WO2001008369A1 - Adaptive ofdm transmitter - Google Patents

Abstract

A range of Adaptive Orthogonal Frequency Division Multiplex (AOFDM) video systems are proposed for interactive communications over wireless channels. The proposed constant target bitrate subband adaptive OFDM (CTBR-AOFDM) modems can provide a lower BER, than a corresponding conventional OFDM modem. The slightly more complex switched time-variant target bitrate adaptive OFDM TVTBR-AOFDM modems can provide a balanced video quality performance, across a wider range of channel SNRs. The main advantage of the proposed technique is that irrespective of the prevailing channel conditions, the transceiver achieves always the best possible source-signal representation quality - such as video or audio quality - by automatically adjusting the achievable bitrate and the associated multimedia source-signal representation quality in order to match the channel quality experienced. This is achieved on a near-instantaneous basis under given propagation conditions in order to cater for the effects of path-loss, fast-fading, slow-fading, dispersion, etc. Furthermore, when the mobile is roaming in a hostile outdoor propagation environment, typically low-order, low-rate modem modes are invoked, while in benign indoor environments predominantly the high-rate, high source-signal representation quality modes are employed.

Description

Title of the Invention

ADAPTATIVE OFDM TRANSMITTER

₃ 1 Background of the Invention

The invention relates to adaptive Orthogonal Frequency Division Multiplexing (OFDM) based transmis-

₅ sion of multimedia signals, such as interactive video or audio, speech etc

₆ In contrast to the burst-by-burst reconfigurable wideband multimedia transceivers descπbed in this doc-

7 ument, the term statically reconfigurable found in this context in the literature refers to multimedia β transceivers that cannot be near-instantaneously reconfigured More explicitly, the previously proposed ₉ statically reconfigurable video transceivers were reconfigured on a long-term basis under the base sta- ιo tion's control, invoking for example in the central cell region - where benign channel conditions prevail it - a less robust, but high-throughput modulation mode, such as 4 bit/symbol Quadrature Amplitude Mod- i₂ ulation (16QAM), which was capable of transmitting a quadruple number of bits and hence ensured a

₁₃ better video quality By contrast, a robust, but low-throughput modulation mode, such as 1 bit/symbol

14 Binary Phase Shift Keying (BPSK) can be employed near the edge of the propagation cell, where hostile i_s propagation conditions prevail This prevented a premature hand-over at the cost of a reduced video is quality

₁₇ The philosophy of the fixed, but programable-rate propnetary video codecs and statically reconfigurable is multi-mode video transceivers presented by Streit et al for example in References [1] was that lrrespec- i₉ tive of the video motion activity expeπenced, the specially designed video codecs generated a constant

₂₀ number of bits per video frame For example, for videophony over the second-generation Global System

₂1 of Mobile Communications known as the GSM system at 13 kbps and assuming a video scanning rate of

₂₂ 10 frames/s, 1300 bits per video frame have to be generated Specifically, two families of video codecs

₂₃ were designed, one refraining from using error-sensitive run-length coding techniques and exhibiting the

₂4 highest possible error resilience and another, aiming for the highest possible compression ratio This

₂₅ fixed-rate approach had the advantage of requiπng no adaptive feedback controlled bitrate fluctuation

₂₆ smoothing buffering and hence exhibited no objectionable video latency or delay Furthermore, these

₂ video codecs were amenable to video telephony over fixed-rate second-generation mobile radio systems,

28 such as the GSM

₂₉ The fixed bitrate ot the above propnetary video codecs is in contrast to existing standard video codecs, ₀ such as the Motion Pictures Expert Group codecs known as MPEG 1 and MPEG2 or the ITU's H 263 3i codec, where the time-vaπant v ideo motion activity and the variable-length coding techniques employed 2 result in a time-vaπant bitrate fluctuation and a near-constant perceptual video quality This time-vaπant 3 bitrate fluctuation can be mitigated bv employing adaptive feed-back controlled buffering, which po- 4 tentially increases the latency or delay of the codec and hence it is often objectionable for example in 5 interactive videophony The schemes presented by Streit et al in References [1] result in slightly vaπable 6 video quality at a constant bitrate, while refraining from employing buffering, which again, would result in latency in interactive videophony A range of techniques, which can be invoked, in order to render the 8 family ot variable-length coded, highly bandwidth-efficient, but potentially error-sensitive class of stan- 9 dard video codecs, such as the H 263 arrangement, amenable to error-resilient, low-latency interactive 0 wireless multimode videophony was summarised in [2] The adaptive video rate control and packetisa- 1 tion algoπthm of [2] generates the required number of bits for the burst-by-burst adaptive transceiver. 2 depending the on the capacity of the current packet, as determined by the current modem mode. Fur- 3 ther error-resilient H 263-based schemes were contπved for example by Farber, Steinbach and Girod at Erlangen University [3], while Sadka, Eryurtlu and Kondoz [4] from Surrey University proposed a 5 range of improvements to the H.263 scheme Following the above portrayal of the pπor art in both video 6 compression and statically reconfigurable narroband modulation, let us now consider the philosophy of 7 wideband burst-by-burst adaptive quadrature amplitude modulation (AQAM) in more depth. 8 In burst-by-burst adaptive modulation a higher-order modulation scheme is invoked, when the channel 9 is favourable, in order to increase the system's bits per symbol capacity and conversely, a more robust o lower order modulation scheme is employed, when the channel exhibits inferior channel quality, in order i to improve the mean Bit Error Ratio (BER) performance A practical scenano, where adaptive modula- 2 tion can be applied is, when a reliable, low-delay feedback path is created between the transmitter and 3 receiver, for example by supeπmposing the estimated channel quality perceived by the receiver on the 4 reverse-direction messages of a duplex interactive channel. The transmitter then adjusts its modem mode 5 according to this perceived channel quality 6 Recent developments in adaptive modulation over a narrow-band channel environment have been pi- 7 oneered by Webb and Steele [5], where the modulation adaptation was utilized in a Digital European β Cordless Telephone - like (DECT) system The concept of vaπable rate adaptive modulation was also 9 advanced by Sampei et al [6], showing promising advantages, when compared to fixed modulation in 0 terms of spectral efficiency, BER performance and robustness against channel delay spread In another 1 paper, the numeπcal upper bound performance of adaptive modulation in a slow Rayleigh flat-fading 2 channel was evaluated by Torrance et al[l] and subsequently, the optimization of the switching threshold 3 levels using Powell minimization was used in order to achieve a targeted performance [8, 9] In addition. b adaptive modulation was also studied in conjunction with channel coding and power control techniques

65 by Matsuoka et al [6] as well as Goldsmith et al [ 10] ee In the narrow-band channel environment, the quality ot the channel was determined bv the short term

6- Signal to Noise Ratio (SNR) ot the received burst, which was then used as a cπteπon in order to choose ea the appropπate modulation mode for the transmitter, based on a list of switching threshold levels, /„ [5, 9] 9 However, in a wideband environment, this cπteπon is not an accurate measure for judging the quality ot

70 the channel, where the existence ot multi-path components produces not only power attenuation of the

71 transmission burst, but also intersymbol interference Subsequently, a new cπteπon has to be defined to

72 estimate the wideband channel quality in order to choose the appropπate modulation scheme

7 2 Summary of the Invention

74 Particular and preferred aspects of the invention are set out in the accompanying independent and depen-

75 dent claims Features of the dependent claims may be combined with those of the independent claims as

76 appropπate and used in combinations other than those explicitly set out in the claims

77 The performance benefits of OFDM symbol-by-symbol adaptive modulation are descπbed, employing a

78 higher-order modulation mode on those OFDM subcarπers, where the frequency-domain channel trans-

79 fer function is favourable, le does not exhibit a high attenuation, or at subchannel frequencies, where so the signal is unimpaired by co-channel interferers This procedure is employed, in order to increase the βi system's bits per symbol (BPS) capacity and conversely, invoking a more robust, lower order modulation

82 mode, when the channel exhibits infeπor channel quality

83 Two specific embodiments are descπbed, a fixed bitrate and a time-vaπant bitrate system The fixed-rate

8 system allocates a fixed number of bits to each OFDM symbol, mapping the bits on to the highest-quality ss subcarπers Hence this system optimises the bit allocation across the frequency domain, but ignores the βe time-vaπant nature of the channel quality Therefore the associated bit error rate (BER) will be time-

87 vaπant By contrast, the time-vaπant bitrate system adjusts the number of bits mapped to the OFDM

88 symbol on a time-vaπant basis, depending on the instantaneous channel quality Hence it endeavours

89 to optimise the bit allocation versus both time and frequency This bit allocation policy allows us to 30 maintain a near-constant BER versus time

9i It is shown that due to the descπbed adaptive modem mode switching regime a seamless multimedia

92 source-signal representation quality - such as video or audio quality - versus channel quality relationship

93 can be established, resulting in a near-unimpaired multimedia source-signal quality πght across the oper-

94 ating channel Signal-to-Noise Ratio (SNR) range The main advantage of the descπbed technique is - in n particular in the context ot the time variant bitrate embodiment - that irrespective ot the prevailing chan-

96 nel conditions, the transceiver achieves always the best possible source-signal representation quality -

97 such as video or audio quality - by automatically adjusting the achievable bitrate and the associated mul-

98 timedia source-signal representation quality in order to match the channel quality expenenced This can 39 achieved on a near-instantaneous or OFDM symbol-by-symbol adaptive basis under given propagation oo conditions in order to cater for the effects ot path-loss fast-fading, slow-fading, dispersion, co-channel i interference, etc Furthermore, when a mobile is roaming in a hostile out-doors - or even hilly terrain ιo2 - propagation environment, typicallv low-order, low-rate modem modes are invoked, while in benign

103 indoor environments predominantly the high-rate, high source-signal representation quality modes are

104 employed

IDS 3 Brief Description of the Drawings

me For a better understanding of the invention and to show how the same may be earned into effect reference ιo7 is now made by way of example to the accompanying drawings, in which

,08 List of Figures

ιo9 1 Signalling scenaπos in adaptive modems 26 o 2 Reconfigurable transceiver schematic 27 in 3 The micro-adaptive nature of the subband-adaptive modem The top graph is a contour

112 plot of the channel SNR for all 512 subcarπers versus time The bottom two graphs i,3 show the modulation modes chosen for all 16 32-subcamer subbands for the same peπod

1,4 of time The middle graph shows the performance of the 3 4Mbps subband-adaptive

,,5 modem, which operates at the same bitrate as a fixed BPSK modem The bottom graph

1.6 represents the 7 0Mbps subband-adaptive modem, which operated at the same bitrate as

1.7 a fixed QPSK modem The Average channel SNR was 16dB 28 nβ 4 Instanteous Channel SNR for all 512 subcarπers versus time, for an average channel ng SNR of 16dB over the channel characteπsed by the impulse response ot Figure 5 29 i2o 5 Indoor three-path WATM channel impulse responce 29 FER or video packet loss ratio (PLR) versus channel SNR for the BPSK and QPSK fixed modulation mode OFDM transceivers and for the corresponding subband-adaptive μAOFDM transceiver, operating at identical effective video bitrates, namely at 3.4 and

7.0 Mbps, over the channel mode of Figure 5 at F = 7.41 x 10^-2 30 Effective throughput bitrate versus channel SNR for the BPSK and QPSK fixed modulation mode OFDM transceivers and that of the corresponding subband-adaptive or μAOFDM transceiver operating at identical effective video bitrates of 3.4 and 7.0 Mbps, over the channel of Figure 5 at FD = 7.41 x 10^~2 31 Average video quality in PSNR versus channel SNR for the BPSK and QPSK fixed modulation mode OFDM transceivers and for the corresponding μAOFDM transceiver operating at identical channel SNRs over the channel mode of Figure 5 at FD — 7.41 x 10^-2. 32 FER or video packet loss ratio (PLR) versus channel SNR for the subband-adaptive OFDM transceivers of Table 2 operating at four different target bitrates, over the channel model of Figure 5 at F = 7.41 x 10^~2 33 Average video quality expressed in PSNR versus channel SNR for the subband-adaptive OFDM transceivers of Table 2, operating at four different target bitrates, over the channel model of Figure 5 at F_D = 7.41 x 10^~2 34 Video quality and packet loss ratio (PLR) performance versus video frame index (time) comparison of subband-adaptive OFDM transceivers having target bitrates of 1.8, 3.4 and 7.0Mbps, under the same channel conditions, at 16dB SNR over the channel of Figure 5 at F_D = 7.41 x 10^-2. (a)- top; (b) - middle (c) - bottom 35 Illustration of mode switching for the switched subband adaptive modem. The figure shows the estimate of the bit error ratio for the four possible modes. The large square and the dotted line indicate the modem mode chosen for each time interval by the mode switching algorithm. At the bottom of the graph the bar chart specifies the bitrate of the switched subband adaptive modem on the right-hand axis, versus time. Using the channel model of Figure 5 at o = 7.41 x 10^-2 36 The micro-adapti e nature ot the time-vaπent target bitrate subband-adaptive TVTBR- AOFDM) modem The top graph is a contour plot of the channel SNR tor all 512 sub- earners versus time The bottom graph shows the modulation mode chosen tor all 16 subbands tor the same penod ot time Each subband is compπsed of 32 subcamers The TVTBR AOFDM modem switches between target bitrates of 2, 3 4, 7 and 10Mbps, while attempting to maintain an estimated BER of 0 1 % before channel coding Average Channel SNR is 16dB over the channel of Figure 5 at Fp = 7 41 x 10^~2 , (a) - top graph (b) bottom graph 37 FER or video packet loss ratio versus channel SNR for the TVTBR- AOFDM modem for a vaπety of BER switching thresholds The switched modem uses four modes, with target bitrates of 1 8, 3 4, 7 and 10Mbps The un-switched 1 8 and 10Mbps results are also shown on the graph as solid markers The channel model of Figure 5 at FD = 7 41 x 10^-2 38 Transmitted bitrate of the switched TVTBR- AOFDM modem, for a vaπety of BER switching thresholds The switched modem uses tour modes, having target bitrates of

1 8, 3 4, 7 and 10Mbps, over the channel model of Figure 5 at FQ = 7 41 x 10^-2 39 Effective throughput bitrate of the switched TVTBR- AOFDM modem for a vaπety of BER switching thresholds The switched modem uses four modes, with target bitrates of

1 8, 3 4. 7 and 10Mbps The channel model of Figure 5 is used at F_D — 7 41 x 10^-2 40 Video quality and packet loss ratio performance versus video frame index (time) com- pansion between switched TVTBR-AOFDM transceivers with different BER switching thresholds, at an average of 16dB SNR, using the channel model of Figure 5 at D = 7 41 x 10^~2 (a) - top graph, (b) - middle graph, (c) - bottom graph 41 Average PSNR versus channel SNR performance for switched- and un-switched subband adaptive modems Figure (a) compares the four un-switched CTBR subband adaptive modems with switched TVTBR subband adaptive modems (using the same four modem modes) for switching thresholds of BER=3, 5 and 10% Figure (b) compares the switched TVTBR AOFDM modems for switching thresholds of BER=0 1 , 1 , 2, 3, 5 and 10% 42 ι , 4 Detailed Description

i76 4.1 State-of-the-art

7 Burst-bv -burst adaptive quadrature amplitude modulation (AQAM) was contπved by Steele and Webb [5], ι-8 in order tor the transceiver to cope with the time-vaπant channel quality of naσowband fading channels

179 Further related research was conducted at the University of Osaka bv Sa pei and his colleagues, mvesti-

,80 gating v ariable coding rate concatenated coded schemes [6], at the University of Stanford by Goldsmith

,31 and her team, studying the effects of variable-rate, vaπable-power arrangements [ 10] and at Southamp-

,82 ton University in the UK. investigating a vaπety of practical aspects ot AQAM [1 1 , 12] The channel's

183 quality is estimated on a burst-by-burst basis and the most appropπate modulation mode is selected in or-

,β4 der to maintain the required target bit error rate (BER) performance, whilst maximizing the system's Bit

,85 Per Symbol (BPS) throughput Using this reconfiguration regime the distπbution of channel errors be-

,66 comes typically less bursty, than in conjunction with non-adaptive modems, which potentially increases

,87 the channel coding gains Furthermore, the soft-decision channel codec metπcs can be also invoked in las estimating the instantaneous channel quality, irrespective of the type of channel impairments

189 A range of coded AQAM schemes were analysed by Matsuoka et al [6], Lau et al [13] and Gold- i9o smith et al [ 10] For data transmission systems, which do not necessaπly require a low transmission

,9, delay, vanable-throughput adaptive schemes can be devised, which operate efficiently in conjunction

,92 with powerful error correction codecs, such as long block length turbo codes However, the acceptable

,93 turbo interleaving delay is rather low in the context of low-delay interactive speech Video communica-

,94 tions systems typically require a higher bitrate than speech systems and hence they can afford a higher

,95 interleaving delay

,96 The above pnnciples - which were typically investigated in the context of narrowband modems - were

,97 further advanced in conjunction with wideband modems, employing powerful block turbo coded wide-

198 band Decision Feedback Equaliser (DFE) assisted AQAM transceivers [ 14] A neural-network Radial

199 Basis Function (RBF) DFE based AQAM modem design was proposed in [15], where the RBF DFE

200 provided the channel quality estimates for the modem mode switching regime This modem was capa- 20, ble of removing the residual BER of conventional DFEs, when linearly non-separable received phasor

202 constellations were encountered

203 The above burst-by-burst adaptive principles can also be extended to Adaptive Orthogonal Frequency

204 Division Multiplexing (AOFDM) schemes [ 16] and to adaptive joint-detection based Code Div ision

205 Multiple Access (JD-ACDMA) arrangements [ 17] The associated AQAM pnnciples were invoked in

206 the context ot parallel AOFDM modems also by Czylwik et al [ 18], Fischer [ 19] and Chow et al [20] o- Adaptiv e subcarπer selection has been advocated also bv Rohhng et al [21 ] in order to achieve BER per-

208 tormance improv ements Due to lack of space without completeness, further significant advances over

209 benign, slowly v arying dispersive Gaussian fixed links - rather than over hostile wireless links - are due 'io to Chow, Cioffi and Bingham [20] trom the US A., rendeπng OFDM the dominant solution for asymmet- 2,, πc digital subscπber loop (ADSL) applications, potentially up to bitrates of 54 Mbps In Europe OFDM ',2 has been favoured for both Digital Audio Broadcasting (DAB) and Digital Video Broadcasting [22, 23]

2.3 (DVB) as well as for high-rate Wireless Asynchronous Transfer Mode (WATM) systems and for the

2.4 new HIPERLAN standard due to its ability to combat the effects ot highly dispersive channels The

2.5 idea of 'water-filling' - as allocating different modem modes to different subcarners was referred to -

2.6 was proposed for OFDM by Kalet [24] and later further advanced by Chow et al [20] This approached

2.7 was rendered later time-vaπant for duplex wireless links for example in [16] Lastly, the co-channel

2.8 interference sensitivity of OFDM can be mitigated with the aid of adaptive beam-forming in multi-user _^,9 scenaπos

220 Our main contπbution is that upon invoking the technique advocated - irrespective of the channel con-

22, ditions expeπenced - the transceiver achieves always the best possible video quality by automatically

222 adjusting the achievable bitrate and the associated video quality in order to match the channel quality ex-

223 penenced This is achieved on a near-instantaneous basis under given propagation conditions in order to

224 cater for the effects of path-loss, fast-fading, slow-fading, dispersion, co-channel interference, etc Fur-

225 thermore, when the mobile is roaming in a hostile outdoor propagation environment, typically low-order,

226 low-rate modem modes are invoked, while in benign indoor environments predominantly the high-rate,

227 high source-signal representation quality modes are employed

228 4.2 AOFDM Signalling Scenarios

229 AOFDM transmission parameter adaptation is an action of the transmitter in response to time-varying

230 channel conditions It is only suitable for duplex communication between two stations, since the trans-

23, mission parameter adaptation relies on some form ot channel estimation and signalling In order to

232 efficiently react to the changes in channel quality, the following steps have to be taken

233 • Channel qualitv estimation In order to appropπately select the transmission parameters to be

234 employed for the next transmission a reliable prediction of the channel quality dunng the next

235 active transmit timeslot is necessary

236 • Choice of the appropriate parameters for the next transmission Based on the prediction of the

2 7 expected channel conditions during the next timeslot. the transmitter has to select the appropriate 238 modulation schemes tor the subcarners

2 9 • Signalling or blind detection of the emplo\ ed parameters The receiv er has to be informed, as

240 to which set ot demodulator parameters to employ for the received packet This information can

24, either be conveyed within the packet, at the cost ot loss of useful data bandwidth, or the receiver 42 can attempt to estimate the parameters employed at the transmitter by means ot blind detection

2 3 mechanisms

2 Depending on the channel charactenstics, these operations can be performed at either of the duplex

245 stations, as shown in Figures 1 (a), 1 (b) and 1 (c) If the channel is reciprocal, then the channel quality

246 estimation tor each link can be extracted from the reverse link, and we refer to this regime as open-

2 7 loop adaptation In this case, the transmitter needs to communicate the transmission parameter set to 2 β the receiver (Figure 1 (a)), or the receiver can attempt blind detection of the transmission parameters

249 employed (Figure 1(c))

250 If the channel is not reciprocal, then the channel quality estimation has to be performed at the receiver

25, of the link. In this case, the channel quality measure or the set of requested transmission parameters is

252 communicated to the transmitter in the reverse link (Figure 1 (b)) This mode is referred to as closed-loop

253 adaptation.

254 4.3 Video Transceiver

255 The schematic of the whole system is depicted in Figure 2 The multimedia source signal generated

256 by the video encoder of Figure 2 is assembled into transmission packets constituting an OFDM symbol

257 and the bits may be additionally mapped by the Mapper of Figure 2 to n number of different Forward 25β Error Correction (FEC) protection classes These bits are then convenveyed to the optional Time Division

259 Multiplex (TDMA) scheme of Figure 2, before they are assigned to the OFDM subcarners of the adaptive

260 QAM modem seen in Figure 2

261 As a particular embodiment of the proposed system concept, in this study we investigate the transmission

262 of 704x576 pixel Four-times Common Intermediate Format (4CIF) high-resolution video sequences at 30

263 frames/s using a subband-adaptive turbo-coded OFDM transceiver The transceiver can modulate 1 , 2 or 2 4 4 bits onto each OFDM sub-earner, or simply disable transmissions for sub-camers which exhibit a high

265 attenuation, or phase distortion due to channel effects We note, however that the proposed pnnciples are

266 applicable to arbitrary multimedia source signals, bit rates, source signal representation quality, or even

26 different channel codecs

26β The main advantage of the proposed technique is that irrespective of the prevailing channel conditions, bj the transceiver achieves alvvavs the best possible source-signal representation quahtv - such as video or

-o audio quahtv - bv automaticallv adjusting the achievable bitrate and the associated multimedia source-

^■>? signal representation quahtv in order to match the channel quality expeπenced This is achieved on a

2-_' near-instantaneous basis under given propagation conditions in order to cater for the effects of path-

-3 loss, fast-fading, slow-fading, dispersion, co-channel interference, etc Furthermore, when the mobile

•>74 is roaming in a hostile out-doors - or even hillv terrain - propagation environment, typically low-order,

2-5 low-rate modem modes are invoked while in benign indoor environments predominantly the high-rate,

2-6 high source-signal representation quality modes are employed

277 The H 263 video codec exhibits an impressive compression ratio, although this is achieved at the cost ot a

278 high vulnerability to transmission errors, since a run-length coded bitstream is rendered undecodable by

2 9 a single bit error In order to mitigate this problem, when the channel codec protecting the video stream 230 is overwhelmed by the transmission errors, we refrain from decoding the corrupted video packet, in order 28i to prevent error propagation through the reconstructed video frame buffer [2] We found that it was more

282 beneficial in video quality terms, if these corrupted video packets were dropped and the reconstructed

283 frame buffer was not updated, until the next video packet replenishing the specific video frame area

284 was received The associated video performance degradation was found perceptually unobjectionable 2β5 tor packet dropping- or transmission frame error rates (FER) below about 5% These packet dropping

286 events were signalled to the remote video decoder by supeπmposing a strongly protected one-bit packet

287 acknowledgement flag on the reverse-direction packet, as outlined in [2] Turbo error correction codes

288 were used The associated parameters will be discussed in more depth dunng our further discourse

289 4.4 Comparing subband-adaptive to fixed modulation mode transceivers

290 In order to show the benefits of the proposed subband-adaptive OFDM transceiver, we compare its per-

291 formance to that of a fixed modulation mode transceiver under identical propogation conditions, while

292 having the same transmission bitrate The subband-adaptive modem is capable of achieving a low bit

293 error ratio, since it can disable transmissions over low quality sub-earners and compensate for the lost

294 throughput by invoking a higher modulation mode, than that of the fixed-mode transceiver over the high-

295 quality sub-earners

296 Table l shows the system parameters for the fixed BPSK and QPSK transceivers, as well as for the

297 corresponding AOFDM transceivers The system employs constraint length 3, l/2-rate turbo coding, 98 using octal generator polynomials of 5 and 7, and random interleavers Hence the unprotected bitrate is

299 about half the channel coded bitrate The protected to unprotected bitrate ratio is not exactly half, since ••oo two tailing bits are required to reset the convolutional encoders' memory to their default state in each

Table 1. System parameters for the fixed QPSK and BPSK transceivers, as well as for the corresponding subband-adaptive OFDM (AOFDM) transceivers for Wireless Local Area Networks (WLANs).

oι transmission burst. In both modes a 16-bit CRC is used for error detection, and 9 bits are used to encode

302 by simple repetition coding the reverse link feedback acknowledgement information. The feedback flag

303 decoding ensues using majoπty logic decoding. The packetisation requires a small amount of header

304 information added to each transmitted packet, which is 1 1 and 12 bits/packet for BPSK and QPSK,

305 respectively The effective video bitrates for the BPSK and QPSK modes are then 3 4 and 7 0 Mbps.

306 The fixed mode BPSK and QPSK transceivers are limited to one and two bits per symbol, respectively so? However, the AOFDM transceivers operate at the same bitrate, as their corresponding fixed modem

308 mode counterparts, although they can vary their modulation mode on a sub-earner by sub-earner basis

309 between 0, 1 , 2 and 4 bits per symbol Zero bits per symbol implies that transmissions are disabled for 3io the sub-earner concerned

3i, The "micro-adaptive" nature of the subband-adaptive modem is characteπsed by Figure 3, portraying 3, at the top a contour plot ot the channel SNR for each subcarπer versus time At the centre and bottom n ot the figure the modulation mode chosen for each 32 subcamer subband is shown versus time for the

3.4 3 4 and 7 0 Mbps subband-adaptive modems respectiv ely The channel SNR is also shown in a three- .5 dimensional form in Figure 4 which mavbe more convenient to visualise It can be seen that when the

3.6 channel is of high quality - like tor example at about frame 1080 - the subband-adaptive modem used 3, the same modulation mode as the equivalent fixed rate modem in all subcarners When the channel is

3.8 hostile - like around frame 1060 - the subband-adaptive modem used a lower-order modulation mode

3.9 in some subbands, than the equivalent fixed mode, or in extreme cases disabled transmission for that 2o subband In order to compensate tor the loss of throughput in this subband a higher-order modulation

32, mode was used in the higher quality subbands

322 One video packet is transmitted per OFDM symbol, theretore the video packet loss ratio is the same, as

323 the OFDM transmission frame error ratio The video packet loss ratio is plotted versus the channel SNR

324 in Figure 6 It is shown in the graph that the subband-adaptive transceivers - or svnonymouslv termed

325 as microscopic-adaptive (μAOFDM) in contrast to OFDM symbol-by-symbol adaptive transceivers -

326 have a lower packet loss ratio at the same SNR compared to the fixed modulation mode transceiver

327 Note in Figure 6 that the subband-adaptive transceivers can operate at lower channel SNRs, than the 32β fixed modem mode transceivers, while maintaining the same required video packet loss ratio Again, the

329 figure labels the subband-adaptive transeivers as μAOFDM, implying that the adaption is not noticable

330 from the upper layers of the system A macro-adaption could be applied in addition to the microscopic

33, adaption by switching between different target bitrates, as the longer-term channel quality improves and

332 degrades This issue is the subject of Section 4 6

333 Having shown how the subband-adaptive transceiver achieved a reduced video packet loss, in compar-

334 ison to fixed modulation mode transceivers under identical channel conditions, we now compare the

335 effective throughput bitrate of the fixed and adaptive OFDM transceivers in Figure 7 The figure shows

336 that when the channel quality is high, the throughput bitrate of the fixed and adaptive transceivers are

337 identical However, as the channel degrades, the loss of packets results in a lower throughput bitrate The 3 β lower packet loss ratio ot the subband-adaptive transceiver results in a higher throughput bitrate than that

3 9 of the fixed modulation mode transceiver

3 0 The throughput bitrate performance results translate to the decoded video quality performance results 4, evaluated in terms of PSNR in Figure 8 Again, for high channel SNRs, the performance of the fixed

342 and adaptive OFDM transceiv ers is identical However, as the channel quality degrades the video qual-

3 3 ltv of the subband-adaptive transceiver degrades less dramatically than that of the corresponding fixed

344 modulation mode transceiver »= 4.5 Comparing subband-adaptive transceivers having different target bitrates

3 6 As mentioned before, the subband-adaptive modems employ different modulation modes tor different

47 subcarπers in order to meet the target bitrate requirement at the lowest possible channel SNR This is 8 achieved by using a more robust modulation mode or eventually by disabling transmissions over subcar-

349 πers having a low channel quality Bv contrast, the adaptive system can invoke less robust, but higher

350 throughput modulation modes over subcarners exhibiting a high channel quality In the examples we J5, have previously considered we chose the AOFDM target bitrate to be identical to that ot a fixed mod- 52 ulation mode transceiver In this section we comparatively study the performance of vaπous μAOFDM

353 systems having different target bitrates

354 The previously descπbed μAOFDM transceiver of Table 1 exhibited a FEC-coded bitrate of 7 2Mbps, 35= which was also equivalent to that of a fixed BPSK transceiver and provided an effective video bitrate 35D of 3 4Mbps If the video target bitrate is lower than 3 4Mbps. then the system can disable transmission 3=7 in more ot the subcarners, where the channel quality is low Such a transceiver would have a lower

358 bit error rate, than the previous BPSK-equivalent μAOFDM transceiver, and therefore could be used at

359 lower average channel SNRs, while maintaining the same bit error ratio target By contrast, as the target

360 bitrate is increased the system has to employ higher-order modulation modes in more subcarners, at the 36, cost of an increased bit-error ratio Therefore high target bitrate μAOFDM transceivers can only perform

362 within the required bit error ratio constraints at high channel SNRs, while low target bitrate μAOFDM

363 systems can operate at low channel SNRs without inflicting excessive BERs Therefore a system, which

364 can ad_just its target bitrate, as the channel SNR changes, would operate over a wide range ot channel

365 SNRs, providing the maximum possible throughput bitrate, while maintaining the required bit error ratio 66 Hence below we provide a performance compaπson of vaπous μAOFDM transceivers having tour dif-

367 ferent target bitrates, of which two are equivalent to that of the BPSK and QPSK fixed modulation mode 36β transceivers of Table 1 The system parameters for all four different bitrate modes are summaπsed in _^69 Table 2 The modes having effective video bitrates of 3 4 and 7 0Mbps are equivalent to the bitrates of a 370 fixed BPSK and QPSK mode transceiver, respectively

3-, Figure 9 shows the FER or video packet loss ratio (PLR) performance versus channel SNR for the four

372 different target bitrates of Table 2 The results demonstrate - as expected - that the higher target bitrate

.,73 modes require higher channel SNRs in order to operate within given PLR constraints For example

37 the mode having an effective video bitrate of 9 8Mbps can only operate for channel SNRs in excess of

3-5 19dB under the constraint of a maximum PLR of 5% However, the mode having an effective video

3-6 bitrate of 3 4Mbps can operate at channel SNRs of 1 l dB and above whilst maintaining the same 5%

Table 2 System parameters tor the four different target bitrates of the vanous subband-adaptive OFDM transceivers (μAOFDM) 3-7 PLR constraint, albeit at about halt the throughput bitrate, and hence at a lower video quahtv 78 The tradeoffs between video quality and channel SNR for the vaπous target bitrates can be _judged from

379 Figure 10 suggesting - as expected - that the higher target bitrates result in a higher video quality,

380 provided that channel conditions are sufficiently favorable However as the channel quality degrades.

38, the video packet loss ratio increases, thereby reducing the throughput bitrate, and hence the associated

382 video quality The lower target bitrate transceivers operate at an inherently lower video quality, but they 83 are more robust to the prevailing channel conditions and hence can operate at lower channel SNRs, while

384 guaranteeing a video quality, which is essentially unaffected by channel errors It was found that the

385 perceived video quality became impaired for packet loss ratios in excess of about 5%

3β6 The tradeoffs between video-quality, packet loss ratio and the target bitrate are further augmented with sβ? reference to Figures 1 1(a), (b) and (c) The figure shows the video quality measured in PSNR versus

38β video frame index at a channel SNR of 16dB and also for an error free situation At the bottom of each

389 graph the packet loss ratio per video frame is shown The three figures indicate the tradeoffs to be made

390 in choosing the target bitrate for the specific channel conditions expeπenced - in this specific example for

39, a channel SNR of 16dB Note that under error free conditions the video quality improved upon increasing

392 the bitrate

393 Specicially, video PSNRs of about 40, 41 5 and 43dB were observed for the effective video bitrates of

394 1 8, 3 4 and 7 0Mbps Figure 11(a) shows that for the target bitrate of 1 8Mbps, the system has a high

395 grade of freedom in choosing, which subcarπers to invoke and therefore it is capable of reducing the

396 number of packets that are lost The packet loss ratio remains low and the video quality remains similar

397 to that of the error free situation The two instances, where the PSNR is significantly different from 39β the error free performance correspond to video frames, in which video packets were lost However, the

399 system recovers in both instances in the following video frame

400 As the target bitrate of the subband-adaptive OFDM transceiver is increased to 3 4Mbps (see Fig-

40, ure 11(b)), the subband modulation mode selection process has to be more "aggressive", resulting in 02 increased video packet loss Observe in the figure that the transceiver having an effective video bitrate of 03 3 4Mbps, exhibits increased packet loss, and in one frame as much as 5% of the packets transmitted for

404 that video frame were lost, although the average PLR was only 0 4% Due to the increased packet loss

405 the video PSNR curve diverges from the error-free performance curve more often However, in almost

406 all cases the effects ot the packet losses are masked in the next video frame, indicated by the re-merging 07 PSNR curves in the figure, maintaining a close to error-free PSNR The subjective effect of this level of

408 packet loss is almost inperceivable

409 When the target bitrate is further increased to 7 0Mbps (see Figure 1 1 (c)), the average PLR is about 5% j, > under the same channel conditions and the effects ot this packet loss ratio are becoming ob_jectionable in

4i perceived video quahtv terms At this target bitrate, there are several video frames, where at least 10% ot

4,2 the video packets have been lost The video quality measured in PSNR terms rarely reaches its error-tree ι,3 level, due to the tact that everv video frame contains at least one lost packet The perceived video quality

4. remains virtuallv unimpaired, until the head movement in the 'Suzie ' video sequence around frames

4.5 40-50 where the effect ot lost packets becomes obvious, and the PSNR drops to about 30dB

.6 4.6 Modifying the target bitrate based on channel conditions

4.7 Bv using a high target bitrate, when the channel quality is high, while a reduced target bitrate, when the 4,a channel quality is poor, such an adaptive system is capable ot maximising the average throughput bitrate

419 over a wide range of channel SNRs, while maintaining a given quality constraint This quality constraint

420 tor our video system could be a maximum packet loss ratio

42i However there is a substantial processing delay associated with evaluating the packet loss information

422 and therefore modem mode switching based on this metπc would be less efficient due to this latency

423 Therefore we decided to invoke an estimate of the bit error ratio (BER) for mode switching The channel

424 quality estimator can estimate the expected bit error ratio based on each specific modulation mode chosen

425 for each subband We decided to use a quadruple-mode switched subband-adaptive modem, using the

426 four target bitrates of Table 2 The channel estimator can then estimate the expected bit error ratio of

427 the four possible modem modes The modem mode for the next OFDM symbol is then choosen based

428 upon the estimate of BER for each of the four modes Our switching scheme opted for the modem mode,

429 whose estimated BER was below the required threshold This threshold could be vaπed in order to tune

430 the behaviour ot the switched subband-adaptive modem tor a high or a low throughput The advantage of 43i a higher throughput was a higher error-free video quality at the expense of increased video packet losses, 32 which could reduce the perceived v ideo quality 33 Figure 12 demonstrates, how the switching algonthm operates for a 1 % estimated BER threshold Specif-

434 ically, the figure portrays the estimate of the bit error ratio for the four possible modem modes versus 35 time The large square and the dotted line indicates the mode chosen for each time inter al by the mode 6 switching algorithm The algonthm attempts to use the highest bitrate mode, whose BER estimate is less ,37 than the target threshold namely 1 % in this case However, if all the four modes' estimate of the BER 38 is above the 1 % threshold, then the lowest bitrate mode is chosen, since this will be the most robust to 9 channel errors An example of this is shown around frames 1035-1040 At the bottom of the graph a bar , 0 chart specifies the bitrate of the switched subband adaptive modem versus time, in order to emphasise 44 when the switching occurs 2 An example ot the algorithm when switching amongst the target bitrates ot 1 8, 3 4, 7 and 10Mbps is

443 shown in Figures 13(a) and (b) Figure 13(a) portrays the contour plot of the channel SNR tor each 44 subcarπer versus time Figure 13(b) displays the modulation mode chosen tor each 32-subcarπer sub-

445 band versus time tor the time-vaπent target bitrate (TVTBR) subband adaptive modem It can be seen

446 at frames 1051- 1055 that all the subbands employ QPSK modulation, therefore the TVTBR-AOFDM

447 modem has an instantaneous target bitrate of 7Mbps As the channel used by the 3 4 Mbps QPSK mode 4 β degrades around frame 1060, the modem has switched to the more robust 1 8Mbps BPSK mode When

449 the channel quality is high around frames 1074- 1081 , the highest bitrate 10Mbps 16QAM mode is used 50 This demonstrates that the TVTBR-AOFDM modem, can reduce the number ot lost video packets, by 45i using reduced bitrate but more robust modulation modes, when the channel quality is poor However, 52 this is at the expense ot a slightly reduced average throughput bitrate Usually a higher througput bitrate 53 results in a higher video quality, however a high bitrate associated with a high packet loss ratio, is usually

45 less attractive in terms of perceived video quality than a lower bitrate, lower packet loss ratio mode

455 Having highlighted how the time-domain mode switching algonthm operates, we will now characteπse

456 its performance for a range of different BER switching thresholds A low BER switching threshold 57 implies that the switching algonthm is cautious about switching to the higher bitrate modes, and therefore

458 the system performance is charactensed by a low video packet loss ratio, and a low throughput bitrate

459 A high BER switching threshold results in the switching algonthm attempting to use the highest bitrate

460 modes in all but the worst channel conditions This results in a higher video packet loss ratio However, 6, if the packet loss ratio is not excessively high, a higher video throughput is achieved

462 Figure 14 portrays the video packet loss ratio or FER performance of the TVTBR-AOFDM modem for

463 a vanety of BER thresholds, compared to the minimum and maximum rate un-switched modes It can 64 be seen that for a conservative BER switching threshold of 0 1 %> the time-vaπent target bitrate subband

465 adaptive (TVTBR-AOFDM) modem has a similar packet loss ratio performance to that of the 1 8Mbps

466 non-switched or constant target bitrate (CTBR) subband adaptive modem However, as we will show,

467 the throughput of the switched modem is always better or equal to that of the un-switched modem, 68 and becomes far supeπor, as the channel quality improves Observe in the figure that the "agressive"

469 switching threshold of 10% has a similar packet loss ratio performance to that of the 9 8Mbps CTBR-

470 AOFDM modem We found that in order to maintain a packet loss ratio of below 5%, the BER svv itching

47, thresholds of 2 and 3% offered the best overall performance since the packet loss ratio was fairly low, 72 while the throughput bitrate was higher, than that ot an un-switched CTBR-AOFDM modem

47 A high BER switching threshold results in the switched subband adaptive modem transmitting at a high

474 average bitrate However, we have shown in Figure 14 how the packet loss ratio increases, as the BER ₅ switching threshold is increased Therefore the overall useful or effective throughput bitrate. le the

1 6 bitrate excluding lost packets can be reduced in conjunction with high BER switching thresholds

4 - Figure 15 demonstrates how the transmitted bitrate ot the switched TVTBR-AOFDM modem increases

4 ₈ with higher BER switching thresholds However, when this is compared to the effective throughput

479 bitrate. where the effects ot packet loss are taken into account, the tradeoff between the BER switching

480 threshold and the etfective bitrate is less apparent 8, Figure 16 portrays the corresponding effective throughput bitrate versus channel SNR for a range ot

482 BER switching thresholds The figure demonstrates that for a BER switching threshold of 10% the 83 etfective throughput bitrate performance was reduced incompaπson to some of the lower BER switching

484 threshold scenanos Therefore the BER=10% switching threshold is clearly too aggressive, resulting 85 in a high packet loss ratio, and a reduced effective throughput bitrate For the switching thresholds

486 considered, the BER=5% threshold achieved the highest effective throughput bitrate However, even

487 though the BER=5% switching threshold produces the highest effective throughput bitrate, this is at the

488 expense of a relatively high video packet loss ratio, which - as we will show - has a detπmental effect

489 on the perceived video quality

490 We will now demonstrate the effects associated with different BER switching thresholds on the video 49i quality represented by the peak-signai-to-noise ratio (PSNR)

492 Figures 17(a)- 17(c) portray the PSNR and packet loss performance versus time for a range of BER

493 switching thresholds 9 Figure 17(a) indicates that for a BER switching threshold of 1 % the PSNR performance is very similar to

495 the corresponding error-free video quality However, the PSNR performance diverges from the error-free

496 curve, when video packets are lost, although the highest PSNR degradation is limited to 2dB Further-

497 more, the PSNR curve typically reverts to the error-free PSNR performance curve in the next frame In

498 this example about 80% of the video frames have no video packet loss

499 When the switching threshold is increased to 2%, as shown in Figure 17(b), the video packet loss ratio

500 has increased, such that now only 41 % of video frames have no packet loss The result of the increased sol packet loss is a PSNR cur\ e, which diverges from the error-tree PSNR performance curve more regularly, 502 with PSNR degradations of upto 7dB It is worth noting that when there are video frames with no packet =03 losses, the PSNR typicallv recovers, achieving a similar PSNR performance to the error-free case When

504 the BER switching threshold was further increased to 3%, which is not shown in the figure, the maximum

505 PSNR degradation increased to 10 5dB, and the number ot video frames without packet losses was

506 reduced to 6%

507 Figure 17(c) portrays the PSNR and packet loss performance tor a BER switching threshold of 5% The 5oβ PSNR degradation in this case ranges from 1 8 to 13dB and all video frames contain at least one lost

509 video packet. Even though the BER=ό% switching threshold provides the highest effective throughput

=,o bitrate, the associated video quality is poor. The PSNR degradation in most video frames is about l OdB

5, 1 Clearly, the highest effective throughput bitrate does not guarantee the best video quality. We will now

5i2 demonstrate that the switching threshold ot BER= 1 % provides the best video quality, when using the

5,3 average PSNR as a performance metπc.

S Figure 1 8(a) compares the average PSNR versus channel SNR performance for a range of switched sis (TVTBR) and un-switched (CTBR) AOFDM modems. The figure compares the four un-switched, le.

5,6 CTBR subband adaptive modems with switching, le. TVTBR subband adaptive modems, which switch

517 between the four fixed-rate modes, depending on the BER switching threshold. The figure indicates

5.8 that the switched TVTBR subband adaptive modem having a switching threshold of BER= 10% results

5.9 in similar PSNR performance to the un-switched CTBR 9.8Mbps subband adaptive modem. When 520 the switching threshold is reduced to BER=3%, the switched TVTBR AOFDM modem outperforms

52, all of the un-switched CTBR AOFDM modems. A switching threshold of BER=5% achieves a PSNR

522 performance, which is better than the un-switched 9.8Mbps CTBR AOFDM modem, but worse than the

523 un-switched 7.0Mbps modem, at low and medium channel SNRs.

524 A comparsion of the switched TVTBR AOFDM modem employing all six switching thresholds that

525 we have used previously is shown in Figure 18(b). This figure suggests that switching thresholds of

526 BER=0.1 , 1 and 2% perform better than the BER=3% threshold, which outperformed all of the un-

527 switched CTBR subband adaptive modems. The best average PSNR performance was acheieved by 52β a switching threshold of BER=1 %. The more conservative BER=0.1 % switching threshold results in

529 a lower PSNR performance, since its throughput bitrate was significantly reduced. Therefore the best

530 tradeoff in terms of PSNR, throughput bitrate and video packet loss ratio was achieved with a switching

53, threshold of about BER=1 %.

5 2 4.7 Summary of Embodiments

533 We have outlined an adaptive modulation technique, which can be applied to OFDM systems In prac-

534 heal terms the subcarner modem mode cannot be independently chosen for each subcarner, since the

535 associated modem mode side-information would be prohibitively high. Hence we divided the subcarn-

536 ers into subband and controlled the modulation modes on a subband-by-subband basis, which resulted in

537 an acceptable side information requirement.

53β In Section 4.4 we compared the performance of subband adaptive OFDM modems to conventional

539 OFDM modems, operating at the same bitrate The subband adaptive modem could invoke BPSK. QPSK. 540 or 16Q AM modulation for each subband or disable transmission for a subband, if the channel conditions

54, were poor The subband adaptive modems could provide a lower BER. than the corresponding conven-

542 tion BPSK or QPSK OFDM modems at the same channel SNR This was acheived by transmitting more

54 bits in the higher-quality subbands, and less bits in the lower-quality subbands, thereby reducing the

544 chances of corrupted bits The low er BER of the subband adaptive OFDM modems provided a higher = 5 effective video bitrate tor the video codec w hich in turn provided a higher video quality Additionally 546 the subband adaptive modem could operate at lower channel SNRs, while maintaining the required video 54- quality

548 In Section 4 5 we compared the performance of subband adaptive OFDM modems, operating at different

549 target bitrates This showed that higher target bitrates required a higher channel quality This was further

550 exploited in Section 4 6, where we added another level of adaption, by switching between different target =5, bitrates, based on the prevailing channel conditions This enabled a time-vaπant target bitrate subband 5b2 adative OFDM (TVTBR-AOFDM) modem to provide a higher bitrate, when the overall channel quality

553 was high, and a lower bitrate when the overall channel quality was poor, in order to maintain the required

55 video quality

555 The proposed constant target bitrate subband adaptive OFDM (CTBR-AOFDM) modems can provide sse a lower BER, than a corresponding conventional OFDM modem The slightly more complex switched

55 TVTBR-AOFDM modems can provide a balanced video quality performance, across a wider range of

558 channel SNRs

=59 4.8 Conclusions

560 The proposed burst-by-burst adaptive multimedia OFDM transceiver concept exhibits substantial ad-

56, vantages in companson to conventional fixed-mode OFDM transceivers, which was substantiated in the

562 context of a specific embodiment of the advocated system concept, namely with the aid of a burst-by-

563 burst adaptive video transceiver

564 Specifically, the main advantage of the proposed burst-by-burst adaptive OFDM multimedia transceiver

565 technique is that irrespective of the prevailing channel conditions, the transceiver achieves always the best 5=6 possible source-signal representation quality - such as video, speech or audio quality - by automatically 567 adjusting the achievable bitrate and the associated multimedia source-signal representation quality in or- seβ der to match the channel quality expenenced This is achieved on a near-instantaneous basis under given

569 propagation conditions in order to cater for the effects of path-loss, fast-fading, slow-fading, dispersion,

570 etc Furthermore, when the mobile is roaming in a hostile outdoor propagation environment, typically

57, low-order low-rate modem modes are invoked, while in benign indoor environments predominantly the high-rate high source-signal representation quahtv modes are employ ed The proposed system concept has the following important features

1 A reliable near-instantaneous channel quahtv metric is employed in order to appropnatelv configure the AOFDM modem for maintaining the required target BER and the associated source signal representation quality

2 The perceived channel quality determines the number of bits that can be transmitted in a given OFDM transmission burst which in tum predetermines the number ot bits to be generated by the associated multimedia source codec, such as for example the associated video, audio or speech codec Hence the multimedia source codec has to be capable of adjusting the number of bits generated under the instruction ot the burst-bv-burst adaptive OFDM transceiver

3 The OFDM transmitter mode requested by the receiver, in order to achieve the target performance has to be signalled by the receiver to the remote transmitter Another scenario is where the uplink and downlink channel quality is sufficiently similar for allowing the receiver to judge what transmission mode the associated transmitter should use, in order for its transmitted signal to man- tain the required transmission integπty Lastly, the mode of operation used by the transmitter can also be detected using blind detection techniques, for example in conjunction with the associated channel decoder

4 In practical terms the AOFDM subcarner modem mode cannot be independently chosen tor each subcarner, since the associated modem mode side-information would be prohibitively high Hence we proposed to divide the AOFDM subcarners into subbands and to control the modulation modes on a subband-by-subband basis, which resulted in an acceptable side information requirement

5 The subband adaptive modems may provide a lower BER, than the corresponding conventional BPSK or QPSK OFDM modems at the same channel SNR This was achieved by transmitting more bits in the higher-quality subbands, and less bits in the lower-quality subbands, thereby reducing the chances of corrupted bits The lower BER of the subband adaptive OFDM modems provided a higher effective video bitrate for the video codec in the studied embodiment of the proposed system which in turn provided a higher video quahtv Additionally the subband adaptive modem could operate at lower channel SNRs while maintaining the required video quahtv

6 Higher AOFDM target bitrates required a higher channel quahtv This was further exploited in Section 4 6 where we added another level of adaption bv switching between different target b02 bitrates based on the prevailing channel conditions This enabled a time-vaπant target bitrate 03 subband adative OFDM (TVTBR-AOFDM) modem to provide a higher bitrate when the overall

604 channel quahtv was high and a lower bitrate when the overall channel quality was poor, in order o05 to maintain the required video quality

606 7 The proposed constant target bitrate subband adaptive OFDM (CTBR-AOFDM) modems can pro-

50 vide a lower BER, than a corresponding conventional OFDM modem The slightly more complex

6oβ switched TVTBR-AOFDM modems can provide a balanced video quality performance, across a

609 wider range ot channel SNRs

6,o References

o<> [ 1 ] J Streit and L Hanzo, 'Dual-mode vector-quantised low-rate cordless videophone systems for

6.2 indoors and outdoors applications," IEEE Tr on Vehicular Technology, vol 46, pp 340-357, May

6.3 1997

6. [2] P Chernman and L Hanzo, "Programmable H 263-based wireless video transceivers for

6.5 interference-limited environments," IEEE Trans on Circuits and Systems for Video Technology,

6.6 vol 8, pp 275-286, June 1998

6.7 [3] N Farber, E Steinbach, and B Girod, "Robust h 263 compatible transmission for mobile video

6.8 server access," in Proc of First International Workshop on Wireless Image/Video Communications,

6.9 (Loughborough, UK), pp 122-124, 4-5 September 1996

620 [4] A Sadka, F Eryurtlu, and A Kondoz, "Improved performance H 263 under erroneous transmission

62, conditions," Electronics Letters, vol 33, pp 122-124, Jan 16 1997

622 [5] W T Webb and R Steele, "Vanable rate QAM for mobile radio," IEEE Transactions on Commu-

623 nications, vol 43, no 7, pp 2223-2230, 1995

624 [6] H Matsuoka, S Sampei, N Moπnaga, and Y Kamio "Adaptive modulation system with vaπable

625 coding rate concatenated code for high quality multi-media communications systems, ' in Proceed-

626 tngs of IEEE VTC '96 [25], pp 487-491

527 [7] J M Torrance and L Hanzo, "On the upper bound performance of adaptive QAM in a slow o28 Rayleigh fading," IEE Electronics Letters, pp 169 - 171 , April 1996

τ> 629 [8] J. Torrance and L Hanzo. "Optimisation ot switching levels for adaptive modulation in a slow

630 Rayleigh fading channel.^" Electronics Letters, vol. 32, pp. 1 167- 1 169, 20 June 1996

6 , [9] J. M Torrance and L. Hanzo, "Demodulation level selection in adaptive modulation," Electronics

632 Letters, vol. 32, pp. 175 1 -1752, 12 September 1996.

633 [ 10] A. J. Goldsmith and S G. Chua, "Vaπable Rate Vaπable Power MQAM for Fading Channels,"

63 IEEE Transactions on Communications, vol. 45, pp. 1218 - 1230, October 1997.

635 [ 1 1 ] J M. Torrance and L. Hanzo. "Latency and networking aspects of adaptive modems over slow

636 indoors rayleigh fading channels," IEEE Tr. on Veh. Techn. , vol. 48, no. 4, pp. 1237-1251 , 1998.

637 [ 12] J. M. Torrance, L. Hanzo. and T. Keller, "Interference aspects of adaptive modems over slow

638 rayleigh fading channels," IEEE Tr. on Veh. Techn., vol. 48, pp. 1527-1545, Sept 1999

639 [ 13] V Lau and M. Macleod, "Variable rate adaptive trellis coded QAM for high bandwidth efficiency

640 applications in rayleigh fading channels," in Proceedings of IEEE Vehicular Technology Conference

64, (VTC'98) [26], pp. 348-352.

642 [ 14] C. Wong, T. Liew, and L. Hanzo, "Blind modem mode detection aided block turbo coded burst-by-

643 burst wideband adaptive modulation," in Proceeding of ACTS Mobile Communication Summit '99,

644 (Sorrento, Italy), ACTS, June 8-1 1 1999.

6 5 [15] M. Yee and L. Hanzo, "Upper-bound performance of radial basis function decision feedback

646 equalised burst-by-burst adaptive modulation," in Proceedings of ECMCS'99, (Krakow, Poland),

647 24-26 June 1999.

6 8 [ 16] T. Keller and L. Hanzo, "Adaptive orthogonal frequency division multiplexing schemes," in Pro-

649 ceeding of ACTS Mobile Communication Summit '98, (Rhodes, Greece), pp. 794—799, ACTS, 8-1 1

650 June 1998.

65, [ 17] E. L. Kuan. C. H. Wong, and L. Hanzo, "Burst-by-burst adaptive joint detection CDMA," in Pro-

652 ceeding of VTC'99 (Spring), (Houston, Texas, USA), IEEE, 16-20 May 1999.

653 [ 18] A. Czylwik. "Adaptive OFDM for wideband radio channels," in Proceeding of IEEE Global

654 Telecommunications Conference, Globecom 96 [27], pp. 713-718.

o55 [ 19] R. F. H. Fischer and J B. Huber, "A new loading algonthm for discrete multitone transmission," in ese Proceeding of IEEE Global Telecommunications Conference, Globecom 96 [27], pp. 713-71 8. cs? [20] P S Chow J M Cioffi, and J A C Bingham, A practical discrete multitone transceiver loading

658 algonthm for data transmission over spectrally shaped channels," IEEE Trans on Communications.

659 vol 48. pp 772-775, 1995

660 [21 ] H Roh ng and R Grunheid, 'Peformance of an OFDM-TDMA mobile communication system," 66i in Proceeding of IEEE Global Telecommunications Conference, Globecom 96 [27], pp 1589-1593

662 [22] K Fazel, S Kaiser, P Robertson, and M J Ruf, "A concept of digital terrestnal television broad-

663 casting," Wireless Personal Communications, vol 2, pp 9-27, 1995

664 [23] H San, G Karam, and I Jeanclaude, "Transmission techniques for digital terrestnal tv broadcast- ess ing," IEEE Communications Magazine, pp 100-109, February 1995

eeβ [24] I Kalet, "The multitone channel," IEEE Tran on Comms, vo\ 37, pp 1 19-124, February 1989

667 [25] IEEE, Proceedings of IEEE VTC '96, (Atlanta, GA, USA), 1996

eeβ [26] IEEE, Proceedings of IEEE Vehicular Technology Conference (VTC'98), (Ottawa, Canada), May

669 1998

670 [27] IEEE, Proceeding of IEEE Global Telecommunications Conference, Globecom 96, (London, UK), 6 , 18-22 Nov 1996

Claims

1. A transmitter for transmission of a multimedia source signal over a transmission medium to a remote receiver, the transmitter comprising: an AOFDM modem having an output for transmitting a multimedia source signal; a source codec arranged to supply the multimedia source signal to the modem; and an input for receiving a metric of channel quality indicative of current transmission integrity; wherein the modem and/or the source codec are reconfigurable responsive to the channel quality in order to maintain a required multimedia source signal integrity at a remote receiver receiving the multimedia source signal transmitted by the modem.

2. A transmitter according to claim 1, wherein the required multimedia source signal integrity is defined in terms of a target bit error rate (BER) and/or a target AOFDM symbol error rate (SER) at the remote receiver.

3. A transmitter according to claim 2, further comprising a channel quality estimator for determining BER and/or SER estimates for each of a plurality of transmission modes of the modem, the modem being operable according to a switching scheme whereby the transmission mode is chosen based upon the BER or SER estimates for the transmission modes.

4. A transmitter according to claim 3. wherein the transmission mode is chosen to be the transmission mode with the highest bit rate that has an estimate of the target BER or target AOFDM SER below a threshold.

5. A transmitter according to claim 4. wherein the threshold target BER or SER is variable according to prevailing channel conditions. 16-

6. A transmitter according to any one of claims 3 to 5, wherein the AOFDM modem is operable to transmit using a plurality of subcarrier subbands. the transmission mode being independently selectable for the different subbands. whereby higher transmission rates are achieved in higher-quality subbands and lower transmission rates in lower-quality subbands.

7. A transmitter according to any one of claims 2 to 6, wherein the target BER is set to limit the AOFDM SER to a maximum.

8. A transmitter according to any one of the preceding claims, wherein the required multimedia source signal integrity is defined in terms of an AOFDM SER at the remote receiver.

9. A transmitter according to any one of the preceding claims, wherein the modem is reconfigurable in use by varying the number of bits per AOFDM symbol responsive to the channel quality.

10. A transmitter according to any one of the preceding claims, wherein the channel quality metric is based on the current BER or AOFDM SER detected at the remote receiver and transmitted back to the transmitter.

1 1. A transmitter according to any one of claims 1 to 10, wherein the channel quality metric is based on the current transmission BER or AOFDM SER detected at a receiver local to the transmitter sharing the transmitter's transmission medium.

12. A transmission system for transmission of multimedia source signals over a transmission medium, the system comprising: a first transceiver including a local receiver and a local transmitter according to any one of the preceding claims; and a second transceiver including a remote receiver and a remote transmitter according to any one of the preceding claims.

13. A method of transmitting a multimedia source signal, the method comprising: providing a transmitter comprising a source encoder and AOFDM modulator; generating a multimedia source signal in the source encoder; supplying the multimedia source signal to the AOFDM modulator: transmitting the multimedia source signal from the AOFDM modulator over a transmission medium to a remote receiver: obtaining a channel quality metric indicative of channel quality experienced by the receiver; and controlling the source encoder and/or the AOFDM modulator responsive to the channel quality metric so that the integrity of the signal received by the receiver meets a desired integrity target.

14. A method according to claim 13, wherein the desired integrity target is defined in terms of a bit error rate (BER) and an AOFDM symbol error rate (SER).

15. A method according to claim 13 or 14, wherein the modem is switched between a plurality of transmission modes according to an estimate of the expected BER or AOFDM SER for the individual transmission modes obtained on the basis of the estimated channel quality, thereby selecting the transmission mode having the highest transmission rate that complies with the desired integrity target.

16. A method according to claim 15, wherein the AOFDM modulator transmits using a plurality of subcarrier subbands. the transmission modes being independently selected for the individual subbands.

17. A method according to any one of claims 13 to 16, wherein the source encoder is reconfigured during transmission according to the number of bits per AOFDM symbol to be generated responsive to the channel quality metric.

18. A method according to any one of claims 13 to 17. wherein the channel quality metric is estimated from the signal received at the receiver and transmitted

.8-

back to the transmitter via a feedback path to set the modulation mode of the transmitter to meet the desired integrity target at the receiver.

19. A method according to any one of claims 13 to 17. wherein the channel quality metric is estimated from signals transmitted from a remote transmitter over the transmission medium to a local receiver.

20. A method according to any one of claims 13 to 19, wherein the channel quality predetermines the source-representation quality of the multimedia source signal received by the receiver under error- free channel conditions.