MPEG-4 video transmission over bluetooth links - IEEE Xplore

2 downloads 0 Views 380KB Size Report
This paper addresses this aspect, and proposes three implementations -. MPEG-4 over Bluetooth (MPEG4BT) via HCI, MPEG4BT via. LZCAP and MPEG4BT via ...
MPEG-4 Video Transmission over Bluetooth Links ,Chong Hooi Chia Faculty of Engineering Multimedia University Cyberjaya, Malaysia e-mail : [email protected]

M. Salim Beg Department of Electronics Engineering, Aligarh Muslim University Aligarh, India e-mail : [email protected]

ABSTRACT

emphasis on robustness and low cost. Its implementation is based on a high-performance, yet low cost, integrated radio transceiver. In the latest Bluetooth Specification Version 1.1 [I], there is no profile specified for video transmission over Bluetooth. Although there are some audiohide0 (AV) protocols [21 [31 and profiles [41 specified after Version 1.1,they are not finalized and appear to be very generic. It is still not clear as to how complete video transport solutions can be realized. This opens an opportunity for research into this area and is therefore the motivation for this work.

Video over wireless communication (VOW) has a lot of potential applications in home and office environment. The emergence of two new technologies, namely Bluetooth and MPEG-4, will prove useful in implementing VOW. Video transmission over wireless channel can be made possible by transmitting MPEG-4 bit streams over Bluetooth channels. This paper addresses this aspect, and proposes three implementations MPEG-4 over Bluetooth (MPEG4BT) via HCI, MPEG4BT via LZCAP and MPEG4BT via IP over Bluetooth. I.

INTRODUCTION

Video over wireless communication (VOW)provides access to streaming video and supports a wide range of current and future content forms and applications. In addition, VOW technology is able to support remote applications, such as surveillance and monitoring. The wireless video camera makes the installation easier at any possible area without any cabling issues. VOW also finds many applications in consumer electronics, such as personal videophone, wireless video-intercom, wireless media player, and wireless video camera. These applications are applicable at home, ofiice and even shopping malls. For example, each trolley in the shopping mall has a small wireless video player attached to the trolley handler. When the consumer moves the trolley to certain department, respective video advertisement will be streamed to the wireless video at the trolley panel for advertising the products as well as other sale offers. The wireless video experience also brings the info-kiosk everywhere along with the consumer. This definitely makes the shopping experience more exciting. At home, VOW can find applications such as personal surveillance camera for security monitoring purposes. At office, VOW can.be used as wireless videoconference. Similar applications can be applied to public places such as airports and railway stations. These are possible VOW applications for consumer electronics.

This paper is organized as follows. In section 2, an overview of Bluetooth architecture is described. In section 3, a general description of MPEG-4 is given. In section 4, MPEG4BT system design is presented. In section 5 , 6 and I the implementation of MPEG4BT via HCI, MPEG4BT via L2CAP and MPEG4BT via IP over Bluetooth are discussed. In section 8 the testing scenarios and results are presented. The conclusions are given at the end of this paper. 11. THE BLUETOOTH ARCHITECTURE

The Bluetooth architecture is based on the impression of treating the Bluetooth air interface as a logical link. When Bluetooth devices are within the communication range, the devices are virtually linked together that is analogous to plugging cables between devices. The difference is that the connectivity of Bluetooth devices depends on the communication range rather than being physically plugged for wired devices. Most Bluetooth devices are available in USB and UART dongles, or PC Card for PCMCIA interface. OBEX1 I

I

TCP/IP

I

SDP

I I Audio 1

I

RFCOMM

However, in enabling VOW to the end user level, especially on mobile devices such as cellular phones and PDAs, home appliances and office intercom, there is still no defined standard, protocols or profiles for VOW.Fundamental problems still remain on how to successfully implement MPEG-4 over Bluetooth (MPEG4BT). Figure 1: Bluetooth Protocol Stack

Bluetooth TM is an open industry standard for small form factor, cable replacement, and short-range radio links between portable devices. It has been specified and designed with

280

The Bluetooth protocol stack is outlined in Figure 1. Bluetooth Radio consists of the transceiver, and operates in the ISM (Indusmal, Scientific, Medical) band of 2.4 GHz. The signals are modulated using GFSK (Gaussian Frequency Shift Keying) at a rate of about IMbit/s and transmitted in one of the 79 channels by a pseudo-random hopping sequence hopped through 79 channels at a rate of 1600 hopds. Baseband is the physical layer of the Bluetooth which manages physical channels and links like error correction, packets handling, data whitening, hop selection, paging~andinquiry to access Bluetooth devices, and Bluetooth security. The baseband transceiver applies a time-division duplex (TDD) scheme. Link Manager Protocol (LMP) carries out link setup, authentication and encryption management, link configuration and other protocols. Logical Link and Control Adaptation Protocol (L2CAP) provides connection-oriented and connectionless data services to upper layer protocols with protocol multiplexing capability, segmentation and reassembly operation, and group abstractions. L2CAP only supports ACL (asynchronous connection-less) links and the packet size is up to 64 Kilobytes. Host Controller Interface, HCI provides a uniform interface method to access the Bluetooth hardware capabilities. Service Discovery Protocol, SDP provides the Bluetooth applications to discover available services in Bluetooth environment and to determine the characteristics of those services. Other protocols such as RFCOMM, OBEX and TCPIIP are adopted protocols for Bluetooth to support existing applications. The Bluetooth system operates on a master-slave concept. The master device controls data transmission of up to I slaves through a polling procedure. Slaves can only transmit information after being polled. The device that initiates the connection is defined as master. Bluetooth supports three configurations, point-to-point connection, point-to-multipoint connection (called piconet), and multiple piconet with overlapping coverage areas (called scatternet)

0

master slave

?

Figure 2 Bluetooth configurations: a) point-to-point b) point-to-multipoint, piconet c) scatternet There are two link types between a master and slave(s) SCO (Synchronous Connection-Oriented) link and ACL link. The SCO link can only be a point-to-point link, while ACL link can be a point-to-point link as well as point-to-multipoint link.

111. THE OVERVIEW OF MPEG-4

MF'EG-4 is an ISOnEC standard developed by MPEG (Motion Picture Expert Group) [5]. MPEG-4 builds on the success of MPEG-1 and MPEG-2 in delivering interactive video on CD-ROM, digital television as well as DVD. This new standard also extends to define tools with which to represent audio and video objects, both natural and synthetic, as well as the broad framework for creating, representing, distributing, and accessing digital audiovisual content. Even though MPEG-4 has defined the contents delivery in DMIF (Delivery Multimedia Integration Framework), most applications in the industry are using their own proprietaiy standard in delivering MPEG-4 compressed contents. There are various reasons for choosing MPEG-4 standard as the wireless video codec. MPEG-4's. support for Very Low Bitrate Coding (VLBC) with rates of 5-64kbps can be optimized for the bandwidth constraints of wireless applications. MPEG-4 robust compression method that uses variable length code (VLC) words and predictive frame coding significantly improves efficiency of the compression. MPEG-4's built-in error resiliency tools are beneficial in the transport of video over error-prone 'networks. MPEG-4 also includes Dura Partitioning that maybe useful for wireless transmission. Data Partitioning organizes data differently in the stream to allow each video packet is an .independent entity inside the stream and can be decoded separately from the others. However, not all these benefits of MPEG-4 can be applicable with video transmission over Bluetooth as Bluetooth itself provides mechanism for FEC and ARQ loss packet retransmission. Furthermore, applying MPEG-4 own mechanisms will add up large overhead, which impact the overall performance of video transmission.

IV. MPEG4BT SYSTEM MPEG4BT consists of a few functional components in the system design. On the transmitting side, the video source can be from a pre-compressed MPEG-4 video, or a real-time video captured by video camera and compressed by MPEG-4 encoder. The compressed video can he partitioned into video packets. The video packet is then sent to intermediate protocols for segmentation, which can be L2CAP. HCI or IP over Bluetooth, as will he discussed in the later sections. It then sends the video packets to Bluetooth module for transmission. On the receiving side, the Bluetooth module receives video packets from air, reassembles the video packets in the intermediate protocols, and sends to MPEG-4 decoder for decompression. The decompressed video can be either saved on the storage device, or displayed on screen. Figure 3 shows the functional components in the system design of MPEG4BT. This system design is based on point-to-point Bluetooth connectivity. It is very important to choose the correct intermediate protocols for MPEG4BT in supporting different applications. Various considerations have to be made. The efficiency of the intermediate protocols, such as the size of the overheads,

281

~

segmentation and reassembly processes affects the overall performance of MPEG4BT. Hardware compatibility is also an issue as much as Bluetooth qualification is concerned. The choice of MPEG-4 tile format is concerned in support for various MPEG-4 playback devices. Thus, this paper proposes three existing middleware protocols for MPEG4BT - HCI, LZCAP and IP over Bluetooth, in support for various MPEG4BT applications. Video

Video Display

Device

Intermediate protocols

protocols

Bluetooth Module

Bluetooth Module

transmitter

receiver

Figure 3: MPEG4BT video transmission system V.

MPEG4BT VIA HCI

Most of the video capture devices such as digital video camera and web-cam available in the market are equipped with USB to connect to host PC to stream or download captured frames to the host PC. This proves the reliability and effectiveness of using USB as the medium to transfer video stream. Bluetooth Specification includes USB as one of the Host Controller Interface (HCI). Thus the implementation of MPEG4BT via HCI proves to be an advantage in achieving wireless video transfer with minimal hardware design changes Bluetonth Host

I

.._. -. . ._. Physical Bus Driver (USE)

packets to send down the layers. The physical bus hardware (USB) can be omitted completely, as the .HCI firmware can communicate direct with the HCI driver at the Bluetooth Host. UART is not used because of insufficient bandwidth. USB however, provides generous bandwidth possible for video transport application. Although using HCI is for maximizing the bandwidth usage, there are few issues, which can be disadvantageous to the system design. The transmission of data using HCI across the physical interface is regulated by the buffer sizes available on the receiving side. Different Bluetooth device has different buffer sizes. Both the host and the module need to inquire about the buffer size available for receiving data on the opposite side of the interface and adjust their transmissions accordingly. Thus, the size of the HCI packets actually depends on the buffer size allocated by the Bluetooth device itself. This buffer size is also used to determine the video packet size for MPEG-4 Data Partitioning. Moreover, HCI does not have segmentation and reassembly (SAR) function. Therefore, this implementation must perform SAR based on the HCI buffer size before sending data to HCI. This loads the system resources and therefore impacts the overall system performance, as MPEG-4 compression needs very high processing power. Another disadvantage is that using HCI alone does not promise compatibility among Bluetooth devices, as most Bluetooth devices expect L2CAP header is carried by HCI packet right after the ACL header. If the L2CAP header is not found, the HCI packet will be discarded. However, this case does not appear in MPEG4BT via L2CAP. VI. MPEG4BT VIA LZCAP MPEG4BT via LZCAP is defined for a different approach than MPEG4BT via HCI. In contrast to HCI, which exposes the internal operation of the lower transport protocols, L2CAp itself hides the peculiarities of the Biuetooth lower-layers. Many existing applications developed for higher-layer transport protocols can be made to run over Bluetooth link with little, if any, modification.

I

Physical Bus Hardware

Physical Bus Firmware (USB) 4

MF'EG-4 Streams

I

SDP

4

Bluetooth Module HCI Firmware

t

LMP

Baseband Bluetooth Radio

Figure 4 MPEG4BT via HCI Implementation of MPEG4BT via HCI is illustrated in1 Figure 4. The MPEG-4 bitstreams are packetized into HCI

Figure 5 software layers for MF'EG4BT via L2CAP.

The implementation of MPEG4BT via L2CAP is shown in Figure 5. L2CAP mainly facilitates the segmentation, and reassembly of larger-size, higher-layer packets to and from the smaller baseband packets, whereas HCI does not facilitate this as mentioned in the previous section. Another advantage of using L2CAP over HCI is that L2CAP supports point-to-multipoint 282

~

connection. Video broadcast can be made possible by using L2CAP. MPEG-4 bitstreams are packetized into L2CAP packets. The L2CAP packet is larger than HCI packets to allow more MPEG-4 bitstreams to be packetized into single LZCAP data packets for better streaming performance. Nevertheless, this method actually produces more overheads compared to HCI because of extra bits needed for L2CAP packet encapsulation. To compensate this, best effort to encapsulate only one video packet by MPEG-4 Data Partitioning for each L2CAP packet is implemented. VII. MPEG4BT VIA IP OVER BLUETOOTH

VIII.

Bluetooth link offers speeds up to 723.2kbps downlink and simultaneously 57.6 kbps uplink for ACL packets. The summary of ACL packets is listed in Table 1. Due to the overhead introduced by the higher layers, bandwidth available for the video data transmission is reduced. For video transmission, only ACL packets are concerned, as SCO packets do not provide the transfer rate sufficient for video transmission. SCO packets maximum speed offer is only 64kbps. All ACL packets listed in Table 1 support data integrity checks with CRC code. They are subject to ARQ and therefore are retransmitted in the presence of an error. In additions, DM packets have FEC scheme on the data payload to reduce the number of retransmissions.

Due to the popularity of IP (Internet Protocol) as one of the most common networking protocols and enonnous existing applications and services based on TCPAP, IP over Bluetooth is essential to make Bluetooth applications into the league of IP-based applications. Unfortunately, IP over Bluetooth profile was not specified in the Bluetooth Specifications. However, there are Bluetooth network profiles such as LAN access profile, which can be used to give indirect support for TCP/IP over Bluetwth. Another workaround is by creating a virtual serial port connection between two Bluetooth devices by RFCOMM, with the aids of the operating system, TCPAP protocol can be seated on top of RFCOMM above LZCAP, as shown in Figure 7.

I

MPEG-4 Streams

I

t"'"----i

w RFCOMM

HC1

Baseband

I'

Bluetooth Radio

I

Figure 6: MPEG4BT via IPover Bluetooth Witb,TCPhP over Bluetooth, MPEG4BT task can be made simpler by transferring video stream over IP over Bluetooth. This is a software approach of MPEG4BT in contrast to hardware approach-by MPEG4BT via HCI; where in the latter, hardware design is the main concern. There are weaknesses in this method as there will be more overheads added to the video packets, which may penalize video streaming performance because of limited bandwidth provided by Bluetootb. The user has the least control over Bluetooth device for this method.

TESTING SCENARIOS AND RESULTS

Table 1:ACLpackets

(be tes) DM3 DM5

0-121 0- 183 0-224 0-339

E No

108.8 172.8 258.1 390.4 286.7 433.9

108.8 172.8 387.2 585.6 477.8 123.2

108.8 172.8

57.6

End users or higher layer applications are not supposed to determine the packet type to use in the transmission. However, the Bluetooth baseband controller can be configured to determine the packet types based on the packet size of the higher layers (L2CAP layer and above). This approach is applied: ifrhe higher layer packer size exceeds rhe limits, nexf packet of the minimum size is taken. This approach allows full utilization of the link. For example, to transmit 500 bytes in LZCAP packet, DH5 + DH3 packets are used for transmission. This method reduces the number of occupied slots and overheads to the minimum. In the test, MPEG4IP [7] was used as the MPEG-4 codec and player, while BlueZf81 was installed on two Linux-based PCs for Linux Bluetooth protocol stack. MPEG4BT test applications were written based on MPEG4IP and Blue-2 to support the. approach mentioned above, and to modify MPEG4BT for HCI, L2CAP and P o v e r Bluetooth configurations. The Bluetwth devices used for the test were 3Com Bluetooth USB Devices, which only expect L2CAP header after the ACL header in HCI packet, thus it failed to run MPEG4BT via HCI test. The HCI test was performed using HCIemu, a virtual HCI omulator. The details of test video used are as follows: Filename Uncompressed fde size Number of frames Clip length Frame size Colour depth Frame rate

Another advantage of using IP over Bluetooth is that IP supports Real-time Transport Protocol (RTP), which is a popular format to use for video transmission on the Internet. This also allows the MPEG-4 tile format being transmitted to be compatible with many MPEG-4 applications available in the market. '

283

LS26.avi 1,305.3 Megabytes 4500 180 seconds 352x288 ClF 24-bit 25

The test video was compressed using MPEG4IP with different parameters, i.e. bit rate from lOOkbps to 1000kbps, data partition video packet size set to 100, 300, 500 and 1000 bytes. The compressed videos were stored in the hard disk and the file sizes were obtained as shown in Figure 7. Higher bit rate allows more video information packed for better video quality, which however results larger file size. The results obtained suggest that data partitioning has insignificant impact on file size for compression bit rate above 300kbps. The dotted horizontal line at 16,272 kilobytes (723.2kbpd8-bit * 180 seconds) is the theoretical file size if compression bit rate at 723.2kbps was used. This is equivalent to the theoretical maximum Bluetooth transfer rate using DH5 packets. Video with compression rate higher than 723.2kbps will suffer frame drop or buffering at the receiving end during playback. Thus, the optimal compression bit rate to use should be less than 723.2kbps for smooth video transmission.

receiving end, it can be observed that frame drops and buffering started occurring for bit rate above 600kbps, which caused longer time to complete the transmission. Similar tests were performed with different data partitioning packet size. However, results did not show any significant improvements, possibly suggesting that data pa&tioning is not so useful for transferring MPEG-4 video over Bluetooth channel. 7w

....... 7mkbpr

1

Figure 8: Transfer rate utilization of Bluetooth channel depends on video compression rate for MPEG4BT via LZCAP.

VI. CONCLUSION

Compression Blt Rate (kbps) -no data p a m n ~vae~packetlWObyter _Ivaeo packet 3w bytes vaeo packel500 bytes emBm3 vaeopacket 1W b-o meoretea fib w iv~ 723.2~

Figure 7: Relation of compressed file size to compression bit rate and data partition packet size used The compressed video files were transmitted over Bluetooth channel via different implementation - HCI (emulation), L2CAP and IPover Bluetooth. While video was being played back at the receiving end, the transfer rate was obtained by measuring the total transferred data in bits arrived at the receiving end over the timestamp difference between current packet and previous packet arrived. The test was repeated for compressed video with different bit rate and different video packet size.

This paper presented the functional components in the system design of transmission of MPEG-4 video over wireless channel based on Bluetooth standard. Three video transport implementations of using Bluetooth - MF'EG4BT via HCI, MPEG4BT via L2CAP and MPEG4BT via IP over Bluetooth, their respective applications, advantages and disadvantages have been discussed. The testing scenarios and the results of MPEG4BT implementations have shown the feasibility of transferring video over Bluetooth link using MPEG-4 compression standard. With higher compression rate of MPEG-4, the video quality is reasonable good despite the Bluetooth bandwidth limitation.

REFERENCES Bluetooth SIC, Specification of the Bluetooth System, version 1.1,22 February 2001 Bluetooth SIG, Audio Video Control Transport Protocol (AVCTP) V0.95a. 2 April 2002. Bluetooth SIG, Audio Video Distribution Transport Protocol (AVDTP) V0.95b, 2 April 2002. Bluetooth SIG, Generic Audio Video Distribution Profile (GAVDP), 2 April 2002. Rob Koenen, Overview of the MPEG-4 Standard, ISO/IEC JTCI/SC29/WGII, n3747, La Baule, October 2000. B. A. Miller, C. Bisdikian, Bluetooth Revealed, Prentice Hall, 2002 MPEG4IP. http://mpeg4ip.sourceforge.net BlueZ. http:/ibluez.sourceforge.net

Figure 8 shows the actual transfer rate of MPEG4BT 'via LZCAP. In this test, 10 video files with different compression bit rate ranging from lOOkbps to lO00kbps with lookbps increment were prepared. Data partitioning was not used. Each video was transferred over Bluetooth link via L2CAP implementation and the video received was being played back at the receiving end. The transfer rates were tabulated at every 5 seconds interval. The test was repeated 10 times for each compression bit rate to tabulate 10 different sets of transfer rates. The results suggest that the Bluetooth link saturated around 560kbps, much lower than the theoretical 723.2kbps due to mixture of ACL packets being used for the transmission. At the

284