[ieee 2011 7th international conference on wireless communications, networking and mobile computing...

4
Improving Video Delivery over Wireless Multimedia Sensor Networks Based on Queue Priority Scheduling Elham Karimi Department of Electrical and Computer Engineering Qazvin Islamic Azad University Qazvin, Iran [email protected] Behzad Akbari Department of Electrical and Computer Engineering Tarbiat Modares University Tehran, Iran [email protected] Abstract— Using multipath routing protocols results in more efficient data transmission than single path routing protocols in Wireless Multimedia Sensor Networks (WMSNs). In this paper, we propose a new mechanism for improving video transmission over WMSN based on queue priority scheduling. We assume that network has high CBR traffic with no video data in addition to video packets. In our mechanism, intermediate nodes buffer (queue) is scheduled for quick and righteously transmission of video packets compared with the other packets in network. Simulation results show that our proposed scheduling improves important Quality-of-Service (QoS) metrics such as end-to-end delay and frame loss percentage for video streaming. Keywords-Wireless Sensor Networks; Video Transmission; Queue scheduling; Packet scheduling; multipath routing I. INTRODUCTION Wireless sensor networks (WSNs) are self organizing networks having very large number of nodes. These nodes are usually scattered over an environment to collect information from the perimeter. They have properties such as low cost, small size and weight, and due to their limited resources, i.e. energy, memory, and processing power, they must be placed closely to each other [1]. WSNs are subject to advancement because of their simplicity and applicability. Furthermore, new applications have been developed for them. Over the last few years, WMSNs have been employed for multimedia communication. WMSNs are able to retrieve multimedia content such as video and audio streams, still images, and scalar sensor data from the environment [1]. There are many challenges to be addressed to support video communication over these networks. Decreasing delay in real-time video streaming is a crucial problem. In addition, large amount of bandwidth is needed for video transmission since video frames are usually large. In WMSNs, routing protocol plays an important role for optimizing throughput of bandwidth and decreasing end-to-end delay. Several routing protocols are proposed for these goals. Using single path routing protocol is not efficient for this purpose, because it doesn't provide bandwidth and delay requirements for video streaming. In a network that uses only one path for transmission of data, the nodes along this path will lose their energy sooner and are failed. This would result in shortening of the lifetime of the WSN. Thus, there would be no reliability and load balancing guarantee. Now, assume the case in which the content of packets in video frames using one path do not achieve their required bandwidth for streaming. In this state, the queuing delay for intermediate nodes will be high, thereby amount of packet losses and end-to-end delay will be high for the real- time video. In single path routing, video frames are prone to loss and intermediate nodes might have high queuing delays. Multipath routing protocols are used to increase bandwidth efficiency, reliability and decrease end-to-end delay for video transmission. The benefits of multipath video streaming are as follows [2]: Reducing correlation among packet losses Increased channel resources that can support the application’s QoS demands The power consumption is more evenly spread in the network nodes which prevents node failures Ability to adjust to arbitrary congestion occurrences in different parts of the network When there is huge amount of traffic in the network, using multipath routing cannot provide quality of requisite in the perceived video alone. The problem of video streaming with resource constraints in WMSN proves challenging as network traffic increases, thus a scheduling mechanism will be needed. Our contribution is a solution in which intermediate nodes, i.e. nodes between source and destination, are scheduled and perceived video quality can be improved with queue priority scheduling. Since each frame scheduling doesn't have the same effect on the perceived video quality, it would be better to give different priority to each frame. Several scheduling solutions have been proposed for packet scheduling that classify video packets according to their types [3]. In [2], a power efficient multipath video transmission scheme has been proposed that selects less important video packets to be dropped by using a recursive distortion prediction model. It proposes power aware packet scheduling that is able to identify the available paths connecting from video source to the receiver and schedules packets for transmitting among the 978-1-4244-6252-0/11/$26.00 ©2011 IEEE

Upload: behzad

Post on 12-Oct-2016

212 views

Category:

Documents


0 download

TRANSCRIPT

Improving Video Delivery over Wireless Multimedia Sensor Networks Based on Queue Priority Scheduling

Elham Karimi Department of Electrical and Computer Engineering

Qazvin Islamic Azad University Qazvin, Iran

[email protected]

Behzad Akbari Department of Electrical and Computer Engineering

Tarbiat Modares University Tehran, Iran

[email protected]

Abstract— Using multipath routing protocols results in more efficient data transmission than single path routing protocols in Wireless Multimedia Sensor Networks (WMSNs). In this paper, we propose a new mechanism for improving video transmission over WMSN based on queue priority scheduling. We assume that network has high CBR traffic with no video data in addition to video packets. In our mechanism, intermediate nodes buffer (queue) is scheduled for quick and righteously transmission of video packets compared with the other packets in network. Simulation results show that our proposed scheduling improves important Quality-of-Service (QoS) metrics such as end-to-end delay and frame loss percentage for video streaming.

Keywords-Wireless Sensor Networks; Video Transmission; Queue scheduling; Packet scheduling; multipath routing

I. INTRODUCTION Wireless sensor networks (WSNs) are self organizing

networks having very large number of nodes. These nodes are usually scattered over an environment to collect information from the perimeter. They have properties such as low cost, small size and weight, and due to their limited resources, i.e. energy, memory, and processing power, they must be placed closely to each other [1].

WSNs are subject to advancement because of their simplicity and applicability. Furthermore, new applications have been developed for them. Over the last few years, WMSNs have been employed for multimedia communication. WMSNs are able to retrieve multimedia content such as video and audio streams, still images, and scalar sensor data from the environment [1].

There are many challenges to be addressed to support video communication over these networks. Decreasing delay in real-time video streaming is a crucial problem. In addition, large amount of bandwidth is needed for video transmission since video frames are usually large. In WMSNs, routing protocol plays an important role for optimizing throughput of bandwidth and decreasing end-to-end delay. Several routing protocols are proposed for these goals. Using single path routing protocol is not efficient for this purpose, because it doesn't provide bandwidth and delay requirements for video streaming.

In a network that uses only one path for transmission of data, the nodes along this path will lose their energy sooner and are failed. This would result in shortening of the lifetime of the

WSN. Thus, there would be no reliability and load balancing guarantee. Now, assume the case in which the content of packets in video frames using one path do not achieve their required bandwidth for streaming. In this state, the queuing delay for intermediate nodes will be high, thereby amount of packet losses and end-to-end delay will be high for the real-time video.

In single path routing, video frames are prone to loss and intermediate nodes might have high queuing delays. Multipath routing protocols are used to increase bandwidth efficiency, reliability and decrease end-to-end delay for video transmission. The benefits of multipath video streaming are as follows [2]:

• Reducing correlation among packet losses

• Increased channel resources that can support the application’s QoS demands

• The power consumption is more evenly spread in the network nodes which prevents node failures

• Ability to adjust to arbitrary congestion occurrences in different parts of the network

When there is huge amount of traffic in the network, using multipath routing cannot provide quality of requisite in the perceived video alone. The problem of video streaming with resource constraints in WMSN proves challenging as network traffic increases, thus a scheduling mechanism will be needed. Our contribution is a solution in which intermediate nodes, i.e. nodes between source and destination, are scheduled and perceived video quality can be improved with queue priority scheduling.

Since each frame scheduling doesn't have the same effect on the perceived video quality, it would be better to give different priority to each frame. Several scheduling solutions have been proposed for packet scheduling that classify video packets according to their types [3].

In [2], a power efficient multipath video transmission scheme has been proposed that selects less important video packets to be dropped by using a recursive distortion prediction model. It proposes power aware packet scheduling that is able to identify the available paths connecting from video source to the receiver and schedules packets for transmitting among the

978-1-4244-6252-0/11/$26.00 ©2011 IEEE

selected paths according to energy efficiency of the participating wireless video sensor nodes by dropping packets.

In [4], video stream data is partitioned into image and audio streams and two kinds of priority are given to them for using of limited bandwidth and energy in WMSN based on application requirements. The authors of [5] have used path priority scheduling and cross-layer technique for adaptive video coding to path status with frame skipping, reference frame selecting and intra-frame refreshing techniques for H.26L real-time video streaming over WSNs.

The key idea of proposed schema in [6] is based on sending video packets over two disjoint paths besides using buffering technique in special nodes of network. In each of every path there is only one node selected as cache node. The main task of these nodes is to realizing different packet types, buffering some of important video packets, reducing forward traffic rate while detecting loss in network and local error management to overcome high loss rate of video packets.

In this paper, we propose a solution in which intermediate nodes’ buffer (queue) is scheduled and QoS metrics such as frame loss and end-to-end delay are improved for video streaming finally. We execute our solution over multipath routing for analyzing results.

The rest of the paper is organized as follows. Section II describes our proposed queue scheduling model. Simulation results are described in Section III and finally conclusion of paper is in Section IV.

II. PROPOSED QUEUE PRIORITY SCHEDULING

As stated in Section I, when there is huge amount of traffic in the network, using multipath routing cannot provide quality of requisite in the perceived video alone. Thus an additional mechanism such as scheduling will be needed. The role of such an additional mechanism is to give appropriate priority to each video frame in conjunction to the other data in order to avoid frame losses while achieving QoS for video streaming.

We propose a scheduling mechanism for intermediate nodes’ buffer (queue). Such a mechanism can surprisingly improve the video delivery over WMSN. In our queue priority scheduling, intermediate nodes’ buffer is divided into four queues and weighted round robin is executed for them. Every packet that reaches to a node, before being placed in node’s buffer, the free buffer size is considered. Furthermore, the sum of lengths of these four queues is compared with the physical buffer length. If the sum of lengths of these four queues equals the physical buffer length, then the packet is dropped. Otherwise, if this value is lower than the physical buffer length, the packet is placed on one of the queues based on its type. Ordering for placing a packet in queues is as following:

• The packet that contains I-frame is placed in the first queue.

• The packet that contains P-frame is placed in the second queue.

• The packet that contains B-frame is placed in the third queue.

• The packet with no video data is placed in the fourth queue.

Our scheduling scheme operates in this way. The packets of three queues that have higher priorities, because of containing video packet, are transmitted since intermediate nodes’ buffer with round robin scheduling in the first. If these three queues are empty and don't have any packets, then the fourth queue transmits one packet over the network then the third queue is checked for the existence of packet for transmission. If it has any packet for transmission, it will transmit it and the round robin mechanism is executed in three queues that contain video packet again. Furthermore, the fourth queue, when don’t exist any video packet in before queues, transmits its packets. It is worth mentioning that First-In-First-Out (FIFO) discipline is used for each queue.

With this queue priority scheduling, the followings are provided; keeping equality between different types of video frames and giving priority to video packets, comparing other packets. In this way, a video packet is quickly transmitted in comparison with other packets, that is, giving priority to video frames while there is a huge amount of CBR traffic with no video data in network. This queue priority scheduling operates without regarding the type of routing protocol.

Since using multipath routing protocols more efficient than single path routing protocols in WMSNs thus we use this queue priority scheduling over multipath routing and analyze the simulation results.

III. SIMULATION RESULTS

A. Simulation Setup For our simulation we used NS2 network simulator [7] to

simulate packet transmission over the network. Moreover, we generated MPEG4 [8] video traffic by using EvalVid [9, 10, 11] frame work. In this frame work we use ffmpeg [12] encoder. Metrics of interest to us are end-to-end delay, cumulative jitter, percentage of frame loss, etc. Parameters of simulation scenario are listed in Table I.

Our simulation setup comprises for showing improvement and effect of our queue scheduling. We executed it on AOMDV [13], as our multipath routing protocol, and compared it with AOMDV without queue scheduling. Then, Foreman video sequence is transmitted over them and the results are analyzed. Video information for this evaluation is listed in Table II.

TABLE I. PARAMETERS OF THE SIMULATION SENARIO

value Parameters 1000m x 1000mNetwork Dimension

150 Number of Nodes1000 JoulesSensor Initial Energy

Uniformly DistributedNode PlacementTwo nodes send video traffic to the

Base Station Every other node sends 400 kbps

traffic to the Base Station.

Traffic

200 Buffer Size400 Packet size

TABLE II. VIDEO IN FORMATION FOR OUR QUEUE SCHEDULING EVALUATION

Value Information MPEG4 Video Encoding

QCIF, 176 * 144 (pixels/frames)

Format

30 frame/s Frame rate 400 Number of

Video Frames

B. Simulation results 1) PSNR

Fig. 1 shows the overall PSNR of frames in AOMDV [13] that doesn't have queue scheduling and in AOMDV that has our queue priority scheduling. Queue priority scheduling could achieve a high improvement in PSNR (near ideal PSNR) because of queue with scheduling transmits video packets quickly in proportion of other packets that is, giving priority to video frames that there is a huge amount of CBR traffic with no video data in network. Thus in queue priority scheduling with decreasing end-to-end delay and decreasing queuing delay in the buffer of intermediate nodes for video packets, improves overall PSNR of frames.

Figure 1. PSNR of AOMDV and queue scheduling frames

2) Frame Loss Play-out buffer is the time after which video frames can be

played. Whatever play-out buffer time is greater video frames find an opportunity that are received in receiver. We show percent of frame loss in time of play out buffer. Whenever play out buffer time increases, more video frames could receive in receiver.

Fig. 2 shows percent of frame loss with increasing time of play out buffer. Decreasing lost video frame in our scheduling in intermediate node buffers can achieve a high improvement in QoS on demand in video real time. The reason is that the queue scheduling scheme tries to avoid losing video packets in the networks with a huge amount of traffic with no video data. But AOMDV originally does not contain any plans to avoid losing video frames. Fig. 2 shows that in low play out buffer time, 2000 ms, value of loss frame percent for using our queue priority scheduling is %40 whiles value of loss frame percent without queue scheduling is %80 approximately in this time. But in higher play out buffer time, such as 4000 ms, value of loss frame percent for using our queue priority scheduling is

very low and is about %1.75 whiles value of loss frame percent without queue scheduling is about of %57.75 in this time. Thus our proposed solution could improve percent of loss frame about of % 56.

Figure 2. Video frame loss in AOMDV and queue scheduling

3) Cumulative Jitter End-to-end delay and cumulative jitter for per frame in

receiver are generated with EvalVid [9, 10]. In video transmission systems not only end-to-end delay is important for the perceived video quality, but also the variation of the delay, usually referred to as frame jitter, is important for the video quality. Fig. 3 shows cumulative jitter in original AOMDV and our scheduling in intermediate node buffers over AOMDV. Our queue scheduling can achieve a high improvement in cumulative jitter. Our mechanism transmits video packets just and quickly in proportion of the other packets. This is equivalent to giving priority to video frames when there is a huge amount of CBR traffic with no video data in network. Thus, in this way with decreasing end-to-end delay and decreasing queuing delay in the buffer of intermediate nodes for video packets, the cumulative jitter of frames will be improved.

Figure 3. Cumulative jitter for per frame in AOMDV and our queue

scheduling

4) Inter Frame Gap Also, Inter frame gap for each frame is generated using

EvalVid. When video quality in receiver increases that inter frame gap decreases. Fig. 4 shows inter frame gap in AOMDV without queue scheduling and queue priority scheduling over it.

Our scheduling can achieve a high improvement in inter frame gap. Since, in our queue priority scheduling, giving priority video packet in intermediate nodes buffer and quickly and sequentially transmitting them, correlation among packet losses reduces, as well as reducing loss frame. Therefore, inter frame gap is decreased and the quality of received video improves.

Figure 4. Inter frame gap for AOMDV and queue scheduling

Fig. 5 presents video quality in receiver for using queue scheduling and without it. In our queue scheduling correlation among packet losses is reduced and with giving video frame in comparison with other data packet in network, it improves quality of perceive video.

(a) (b)

Figure 5. Foreman sequence (a) using AOMDV without queue scheduling (b) using AOMDV with queue scheduling

IV. CONCLUSION

Video delivery with high quality over WMSN that has limited resources such as energy, etc. is a challenging problem. In this paper, we first compare single and multipath video streaming over WMSN. Then we proposed a new mechanism for improving video delivery over WMSN in which all of nodes participate in scheduling. Using queue priority

scheduling, different types of video frames will be treated righteously and the correct priority will be given to video packet in proportion of other packets in network that has high traffic of CBR with no video packet. Simulation results show that our proposed solution achieves an improvement in some of QoS metrics such as end-to-end delay and percentage of lost frames.

REFERENCES [1] I.F. Akyildiz, T. Melodia, K.R. Chowdhury, “Wireless multimedia

sensor networks: A survey,” IEEE Wireless Communications 14, no. 6, pp. 32–39, 2007.

[2] I. Politis, M. Tsagkaropoulos, T. Dagiuklas, S. Kotsopoulos, “Power Efficient Video Multipath Transmission over Wireless Multimedia Sensor Networks,” Mob. Netw. Appl. 13, no. 3–4, pp. 274– 284, 2008.

[3] A.R. Lari, B. Akbari, “Network-Adaptive Multipath Video Delivery over Wireless Multimedia Sensor Networks Based on Packet and Path Priority Scheduling,” IEEE International Conference on Broadband, Wireless Computing, Communication and Applications, 2010.

[4] L. Zhang, M. Hauswirth, Z. Zhou, V. Reynolds, and G. Han, “Multi-priority Multi-Path Selection for Video Streaming in Wireless Multimedia Sensor Networks,” In the fifth International conference on Ubiquitous Intelligence and Computing (UIC 2008), Oslo, Norway, pp. 23–25, June 2008.

[5] Min Chen, Victor C.M. Leung, Shiwen Mao, Ming Li, “Cross-layer and Path Priority Scheduling based Real-time Video Communications over Wireless Sensor Networks,” IEEE vehicular technology conference, Singapore, VTC spring, pp. 11–14, 2008.

[6] P. Panahi, “The Feedback Based Mechanism for Video Streaming Over Multipath Ad Hoc Networks,” Journal of Sciences, Islamic Republic of Iran 21(2), pp. 169–179, 2010.

[7] Information Science Institute, NS-2 network simulator, Software Package, 2003, http://www.isi.edu/nsnam/ns/

[8] MPEG4 Encoder, http://megarea.ee.nctu.edu.tw/mpeg. [9] J.Klaue, C. K. Shieh, W. S. Hwang, and A Ziviani, “An Evaluation

Framework for More Realistic Simulations of MPEG Video Transmission,” Journal of Information Science and engineering, vol. 24, no. 2, pp. 425–440, March 2008.

[10] Jirka Klaue, Berthold Rathke, and Adam Wolisz, “EvalVid – A Framework for Video Transmission and Quality Evaluation,” in Proc. of the 13th International Conference on Modelling Techniques and Tools for Computer Performance Evaluation, Urbana, Illinois, USA, September 2003.

[11] Integrating EvalVid with NS2, http://140.116.72.80/~smallko. [12] ffmpeg's Official Webpage, http://ffmpeg.mplayerhq.hu. [13] Mahesh K. Marina, Samir R. Das, “On-demand multipath distance

vector routing in ad hoc networks,” Ninth International Conference for Network Protocols (ICNP), Nov 2001.

[14] AOMDV in NS2, http://wpage.unina.it/marcello.caleffi/ns2/aomdv.