A Hybrid Scheme for Video Transmission Over Wireless ... - IEEE Xplore

3 downloads 268 Views 211KB Size Report
routing of multimedia content over Wireless Multimedia. Sensor Networks. The proposed scheme comprises the advantages of both energy efficient hierarchical ...
17th Mediterranean Conference on Control & Automation Makedonia Palace, Thessaloniki, Greece June 24 - 26, 2009

A Hybrid Scheme for Video Transmission over Wireless Multimedia Sensor Networks Dionisis Kandris∗ , Michail Tsagkaropoulos †, Ilias Politis† , Anthony Tzes † and Stavros Kotsopoulos † ∗ Dept.

of Electronics, Technological Educational Institution of Athens, Greece 12210 Email: [email protected] † Dept. of Electrical &Computer Engineering, University of Patras, Patras, Greece 26500 Email: {mtsagaro,ipolitis,tzes,kotsop} @ece.upatras.gr

Abstract—The introduction of multimedia capable sensors has posed new challeges for classic Wireless Sensor Networks that are henceforth required to provide both power and quality efficent routing. This paper proposes an innovative protocol for power and perceived QoS aware routing of multimedia content over Wireless Multimedia Sensor Networks. The proposed scheme comprises the advantages of both energy efficient hierarchical routing and video packet schedulling according to the packet’s importance. The hierarchical routing algorithms are often utilized for bandwidth efficiency and power control over Wireless Sensor Networks. In addition, packet scheduling is widely adopted as an effective rate adaptation technique in wireless multimedia communications. Both these techniques are combined in a proposed scheme which firstly, utilizes nodes with the highest residual energy and low power-cost paths, in order to perform the routing and secondly, predicts the distortion of video packets and selects to either drop them or transmit them according to the current channel bandwidth limitations. The simulation results prove the efficiency of the proposed combined scheme in terms of power consumption and received video distortion (PSNR).

I. I NTRODUCTION The up to date advances accomplished in the domains of wireless communications, digital electronics and Micro Electro-Mechanical Systems technology, have enabled the inexpensive development of multifunctional sensor nodes with superior features. Such sensor nodes are capable to retrieve video and audio streams, still images and scalar sensor data, while being wirelessly interconnected with each other, form Wireless Multimedia Sensor Networks (WMSNs) [1]. In particular, sensor devices equipped with miniature battery powered cameras and wireless low-power transceivers capable of transmitting, receiving and processing video streams, can compose wireless video sensor networks that will complement the current surveillance systems. These networks will have to support reliable, bandwidth efficient video transmission with minimum power consumption. The sensor node makes use of its communicating components in order to transmit the data, over a wireless channel, to a base station which, collects all the data transmitted to it in order to

978-1-4244-4685-8/09/$25.00 ©2009 IEEE

act as a supervisory control processor or an access point for human interface or even a gateway to other networks. The design and development of WMSNs has to take into consideration several factors including strict limitations in energy consumption, while guarantying high level of QoS [1], [2]. In fact, the most of energy expenditure of a sensor node takes place during communication and the rest whilst sensing and data processing. Therefore, there is a need for eradication of energy inefficiencies at all layers of the protocol stack of sensor nodes [3], [4]. This study proposes a inovative routing scheme named Power Efficient Multimedia Routing (PEMuR) spesialized for WMSNs aiming at achieving considerable reduction of energy consumption during routing along with high perceived video QoS. The development of PEMuR is based on the Scalable Hierarchical Power Efficient Routing (SHPER) scheme [5]. According to this scheme the data transmission is performed via the nodes with the higher residual energy and the routes with the minumum energy cost. Moreover, it considers the H.264/AVC codec [6] that can achieve higher compression efficiency than any of the previous standards, by using previously encoded frames as reference for the motion-compensated prediction of each inter-macroblock or macroblock partition. Since motion estimation functionalities require higher complexity and powerful processing encoders, PEMuR incorporates an efficient packet scheduling algorithm. This algorithm compensates for this increase in energy consumption by selectively dropping packets prior to transmission in order to reduce the amount of transmitting data, without increasing significantly the video distortion in the receiving end. The increase in the received video distortion is kept to a minimum due to a video distortion prediction model that identifies the least important video packets that can be omitted prior to transmission [7], [8]. The rest of this work is organized as follows. In Section II, a discussion on some of the most popular hierarchical routing protocols is made. Furthermore, in Section III a detailed description of the proposed hy-

964

brid scheme, is extended over the analysis of distortion prediction, packet scheduling and hierarchical routing. In Section IV, the simulation scenarios and parameters are analyzed, while Section discusses the results. Finally, Section VI concludes the paper. II. H IERARCHICAL ROUTING P ROTOCOLS FOR WSN S LEACH (Low Energy Adaptive Clustering Hierarchy) introduced in [9] is one of the most popular energy efficient hierarchical routing protocols proposed for sensor networks. In LEACH, adjacent nodes dynamically form clusters in a distributed manner. One node per cluster is randomly chosen to be the cluster head. Each cluster head collects and aggregates all data it receives from its cluster nodes and forwards the aggregated data directly to the base station. The election of cluster heads is rotated among the cluster nodes according to an a priori suggested proportion of cluster heads and the number of times each node has been elected to be a cluster head so far. LEACH is supposed to be more suitable for constant proactive monitoring because the data collection is performed periodically. However, in numerous cases, periodic data transmissions are needless, thus causing ineffectual expenditure of energy. Moreover, LEACH supposes that at every election round all nodes have equal quantities of energy capacity. Nevertheless, this hypothesis is unrealistic. Furthermore, it is also assumed that every node has enough transmission power to directly transmit to the base station. In most cases however this asumption is impactical. TEEN (Threshold sensitive Energy Efficient Network) [10] is another extensively used energy efficient hierarchical routing protocol. TEEN uses a hierarchical scheme along with a data centric mechanism. More precisely, in a way alike to that adopted by LEACH, neighboring nodes are dynamically grouped into clusters. Within every cluster, one of the cluster nodes is elected to be the cluster head for a time interval called cluster period. Each cluster head aggregates the data it receives from its cluster nodes and forwards the aggregated data either directly or indirectly to the base station. Each new elected cluster head broadcasts the values of two thresholds which refer to the sensed attribute. The first of them is called hard threshold. It represents the minimum value of the sensed attribute required to force a sensor node to activate its transmitter and transmit to the cluster head. The second one is called soft threshold. It represents a small change in the value of the sensed attribute that prompts the node to turn on its transmitter and transmit. The use of the hard and soft thresholds reduces considerably the number of transmissions, for the reason that the nodes transmit only when either the

sensed attribute exceeds a critical value or there is a sizeable alteration in the sensed attribute. TEEN is appropriate for reactive monitoring of time critical applications. However, it is not suitable wherever periodic reports are required for the reason that if the thresholds are not reached the nodes may not transmit at all. An extension of TEEN protocol, named APTEEN (Adaptive Threshold sensitive Energy Efficient sensor Network) was proposed in [11]. Although APTEEN adopts a hierarchical scheme similar to that used in TEEN, it however enables both reactive and proactive ways of operation. More precisely, in APTEEN sensor nodes are required not only to respond instantly to time critical situations but to periodically send data too. The periodicity and the values of thresholds are adapted to the user needs and the type of the application. Experimentally has been shown that even though APTEEN surpasses LEACH in terms of energy dissipation and network lifetime, TEEN however outperforms both protocols [11]. III. PEM U R ALGORITHM DESCRIPTION The proposed PEMuR routing algorithm for efficient video communication on WMSNs performs two different tasks: video packet scheduling and hierarchical routing. The combination of these two functions ensures low power consumption over all sensor nodes and high perceived video QoS. The detailed description of the functionalities involved in PEMuR are described below. A. Video distortion model The random nature of the wireless channel among video sensor nodes in a WMSN, will result into losses of the transmitted video data. It has been suggested in [12] that these losses can be characterized as single or isolated losses, burst of losses and losses separated with a small lag. Apparently, the effect that each of these three types of losses has on the received video sequence is very different because of the inter-frame and intraframe dependences in the encoded video. Therefore, the model includes all the pre-mentioned parameters, thus can accurately predict the resulted video distortion due to any error pattern that may include one or more of isolated, burst or errors with lag. In order to analytically express the distortion model, a list of previously encoded reference frames with size MR that is used during the encoding and decoding processes for motion-compensated prediction, is defined. This parameter accounts for the impact of the number of reference frames on the distortion propagation. Moreover, each frame is coded into a number of video packets according to each size. Finally, a simple error concealment mechanism, which replaces a lost frame

965

with its previous at the decoder, is applied. The proposed model includes analytical models for a single frame loss, a burst of losses with variable burst length B (where B ≥ 2) and frame losses separated by a lag. In (1), D n is the total video distortion due to an erroneous pattern of n-frames, where n ≥ 1 and F n is the frame number of the nth erroneous frame. Moreover, Λ(i) represents the distortion propagation effect until the end of the intra period N due to an error occurred at frame k. The frame number of the n th erroneous frame is denoted by F n . Additionally, the error power introduced in a single frame Fn is denoted by σs2 (k) and the total video distortion due to error frame k and its error power propagation to the following frames is denoted by D s (k). Correspondingly, σ 2 (k) and D are the MSE and the sum of the MSE values over all frames in the intra frame period, of more general loss patterns, respectively. ı = N − 1 + Fn−1

Dn =Dn−1 −



Λ(i) ·σ 2 (Fn−1 ) + σ 2 (Fn−1 )+

ı =0

+

               

C. Hierarchical routing

ı = N − 1 + Fn



Λ(i) ·σs2 (Fn ), uncorrelated

ı = 0

ı = N − 1 + Fn



Λ(i) ·σ 2 (Fn ), burst   ı = 0        ı = Fn − 1 − Fn−1 ı = N − 1 + Fn      2  Λ ·σ (F ) + Λ(i) ·σ 2 (Fn ), lag  n (i)  ı = 0

video transmission by the video sensor node. Dropping a video packet imposes a distortion that affects not only the current frame but all the correlated frames. The intelligence of the packet scheduling algorithm is that utilizes the distortion prediction model, which considers the correlation among the reference frames, thus it selects the optimum pattern of packets/frames to drop in each transmission window. In each transmission window the sender calculates all the possible combinations of packets to drop and the respective distortion imposed by each combination. This process is neither time nor power consuming as the transmission window is generally small and the mathematical calculations are not of high complexity. Therefore, the proposed packet scheduler allows the video sensor node to determine in the current transmission window several combinations of packets to drop, suitable for different ranges of transmission rates that will be possible imposed by the network at the next transmission window.

ı =0

(1) This recursive formula calculates the total distortion for the first error frame and depending on whether the next error frame is correlated or not with the previous error frame, it combines the above formulas and estimates the total distortion of the resulted error pattern. Detailed analysis of all the parameters and extensive simulation results that prove the accuracy of the distortion prediction model, for different video sequences, is presented in [7] and [8]. B. Video packet scheduling Each video packet in the video stream is characterized by its importance in the overall video distortion according to the previous video distortion model. There may be cases when the transmission bandwidth required from the sensor node exceeds the capacity limit of the shared wireless channel. In such a case the sensor node decides which video packets will be optimally dropped in order to reduce its current transmission rate. The packets to be dropped are selected according to their impact to the overall video distortion. A combination of one or more video packets may be omitted prior to the

The great advantage that hierarchical routing offers is the achievement of high scalability. This means that no matter how many the sensor nodes are, the routing operations remain unaffected. PEMuR utilizes SHPER protocol in order to perform hierarchical routing because contrary to the protocols described in Section II, in SHPER protocol the election of the cluster heads is not randomized but it is based on the residual energy of the nodes. That is to say, the member of each cluster having the highest residual energy is the one selected as the cluster head. The cluster heads which are close enough to the base station have the ability to communicate directly with the base station with reasonable power consumption. These cluster heads are considered to be the highest level cluster heads. Similarly, cluster heads which are located far away from the base station are considered to be lower level cluster heads. The operation of SHPER protocol, which is analytically explained in [5] is divided in two phases, which namely are the initialization phase and the steady state phase. 1) Initialization Phase: During the initialization phase the base station creates a TDMA according to which the nodes advertise themselves, thus the relative distances among them are identified. The base station randomly elects a predefined number of high- and lowlevel cluster heads and broadcasts their identities along with the values of the thresholds. Each non cluster head node decides to belong to the cluster of the cluster head, whose advertisement message has the largest signal strength. The lower level cluster head nodes cannot transmit directly to the base station, thus it is necessary to route their messages via an upper level cluster head node

966

and possibly other lower level cluster head nodes located closer to the network upper level. However, there are various alternative paths that may me followed. According to SHPER algorithm, each path p = (c 1 , c2 , ..., cn ) between a lower level cluster head node c 1 and the base station cn that spans n − 2 intermediate cluster head nodes c 2 , ..., cn−1 , is assigned with a corresponding value of the Routing Index RI(r), which is shown in the following equation: RI(p) =

n−1  ı=2

Erı −

n−1 

E(cı , cı+1 )

(2)

ı=1

where, Eri denotes the residual energy of the cluster head node c i , and E(ci , ci+1 ) denotes the energy consumed for a message to be routed between two sequential cluster head nodes c i and ci+1 . If A is the set of all possible paths pl that can be followed for the transmission of the messages of a lower level cluster head to the base station, then the path p k selected is the one which satisfies the property given in: pk = min{RI(pı ) : l ∈ A} .

Fig. 1. Simulation sensor network field comprising 100 randomly deployed video sensor nodes

(3)

After each node has decided which cluster it belongs to, it informs its cluster head that it will be a member of its cluster by using a CSMA MAC protocol. Each cluster head informs its cluster nodes when they can transmit according to the TDMA schedule which it broadcasts back to them. Each node transmits its residual energy to its cluster head while the node having the most excessive value of the sensed attribute (thus the node located closest to the center of the event area), transmits its video data. Each cluster head merges the data received into a composite message containing: i) the identity (ID) of the node with the peak residual energy, ii) the video data and iii) the ID of the video source node. Each cluster head, transmits its composite message either directly to the base station or via intermediate upper level cluster heads, according to the selected path. The initialization phase is completed as soon as the base station collects the composite messages from all cluster heads. 2) Steady state phase: During the steady state phase the base station, according to the data received, elects the node having the highest residual energy in each cluster as the new cluster head, and defines the new soft and hard thresholds. The base staion then, broadcasts the IDs of the newly elected cluster heads and the new values of the thresholds. IV. S IMULATION SETUP The simulation was carried out between SHPER and TEEN protocols. The simulation was performed by using TrueTime 1.5, which is a Matlab/Simulink based simulator for real time control systems [13], [14]. The network

used in the simulation scenario adopted consists of 100 stationary sensors randomly deployed within an area of 100m x 100m, capable of capturing, encoding and broadcasting live video sequences to a receiving point, as shown in Fig. 1 and a base station located at a position out of this area. The nodes transmit data during their time slots, only when an event is detected. Events are generated at random locations within the network field, each of them concerning an area of arbitrary surface. Each path between the sensor nodes and the cluster leaders, is subjected to variable background traffic with constant bit rate (CBR) over UDP in order to increase the virtual collisions. The video sequence is encoded according to H.264/AVC standard with a reference frame list of size 5 frames for compensated prediction [6]. The video testing sequence Foreman is used at QCIF resolution with 300 video frames at a frame rate of 30 fps with a constant quantization step that results at an average PSNR of 38dB. The inter-frame period is 36 frames and is set to be equal with the transmission window. The video frames are encapsulated into RTP packets of size 1024 bytes [15]. The generated video packets are delivered through the 802.11g system [16] at the form of UDP/IP protocol stack in a unicast transmission mode (one video source – one sink). The radio model that is used is the same adopted in [10]. By using this approach, an energy loss d 2 due to channel transmission is assumed. The energy ET x (k, d) that a node dissipates for the radio transmission of a message of k bits over a distance d is due to running both the transmitter circuitry ET x−elec(k) and the transmitter

967

45

100

40

80

35 30

60

PSNR (dB)

Number of Nodes Alive

PEMuR TEEN

40

25 20 15

20

10 Decoded video during errorless transmission Decoded video during TEEN Decoded video during PEMuR

5

0 0

50

100 150 Simulation Rounds

200

0 0

Fig. 2. Comparative depection of nodes remaining alive over the simulation time

100 150 200 250 300 350 400 450 500 550 600 Frame index

Fig. 3. PSNR per video frame during errorless transmission, PEMuR and TEEN algorithms TABLE I C OMPARISON B ETWEEN PEM U R AND TEEN IN N ODE L IFETIME A CHIEVEMENT

amplifier ETxamp (k, d) and is given by: ET x (k, d)

50

= ET x−elec (k) + ET x−amp (k, d) = Eelec · (k) + εamp · k · d2

(4)

where, Eelec is the transmitter circuitry dissipation per bit, which is considered to be equal to the corresponding receiver circuitry dissipation per bit and ε amp is the transmit amplifier dissipation per bit per square meter. Similarly, the energy ER x(k) dissipated by a node for the radio reception of a message of k bits is due to running the receiver circuitry ER x−elec(k) and is given by: ERx (k) = ERx−elec (k) = Eelec · (k)

(5)

Moreover, a node dissipates additional sums of energy while being in idle state (equal to E elec ) and while sensing (equal to 0.1·ε elec ). The radio channel is supposed to be symmetrical. Thus, the energy required to transmit a video packet from a source node to a destination node is the same as the energy required to transmit the same video packet from the destination node back to the source node for a given SNR (Signal to Noise Ratio). Moreover, it is assumed that the sensor nodes do not retransmit lost or delayed video packets. Each node has the ability of monitoring its residual energy. The initial energy of all nodes takes random values in the range of 10 Watts to 50 Watts. It is further assumed that Eelec = 50nJ/bit and εamp = 100pJ/bit/m2 . V. S IMULATION RESULTS The simulation over different topologies showed that PEMuR, which utilizes SPHER protocol, surpasses TEEN by reducing the energy dissipation and thus extending the overall network lifetime, while achieving better QoS performance. More precisely, referring to the topology illustrated in Fig. 1, simulation results are graphically displayed in Fig. 2 and arithmetically

Round Round Round Round Round

first node dies 30% of nodes die 50% of nodes die 70% of nodes die last node dies

TEEN 2 66 105 119 143

PEMuR 7 101 126 138 160

Change (%) +250 +53.03 +20.00 +15.97 +11.89

presented in Table I, where the enhancement provided by PEMuR algorithm becomes profound. Particularly, Fig. 3 illustrates the Peak Signal to Noise ratio of the decoded Foreman sequence under PEMuR and TEEN algorithms and compares it against the PSNR when the transmission is errorless. It is clear that the proposed algorithm improves the perceived QoS of the received video sequence significantly. Furthermore, it can be seen that the quality variations due to the dropped packets are limited and smoother in the case of the PEMuR. This is very important in terms of perceived video quality since it improves the Mean Opinion Score (MOS) of the video sequence. VI. C ONCLUSIONS The paper presented PEMuR, a hybrid scheme for efficient video communications over WMSNs that comprises a hierarchical routing protocol based on SPHER, which ensures that only the nodes with the highet residual power and the paths with the lowest power costs, are used durring the routing. Moreover, the proposed scheme utilizes an intelligent video packet scheduling algorithm that in case of limitted available channel bandwidth, it sellectively drops non significant packets prior to their transmission, hence it reduces the transmission video rate. This selection is based on an analytical distortion prediction model. The simulation results indicated that

968

PEMuR achieves significant improvement compared to TEEN, in terms of both power consumption and perceived video quality measured as PSNR. R EFERENCES [1] Ian F. Akyildiz, Tommaso Melodia, and Kaushik R. Chowdhury, A survey on wireless multimedia sensor networks, Comput. Netw., 51(4):921960, 2007. [2] I.F. Akyildiz, I.H. Kasimoglu, Wireless sensor and actor networks: Research challenges, Ad Hoc Networks (Elsevier), 2 (4) (2004) 351367. [3] K. Akkaya and M. Younis, A Survey of Routing Protocols, In Wireless Sensor Networks, Elsevier Ad Hoc Network Journal, vol. 3, no. 3, pp. 325-349, 2005. [4] J.N Al-Karaki, and A.E. Kamal, Routing techniques in wireless sensor networks: a survey, IEEE Wireless Communications, vol. 11, no. 6, pp.6-28, December 2004. [5] D. Kandris, P. Tsioumas, A. Tzes, N. Pantazis and D. Vergados, Hierarchical Energy Efficient Routing in Wireless Sensor Networks, In Proceedings of 16th Mediterranean Conference on Control and Automation, Assacio, France, pp.1856-1861, June 2008. [6] ISO/IEC 14496-10 and ITU-T Rec., H.264 Standard, Advanced Video Coding, 2003. [7] I. Politis, M. Tsagkaropoulos, T. Pliakas, T. Dagiuklas and S. Kotsopoulos, Intelligence packet scheduling for optimized video transmission over wireless networks, Proc. MobiMedia 07, Nafpaktos, Greece, Aug, 2007 [8] I. Politis, M. Tsagkaropoulos, T. Pliakas and T. Dagiuklas, Distortion Optimized Packet Scheduling and Prioritization of Multiple Video Streams over 802.11e Networks, Advances in Multimedia, vol. 2007, Article ID. 76846, pp. 1-11, 2007. [9] W. Heinzelman, A. Chandrakasan and H. Balakrishnan, EnergyEfficient Communication Protocol for Wireless Mi-crosensor Networks, In Proceedings of the 33rd Hawaii International Conference on System Sciences (HICSS ’00), January 2000. [10] A. Manjeshwar and D. P. Agarwal, TEEN: a routing protocol for enhanced efficiency in wireless sensor networks, In 1st International Workshop on Parallel and Distributed Computing Issues in Wireless Networks and Mobile Computing, San Francisco, USA, April 2001. [11] A. Manjeshwar, D.P. Agrawal, APTEEN: a hybrid protocol for efficient routing and comprehensive information retrieval in wireless sensor networks, In Proceedings of the 2nd International Workshop on Parallel and Distributed Computing Issues in Wireless Networks and Mobile computing, Ft. Lauderdale, FL, USA, April 2002. [12] Y. Liang, J. Apostolopoulos and B. Girod, Analysis of packet loss for compressed video: Does burst-length matter, Proc. Int’l Conf. Acoustics, Speech, and Signal Processing, Hong Kong, China, April, 2003. [13] TrueTime 1.5 - Reference Manual, Department of Automatic Control, Lund University, Sweden, January 2007. Available at: http://www.control.lth.se/documents/2007/ohl+07tt.pdf. [14] A. Cervin, M. Ohlin, and D. Henriksson, Simulation of Networked Control Systems Using TrueTime, In Proc. 3rd International Workshop on Networked Control Systems: Tolerant to Faults, Nancy, France, June 2007. [15] H. Schulzrinne, RTP: A Transport Protocol Real-Time Applications, RFC 1889, 1996 [16] 802.11g IEEE 2003, Part11 Wireless LAN Medium Access Control (MAC) and Physical Layer (PHY) Specification Band, Supp. IEEE 802.11, 2003.

969