, 2, . . . , N }, capable of transmitting,
receiving and relaying data. The sensors are deployed
randomly and uniformly in a unit square area toperiodically
monitor data at a pre-defined rate, and to disseminate the
acquired information
to the sink. For each sensing period, the sink is responsible
for obtaining an accurate reconstruction of the monitored
field, i.e., to recover the readingsduringthesensingperiodof
all N sensors. All the nodes are assumed to have an identical
transmission radius r, and thus anytwonodesareconnected
if their distance is smaller than r. It is further assumed that
the condition r2 > ln (N )/(π N ) is satisfied, which
guarantees the connectivity of the whole work with high
probability. The sensor observations are assumed to
encompass both spatial and temporal correlation,typical for
various environmental sensing applications in densely
deployed WSNs. For formation of wireless sensor network
we consider the 64 sensor nodes forinitialize.Asweproceed
the source node selects the nearest neighbor node, then the
CS measurement is made in the neighbor node and the
compressed sensing data is passed into the sink node.
VI. PERFORMANCE COMPARISON
Performance comparison is a final simulation of network to
compare the data gathering performance of other models
like KSRM-0,KSRM-1, KGAU-0, KGAU-1, GAU-G with the
proposed system model of NACS-0, NACS-1 to show as
shown in fig.6.1 that the performanceof NACSismuchbetter
than other model like global Gaussian (GAU-G), the KCS
model with sensing matrices of each signal dimension as
Gaussian matrices (KGAU-0) represents the optimal

performance of conventional KCS model Note that the
analysis was taken directly without changing of sensing
parameters, In other words, the NACS model achieved
superior sensing performance with much fewer
transmutations (energyconsumption)thantheconventional
KCS model. The total numbers of transmitted packets of
these three models were counted with the sensing period
increasing from 10 to 500. The resultant graph ofsimulation
is based on the axis of signal sparsityandprobabilityofexact
recovery.
Chart 1: Performance curves: probability of exact recovery
versus sparsity K. The curves indicate that the performance
of NACS model outperforms than that of the conventional
KCS model.
VII. CONCLUSION AND FUTURE WORK
NACS is proposed in this paper under the framework of KCS
for efficient compressive data gathering of spatial and
temporal correlated WSNs. NACS has improved the
performance of conventional KCS, meanwhile, greatly
decreased the data loss in WSNs. However, the recover
complexity might be slightly increased. Fortunately, the
decoder side always haspowerful computeabilityinrealistic
scenarios. We will take the fast and paralleled recovery
algorithm as our future work.
VIII. REFERNCES
1. W.Wang,M. Garofalakis, and K.Ramachandran,
“Distributed Sparse Random Projections for Refinable
Approximation,” Proc.ACM/IEEE 6th Int. Conf. Inf.
Process. Sensor Netw. (IPSN ’07),Apr.2007.
2. Haifeng Zheng, Feng Yang, Xiaohua Tian, Member, IEEE,
Xiaoying Gan, Xinbing Wang, Senior Member, IEEE, and
Shilin Xiao, Member, IEEE “Data Gathering with
Compressive Sensing inWirelessSensorNetworks ”Jan-
2015
3. Ciancio, S.Pattem, A. Ortega and B. Krishnamachari,
“Energy-Efficient Data Representation and Routing for
Wireless Sensor Networks Based on a Distributed
Wavelet Compression Algorithm,” Proc. ACM/IEEE 5th
Int. Conf. Inf. Processing Sensor Netw. (IPSN ’06), pp.
309-316, 2006
4. Xuangou Wu, Yan Xiong, Mingxi Liy and Wenchao
Huang“Distributed Spatial-Temporal CompressiveData
Gathering for Large-scale WSNs“ School of Computer
Science and Technology, University of Science and
Technology of China,Hefei,China StateKeyLaboratories
of Transducer Technology, Shanghai,
Email:wxgou@mail.ustc.edu.cn,yxiong@ustc.edu.cn
keeper@gmail.com,wchuang@ustc.edu.co,IEEE,2013
5. Riccardo Masiero, Giorgio Quer, Michele Rossi and
Michele Zorzi “A Bayesian Analysis of Compressive
Sensing Data Recovery in Wireless Sensor Networks”

Compressive Data Gathering using NACS in Wireless Sensor Network

More Related Content

What's hot (17)

Similar to Compressive Data Gathering using NACS in Wireless Sensor Network (20)

More from IRJET Journal (20)

Recently uploaded (20)

Compressive Data Gathering using NACS in Wireless Sensor Network