- Digital multimedia transmission is one of the most significant technological challenges of our time. - In this chapter we will face the problem of the dynamic optimization of Variable Bit Rate (VBR) video transmission in wireless environments. - In the UMTS scenario data transmission will be varied by taking into account the user interactivity. - Since the Terrestrial DVB (DVB-T) infrastructure is able to serve both fixed and mobile terminals, DVB-T and DVB-H systems share almost the same common physical layer of the ISO-OSI protocol stack. - nevertheless, DVB-H introduces additional features in the Physical Layer, Data Link Layer and Network Layer of the stack (ISO/IEC 7498-1), as represented in Fig. - This system is designed for VHF and UHF frequency bands, in the same spectrum assigned to analogue TV.. - In the SFNs all transmitters carry the signal in the same frequency range, and in small SFNs there is a reduced distance among transmitters. - different MPEG streams are multiplexed in the modulated signal: a High Priority (HP) stream and a Low Priority (LP) stream.. - In the Outer Coder a Error Correction Code is applied. - The so obtained data are then interleaved in the Outer Interleaver. - a bit coding in the Inner Coder allows a further protection against transmission errors in the aerial channel. - Service Bursts are interspaced by “off-times”, i.e., time intervals between two consecutive bursts of the same service, where no service data are decoded but data of other multiplexed services are transmitted. - The energy saving is achieved because receiver remains active only for a fraction of time, during the reception of the chosen service, and switches off during that service off-times. - The Burst Bitrate (BB) is the constant bit rate of the time-sliced stream during the burst transmission. - In the sequel, we will define the Burst Cycle (BC) as the time interval between the beginning of two burst of the same service. - The burst size is in fact buffered in the client memory during burst durations, and consumed during the subsequent Off-times. - The Core Network (CN) is the part of the UMTS network that provides services to final users. - The streaming server providing multimedia services can reside either in the external packet data network, or just before the GGSN.. - All radio resource management functionality is pooled in the Common Radio Resource Manager (CRRM). - All processing related to the radio interface of the RNC is relocated into the Node B+, whose main functions are, among others: macro diversity control, handover related functions, power control and frame scheduling. - A very good way for scheduling optimization at transmission side is that the streaming server has knowledge of the “status”. - of the network and the client terminal. - RTCP main role is to provide a feedback information on the quality of the stream delivery.. - RTCP can be a solution for the wireless multimedia streaming over IP networks, since it could help improving the quality of the delivered data, that depends on the variable conditions of the wireless links and the user equipments limited amount of memory. - HTSN (Highest Trasmitted Sequence Number), that is the sequence number of the last packet sent by the server;. - HRSN (Highest Received Sequence Number), that is the sequence number of the last packet arrived at the client buffer;. - NSN (Next Sequence Number), that is the sequence number of the next packet to be decoded by the client.. - SSRC, that is the source identifier of the stream this packet belongs to;. - NSN (Next Sequence Number), the sequence number of the next packet to be decoded;. - Through a NADU packet a high number of information can be derived, such as the number of packets that reached the client, the number of packets stored in the client buffer and decoded by the client. - If the streaming server knows the size of each transmitted packet, the HRSN and NSN parameters provide the occupancy level of the client buffer. - In this section we only give a brief overview of the MPEG (further distinguished in the MPEG-1, MPEG-2 and MPEG-4) and H.264/AVC compression techniques, actually the most widely adopted standards.. - MPEG video compression is applied to temporally consecutive images (the video frames) of a multimedia stream, exploiting the concept of the similarities between temporally adjacent frames. - Let us note that MPEG sometimes uses information from future pictures in the sequence. - so the order in which compressed pictures are found in the bitstream, called “coding order”, is not the same as the “display order”, the order in which pictures are presented to a viewer.. - Motion Estimation is simply the process of determining the motion vectors in the encoder, to describe direction and the amount of motion of the macroblocks (Mitchell et al., 1996).. - The H.264/AVC standard (ISO/IEC JTC 1, 2003) is the new state of the art of video coding.. - Since the standard is the result of collaborative effort of the VCEG and MPEG standards committees, it is informally referred to as Joint Video Team (JVT) standard as well. - SP slices aim at efficient switching between different versions of the same video sequence whereas SI slices aim at random access and error recovery.. - The accuracy of Motion Compensation is in units of one quarter of the distance between luminance samples. - Prediction values are obtained by applying specific filters horizontally and vertically in the image. - Multiple reference pictures not only contribute to the improvement of the compression efficiency, but also to error recovery.. - Nevertheless, this assignment surely would bring to a bandwidth waste for almost all the video duration, because of the high video bit rate variability (Zhang et al., 1997).. - Let us now consider the client buffer model in the k th discrete frame time. - represents the smoothed stream bit rate in the i th frame time. - Bandwidth dependent smoothing considers the additional available bandwidth information for improving the efficiency of the stream schedule.. - The basic smoothing principles previously mentioned are the starting point for the implementation of the novel transmission techniques proposed in this chapter. - 4.1 Optimal scheduling of multiservice VBR video transmission in DVB-H systems As stated in Section 2.1.1, service data are stored in bursts at transmission side and buffered in the client memory during burst times for continuous playing also during the subsequent off-times. - We conventionally say that a burst “covers” the off-time if the burst size guarantees the continuous playback on the DVB-H terminal. - In fact the burst size must always be less than the memory available in the receiver and the burst bitrate and burst duration can be set accordingly. - In these cases, since the video bit rate is highly variable in time, data stored in the client buffer can heavily vary during video reproduction. - Statically setting the burst size for the whole video transmission could easily bring to losses at receiving side, because of the insufficient amount of buffered data and/or because of the relatively small receiving buffer size.. - Transmission optimization is performed by dynamically and simultaneously varying the burst durations of the whole set of services transmitted in a temporal window of fixed size, sliding in time. - The goal is the loss minimization of the whole service set. - In the specific DVB-H scenario, a service buffer underflow occurs if the burst size is not enough to cover off-time. - This is the amount of data stored in the client buffer during the previous burst duration and not yet consumed by client at the end of the previous burst cycle. - The first N B burst cycles in the previous step are transmitted and N B new burst cycles are introduced in the optimization process of the following step. - In fact, the service data to be scheduled in the last N B new burst cycles included in the. - 11 refine also the calculation of the service burst durations of the previous W S − N B burst cycles already calculated in the n th step, even if the computational overhead increases. - i j is the i th service burst duration in the j th burst cycle. - i j is a positive integer multiple of the frame time unit. - T = T T that minimizes all service losses a Total Loss Function (TLF) is introduced, that considers all the service losses for buffer underflow in the TOW. - i j at the end of the j th burst cycle (see Fig. - i j the i th service burst duration in the j th burst cycle. - Cumulative data leaving the buffer at the end of the burst cycle are instead:. - Let us note that the N S burst durations of the last burst cycle in a TOW are only used to evaluate the N S streams off-times for losses evaluation in the ( W S − 1 ) th burst cycle.. - i j of the i th service off-time in the. - and depends on the burst durations of the other services.. - To guarantee fairness among services, the TLF normalizes each service losses to the amount of service data transmitted in the TOW. - is the available bandwidth vector of the N S services in a TOW.. - Video streams Jurassic Park Video Clip Star Wars IV The Silence of the Lambs Compression ratio. - The first proposed experiment shows the influence of the TOW length in VBT losses calculation for three different N B . - Let us note that CBT losses found in the best scheduling conditions are approximately an order of magnitude higher than the maximum amount of VBT losses (observed for W S = 4 and N B = 3. - 13 shows the impact of the available bandwidth over losses for CBT and VBT. - 14 shows the impact of the receiving buffer size over losses for CBT and VBT schedules. - VBT losses have been evaluated for two different available bandwidth values (3 and 4 Mbps), while the available bandwidth for CBT has been set to 5 Mbps to exclude the influence of the bandwidth limitation in loss calculation.. - UMTS systems should guarantee lossless data delivery despite of the highly variable bandwidth conditions of wireless channel and the high fluctuating data bit rates. - In this work we focus on the effects of the user interactivity that modifies the status of the mobile terminal, ignoring the influence of other aspects like the fluctuating channel bandwidth.. - In each step, the first α frames scheduled in the previous step are sent to the client, while α new frames are scheduled together with the remaining N − α frames already scheduled in the previous step. - In the on-line smoothing algorithms with a statically assigned α analyzed in (Rexford et al., 2000), it has been experimented that α = N / 2 is a good compromise between an optimized schedule and a reduced computational overhead. - In this analysis, we suppose that there is enough available bandwidth in the UMTS network, so that delays in data transmission and control information across the network can be assumed almost null. - In fact in the generic k th frame time it holds:. - Let us now suppose that a NADU packet containing the FBS information arrives to the server in the k th frame time. - The frequency of the feedback information must be increased as a function of the. - Nevertheless, whenever the client buffer underflows because of a fast forward action, when the server performs a rescheduling in k it knows exactly the number of the last decoded packet HTSN k. - In the case of a buffer overflow instead losses surely will occur because the buffer is full and the server continuously sends data. - This suggests the implementation of the relationship ( α Δ B. - The maximum slide length has been chosen α = N / 2 since, as previously mentioned, it is the best trade-off between the optimality of the schedule and the algorithm computational overhead. - When a RTCP packet comes to the server in k , the server compares FBS k C. - B ) k α of the RTCP packet through (21), and sends this information back to the client through specific RTCP packets (Schulzrinne et al., 2003), supposed to be immediately available to client;. - calculated in the step 2 as the first frame to be decoded. - Comparison between S-BDSA and D-BDSA has been made by simulating the transmission of 70.000 video frames of the “Jurassic Park” video, MPEG-4 coded with high quality, with a sequence of the simulated user external actions summarized in Table 2.. - It shows the influence of the TOW length N on losses for D-BDSA and S-BDSA. - D-BDSA losses are always smaller than S-BDSA ones, thanks to the D-BDSA dynamic change of the slide length. - Losses have been calculated by choosing the same piece of video stream used in the previous simulation and the same sequence of user actions listed in Table 2, with N=600 frame times. - Sequence of the user actions on the client terminal. - Further work in this direction can be done in the improvement of the optimization method that finds the minimum of a nonlinear function of several variables. - Simulation have been performed over different types of video streams, TOW lengths and buffer sizes, testifying the effectiveness of the proposed method.. - Performance results are strongly influenced by the sequence of the user actions and especially by the slide length values calculated by the server. - The real status of the UMTS core network, together with its buffers, could also be modeled to analyze the network behavior towards losses.