专利摘要:
method for scheduling transmission of data streams from a node (bs) to a receiver (ue2) on a communication network, communication node, and computer program product. data streams are transmitted from a node (bs) to a receiver on a communication network in the form of data packets for representation by a reproduction buffer in the receiver. the data packets are arranged in a programming queue (sq) and deleted from the programming queue (sq) if their dwell time (t) in the queue (sq) exceeds a given fall deadline (d). the reproduction buffer (b) is emulated on the node (bs) in order to determine respective representation values (po) for the data packets that are indicative of expected representation times (p) for the data packets by the buffer the receiver. the drop deadlines (d) are assigned to the data packets as a function of the respective representation values (po) determined by the reproduction buffer as emulated on the node (bs).
公开号:BR112012015452B1
申请号:R112012015452-4
申请日:2009-12-24
公开日:2020-12-15
发明作者:Dario Sabella;Marco Caretti;Matteo Maria Andreozzi;Giovanni Stea
申请人:Telecom Italia S.P.A;
IPC主号:
专利说明:

Field of the Invention
[001] This exhibition relates to techniques for programming data transmission in communication networks.
[002] This exhibition was developed with attention paid to its possible use in programming data packets from a Voice data flow over IP (VoIP) in a point-to-multipoint communication node. Description of the Related Art
[003] The Internet was originally designed with secure packet transfer in mind. In reality, all the lower layers of the protocol stack include mechanisms to ensure secure communication, such as sequence numbers, intervals, checksums (such as Cyclic Redundancy Check), Front Error Correction (FEC), Automatic Retry Request (ARQ), etc. This is because the Internet was initially conceived as a computer data network, where accuracy and integrity of computer data are fundamental attributes.
[004] Today, the Internet is well established as a global multiservice infrastructure, providing ever-increasing support for real-time services. Packetized voice, such as Voice over IP (VoIP), is among the most important emerging real-time applications for the Internet, acting as a primary motivator for the development of advanced Quality of Service (QoS) solutions.
[005] Despite the massive presence of real-time traffic, for which punctuality is as important as integrity, few Quality of Service (QoS) solutions actually use time as a parameter to make decisions, except at an application level. This is probably due to the historically well-known problem of clock synchronization over the Internet, which makes it difficult to have clocks globally synchronized to two remote points.
[006] The lack of reliable clock synchronization is often a problem for performance measurement, where long measurement intervals (days to weeks) and high accuracy are required (thus making even small clock deviations extremely harmful).
[007] Progress in clock synchronization is currently being made, which enables achieving sub-millisecond accuracy in such a long term. However, voice conversations last for minutes to a few hours, and synchronization errors in the order of a few milliseconds are unlikely to have any noticeable effect on human perception.
[008] The precision required to keep watches synchronized for a couple of hours with up to a few milliseconds of absolute difference is already achievable at little cost with available techniques, such as chips of Global Position System (GPS), which are cheap and already assembled on most portable devices, the same result is also achieved by proprietary solutions available on the Universal Mobile Telecommunications System (UMTS) network (where much tighter synchronization is really a major requirement) or the famous Network Time Protocol.
[009] Figure 1 schematically represents a typical VoIP scenario in which two users U1 and U2 communicate through two user terminals UE1 and UE2 that are connected to a common network N (for example, the Internet). On the sender side (i.e., UE1), the first component is an encoder 100, which periodically samples the voice signal from the first user U1 (i.e., the speaker). A wide variety of encoders are available, representing different points of compromise in balancing bandwidth consumption and quality. Typically, encoders are either sample-based or frame-based. Previous encoders (for example, based on the G.711 codec ) encode individual speech samples periodically, while later encoders (for example, based on the G.729 codec) group a number of samples within a time window (ie, a frame) of a few milliseconds. For this reason, frame-based encoders often achieve higher compression rates and lower data rates, however their encoding / decoding complexity is usually higher.
[0010] The generation of speech frames can either occur at periodic intervals, or, more generally, be modulated by Voice Activity Detection (VAD). VAD capitalizes on the natural alternation of voice bursts and periods of silence in a single (unidirectional) flow of a bidirectional conversation. During periods of silence, either no speech frame is produced at all, or they are produced at a reduced rate and / or using a reduced number of bits, to bring some comfort noise to the listener. Comfort noise allows one conversation party to feel the liveliness of the other, and can also be generated locally. For example, the G.711 codec normally generates 172-byte packets during voice bursts and does not send information during periods of silence. On the other hand, the Global System for Mobile Communications (GSM) Adaptive Multi-Rate (AMR) codec typically generates 31-byte payload packets during voice surges, and either 0-byte synchronization information or noise packets of 5-byte comfort during periods of silence.
[0011] Several speech frames can also be multiplexed in the same packet payload through a 102 multiplexer, to reduce the cost of transport, network headers and Media Access Control (MAC), however at the expense of increasing the transmission delay. This can also be done to increase the level of redundancy, to hide losses. The level of speech multiplexing may also vary over time, depending on network conditions.
[0012] Finally, the VoIP payload is encapsulated in data packets, such as RTP / UDP / IP packets, that is, packets being communicated at the application layer by the Real Time Transport Protocol (RTP), at the layer transport by the User Datagram Protocol (UDP), and at the Internet layer by the Internet Protocol (IP).
[0013] On the receiver side (i.e., at the UE2 terminal), speech frames are demultiplexed by means of a demultiplexer 104 and fed into a representation buffer 106. The representation buffer 106 causes speech frames to be decoded at the same time. same interval at which they were generated by the encoder. For this reason, it may be necessary to reorder, delay or even suppress speech frames if they arrive after their expected reproduction time.
[0014] Temporary representation memories can be fixed or adaptable. Fixed temporary memories assume the network delay is constant during a conversation, and therefore delay the first packet of a fixed-duration voice surge. Adaptive temporary memories, strive to dynamically adapt the reproduction point to changing network conditions, usually on a per-voice basis (although some may also vary the representation delay within a voice outbreak, at the risk of generating effects of spurious frequency).
[0015] The representation buffer delivers speech frames to a decoder 108, which actually reproduces them for the U2 user (ie, the listener). Some decoders may implement Packet Loss Masking (PLC) techniques, which allow lost speech frames to be reconstructed somewhat by interpolating surrounding frames that were received correctly. While PLC techniques can mask only a limited number of losses, they effectively reduce the loss due to the loss as perceived by a listener.
[0016] The colloquial quality assessment of VoIP was the subject of extensive investigation. The most widely used evaluation structure for VoIP is the so-called model E, standardized by the Telecommunication Standardization Sector (ITU-T) of the International Telecommunication Union (ITU). This evaluation structure involves computing a predictive estimate of the subjective quality of voice in packages of transmission parameters.
[0017] The output of a model E computation is a scalar number, called the “R factor”, computed as a function of delays, packet loss, equipment deterioration factors, and user quality call expectation: R = Ro - Is - Id - Ie, eff + A (1) where R0 is the signal to basic noise ratio (level of speech received relative to circuit and acoustic noise), Is accounts for the deteriorations that occur with the voice, Id adds all deteriorations due to delay and echo effects, ie, eff is the effective equipment deterioration factor, taking into account the codec and its tolerance for random packet losses.
[0018] In addition, A is a “bonus” factor that models the user's expectation of the technology used. For example, the value of A is higher in satellite networks than in classic circuit-switched networks, because user expectations in satellite networks are lower than those in wired networks. The typical range for factor A is [0, 20] and exemplary values as proposed by the ITU are reported in the following table:

[0019] Finally, Ie, eff can be computed as:
where Ie is an equipment deterioration factor, which is used to characterize codec behaviors with a low bit rate, Ppl is the packet loss probability, and Bpl is the codec packet loss robustness factor.
[0020] Once an R factor has been obtained, it can be mapped directly to an estimated Average Opinion Score (MOS). For example, the ITU-T G107 specification (Recommendation ITU-T G.107 (042009), “The E-model, a computational model for use in transmission planning”) describes a possible relationship between the R factor and the estimated MOS. The Average Opinion Count can thus be used to improve Quality of Experience (QoE) or QoS solutions. Usually such optimizations are performed at the application level, where QoE can be measured.
[0021] For example, the article by L. Atzori, M. L. Lobina, and M. Corona, “Playout buffering of speech packets based on a quality maximization approach”, ‘IEEE Transactions on Multimedia’, Vol. 8 no. 2, pages 420-426, 2006, proposes the use of an optimal representation buffer at the receiver in order to predict the representation delay for an incoming voice surge. Specifically, the instant of representation for the next voice surge is fixed to what would have maximized the quality of the last N voice attacks.
[0022] Also, the article by A. Bacioccola, C., Cicconetti, and G. Stea, “User level performance evaluation of VoIP using ns-2”, 'Proceedings of NSTOOLS'07', First International Seminar on Simulation Tools de Rede, Nantes (FR), October 22, 2007, describes a possible optimal reproduction algorithm.
[0023] Z. Qiao, L. Sun, N. Heilemann, and E. Ifeachor, “A new method for VoIP quality of service control use combined adaptive sender rate and priority marking”, IEEE ICC 2004, Paris, France, 20- June 24, pages 1473 - 1477, proposes a QoS control scheme to adapt the rate of a GSM AMR codec to adapt the sending rate to current network conditions. The adaptation is based on the MOS measured by the VoIP decoder, which reports feedback information to the VoIP encoder via Real-Time Control Protocol (RTCP) messages. Purpose and Summary of the Invention
[0024] The inventors noted that optimizing QoE at an application level based on feedback information has several disadvantages. For example, such feedback-based techniques make the application unduly complex, and either require synchronization or rely on round-trip times that are not symmetrical and are therefore much less accurate than one-way delays.
[0025] The need is therefore felt for improved solutions that can dispense with such disadvantages, and can allow effective allocation of network resources, for example in programming downlink voice flows in a centralized point-to-multipoint network (by wires or wireless).
[0026] In accordance with the present invention, this objective is achieved by means of a method having the characteristics published in the following claims. The invention also relates to a corresponding communication node, as well as a downloadable computer program product in the memory of at least one computer and including portions of software code to perform the steps of the method of the invention when the product is run on a computer. As used herein, reference to such a computer program product is intended to be equivalent to reference to a computer-readable medium containing instructions for controlling a computer system to coordinate the performance of the method of the invention. Reference to “at least one computer” is intended to highlight the possibility for the present invention to be implemented in a distributed / modular mode.
[0027] The claims are an integral part of the display of the invention provided here.
[0028] Several embodiments described here can be used to program the transmission of data streams from a communication node (for example, a base station) to a receiver (for example, a mobile terminal), where the data packets are intended for representation by a temporary playback memory on the receiver.
[0029] In various embodiments, data packets are arranged at the node in a programming queue and deleted from the programming queue if their short stay in the queue exceeds a given fall deadline.
[0030] In various embodiments, the receiver's reproduction buffer is emulated at the node in order to determine respective expected representation times for the data packets. Subsequently, the data packet drop deadlines are assigned to the data packets as a function of the respective expected representation times determined by the reproduction buffer as emulated by the node.
[0031] Several embodiments are applicable to various access technologies, such as cellular networks including Long Term Evolution of UMTS (LTE) or High Speed Downlink Packet Access (HSDPA).
[0032] Several embodiments rely on the recognition that in these cellular network technologies, absolute time is normally known for out-of-band network synchronization procedures.
[0033] Several embodiments are also applicable in broadband wireless networks with centralized control at the MAC level, such as 802.11e Hybrid Coordination Function (HCF) Centralized Control Access (HCCA), 802.16, or TDMA systems ( for example, 802.11 WLANs with twisted TDMA drivers), when an absolute time reference is added.
[0034] Several embodiments are also applicable to wired networks, where a central key sends data packets to customers.
[0035] Several achievements rely on the recognition that the Average Opinion Score (MOS) can be used directly at a communication network level.
[0036] Several embodiments are applied to downlink flows, which rely on a prediction of the behavior (loss and delay) of the segment downstream from the end-to-end path (including the application part).
[0037] Several embodiments involve making programming decisions in a cross-layered approach, based on (absolute) timestamps, for example in the form of RTP timestamps, which was introduced in the data packets at the data source. Such absolute time stamps can be derived with various means, including GPS synchronization, UMTS network synchronization, network protocols and / or time servers.
[0038] Several embodiments exploit the absolute time to program downlink flows in a point to multipoint network (for example, an LTE or UMTS cell, or a TDMA network), that is, a communication network in which the Network resources are shared between a plurality of users or terminals.
[0039] The exemplary embodiments described here mainly refer to downlink voice flows, where: there is a well-established way to estimate the (subjective) quality perceived by the user of end-to-end network level measurements (objective), such as loss and delay, that is, model E or Average Opinion Counting (MOS); and the behavior of a voice application on the receiver side (specifically, its temporary representation storage mechanism) is quite predictable, so that it is possible to deduce the (not negligible) contribution of the voice receiver to loss and delay.
[0040] Those of skill in the art will appreciate that various embodiments can also be applied to other traffic streams in real time, such as video streams. For this purpose, a structure similar to the E model for video tracks can be developed.
[0041] In several embodiments, each package is assigned a deadline equal to the presumed or expected reproduction time; in doing so, recent packets, that is, those that would be deleted at the receiver, can be deleted directly at the programmer, thus saving bandwidth and optimizing the allocation of network resources.
[0042] In various embodiments, the programmer can also delay packages early to their current playing point without affecting user perception, thus increasing the chance that packages with tighter deadlines can be delivered to the receiver in time.
[0043] In several embodiments, the reproduction point on the receiver is estimated by emulating the connection to a simplified receiver and / or temporary representation memory ("temporary reproduction memory").
[0044] For example, in the case of a VoIP stream, a simplified adaptive representation buffer can also be emulated, that is, a buffer whose reproduction point is dynamically varied for each voice surge. For example, at the end of a voice surge, computations allow us to deduce the optimal reproduction moment later, that is, what would have guaranteed the highest MOS for that voice surge. The history of these optimal breeding moments can then be used to deduce the new assumed breeding time for the subsequent burst of voice.
[0045] In various embodiments in order to compute the optimum moments of posterior reproduction, the formulas of model E are used. These take into account mouth-to-ear delay, which can be computed if the package generation and reproduction times ( presumed) can be determined based on clock synchronization. For this reason, packages can be timed in source applications with absolute times sampled from an accurate clock, and the programmer is aware of the absolute time when packages are delivered to the receiver, again reading an accurate absolute clock. In various embodiments, the RTP protocol, in particular the RTP time stamp field, is used for this purpose: for example, the RTP header can be read twice per packet, that is, on packet arrival, when sequence numbers and packet time marks can be used to compute the packet deadline, and at packet start, when the programmer can estimate the mouth-to-ear delay by comparing the clock time (that is, the absolute time when the package leaves the programmer's queue) at the package time stamp.
[0046] The embodiments here have several advantages, such as: they are inexpensive, because the spatial cost of the reproduction buffer emulation is quite accessible, as the entire implementation structure normally requires less than 100 bytes per stream; time cost is negligible, because normally a constant number of operations is required for each flow; the most complex computation (that is, computing the optimal reproduction instant a posteriori) typically involves only a few memory accesses, and can be completed during periods of silence, that is, several hundred milliseconds; the resulting arrangement is robust: even very gross clock errors (in the order of ± 10ms) are tolerated without any appreciable quality degradation, so that temporary loss of synchronization (even for a long time) hardly affects the programming process ; the resulting arrangement can work together with any deadline-based programmer, such as the famous Early Due Date (EDD), or for wireless networks with user-dependent and time-dependent channel conditions with a programmer that takes into account both deadlines and channel status; the structure can be applied to various communication technologies, including i communication scenarios where the following conditions are met: a) the package source and the programming node are synchronized (for example, by GPS), b) each package contains one time stamp, c) the programming component can read the package time marks on arrival and the absolute time on departure, and d) the downstream segment only includes a representation buffer plus additional delays that can be measured online (for example, delay due to H-ARQ retransmissions) and / or estimated (for example, propagation and processing delays).
[0047] Therefore, several embodiments can be implemented in any access network with a point-to-multipoint programming coordinated under the previous assumptions. Brief Description of Attached Representations
[0048] The invention will now be described, by way of example only, with reference to the included representations, in which: Figure 1 has already been described in the foregoing; Figure 2 shows an exemplary communication scenario; Figure 3 is a block diagram of an embodiment; and Figure 4 is an operation timing diagram of an embodiment. Detailed Description of Achievements
[0049] In the following description, numerous specific details are given to provide a complete understanding of embodiments. The embodiments can be practiced without one or more of the specific details, or with other methods, components, materials, etc. In other examples, well-known structures, materials, or operations are not shown or described in detail to avoid obscuring aspects of the embodiments.
[0050] Reference throughout this specification to "an embodiment" or "the embodiment" means that a particular aspect, structure, or feature described with respect to the embodiment is included in at least one embodiment. Thus, the appearances of the phrases "in one embodiment" or "in the embodiment" in various places throughout this specification are not necessarily all referring to the same embodiment. In addition, the particular aspects, structures, or characteristics can be combined in any satisfactory manner into one or more embodiments.
[0051] The titles provided here are for convenience only and do not interpret the extent or meaning of the embodiments.
[0052] Figure 2 shows an exemplary embodiment, in which a voice flow organized in data packets is transmitted between two mobile LTE devices, UE1 and UE2, through the RTP protocol.
[0053] In the considered embodiment, the first mobile device UE1 sends data to a base station BS (i.e., the eNodeB in an LTE network) with which the mobile device UE2 is associated by a communication network N.
[0054] Each data packet can include n voice frames and each packet can be transmitted by m Protocol Data Units (PDU). In addition, the value of m may be different for different connections, and may change over time, for example when the feedback from the H-ARQ process indicates that the number of retransmissions has increased / decreased significantly. The value of m can also be fixed and pre-configured at installation.
[0055] Some simplified assumptions will be made in principle to facilitate understanding of the description; furthermore, reading the description will make it clear that such assumptions are in no way mandatory. Specifically, it will be assumed that each RTP packet includes only one voice frame, and that voice packets are not fragmented in the eNodeB, so that a PDU takes a voice frame.
[0056] In the considered embodiment, packages are marked in time with the frame generation time at creation, where gi, k is the generation time of the k-th packet in the i-th speech surge, with i, k> 0, and ai, k is the arrival time at eNóB.
[0057] Packets can accumulate a variable delay on the N network, and some packets can also be suppressed and / or delivered out of sequence.
[0058] Figure 3 shows a block diagram of a possible implementation of the BS base station according to the invention.
[0059] In the considered embodiment, when packets arrive at the base station BS at a time ai, k, they are assigned a deadline di, k to a control module 200.
[0060] The packages, leaving control module 200 with the deadline named di, k are then programmed by a programmer S, who can take into account their deadlines so that programmer S can suppress packages if their length of stay in the eNodeB queue it exceeds its deadline di, k, because these packets are likely to be deleted in the receiver's temporary memory, therefore transmitting them would just waste radio resources.
[0061] In the considered embodiment, the programmer S includes a plurality of SQ queues, where the data packets of the respective data streams are stored. For example, the SQ queues can be implemented through First In First Out (FIFO) memories ). Subsequently, an SM programming module decides which packet will be transmitted or dropped. For example, in the considered embodiment, a packet will be suppressed if the scheduled transmission time of the ti, k packet exceeds the packet's di, k deadline, i.e., ti, k> di, k.
[0062] In several embodiments, a package selected for transmission by the programmer S in time ti, k is taken care of through H-ARQ processes, and, after a measurable delay d1i, k, is available for representation on the mobile device UE2.
[0063] In the considered embodiment, the delay d1i, k includes possible H-ARQ retransmissions of that packet and an estimate of the physical delays at the air interface and at the mobile node, which are normally substantially constant.
[0064] In several embodiments, the estimated delay d1 is fixed as follows:
where Tproc represents an estimate of the processing delay at the physical and MAC layers at the receiver, TOW is the time it takes to send a PDU over the downlink segment (that is, the time it takes an H-ARQ process to deliver the packet to the receiver, excluding the generation of ACK / NACK and information), TH-ARQ is the time for an H-ARQ transmission cycle and is the number of transmissions taken into account for a packet.
[0065] In addition to the delay d1i, k, the only additional delay that the packet suffers before the voice sample is represented in time pi, k is the representation buffer in the receiver.
[0066] Therefore, the arrival time qi, k in the receiver representation buffer is:
and the global network delay of frame i can be defined as:

[0067] In several embodiments, the sequence of δi, k is fed as an entry to a buffer of emulated representation B that allows to determine the deadline di, k. Specifically, the emulated optimum representation buffer B emulates the representation buffer of the UE2 mobile device, and its purpose is to identify a posteriori (that is, once the voice surge i ended) what the optimal representation moment should be that is, what would have guaranteed the highest MOS for that outbreak of voice.
[0068] In the considered embodiment, the deadline di, k is computed as follows:
where poi is the representation delay estimated at the receiver for the i-th speech surge, where the representation delay is defined as the interval between the generation and representation time of any packet successfully represented in the speech surge, and dl it is the estimate of the delay due to the radio and processing interface at the receiver (different from that generated by its temporary representation memory).
[0069] The history of optimal representation delays past poi, with 0 <j <i, is then used to deduce the new representation delay poi + i for the incoming voice surge.
[0070] In several embodiments, an optimal non-causal reproduction algorithm is used. More specifically, in various embodiments, the representation buffer waits for the entire set of frames of a given speech surge to arrive at the programmer and be transmitted (or suppressed), and then selects the playback delay - based on the pattern of arrivals, transmissions and falls - according to what the best possible quality of speech would have been achieved. The purpose of including such an optimal representation buffer is to predict what an intelligent adaptive representation storage algorithm would do. For example, this allows you to delay packages early and / or suppress packages that have a low chance of being represented.
[0071] In various embodiments, the control module 200 and the emulated buffer B are implemented by means of software code portions, which are executed by a processor at the BS base station. However, the modules can also be implemented in the form of a dedicated digital circuit.
[0072] Figure 4 shows an exemplary scenario including a transmission phase 400, a temporary storage phase 402 and a representation phase 404 of a voice surge with five VoIP frames being generated with time difference P at the time of time gi, 1, gi, 2, gi, 3, gi, 4 and gi, 5.
[0073] In the considered embodiment, the k-th speech surge frame i is passed (virtually) to the decoder at time Pi, k, that is, represented, and the representation delay poi for all packets in a voice surge that are actually represented is assumed to be constant, that is, Pi, k- Pi, j = gi, k-gij, for each k, j that is actually represented.
[0074] Without loss of generality, it can be assumed that qi, k = ^ for each frame i that is lost due to the network. Similarly, it can be assumed that the receiver's representation buffer directly discards all frames that are received too late, that is, packets with δi, k> poi, which implies that all frames with δi, k = qi, k = TO are discarded, regardless of the poi value. Frames that are dropped by the representation buffer contribute to the Li loss rate, which is defined as the relationship between the number of dropped frames and the number of frames in the voice surge.
[0075] In the example in Figure 4, Li = 1/5 because all frames, except the first, are received "on time".
[0076] Therefore, a degree of freedom of representation buffer B is the delay of representation poi.
[0077] In several embodiments, an optimal delay of representation piopt and {δi, k} is determined based on a MOS function that considers QoS and / or QoE criteria. For example, in the considered embodiment, the piopt value is calculated as the value of possible representation delays that maximizes the R factor. However, those of skill in the art will appreciate that the current formula used to compute the optimal piopt representation delay may depend specific codec used.
[0078] In several embodiments, the piopt value is calculated by merging equations (1) and (2) described here, and using equations (2) from Atzori et al .:
R = Ro - Is - [0.024. poi + 0.11. (poi - 177.3). (poi - 177.3). 1 (poi> 177.3)] - Is - (95 - Is). Pp1 / (Pp1 + Bp1) + A (7)
[0079] In several embodiments, the probability of Pp1 packet in equation (7) is replaced with:
where Ni is the number of packets within the i-th speech surge (which can be measured, for example, by inspecting sequence numbers), and the notation 1 (x) means that the result is 1 if x is true, and otherwise 0. Therefore, the sum provides the count of packets in the i-th speech surge that has a delay δx, i being greater than poi, that is, those that were both suppressed within the network (that is, δk , i = + ») or in the representation's own temporary memory (that is, δk, i> poi).
[0080] Therefore, equation (7) can be written as:

[0081] Those of skill in the art will appreciate that the previous equations are just an exemplary implementation, and that other calculations of the R factor can be used. In any case, as already mentioned in the foregoing, piopt can be calculated as the value of possible representation delays that maximizes the R factor:

[0082] Normally, MOS is obtained by a non-increasing function of poi and Li, and Li itself is a non-increasing function of poi. Therefore, there will be an optimum piopt value that allows maximizing the R factor for the voice surge . For example, such an optimal piopt value can be computed by searching the set of possible network delays in a voice surge {δi, k}.
[0083] Therefore, the representation buffer is adaptable, as long as it fixes the reproduction point in one per base voice burst, while it is non-causal, because the delay of all speech frames has to be known before selecting the breeding point.
[0084] In various embodiments, the optimal representation buffer is used as follows: after a voice surge i, the optimal piopt representation delay is computed, then an exponential average is used to deduce a likely representation delay for the next voice surge (i + 1):
poi + 1 is then used to set the deadlines for the packages in the next voice surge according to equation (6).
[0085] Thus, moments of representation are selected based on past history. This leaves the problem of selecting the representation delay for the first voice surge, since no estimate is available at that time. To overcome this problem, a preset value can be selected for the initial value, where the initial value can also be based on information related to the call end points. For example, a typical preset value for the initial representation delay can be 200 ms. In addition, the weight parameter α can also be optimized heuristically.
[0086] As mentioned in the background, an RTP packet can also include n voice frames, with n> 1.
[0087] Several embodiments assume that the RTP time stamp refers to the generation time of either the first or the last voice frame. In this case, the generation time for each frame can be computed for example based on the codec period.
[0088] In several embodiments, the package deadline and disposal time (which were assumed to be the same until now) are considered separately. For example, the deadline can be computed based on the moment of generation of the first voice frame, while the time of discarding the package can be fixed to the moment of presumed representation of the last voice frame. This means that the package can be transmitted as long as it still contains some useful content.
[0089] Therefore, when inserting frames in the temporary representation memory, each frame can be associated with the correct generation time. In addition, depending on how large RTP packets are and how the PDU length is fixed in the LTE cell, a packet may end up being fragmented into multiple PDUs in the eNodeB. In this case, all PDUs can be associated with the same deadline and disposal time, and the packet's network delay can be fixed based on the transmission time of the last PDUs of a packet.
[0090] Due to the fact that the network can suppress packets and it is not assumed to preserve its sequence, it could be possible that the first packet of a voice surge i, which arrives in time ai, k0 at the base station BS is actually the ko-th, with ko> 0.
[0091] Therefore, several embodiments provide for packages to carry sequence numbers. In this way, the programmer can determine that k0 other packages are lost, regardless of their generation time.
[0092] In several embodiments, the programmer thus computes a lower limit on the network delay of previous packets directly on arrival of packet k0. Such packages will inevitably have an earlier deadline than packages already queued in the system.
[0093] In several embodiments, the programmer tries to realign the PDUs. For example, if a k <k0 packet arrives, the associated PDUs can be placed in temporary memory before later packets.
[0094] As this may not be possible for programmers using FIFO memories as temporary memories, the associated PDUs can be placed in FIFO order, but with their correct deadline and disposal time, computed according to equation (6) . In this case, deadlines are increasing not monotonically for the same flow. However, this will not affect scheduling, and if later packages are transmitted well before their deadline, there may still be room to transmit the previous ones.
[0095] The exemplary solution described above distinguishes between voice bursts. For most - if not all - codecs, the packet generation period is constant, which makes such an easy task. In reality, RTP packets carry both time stamps and sequence numbers. Thus, for those codecs, which do not generate and send any information during periods of silence, it is possible to observe that any two packets that arrive consecutively at the BS base station, carrying sequence numbers k1, k2 and with generation time marks such that outbreaks of voice ts with ts (k2)> ts (k1) + P. (k2 - ki) certainly belong to different voice outbreaks. In practical cases, a safety margin (for example, a period) can be added to the right side of the previous inequality to account for unstable sources.
[0096] Conversely, periods of silence should only be estimated a posteriori, that is, when a new outbreak of voice begins, because a salvo of successive losses can be confused with a period of silence.
[0097] For those codecs that send reduced information during periods of silence, a silence period can be detected immediately by examining the packet size.
[0098] Computation of the optimal representation delay can be performed by ordering the network delays, and computing the R factor for each possible network delay. Assuming M packets in a voice surge, this would lead to an O (M) cost of time. Although M may not be expected in a very large number in practical cases (the average voice surge length being in the order of 1 s), the cost of time can also be limited without relying on this assumption.
[0099] For example, network delays can be quantized to a very coarse resolution, for example, 10 ms that normally does not introduce degradation of appreciable quality.
[00100] In addition, instead of storing all possible network delays, a limited number of B + 1 entire delay counters can be used.
[00101] For example, in several embodiments, counter 1, with 0 <1 <B-1, is related to a network delay range being equal to [i. Q + C, (i + 1). Q + C], where Q is the quantization interval and C is a delay offset that can be fixed to the fixed delay component along the path if the latter is known (otherwise C can also be set to zero as an reliable estimate).
[00102] Conversely, the last counter, that is, the B-th counter, can be related to the delay range [B .Q + C, »]. In this way, whenever a package arrives at the optimal representation buffer, the related counter can be increased in a constant time, and the cost is reduced to 0 (B). In reality, when the delay is divided into Q buckets (for example buckets having a delay range from 0 to 10 ms, from 11 to 20 ms, from 21 to 30 ms, etc.), a package arrival with a given delay can be immediately associated with the respective bucket. Therefore, it is not necessary to scan an arrangement for possible delays, but only a single operation is required. Therefore, the time required to perform this operation is constant for all possible delays.
[00103] For example, in the case that the quantization interval is fixed at 10 ms, and the maximum delay is 500 ms, B would be 50, which is quite accessible.
[00104] Generally, bucket delay ranges do not need to be the same size, and / or may also vary over time. In reality, progressive delay ranges can be a viable solution, especially for long-distance calls.
[00105] In addition, optimal values of Q and C for a flow can be estimated dynamically by examining the delay distribution when the connection progresses. This can furthermore decrease the amount of operations required to compute the optimal representation delay.
[00106] As already stated, the solution described here can be used with any programmer that takes into account deadlines in order to select which PDUs to transmit. For example, the Earlier Due Date (EDD) algorithm can be used.
[00107] In cellular networks, however, where mobility is a fundamental requirement and channel conditions vary rapidly, advanced programmers also take into account other information such as a channel quality feedback from mobile users. For example, the Hybrid Aware and Real Time Channel (HY-CART) programmer, designed for UMTS High Speed Packet Access, orders users according to their priority, which is computed by mixing channel information and deadline. In such cases, the role of the solution described here can provide significant deadlines, which also allow you to make better decisions directly at the programmer.
[00108] Without prejudice to the underlying principles of the invention, details and embodiments may vary, even appreciably, with respect to what has been described by way of example only, without departing from the extent of the invention as defined by the appended claims.
权利要求:
Claims (12)
[0001]
1. Method for scheduling transmission of data streams from a node (BS) to a receiver (UE2) on a communication network, where the data streams include data packets for representation by a receiver's playback buffer (106) ), at the receiver (UE2), the method including: arranging said data packets in at least one programming queue (SQ) at the node (BS); suppress data packets from the programming queue (SQ) if their dwell time (t) in the programming queue (SQ) exceeds a given drop deadline (d); determining respective representation values (po) for the data packets indicative of the expected representation time (p) for the data packets by the receiver's reproduction buffer (106); and assign (200) the drop deadlines (d) for the data packages as a function of the respective representation values (po); the method characterized by: in which the step of determining respective representation values (po) for the data packets indicative of the expected representation time (p) for the data packets by the receiver's reproduction buffer (106) comprises emulating, on the node (BS), the receiver's reproduction buffer (106) for the data with an emulated reproduction memory (B), in which to emulate the reproduction buffer of the receiver (106) comprises emulating an optimal representation buffer algorithm, in which the optimal representation buffer algorithm determines the representation values (po) as a function of the Average Opinion Count (MOS), which considers Quality of Service (QoS) and / or Quality of Experience (QoE) criteria.
[0002]
2. Method according to claim 1, characterized by the fact that said data packets include time marks (g) indicative of the generation time of the data packets.
[0003]
3. Method according to claim 1 or 2, characterized by the fact that determining the respective representation values (po) for the data packages indicative of the expected representation time (p) for the data packages by the temporary reproduction memory of the receiver (106) comprises emulating, on the node (BS), the reproduction buffer of the receiver (106) for the data with an emulated reproduction memory (B), in which to emulate the reproduction buffer of the receiver (106) comprises emulating an optimal representation buffer algorithm, includes determining the respective representation values (po) as a function of an estimated transmission delay (δ) between the generation time (g) of the packets on a sender (UE1) in communication network and the arrival time (q) of the packets at the receiver (UE2).
[0004]
4. Method according to claim 3, characterized by the fact that the estimated transmission delay (δ) is determined as a function of: a first value indicating the time (g) that the data packets are generated at the sender (UE1 ), a second value indicating the time (a) that the data packets are received at the node (BS), a third value indicating the expected time (t) that the data packets are transmitted from the node (BS) to the receiver ( UE2), and a fourth value indicating the estimated delay to transmit the data packets from the communication node (BS) to the receiver (UE2).
[0005]
Method according to claim 3 or 4, characterized by the fact that it includes emulating the reproduction buffer of the receiver (106) on the node (BS) as an adaptive representation buffer.
[0006]
6. Method according to any one of claims 1 to 5, characterized by the fact that the data flow is a flow of Voice data over IP, and in which the Average Opinion Count (MOS) is calculated using the model E.
[0007]
Method according to any one of claims 1 to 6, characterized in that the respective representation values (po) are determined later, determining an optimal representation delay (po) as a function of previous transmission delay (δ ).
[0008]
8. Method according to claim 7, characterized by the fact that the data flow is a flow of Voice data over IP, and in which the respective representation values (po) are determined as indicative for the instant of representation expected (p) of all data packets of a given voice surge, determining the optimal representation delay (po) of the previous voice surge.
[0009]
9. Method according to claim 7 or 8, characterized by the fact that the respective representation values (po) are determined by means of an exponential average also taking into account previous values (po) for the respective representation values.
[0010]
10.Communication node (BS) for the programmed transmission of data streams to a receiver (UE2) in a communication network, in which the data streams include data packets for representation by a receiver's reproduction buffer (106) ), at the receiver (UE2), the communication node (BS) characterized by the fact that it comprises: at least one programming queue (SQ), where the data packets of the data streams are stored; and a control module (200) configured to perform all steps of the method as defined in any of claims 1 to 9.
[0011]
11. Communication node (BS) according to claim 10, characterized by the fact that the communication node (BS) is a communication node of Long Term Evolution of Universal Mobile Telecommunication System and / or Access Communication High Speed Downlink Packet.
[0012]
12. Computer-readable memory, characterized by the fact that it includes instructions executable by computer to perform the steps of the method as defined in any one of claims 1 to 9, when the instructions are executed on a computer.
类似技术:
公开号 | 公开日 | 专利标题
BR112012015452B1|2020-12-15|METHOD FOR PROGRAMMING DATA FLOWS TRANSMISSION FROM A NODE TO A RECEIVER IN A COMMUNICATION NETWORK, COMMUNICATION NODE, AND COMPUTER-READABLE MEMORY
Bacioccola et al.2007|User-level performance evaluation of voip using ns-2
US7924711B2|2011-04-12|Method and apparatus to adaptively manage end-to-end voice over internet protocol | media latency
US20050094628A1|2005-05-05|Optimizing packetization for minimal end-to-end delay in VoIP networks
TWI390503B|2013-03-21|Dual channel voice transmission system, broadcast scheduling design module, packet coding and missing sound quality damage estimation algorithm
US10680657B2|2020-06-09|Media controller with jitter buffer
KR20040111669A|2004-12-31|Protocol, information processing system and method, information processing device and method, recording medium, and program
WO2013097783A1|2013-07-04|Method and device for evaluating media transmission quality
WO2010142312A1|2010-12-16|Monitoring of delay in packet-switched networks
Hoene et al.2001|Voice over IP: improving the quality over wireless LAN by adopting a booster mechanism: an experimental approach
US9078166B2|2015-07-07|Method for determining an aggregation scheme in a wireless network
Sengupta et al.2006|Improving R-score of VoIP streams over WiMAX
Chhabra et al.2011|Performance evaluation and delay modelling of VoIP traffic over 802.11 wireless mesh network
Andreozzi et al.2010|OptiMOS: Optimal MOS-based scheduling of downlink voice flows in point-to-multipoint access networks
Dolezal et al.2012|Improving QoE of SIP-based automated voice interaction in mobile networks
Elliott2004|Stream synchronization for voice over IP conference bridges
Jurca et al.2006|Media streaming with conservative delay on variable rate channels
Gholibeigi2011|Cross-layer per-flow QoE evaluation for VoIP in wireless system
Anis2006|Packet Voice Transmission over Data Networks
Li et al.2003|Improving perceived speech quality for wireless VoIP by cross-layer designs
Darmani2004|Lost VOIP packet recovery in active networks.
AlBahouth2010|Development and Performance Study of Sender-Based Loss Recovery Techniques for VoIP systems
Parab0|Assessing the quality of Voice over IP and Video
Ramos2004|Robust and reliable multimedia transmission over the Internet
Zhang2007|Performance of voice over IP in mobile ad hoc networks
同族专利:
公开号 | 公开日
KR20120109574A|2012-10-08|
KR101590972B1|2016-02-02|
US20120250678A1|2012-10-04|
CN102668466A|2012-09-12|
AR079736A1|2012-02-15|
BR112012015452A2|2016-04-19|
JP5632486B2|2014-11-26|
CN102668466B|2015-07-22|
EP2517419B1|2016-05-25|
EP2517419A1|2012-10-31|
WO2011076239A1|2011-06-30|
US9036624B2|2015-05-19|
JP2013516093A|2013-05-09|
引用文献:
公开号 | 申请日 | 公开日 | 申请人 | 专利标题

US6259677B1|1998-09-30|2001-07-10|Cisco Technology, Inc.|Clock synchronization and dynamic jitter management for voice over IP and real-time data|
EP1142257A1|1999-01-14|2001-10-10|Nokia Corporation|Response time measurement for adaptive playout algorithms|
US6512761B1|1999-02-02|2003-01-28|3Com Corporation|System for adjusting billing for real-time media transmissions based on delay|
US6973184B1|2000-07-11|2005-12-06|Cisco Technology, Inc.|System and method for stereo conferencing over low-bandwidth links|
US20040204935A1|2001-02-21|2004-10-14|Krishnasamy Anandakumar|Adaptive voice playout in VOP|
US6965597B1|2001-10-05|2005-11-15|Verizon Laboratories Inc.|Systems and methods for automatic evaluation of subjective quality of packetized telecommunication signals while varying implementation parameters|
US20030112758A1|2001-12-03|2003-06-19|Pang Jon Laurent|Methods and systems for managing variable delays in packet transmission|
US20030115320A1|2001-12-19|2003-06-19|Yarroll Lamonte H.P.|Method for tuning voice playback ratio to optimize call quality|
US7245608B2|2002-09-24|2007-07-17|Accton Technology Corporation|Codec aware adaptive playout method and playout device|
US7269141B2|2002-09-24|2007-09-11|Accton Technology Corporation|Duplex aware adaptive playout method and communications device|
US8400932B2|2002-10-02|2013-03-19|At&T Intellectual Property Ii, L.P.|Method of providing voice over IP at predefined QoS levels|
US7051246B2|2003-01-15|2006-05-23|Lucent Technologies Inc.|Method for estimating clock skew within a communications network|
DE60331008D1|2003-11-11|2010-03-04|Ericsson Telefon Ab L M|ADJUSTING A PLAYOUT BUFFER BASED ON THE AUDIO BURST LENGTH|
KR100938034B1|2004-08-30|2010-01-21|퀄컴 인코포레이티드|Adaptive de-jitter buffer for voice over ip|
US7903688B2|2005-03-11|2011-03-08|Alcatel-Lucent Usa Inc.|VoIP encoded packet prioritization done per packet in an IP communications network|
US7675946B2|2005-08-22|2010-03-09|Infosys Technologies, Ltd.|System and method for managing playout time in packet communication network|
EP2027536B1|2006-06-13|2012-07-11|BRITISH TELECOMMUNICATIONS public limited company|Quality based service selection in a peer to peer network|
FR2908576A1|2006-11-14|2008-05-16|Canon Kk|METHOD, DEVICE AND SOFTWARE APPLICATION FOR SCHEDULING A PACKET TRANSMISSION OF A DATA STREAM|
US8111720B2|2007-01-09|2012-02-07|Telefonaktiebolaget Lm Ericsson |Method and apparatus to indicate maximum scheduling delay for jitter buffer implementations|
US20080267224A1|2007-04-24|2008-10-30|Rohit Kapoor|Method and apparatus for modifying playback timing of talkspurts within a sentence without affecting intelligibility|
JP4911010B2|2007-12-11|2012-04-04|富士通株式会社|Packet capture device, packet capture method, and packet capture program|
EP2099176A1|2007-12-18|2009-09-09|Nokia Corporation|Method and device for adapting a buffer of a terminal and communication system comprising such device|
WO2009109069A1|2008-03-07|2009-09-11|Arcsoft Technology Company, Ltd.|Implementing a high quality voip device|
US7953004B2|2009-01-06|2011-05-31|Alcatel Lucent|Minimizing effects of packet delay variation in time-division multiplexing pseudowire services|
WO2010112975A2|2009-03-31|2010-10-07|Freescale Semiconductor, Inc.|Receiving node in a packet communications system and method for managing a buffer in a receiving node in a packet communications system|
US8345680B2|2009-05-07|2013-01-01|Alcatel Lucent|Handling out-of-sequence packets in a circuit emulation service|CN102474876B|2009-06-29|2015-07-15|皇家飞利浦电子股份有限公司|A method for communicating in a network|
IN2014KN00967A|2011-10-07|2015-10-09|Ericsson Telefon Ab L M|
CN103716114B|2012-09-28|2018-02-23|华为技术有限公司|Parameter setting method, terminal and base station in data transmission service|
KR20140067512A|2012-11-26|2014-06-05|삼성전자주식회사|Signal processing apparatus and signal processing method thereof|
US9686201B2|2013-01-25|2017-06-20|Cable Television Laboratories, Inc.|Predictive management of a network buffer|
US20140241445A1|2013-02-28|2014-08-28|Univerza V Ljubljani, Fakulteta Za Elektrotehniko|Method for providing quality of service in a multiuser orthogonal frequency division multiplexsystem|
US9232048B2|2013-12-04|2016-01-05|International Business Machines Corporation|Quality of experience determination for multi-party VoIP conference calls that account for focus degradation effects|
HUE032992T2|2014-01-17|2017-11-28|Deutsche Telekom Ag|Method for an enhanced communication between a first network node and a second network node of a telecommunications network, and telecommunications network|
EP2963875B1|2014-07-02|2018-02-28|ABB Schweiz AG|Method for processing data streams including time-critical messages of a power network|
KR102176653B1|2014-09-04|2020-11-09|삼성전자주식회사|Apparatus and method for rate control in wireless communication system|
CN106688207A|2014-09-10|2017-05-17|西门子公司|Method for determining a propagation time of a telegram in a communication network, and corresponding network components|
EP3070893B1|2015-03-20|2017-10-04|Alcatel Lucent|Scheduling of packets in network devices|
KR20170028798A|2015-09-04|2017-03-14|삼성전자주식회사|Playout delay adjustment method and apparatus and time scale modification method and apparatus|
US10721178B2|2016-01-22|2020-07-21|Medtronic, Inc.|Systems, apparatus and methods facilitating data buffering and removal|
US10432553B2|2016-02-23|2019-10-01|Microsemi Solutions , Inc.|Systems and methods for transportation of multiple constant bitrate data streams|
CN108173784B|2017-12-29|2021-12-28|湖南恒茂高科股份有限公司|Aging method and device for data packet cache of switch|
EP3790203A1|2019-09-03|2021-03-10|Imec VZW|Device and method for application-requirement aware medium access control|
法律状态:
2018-03-27| B15K| Others concerning applications: alteration of classification|Ipc: H04L 12/875 (2013.01), H04L 12/801 (2013.01), H04L |
2019-01-15| B06F| Objections, documents and/or translations needed after an examination request according [chapter 6.6 patent gazette]|
2019-11-05| B06U| Preliminary requirement: requests with searches performed by other patent offices: procedure suspended [chapter 6.21 patent gazette]|
2020-10-06| B09A| Decision: intention to grant [chapter 9.1 patent gazette]|
2020-12-15| B16A| Patent or certificate of addition of invention granted [chapter 16.1 patent gazette]|Free format text: PRAZO DE VALIDADE: 10 (DEZ) ANOS CONTADOS A PARTIR DE 15/12/2020, OBSERVADAS AS CONDICOES LEGAIS. |
优先权:
申请号 | 申请日 | 专利标题
PCT/EP2009/009266|WO2011076239A1|2009-12-24|2009-12-24|A method of scheduling transmission in a communication network, corresponding communication node and computer program product|
[返回顶部]