A Dynamic Rate Control Scheme for Video Streaming on Inter-Building Wireless Networks Takeyuki SUGIMOTO Yuichi TAKAHASHI Saneyasu YAMAGUCHI and Koichi ASATANI Graduate School of Electric & Electronic Engineering. Kogakuin University Dept. of Comp Science & Comm., Kogakuin University 1-24-2 Nishishinjuku, Shinjukuku, Tokyo, 163-8677 Japan {cm623, cm732}@ns.kogakuin.ac.jp, {sane, asatanik}@cc.kogakuin.ac.jp Abstract 2.4 GHz radio band is commonly used by many systems and appliances such as IEEE82.11b, Bluetooth, RFID, and even microwave ovens. Radio interference among these is often observed. Therefore, in 2.4 GHz inter-building wireless networks, available throughput temporally changes, which could yield serious packet error and/or packet loss. In this paper, we propose a new dynamic source rate control scheme by using ICMP (Internet Control Message Protocol) echo success rate, RTT (Round Trip Time) and scalable-coded-video source to reduce packet error and packet loss for video streaming applications. We apply the scheme to the inter-building wireless system installed between Kogakuin University and Waseda University in Tokyo. It is shown that the proposed scheme significantly reduces PER and increases effective throughput. IEEE82.11b inter-building wireless network system, which we built between Kogakuin University and Waseda University [3]. The results show that the proposed scheme decreases packet loss rate and increases effective throughput. This paper is organized as follows. Section 2 describes the inter-building wireless system which we installed. Section 3 explains a scalable video coding scheme with independent layers, which is deployed by the multi-layer transcoder. Section 4 shows an available throughput of the inter-building wireless system and its issues. Section 5 describes a new proposed dynamic source bit control scheme using ICMP echo success rate and RTTs. It also presents experimental evaluation results. Section 6 summarizes the results and describes a future issue. Index terms Video Streaming,Inter-Building Wireless Systems, Rate Control,Wireless Networks in Urban Environments I. INTRODUCTION The 2.4 GHz radio band is called ISM (Industrial Scientific and Medical) band, and used for household electrical appliances, medical equipment, Bluetooth, VICS (Vehicle Information and Communication System), and RFID (Radio Frequency Identification), as well as wireless LAN. Radio waves from these equipments may cause packet losses, delays and jitters in the wireless LAN [1] [2].Especially in inter-building wireless networks in urban environments, packet losses, jitters and delays are often caused by the interference. Packet losses deteriorate qualities of received real-time streaming videos signals. In addition, large delays and jitters may stop playing the video streams [3]. Therefore, it is important to control source video bit rate according to the communication channel condition. In this paper, we propose a new dynamic video source rate control scheme. The purpose is to achieve the high quality video streaming by adjusting video source bit rate according to the communication channel condition of wireless LANs under urban environments. The proposed method uses ICMP echo success rate and RTTs, as control indices, and the scalable transcoder [4] in order to maximize qualities of real-time streaming videos. We evaluated the proposed scheme on the II. THE INTER-BUILDING WIRELESS NETWORK SYSTEM The inter-building wireless system is installed between Shinjuku building of Kogakuin University and No. 55 building in Okubo campus of Waseda University in Tokyo. The system deploys IEEE82.11b. The wireless system configuration is shown in Figure1. The distance is approximately 2. km. There are railways, stations, office buildings, shopping malls, and hospitals between them. It is observed that radio interference from ISM equipments in these buildings cause packet losses and fluctuations of available throughput [3]. III. SCALABLE CODING SCHEME WITH INDEPENDENT LAYERS A video stream is divided into five layers by the scalable coding scheme with independent layers [4]. Each layer is independent. With multiple layer coding schemes using non-independent layers, the upper layers depended on the lower layers. Therefore, if a lower layer is lost, the upper layers cannot de decoded. However, in the scalable coding scheme with independent layers, each layer can be decoded independently. This encoding scheme is realized by an extended DCT (Discrete Cosine Transform) index transcoder method [4]. Therefore, it is possible to decode all received layers, even if any layer is lost. The quality of the decoded video streaming improves as number of received layers increases [5]. 978-1-4244-222-9/8/$25. 28 IEEE
Table 1 Minimum, Maximum and Average Bit Rates Number of Source Layer Min.rate Max.rate rate Figure.1 The Inter-building Wireless System Configuration 6 Available throughput[kbps] 5 4 3 2 1 1 2 3 4 Figure 2 Temporal Fluctuation of the Available Throughput. 1 1 1.5 1.4 2 1.2 3.2 2.8 3 1.4 4.8 4.1 4 1.5 6.3 5.3 5 2.6 7.6 6.5 Table 2 Average Sent and Received Bit Rates and Packet Loss Rate for Fixed Source Layer Send bit rate [%] Fixed source layer number = 1. 1.4 1.4 Fixed source layer number = 2 7.4 2.7 2.5 Fixed source layer number = 3 1.7 3.9 3.5 IV. AN EXPERIMENT WITH FIXED SOURCE RATE SCHEME IN THE INTER-BUILDING WIRELESS SYSTEM IV-1 Fluctuations of Available Throughput The available throughput was measured of the inter-building wireless system. The temporal fluctuation of the available throughput changes between 2. Mbps and 5. Mbps, as shown in Figure 2. IV-2 Experimental Evaluation of Fixed Number of Source Layers Scheme We measured the received throughput with the fixed source layers scheme using the multi-layer transcoder on the inter-building wireless system. The experimental block diagram is shown in Figure 3. The sender has video data and sends them to the controller. The video data are encoded into five layers. The controller determined the number of source layers to be sent. The selected layers are transmitted from Kogakuin University to Waseda University on the wireless link. Minimum, maximum and average bit rates at sending one layer to five layers are summarized in Table 1. Transition of sent and received bit rates and packet loss rate with the fixed source layers scheme are shown in Figures 4, 5 and 6. Fixed number of source layers is 1, 2 and 3 in each figure. Average sent and received bit rates and packet loss rate with the fixed source bit rate scheme is summarized in Table 2. The experimental results show that packet loss rate is low for the fixed number of source layers of 1. On the other hand, they show that the bit rate of the source stream sometimes exceeds the available throughput for number of the source layers of 2 and 3. is accordingly high. Bit rate[kbps] Figure 3 Experimental Block Diagram 6 1 9 5 8 4 7 6 3 5 4 2 3 1 2 1 2 4 6 8 1 Figure 4 Sent and Received Bit Rates and Packet Loss Rate (Fixed Number of Source Layer=1) IV-3 Issue of the Fixed Number of Source Layers The available throughput of the wireless system fluctuates between 2.-5. Mbps. The experimental results show that the available throughput fluctuation causes packet losses, when the number of source layers is large. The available throughput is not used effectively for 1 layer source. Consequently, transmitting video source with fixed number of layers causes deterioration of quality of the streaming video, or does not use available throughput effectively. Therefore, source bit rate adjustment to the available throughput is useful. [%]
Bit rate[kbps] 6 5 4 3 2 1 2 4 6 8 1 Time[m] Figure 5 Sent and Received Bit Rates and Packet Loss Rate (Fixed Number of Source Layer=2) 6 1 9 5 8 4 7 6 3 5 4 2 3 1 2 1 2 4 6 8 1 Figure 6 Sent and Received Bit Rates and Packet Loss Rate (Fixed Number of Source Layer=3) V. A PROPOSAL OF A NEW DYNAMIC SOURCE BIT RATE CONTROL SCHEME USING ICMP ECHO SUCCESS RATE AND RTT V-1 Relationships among ICMP Echo Success Rate, RTTs and Available Throughput In the proposed dynamic source bit rate control scheme, ICMP echo success rate and RTTs are used as indexes. The ICMP echo success rate and RTTs are measured for a fixed bit rate source stream transmission on the wireless system. The experimental result is shown in Figure 7. The received bit rate is equal to the available bit rate. The result shows that ICMP echo success rate decreases as received bit rate decreases, and the RTTs are large when the received bit rate is low. Therefore, in the proposed scheme, ICMP echo success rate and RTTs are used as indices in order to extrapolate the available throughput. Detailed discussion is described the Section V.4. V-2 A New Source Rate Control Scheme A new source bit rate control scheme uses ICMP echo success rate and RTTs as feedback information in order to estimate available throughput. Block diagram of the proposed dynamic rate control scheme is shown in Figure 8. The proposed control scheme behaves as follows. Firstly, the transcoder in the sending part encodes a video stream into five layers, and input all layers to the receiving thread of the controller. 1 9 8 7 6 5 4 3 2 1 [%] [%] Bit rate[kbps] RTTaverage[msec] 6 5 4 3 2 1 5 1 15 2 25 3 RTTaverage ICMP echo success rate Figure 7 Relationships among RTT, Received Bit Rate and ICMP Success Rate Figure 8 Block Diagram of the Proposed Source Rate Control System for Video Streaming over the Wireless System Secondly, the controller determines the number of layers to be sent, from one to five, according to the throughput information obtained from the link monitor. Finally, the transmitting thread in the controller sends the selected layers to the wireless link, and the receiver receives these layers and decodes them into one video stream. The link monitor extrapolates the available throughput with use of ICMP echo packets and reply packets. ICMP packets are issued by a ping tool. If the source bit rate exceeds the available throughput, the echo and reply packets are delayed or lost. If the available throughput is higher than the source bit rate, delays and packet loss rate are low. Thus, the controller finds the suitable source bit rate by observing ICMP echo success rate and RTTs. V-3 Determination of Source Layer Numbers The number of source layers is determined as follows. PingSR (ping success rate) and RTTavr (RTT average) are periodically calculated by the following equations. PingSR = ( Pings / Pingtotal ) RTTavr = RTTtotal / Pings 12 1 8 6 4 2 ICMP echo success rate[%] (4-1) (4-2)
where Pings is ICMP echo/reply success times, Pingtotal is defined as the total number of transmitted ICMP echo packets in the calculation interval. The condition for decrease in number of source layers is given by PingSR Pingthd (4-3) or RTT > RTTthd. (4-4) The condition for increase in number of source layers is given by PingSR Pingthu (4-5) and RTTavr < RTTthu. (4-6) The RTTthd is a threshold for decrease in the number of source layers. The RTTthu is a threshold for increase in the number of source layers. If PingSR is smaller than Pingthd or RTT is larger than RTTthd, the number of source layers is decreased. If PingSR is larger than Pingthu, and RTTavr is smaller than RTTthu, the number of source layers is increased. This scheme reduces the number of source layers without waiting for the calculation interval by comparing last RTT and RTTthd individually. Increase in the number of source layers is achieved according to RTTavr and Pingthu, which are calculated periodically. The interval length is discussed in Section V.4. V-4 Determination of Thresholds We investigated the relationship between RTTs and received bit rate in order to determine the suitable thresholds on the inter-building environment. The received bit rate and RTTs are measured for 1. Mbps, 2. Mbps, 3. Mbps, 3.7 Mbps, 4. Mbps and 5. Mbps streams. The results are shown in Figures 9 and 1. For 1. Mbps stream, received bit rate is approximately 1. Mbps and RTTs are less than 1ms. For 5. Mbps stream, received bit rates ranges from 4. Mbps to 5. Mbps, and RTTs are mostly larger than 1ms. We determined the threshold for increase based on these experiments. Source bit rate incremental increase of a layer is approximately 1.3Mbps (see Table 1). The number of source layers should be increased only if the available throughput is higher than the source bit rate by 1.3 Mbps. Since, the available throughput cannot be measured directly, thus the available throughput is estimated by Measuring RTTs. These experiments show that the available throughput is approximately 4.8 Mbps. Thus, the boundary of increase or keeping of the number of source layers is 3.5 Mbps. Correlation between the available throughput and each RTT is not strong enough. Hence, we investigated the relationship between average RTT and received bit rate. The relationship is shown in Figure 1. The horizontal axis is average RTT, and the vertical axis is integral of probability density function of RTT. The threshold should be determined to detect whether the source bit rate is larger than 3.5 Mbps. Consequently, we determined RTTthu of 25 ms. Figure 9 Relationships between RTT and Received Bit Rate for Fixed Bit Rate Streams Integrals of probability density function of RTT 1.2 1.8.6 1.Mbps 2.Mbps.4 3.Mbps 3.7Mbps.2 4.Mbps 5.Mbps 5 1 15 2 25 3 RTT average[msec] Figure1 Relationship between Average RTT and Received Bit Rate (Calculation Interval=5 sec) The 99.3 % of RTTs are larger than 16 ms when the transmitted bit rate is 5. Mbps as shown in Figure 9. In this case, the source bit rate is higher than the available bit rate (4.8 Mbps), thus the number of source layers is decreased. On the other hand, 99.9% of RTTs are less than 14 ms when source bit rate is lower than 4.5 Mbps. Therefore, we chose RTTthd of 15ms to detect that the source bit rate is higher than the available bit rate The calculating interval of 5 sec is chosen based on the following investigation. For the calculating intervals of 1 sec, 3 sec, 5 sec, 1 sec, and 15 sec, the probabilities of correct detection are.841,.89,.91,.915, and.913, respectively. The longer interval gives the higher probability of correct detection. The probabilities for 5 sec, 1 sec, and 15 sec intervals are similar. Therefore, the interval of 5 seconds is chosen for source layer number change. V-5 Evaluation of the Proposed Scheme on a Network Emulator The performance evaluation of the proposed scheme is made by using an emulated network. The network emulation uses Dummynet [6], which can emulate available throughput. We emulated throughputs conditions of the inter-building wireless system. The experimental parameters are shown in Table 3. The experimental block diagram is shown in Figure 11.
6 Figure 11 An Experimental Configuration for Performance Evaluation Using Network Emulator Table3 Experimental Parameters Pingthu 1% RTTthu 25[msec] RTTthd 15[msec] Echo packet frequency 5[times/s] PingSR,RTTavr caluculation interval 5[sec] At the network emulator, the available throughput is changed every 1 ms to emulate the wireless system. The available throughput behavior of the network emulator is shown in Figure 12. Average sent and received bit rates and packet loss rate with the proposed scheme and fixed number of source layers (1, 2, and 3 layers) scheme are summarized in Table 4. The measurement results are shown in Figures 13 and 14, which include sent and received bit rates and packet loss rate with the proposed scheme and fixed number of source layers scheme of 2 layers. The available throughput is not used effectively with the fixed source layer scheme of 1 layer. is high with fixed source layer scheme of 2 layers, because this scheme does not follow temporal fluctuation of the available throughput. In the proposed scheme, the average packet loss rate is as low as the fixed source bit rate scheme of 1 layer, and 7% lower than that of the fixed source bit rate scheme of 2 layers. The throughput of the proposed scheme is higher than that of the fixed source bit rate scheme of 1 layer. These results show that the proposed scheme controls source bit rate, and suppress packet loss rate by accommodating the available throughput fluctuation. V-6 Evaluation of the Proposed Scheme on the Inter-Building Wireless Network The performance evaluation configuration for the inter-building wireless system is shown in Figure 3. The experimental parameters are shown in Table 3. Average sent and received bit rates and packet loss rate with the proposed scheme and fixed number of source layers (1, 2 and 3 layers) scheme is summarized in Table 5. The sent and received bit rates and packet loss rate with the proposed scheme are shown in Figure 15. They show that the proposed scheme controls source bit rate accommodating the fluctuation of the available throughput. The average packet loss rate of the proposed scheme is as low as the fixed source bit rate scheme of 1 layer, and is 7% lower than that of the fixed source bit rate scheme of 2 layers. Additionally, the throughput of the proposed scheme is almost same as that of the fixed source bit rate scheme of 2 layers. The results show that the proposed scheme makes source bit rate adapted to the fluctuation of the available throughput in the inter-building wireless system. Available throughput[kbps] 5 4 3 2 1 2 4 6 8 1 12 Time[s] Figure 12 Temporal Fluctuation of the Available Throughput on the Network Emulator Table 4 Average Sent and Received Bit Rates and Packet Loss Rate Send bit rate [%] Proposed scheme.3 1.9 1.9 Fixed source layer number = 1. 1.4 1.4 Fixed source layer number = 2 7.4 2.7 2.5 Fixed source layer number = 3 24 4. 3.1 6 5 4 3 2 1 Available throughput 2 4 6 8 1 12 Time[s] Figure 13 Sent and Received Bit Rates and Packet Loss Rate (Proposed Scheme) 6 5 4 3 2 1 Available throughput 2 4 6 8 1 12 Time[s] Figure 14 Sent and Received Bit Rates and Packet Loss Rate (Fixed Number of Source Layer=2) 9 8 7 6 5 4 3 2 1 1 9 8 7 6 5 4 3 2 1 1 [%] [%]
Table 5 Average Sent and Received Bit Rates and Packet Loss Rate Send bit rate [%] Proposed scheme. 2.5 2.5 Fixed source layer number = 1. 1.4 1.4 Fixed source layer number = 2 7.4 2.7 2.5 Fixed source layer number = 3 2 4. 3.2 Consideration on parameters of Dynamic Rate Control for Streaming Signal over 2.4GHz Band Wireless Systems,IPSJ SIG Notes,AVM-44,pp.25-3,Mar.24. [6]. http://info.iet.unipi.it/~luigi/ip_dummynet/ 6 5 4 3 2 1 2 4 6 8 1 Figure 15 Sent and Received Bit Rates and Packet Loss Rate (Proposed Scheme) 1 9 8 7 6 5 4 3 2 1 [%] VI. CONCLUSION Authors installed an inter-building communication system using 82.11b wireless network in an urban environment, and found that the available throughput fluctuates temporally. A new dynamic source bit rate control scheme is proposed, which controls video source bit rate based on the estimated available bit rate with use of the measured RTT. The proposed scheme applied to video streaming over the inter-building wireless system achieves low packet loss rate and higher throughput as compared with the fixed source rate scheme. The subjective quality evaluation for the received video stream is to be studied. REFERENCES [1]. Axel Sikora, Voicu F. Groza, Coexistence of IEEE82.15.4 with other Systems in the 2.4 GHz-ISM-Band,pp1786-1791 IMTC 25,May 25. [2]. Shinichi Miyamoto, Satoru Harada,Norihiko Morinaga Performance of 2.4GHz-band Wireless LAN System using Orthogonal Frequency Division Multiplexing Scheme under Microwave Oven Noise Environment, pp157-162, IEEE International Symposium on Electromagnetic Compatibility, May 25 [3]. J.Nagata,A.Osawa,Y.Mukai,N.Nakatani,T.Wakahara, M.Matsumoto,K.Asatani, A Dynamic Rate Control of MPEG-4 Streaming for 2.4GHz Band Wireless Systems under Urban Environments,Technical Report of IEICE,CQ23-117, pp17-pp22,feb.24. [4]. Nagayoshi,T.Hanamura,H.Kasai,H.Tominaga, Scalable video transmission by separating and merging of MPEG-2 Bit stream,multimedia and Expo, pp 16-19, ICME Aug. 21 [5]. M.Wakui,I.Nagayoshi,T.Hanamura,H.Tominaga,