Apparatus and method for transmitting a plurality of
information signals in flexible time-division multiplexing
The present invention relates to a concept for transmitting a plurality of information signals
in statistical multiplexing, in particular in statistical time-division multiplexing as may be
employed, for example, for transmitting audio-visual contents in a digital video
broadcasting system.
In a DVB-H system (DVB-H = digital video broadcasting handhelds), several multimedia
services, in particular digital video signals, may be transmitted, in a transport stream, in
time-division multiplexing via a transmission channel having a quasi-constant bit or data
rate. If each video signal is assigned a fixed data or encoding rate, in accordance with an
encoded information signal, a program provider, for example, will be forced to find a
tradeoff between a transmission capacity that is sometimes expensive and a picture quality
that may be achieved for critical scenes. Occasionally, a data or compression rate will not
suffice in this context, and scenes that are rich in detail might suffer an impairment of
quality. On the other hand, with a fixedly assigned encoding rate, it may also happen that
the assigned encoding rate exceeds an encoding rate required for a current scene, and that,
therefore, the encoding rate, or cost, is wasted.
Depending on a current picture content, a video encoder requires different levels of data
and/or encoding rates so as to ensure, for example, a high-quality television transmission,
A sports broadcast, for example, typically requires a higher data rate -because its picture
contents contain a lot of motion - than, e.g., a talk show or a news broadcast, which has
rather static picture contents. Particularly high data rates are required for transmitting
scenes that are rich in detail and contain a lot of motion.
Video encoding and/or compression methods are based, for example, on predictions, such
as the so-called hybrid encoders, which perform, for a picture, a motion-compensated or
picture-internal prediction with subsequent, e.g. entropy-based, compression of the
remainder of the prediction. This means, similarities within a picture (intra) and/or among
the pictures (inter) are exploited for making a prediction. Said predictions works with
varying degrees of success, depending on the picture content. Accordingly, the residual
signal will be larger or smaller, depending on the quality of the prediction. A larger
residual signal requires a larger number of bits for encoding. Conversely, encoding of the
motion compensation also requires, as side information, bits for encoding, so that a more
complex prediction does not necessarily result in an improved compression rate. Overall,

an ideal picture quality, or an ideal tradeoff between rate and quality, may be found for
various data rates available and, therefore, for various compression rates. This relationship
between the rate available and the picture quality achievable is dependent on the signal.
Therefore, different compression rates or data rates are required for encoding for different
scenes in order to achieve the same subjective picture quality.
The larger a number of programs or a number of program providers, the more unlikely it
will be that all of the programs simultaneously require a very high data rate for being
encoded. If several information signals, in particular videos, are transmitted in a transport
stream via a channel having a constant overall data rate, said differences in the data rates
may be exploited in assigning data rates to the individual services.
To this end, the individual data rates of a DVB-H network may be configured dynamically
in accordance with a so-called statistic multiplex. This involves distributing the data rates
such that a ratio between the encoding rate and the picture quality becomes optimal. This
method is cooperative and requires that a sum of data rates of the individual services
always remain smaller than the available overall data rate. Instead of allotting a fixed data
rate to each information signal, the statistical multiplexing analyzes contents of the picture
material to be transmitted and assigns different individual data rates, depending on the
prediction properties, to the plurality of information signals for shared transmission in a
transport stream via the channel having the constant overall data rate. Instead of assigning
a maximally required data rate to each video, it is thus possible to operate with a clearly
reduced average data rate per video without reducing the picture quality perceived.
Therefore, an overall impairment of quality may be reduced in this manner.
Reception of videos or information signals at a mobile terminal, such as a DVB-H receiver,
should obviously not result in its battery being depleted within a very short time. For DVB-
T systems (DVB-T = digital video broadcasting terrestrial), an entire data stream must
always be decoded before access may be made to any of the services contained within the
data stream, such as TV programs, for example. With DVB-H, one uses the so-called time
slicing technique, wherein only part, or a time slice, of the data stream is received which
contains data of a service or program that has just been selected. With DVB-H, merging or
multiplexing of different services is therefore performed in pure time-division
multiplexing, wherein information signals of each service are periodically sent in
compressed data packets or time slices. Thus, an individual service is not emitted
continuously, but only from time to time and at a correspondingly high data rate, and
sometimes it is not emitted at all. Time-division multiplexing of several services will then

yield a continuous data stream with a quasi-constant data rate, as is shown in Fig. 7, for
example.
Fig. 7 shows a continuous data stream 700 with a constant mean data rate BR. The data
stream 700 represents an MPEG transport stream that is made up of MPEG-2-compatible
(MPEG = moving picture experts group) elementary data streams organized into time
slices 702. Fig. 7 reveals that program-specific information and meta data 704 (PSI/SI) are
not subject to the time slicing method. In addition, neither a fixed, e.g. repetitive,
assignment of the individual services to time slices 702, nor a fixed magnitude or duration
of same is prescribed, even though in many DVB-H multiplexes such a fixed structure is
used anyway. The duration of a time slice 702 associated with an information signal
generally depends on the size of the current data packets of the respective service that are
to be transmitted within said time slice. For example, if a video signal currently requires a
comparatively high encoding rate, the time slice 702 that may be associated with the video
signal will have a correspondingly long time duration.
Due to the variable time-slice structure depicted in Fig. 7, a receiver of the data stream 700
must have the precise position and configuration of the time slices 702 transmitted to it, so
that the sequential data flow of the individual broadcasting services may be reconstructed
therefrom. For DVB-H, the so-called delta-T method is used to this end. It includes
transmitting, within each time slice 702, a relative waiting time delta-T informing the
receiver when the next time slice of the same service is receivable. The system allows
signaling waiting times within a range of a few milliseconds up to about 30 seconds (see
Fig. 8).
Within the receiver, incoming time slices 702 are buffered and subsequently read out at a
constant rate (at the average data or encoding rate of the respective service). The duration
of the time slices 702 is typically within a range of several hundred milliseconds, whereas
the switch-off time, in accordance with delta-T, of the receiver between the time slices may
be many seconds (see above). Depending on the ratio of on/off-time, power savings of
more than 90% as compared to DVB-T may result. For this purpose, time slicing
presupposes a sufficient number of services or information signals in order to be as
effective as possible.
In a DVB-H system, information signals or services are transmitted on the basis of the
internet protocol (IP). This approach enables simple connection with other networks. The
MPEG-2 transport stream 700 serves as a physical carrier. Embedding of IP data within the
transport stream is effected by using an existing adaptation protocol, so-called multi-

protocol encapsulation (MPE). To protect the transport stream 700 from interfering effects
of a radio channel, DVB-H additionally comprises resorting to an error protection (MPE-
FEC), which is applied at the level of the IP data stream before the IP data is encapsulated
by means of MPE. By this mechanism, the receiving power is to be generally improved, in
particular the reliability for mobile reception and with strong pulse-shaped interferences as
may occur, for example, due to multipath propagation and resulting destructive
interferences at the point of reception.
MPE-FEC is very similar to time slicing and MPE. These three techniques are directly
tuned to one another, and together form the so-called DVB-H codec. IP data streams from
the various sources are multiplexed as individual elementary streams by the time slicing
method. The error protection MPE-FEC is calculated separately for each individual
elementary stream and is added. This is followed by encapsulating the IP packets into the
so-called sections of the multi-protocol encapsulation and, subsequently, by embedding
into the transport stream.
With regard to temporal behavior, the disadvantage of the delta-T method is that the data
or encoding rate of a DVB-H service can be changed only for the future time slices of the
respective service. An algorithmic delay results, which corresponds to the distance of the
time slices of the individual services. In the event of constant-rate services, this
disadvantage is not relevant. In this case, the payload data of the services may be
immediately encapsulated and combined into time slices.
This is different, however, for services that employ a variable data/encoding rate, such as
in statistical multiplexing, for example. Due to the offset signaling, such services must be
delayed in accordance with the repetition rate of the time slices assigned. To be able to
indicate, in the time slices of a current transmission cycle, the respective relative waiting
times delta-T up to the corresponding time slices of the subsequent transmission cycle, a
time-slice structure, i.e. time-slice starting times and/or time-slice durations, of the
subsequent transmission cycle must ideally have already been recognized. After all, the
data-rate requirements of a service may change from one time slice to its subsequent time
slice, as a result of which a completely different time-slice structure may result for the
subsequent transmission cycle as compared to the current transmission cycle.
In order to know future data-rate requirements or encoding-rate requirements, and, thus,
the subsequent time-slice structure, for statistical multiplexing, the information signals
must already be analyzed in advance. This may result in considerable latency periods. The
time diagram of Fig. 9 shows an illustration of this circumstance.

Fig. 9 shows a temporal sequence of processing of an elementary stream 900 incoming for
a service on the transmitter side. The data stream 900 is partitioned into portions N-1, N,
N+1. The data of the portion N is to be transmitted within a time slice 902-N, and the data
of the portion N+1 is to be transmitted within a subsequent time slice 902-(N+1). It may be
seen from Fig. 9 that a data-rate analysis 904-(N+1) for the data-stream portion N+1 must
have been completed by the time the time slice with the data of the respectively temporally
preceding portion is sent, i.e. time slice 902-N. This is due to the fact that, as was already
described above, the relative waiting time up to the subsequent time slice 902-(N+1) is
integrated in the time slice 902-N. Thus, the data rate analysis for the portion N must
already have been completed by the time-slice starting time TN-1 of the time slice 902-(N-
1), the data-rate analysis of the portion N+1 must already have been completed by the
time-slice starting time TN of the time slice 902-N, etc. This results in a relatively long
latency period TL between the arrival of the data portions N, N+1 of the elementary stream
900 and the corresponding reproduction times or time-slice starting times TN or TN+1.
The above-described long latency period TL actually contradicts the goals of statistical
multiplexing.
It is thus the object of the present invention to provide a concept that enables reduced
latency periods between the data input and the reproduction time as compared to the prior
art. The concept provided is also intended to improve the receiver's energy efficiency.
The object is achieved by an apparatus having the features of claim 1, and by a method as
claimed in claim 14.
The finding of the present invention is that the object mentioned above may be achieved in
that exact data-rate analysis or encoding-rate analysis of the information signal portions to
be transmitted in a subsequent transmission cycle is initially omitted and, instead, on the
basis of highly exact estimated values for said subsequent data or encoding rates, estimated
values for the relative waiting times delta-T are transmitted within the time slices of a
current transmission cycle. Actual data or encoding rates, which may deviate from the
estimated data rates for the individual information signals, may be set in the subsequent
transmission cycle, as a result of which the predicted time-slice boundaries for the
subsequent transmission cycle may shift, however. The possible shift of the time-slice
boundaries or the time-slice starting times is subject to several boundary conditions,
however. It is important that no time slice of the subsequent transmission cycle starts
before its signaled estimated starting time. In this manner, it may be ensured, on the one

hand, that a receiver will not "miss" the time slice designated for it. On the other hand, by
means of a very exact estimation of the time-slice structure one may achieve that a time
slice does not start at a considerably later point in time than was previously signaled, as a
result of which receiving power may be saved since the receiver does not have to bring
forward its reception time on the off chance. This mainly applies to data or encoding rates
of the individual services that are variable only at a relatively slow rate. For constant data
rates, the estimated time-slice structure and the actual time-slice structure are even
identical, so that in this case the concept introduced takes full advantage of the benefits and
the efficiency of time slicing.
In more general terms, one may state that with input quantities for a data-rate estimator that
remain constant for all of the information signals from transmission cycle to transmission
cycle, such as the information signals themselves and/or a ratio between the coding rate
and the quality and/or a ratio between the coding rate and the data-rate cost, the estimated
values determined for the relative waiting times match actual waiting times, so that an
estimated time-slice starting time derived therefrom and an actual time-slice starting time
of the subsequent transmission cycle exactly coincide. In this case, therefore, no power is
wasted on the receiver side by switching on the receiver too early.
A time duration of the incoming information signal portions together with estimated or
actual encoding rates yields, equivalently, estimated or actual data quantities, respectively,
that are transmitted within the various time slices. Thus, one may also speak of estimating
data quantities and of determining actual data quantities.
The redistribution of data rates among the various time slices of a transmission cycle is
effected in dependence on currently actually required data rates for the individual services
as well as in dependence on cost of the individual data rates. This cost may be technical
cost, such as computing power, but also monetary cost. This dependence is necessary,
since otherwise every service would try to extend its time slice - in deviation of the
respective estimation - as much as possible in order to gain the largest benefit possible for
itself. I.e., in reality, unlimited cooperation of the service providers cannot be assumed,
which is why data-rate cost forms an incentive to reduce or increase data rates or time-slice
lengths as required. It is to be noted that in the redistribution all of the time slices are more
or less equal, i.e. each time slice has the same possibilities or degrees of freedom available
to it in the redistribution as the other time slices, irrespective of its position within the
time-slice structure.

Embodiments of the present invention provide an apparatus for transmitting a plurality of
information signals within a plurality of current and subsequent time slices in a current and
a subsequent transmission cycle in accordance with time-division multiplexing. The
inventive apparatus comprises an estimator configured to estimate, for each of the
information signals, a data rate with which the information signal probably will be encoded
within a time slice which follows the current time slice, to obtain an estimated time-slice
structure for the subsequent transmission cycle. In addition, the apparatus comprises a
processor configured to determine, for each of the information signals and on the basis of
the estimated time-slice structure, a relative waiting time which indicates an estimated
starting time of the subsequent time slice of the information signal. A time-slice structurer
is provided which is configured to assign an actual starting time to each of the following
time slices on the basis of an actual data rate for the information signal so as to obtain an
actual time-slice structure, it being possible for the actual data rate to deviate from a data
rate estimated by the data rate estimator in the current transmission cycle. The time-slice
structurer is configured to select the actual starting times of the subsequent time slices to
be larger than or equal to the respectively estimated starting time.
In the event of only slowly variable data rates, or of a good match of estimated and actual
data rates, the estimated and the actual time-slice structures are nearly identical. In the
event of constant data rates or of a perfect match of estimated and actual data rates, the
estimated and the actual time-slice structures are identical, so that a receiver may operate at
maximum power efficiency. I.e., the better the estimator, the smaller the deviations
between estimated and actual values, so that a receiver will not be switched on before, or
only marginally before, a time slice.
Thus, by means of the present invention, the advantages of statistical multiplexing may be
combined with the advantages of time slicing and of the delta-T method without wasting
more receiving power than necessary at a receiver that relies on the accuracy of the delta-T
values signaled.
Advantageous implementations of the present invention are the subject matter of the
dependent claims.
Embodiments of the present invention will be explained below in more detail with
reference to the accompanying figures, wherein:

Fig. 1 shows a method of transmitting a plurality of information signals in flexible
time-division multiplexing in accordance with an embodiment of the present
invention;
Fig. 2 shows a schematic representation of a plurality of information signals in flexible
time-division multiplexing in accordance with an embodiment of the present
invention;
Fig. 3 shows a block diagram of an apparatus for transmitting a plurality of information
signals in flexible time-division multiplexing in accordance with an embodiment
of the present invention;
Fig. 4 shows a schematic representation of a shift of time-slice starting times;
Fig. 5 shows a schematic representation of a flexible multiplex system in accordance
with an embodiment of the present invention;
Fig. 6 shows a timing diagram of a flexible multiplexing technique in accordance with
an embodiment of the present invention;
Fig. 7 shows an example of an architecture of a DVB-H multiplex;
Fig. 8 shows delta-T signaling taken from the European telecommunication standard
ETSI EN 302 304; and
Fig. 9 shows an example of a time flow of a service with a variable data rate in
conventional statistical multiplexing.
Initially, the fundamental concept of the present invention is to be described with regard to
Figs. 1 and 2. To this end, Fig. 1 shows a schematic flowchart of an inventive method 100
for transmitting a plurality of information signals in accordance with statistical time-
division multiplexing. The individual method steps of the method 100 of Fig. 1 will now
be explained in more detail by means of Fig. 2.
A transport stream 200 shall be assumed, wherein time slices Bj[.] (i=1,...,I) are arranged or
transmitted in time-division multiplexing by a plurality I of service providers. The time
slices B,[.] are arranged in successive transmission cycles {...,Z[n-1],Z[n],Z[N+1],...}.

In a first step 102 of the inventive method 100 depicted in Fig. 1, for each of the I time
slices Bi[n] (i=1,...,I) of the nth transmission cycle, a data rate ri[n] (i=1,...,I) is estimated
with which the respective information signal (in accordance with the service i) is expected
to be transmitted within the time slice Bi[n+1] of the (N+1)th transmission cycle which
follows the current time slice Bi[n], This results in an estimated time-slice structure for the
subsequent transmission cycle Z[n+1]. This estimated time-slice structure comprises
estimated time-slice durations and/or time-slice starting times of the subsequent time slices
Bi[N+1](i=1,...,I).
In a further step 104, a relative waiting time delta-Ti[n], which indicates an estimated
starting time Ti[n + 1] of the subsequent time slice Bi[n+1] of the ith information signal, is
determined for the current transmission cycle Z[n] for each of the I information signals on
the basis of the estimated time-slice structure for the subsequent transmission cycle
Z[n+1].
In a subsequent step 106, each of the subsequent time slices Bj[n+1] (i=1,...,I) has an actual
starting time Ti[n+1] and an actual data rate reff,i[N+1] (i=1,...,I) assigned to it so as to
obtain an actual time-slice structure for the subsequent transmission cycle Z[n+1]. The
actual data rate reff,i[N+1] may deviate both in the downward and in the upward directions
from the data rate ri[n] that has been estimated and transmitted in the transmission cycle
Z[n]. In step 106, the actual starting times Ti[n+1] are selected such that each of the actual
starting times Ti[n+1] (i=1,...,I) of the subsequent time slices Bi[n+1] is larger than or equal
to the starting times r,[n + 1] (i=1,...,I) that have been previously estimated in each case.
I.e., Ti[n+1] ≥ Ti[n + 1] = Ti[n] + delta-Ti[n]. In other words, this means that a time slice
Bj[n+1] can never start prior to its starting time Ti [n +1] signaled within the previous time
slice Bj[n].
In the event of only slowly variable data rates reff,i[n] (i=1,...,I) or of a good match of
estimated and actual data rates ri[n], reff,i[n] (i=1,...,I), the estimated and actual time-slice
structures are almost identical, i.e. Ti[n] ≈ Ti[n]. In the event of constant data rates or of a
perfect match of estimated and actual data rates, the estimated and the actual time-slice
structures are identical, i.e. Ti[n] ≈ Ti[n], so that a receiver may operate at maximum power
efficiency. I.e., the better the estimating 102, the smaller the deviations between estimated
and actual values ri[n], reff;i[n] and Ti[n], Ti[n], respectively, so that a receiver will not be
switched on before, or only marginally before, a time slice Ti[n] (i=1 ,...,1). Naturally, Ti[n]
≈ Ti[n] may also apply with rapidly variable data rates, i.e. a change between two

successive cycles Z[n], Z[n+1], if the estimation or the prediction 102 provides
correspondingly accurate estimated values.
Before the inventive concept will be described in more detail in the following paragraphs,
the mathematical symbols used in this specification shall initially be explained:
I number of services in DVB-H multiplexing
{S1, S2,..., Si,..., SI} the quantity of services of DVB-H multiplexing
{B[1], B[2],..., B[n],...} the quantity of the temporally linear sequence of the time
slices of all the services in DVB-H multiplexing
{Bi[1], Bi[2],..., Bi[n],...} the quantity of the temporally linear sequence of all time
slices belonging to the service Si
r[n] signaled rate requirement of the time slice B[n]
reff[n] effective data rate of the time slice B[n]
rw[n] desired rate of the time slice B[n]
Jmax maximum time jitter for the beginning of the time slice
ΔT[n] delay of the start of the time slice B[n] as compared to the signaled start
ΔTmax[n] maximally possible delay as compared to the signaled start of the time slice
B[n]
Δt[n] temporal change in the time slice length of the time slice B[n]
f(i;n) calculates the linear time-slice number of the time slice B;[n]
fs-1(n) determines the number of the service with which the time slice B[n] is
associated

fn-1(n) determines the time-slice number while exclusively considering the time
slices of the service Sfs-1(n).
c cost per second of transmission capacity
u[n] utility function (payout function) for the associated service at the time of the
time slice B[n]
An apparatus 300 for transmitting a plurality of information signals 301-i (i=1,...,I) in a
plurality of current and subsequent time slices Bi[n], Bi[n+1] (i=1,...,I) in a current and a
subsequent transmission cycle Z[n], Z[n+1] in accordance with time-division multiplexing
shall now be described by means of Fig. 3.
The apparatus 300 comprises an estimator or predictor 302 configured to estimate, for each
of the information signals 301-i, a data rate r[f(i;N+1)] with which the information signal
301-i in the time slice Bi[n+1] following the current time slice Bi[n] is presumably encoded
by an associated encoder 304-i (i=1,...,I) so as to obtain an estimated time-slice structure
for the subsequent transmission cycle Z[n+1].
The apparatus 300 comprises a processor 306 configured to determine, for each of the
information signals 301-i, on the basis of the estimated time-slice structure or the estimated
data rates r[f(i;n+1)] (i=1,...,I), a relative waiting time delta-Ti which indicates an estimated
starting time Tf [n +1] of the subsequent time slice Bj[n+1].
What is also provided is a time-slice structurer 308 configured to assign, to each of the
subsequent time slices Bi[n+1], an actual starting time Ti[n+1] on the basis of an actual
data rate reff[f(i;N+1)] for the information signal 301-i so as to obtain an actual time-slice
structure, it being possible for the actual data rate reff[f(i;n+1)] to deviate from a data rate
r[f(i;n+1)] (i=1,...,I) estimated by the data rate estimator 302 in the transmission cycle Z[n].
The time-slice structurer 308 is configured to select the actual starting times Ti[n+1] of the
subsequent time slices Bi[n+1] to be larger than or equal to the starting time Ti[n + 1]
estimated in the preceding transmission cycle Z[n], such that an actual starting time never
comes before an estimated starting time. I.e., Ti[n+1] ≥ Ti[n + 1]. Ideally, i.e. when
r[f(i;n+1)] =reff[f(i;n+1)], then Ti[n+1] = Ti[n + 1] shall apply.
The actual data rates reff[f(i;n+1)] with which the information signals 301-i in the
transmission cycle Z[n+1] are actually encoded by encoders 304-i (i=1,...,I) result from

desired or preferred data rates rw[f(i;N+1)] and various further boundary conditions, which
will be explained below.
In accordance with embodiments, the processor 306, therefore, is configured to assign each
of the current time slices Bj[n] an actual starting time Ti[n] on the basis of an actual data
rate reff[f(i;n)] for an information signal 301-i associated with the current time slice so as to
obtain an actual time-slice structure for the current time slice, it being possible for the
actual data rate reff[f(i;n)] to deviate from a data rate r[f(i;n)] estimated by the estimator
302 in a preceding transmission cycle Z[n-1], so as to determine the relative waiting time
delta-Ti[n] for each of the information signals on the basis of the estimated time-slice
structure the subsequent transmission cycle Z[n+1] and of the actual time-slice structure
for the current transmission cycle Z[n].
In accordance with embodiments, the processor 308 is further configured to incorporate
each of the relative waiting times delta-Tj[n] that have been determined in a current time
slice Bi[n] of an associated information signal 301-i so as to transmit them to a remote-
side receiver for the multiplex signal resulting from the multiplexed time slices B,[n]
(i=1,...,I). As compared to a conventional system wherein delta-T values are transmitted,
this, therefore, is currently about estimated values for actual delta-T values that are not yet
known in the current transmission cycle Z[n] since the corresponding signal portions of the
subsequent cycle Z[n+1] have not yet been analyzed. This results in a shorter latency
period between the input of the signal portions and the transmission of their encoded
versions within the time slices.
Even though the inventive concept enables shifting of the previously estimated time-slice
starting times Tj[n+1] back in time, as will be explained later on, as exact an estimation as
possible of the delta-T values and, thus, of the starting times Ti[n+1] (i=1,...,I) is
advantageous since it is only with estimated values Ti[n + 1] having little deviation that a
receiver is able to work in a power-efficient manner.
In accordance with embodiments, the data rate estimator 302 is configured to estimate a
data rate r[f(i;N+1)] of an information signal 301-i for the subsequent time slice Bi[n+1] at
least on the basis of preceding and/or current actual data rates
{reff [f(i;1)], reff [f(i,2)],...., reff [f(i;n)]}, of preceding and/or current preferred data rates
{rw [f(i;1)], rw [f(i,2)],...., rw [f(i;n)]}, and preceding estimated and/or signaled data rates
{r[f(i;1)], r[f(i,2)],...., r[f(i;n)]} of the information signal 301-i. It may be a causal predictor
that outputs, by observing the past rate requirement of all of the information signals of the
ith information signal 301-i, an estimated value for the requirement needed within the next
time slice Bi[n+1] (i=1,...,I). The estimated value r[f(i;N+1)] may depend on the past data

rates signaled, the past data rates desired and the past actual data rates of the ith information
signal 301-i (i=1 I).
A predictive estimator 302 may be realized, in accordance with an embodiment, by the
following specification, for example:

Here, r[f(i;n)] designates the data rate signaled of the current time slice Bi[n] (that was
already estimated in the cycle Z[n-1]), and r[f(i;N+1)] designates the estimated data rate of
the following time slice of the ith service S,. rw[f(i;n)] is the preferred data rate for the time
slice Bi[n], whereas reff[f(i;n)] is the ultimately allocated or actual data rate for the time
slice Bi[n]. The compensation factor α ϵ [0;1) regulates the degree of error compensation.
A person skilled in the art will readily appreciate that the predictive estimator 302
described by means of equation (1) is merely exemplary and may quite possibly be realized
differently. Algorithms for implementing predictors are known from specialized literature,
which is why they will not be addressed in more detail here. As has already been
emphasized several times, the quality of the estimated values r[f(i;N+1)] (i=1,...,I) is
decisive for the power consumption that may be achieved at the receiver.
Embodiments of the present invention realize flexible time-division multiplexing wherein
it is possible, in principle, to change the data quantity of the individual time slices Bj[n]
within certain limits, even though the time-slice starting times Tj[n] concerned or their
estimated values Ti [n] were already signaled to the receivers by means of the delta-T
method (delta-Ti[n-1]). The inventive concept is based on a possible extension of time
slices Bi[n] into the neighboring following time slices Bi+1[n], Bi+2[n], etc. The inventive
concept allows a redistribution of the data rates of the individual services or information
signals 301-i within certain limits.
The time-slice extension shall be explained below by means of a simple example: a time
slice of a service S2 and S3 shall be contemplated, respectively, that directly follow each
other within the transmission cycle contemplated. If it is found that the data rate of the
service S2 that was determined and signaled previously, i.e. in a preceding transmission
cycle, is too low for the current time slice, and that, equally, that of S3 is too high, the
duration of the time slice of S2 could be extended by a certain amount. In this case, S2

would be able to transmit the desired data, and the data rate of S3 would decrease to a
lower level. However, for S3 this entails the disadvantage that the start of the time slice of
S3 is later than was previously signaled. In this case, the receiver expecting the service S3
would be activated too soon and, thus, receiving power would be wasted.
In a further example it shall be assumed that the data rate of S3 is not to be changed after
all. The additional data rate required by S2 is to be provided, in this example, by S4 instead.
The time slice of S4 immediately follows that of S3. In this case,
• the time slice of S2 could be extended,
• the time slice of S3 could then be shifted, in total, by the extended time interval, and
• the time slice of S4 could be shortened accordingly.
These exemplary circumstances are depicted in Fig. 4. In this example, too, the data rates
may be redistributed successfully. A start of the time slice that is behind schedule by Δt3
and Δt4, respectively, results both for S3 and for S4. It becomes apparent that for an
exchange of data rates between two different services, the time slices of all of the
intermediate services (here: S3) are reproduced behind schedule.
The following paragraphs shall provide an explanation of the conditions underlying the
exchange of data rates between different time slices.
To this end, the chronology of successive time slices B[1],...,B[n],..., whose estimated
starting times were already signaled to the receivers by means of the delta-T method, is
contemplated. For each time slice B[n], the time-slice duration is changed by the value
Δt[n], Thus the time-slice duration is changed by the value Δt[n], This results in the
following deviation from the estimated starting time of the time slice B[n]:

As a boundary condition for further, subsequent observations, the fact that a time slice B[n]
must never start before its estimated or signaled starting time results in the following:


If the above-described redistribution of the data rate is regarded as a game as defined by
game theory, it becomes apparent that in order to obtain maximum benefit all of the
participants in the game must cooperate, since otherwise the optimum strategy for each
participant would consist in extending their time slices by a maximum amount so as to be
able to accommodate as high data rates as possible. To avoid this obligation to cooperate as
a prerequisite for functional flexible multiplexing, the following restricting rules are
therefore introduced for redistributing the data rate:
1. If a time slice B[f(i;n)] is extended by Δt[f(i;n)], it should be possible for the service
Si to compensate for said extension within the subsequent time slice B[f(i;n+1)], i.e.
the scheduled time-slice duration of the subsequent time slice B[f(i;n+1)] should be
longer than At[f(i;n)]. To this end, the time-slice structurer 308 is configured, in
accordance with embodiments, to assign each of the current time slices B[f(i;n)] an
actual starting time and an actual data rate reff[f(i;n)] that may deviate from a
previously estimated data rate r[f(i;n-1)], and to change a duration of a current time
slice by a differential time duration Δt[f(i;n)] only when an estimated duration of
the subsequent time slice B[f(i;n+1)] is longer than the differential time duration
Δt[f(i;n)J.
2. The actual change Δt[f(i;n)] of the current time-slice length, i.e. the differential time
duration, is determined, while observing inequality (3), by the following:

In this context, Δtw[f(i;n)] designates a desired or preferred differential time
duration of the current time slice B[f(i;n)]. The preferred differential time duration
Δtw[f(i;n)], which directly depends on the preferred data rate rw[f(i;n)] for the
current time slice B[f(i;n)], additionally results, in accordance with embodiments,
from the data-rate cost, such as a price per data rate unit, for example. This will be
addressed in more detail below. The preferred differential time duration Δtw[f(i;0)]
for the respectively first time slice B[f(i;0)] of the service Sj shall be assumed to be
zero, i.e. Δtw[f(i;0)] = 0. To realize equation (4), the time-slice structurer 308 is
configured, in accordance with an embodiment, to determine the differential time
duration Δt[f(i;n)] as a function of a differential time duration Δtw[f(i;n)] resulting
from a preferred data rate for the current time slice, of a differential time duration
Δtw[f(i;n-1)] resulting from a data rate that is preferred for the preceding time slice,

and of a delay ΔT[f(i;n)] of the actual starting time T[f(i;n)] as compared to a
starting time Ti [n] estimated in the preceding transmission frame.
Equation (4) describes more or less the data rate exchange between two successive
time slices B[f(i;n-1)] and B[f(i;n)] (or B[f(i;n)] and B[f(i;n+1)]). If the preceding
time slice B[f(i;n-1)] was already extended by Δtw[f(i;n-1)], this extension time
duration is subtracted, in accordance with equation (4), from the currently desired
differential time duration Δtw[f(i;n)]. This may be interpreted to the effect that data
rate debts of the preceding time slice B[f(i;n-1)] are settled in the subsequent time
slice B[f(i;n)]. If the preceding time slice was, therefore, extended (Δtw[f(i;n-1)] >
0), this extension time is subtracted, in accordance with equation (4), from the
desired extension Δtw[f(i;n)] of the current time slice B[f(i;n)] so as to obtain the
actual change Δt[f(i;n)] of the time-slice length. The actual change Δt[f(i;n)] of the
time-slice length may of course also be negative, which corresponds to a shortening
of the time slice B[f(i;n)]. In addition, equation (4) expresses that the current time
slice B[f(i;n)] must not be shortened by more than the overall delay ΔT[f(i;n)] so as
not to infringe upon inequality (3), i.e. Δt[f(i;n)] ≥ -ΔT[f(i;n)]. The time-slice
structurer 308 is therefore configured, in accordance with an embodiment, to select
the differential time duration Δt[f(i;n)] of a time slot B[f(i;n)] of an information
signal to be longer than a negative accumulated overall delay time -ΔT[f(i;N+1)] of
preceding time slices of a transmission cycle.
3. A maximally possible delay ΔTmax[f(i;n)] of the beginning of the time slice, said
delay being known for all n, be defined for each time slice B[f(i;n)]. This
maximally possible delay ΔTmax[f(i;n)] may be determined, for example, from the
capacity utilization of the jitter buffers of the respective service Sj. Thus, no time
slice B[f(i;n)] should be delayed more than is allowed by ΔTmax[f(i;n)]. When
taking into account known and estimated shifts of neighboring time slices, a further
boundary condition results for the actual change Δt[f(i;n)]:

For inequality (5) to be fulfilled, the desired change Δtw[f(i;n)] of the time-slice
duration must be selected accordingly. I.e., on the basis of inequality (5), the
various services Si (i=1,...,I) may trade with data rates among each other so as to
meet the boundary condition expressed by said inequality for each time slice and

each transmission cycle. The function ΔT[m; n] represents an auxiliary function
that estimates future shifts, i.e. shifts that have not yet taken place, of neighboring
time slices. Let n be the number of the current time slice, and let m be the number
of a neighboring time slice, the shift of which is to be estimated. In this case, let the
function be defined by the following:

Thus, inequality (5) in connection with equation (6) states that each time slice may
be extended, at a maximum, only to such an extent that for none of the neighboring
time slices in the transmission cycle, the maximally possible delay ΔTmax thereof is
exceeded. In the estimation of ΔT[m; n], the time-slice changes (1 ≤ n), which have
already been known, of previously contemplated or changed time slices, and time
slices (1 > n and prev(1) < n) - that are still to be changed - of the transmission
cycle are to be taken into account, of course. For the time slices B[m] still to be
contemplated, the corresponding estimated value ΔT[m; n] for the change is
determined from the change of the time slice fi-1(m) of the preceding transmission
cycle, said fi-1 (m) corresponding to the associated service -fs-1 (m). This estimation
is based on the assumption that the data rate that was subtracted or added in the
preceding transmission cycle will be reclaimed or returned in the current
transmission cycle. Thus, in the estimation in accordance with equations (5) or (6),
the principle of data rate borrowing is employed as well.
Each time slice B[f(i;n)] for which max(Δtw[f(i;n-l)];0) < ΔT[f(i;n)] applies may
alternatively also be shortened, i.e. Δtw[f(i;n)] < 0. Here, inequality (3) must again
be contemplated as a boundary condition, the minimum desired temporal change
Δtw[f(i;n)] results from an insertion into the equation (4):


As was already described above, equation (7) expresses that the change Δt[f(i;n)] of
the time-slice length must not be selected such that a time slice B[f(i;n)] may begin
prior to its previously signaled or estimated starting time. Δtw,min in accordance with
equation (7) expresses precisely this.
The above rules 1 - 4 signify
a) that each time slice B[f(i;n)] can bring forward so much data rate from its own
future time slice B[f(i;n+1)] as may be compensated by the other services by means
of their jitter buffers,
b) that the delay Δt[n] that has already been accumulated during each time slice may
be partly or fully compensated for by shortening the time-slice lengths of other
services.
As was already indicated, the desired changes Δtw[f(i;n)] of the time-slice lengths result, in
accordance with embodiments, from an interplay of data rates rw[f(i;n)] desired or
preferred by the encoders 304-i and of data-rate cost, for example in the form of a price per
data rate unit. I.e., in accordance with embodiments, the time-slice structurer 308 is
configured to set each of the actual data rates reff[f(i;n)] of a transmission cycle in
dependence on the information signal 301-i currently to be transmitted, and on data-rate
cost.
The change Δt[f(i;n)] in the time-slice lengths is to be assigned a payout function. For
services Si, which extend a time slice Bi[n] or B[f(i;n)], a negative payout function shall be
defined. This negative payout is paid out (positively) in equal parts to all of the services Sj
(j ≠ i), the time slices of which have been shifted by this extension Δt[f(i;n)]. However, if a
time slice is shortened (Δt[f(i;n)] < 0), the service that now transmits fewer bits will obtain
a payout equivalent for this reduction of the shift.
This shall be explained below for a proportional cost function. A constant price c shall be
specified per second of transmission time. If a time slice B[n] belonging to the service Si is
extended by Δt[n], the other services will receive the following payout for all of the time
slices transmitted between B[f(i;n)] and


If several time-slice extensions are superimposed, the following payout function results
therefrom:

The (negative) payout function for the service whose time slice is extended may be
determined from equation (8):

If a time slice is shortened, i.e. (Δt[n] < 0), and if an existing overall shift is thus fully or
partly compensated for, the service at whose expense the shortening was made will receive
an additional payment for the compensation:

Thus, the overall payout per time slice results in:

In accordance with embodiments of the present invention, the time-slice structurer 308 is
therefore configured to set each of the actual data rates reff[f(i;n)] in dependence on the
information signal 301-i actually to be transmitted within the associated time slice, and on
data-rate cost c. Each of the information signals 301-i may be credited with a value
u[f(i;n)] dependent on the data-rate cost c when the time slice B[f(i;n)] of the information
signal 301-i is shortened (Δt[f(i;n)] < 0), and a value u[f(i;n)] dependent on the data-rate
cost may be subtracted when the time slice B[f(i;n)] of the information signal 301-i is
extended, i.e. Δt[f(i;n)] > 0.
An overall system for transmitting the plurality of information signals 301-i in flexible
multiplexing is shown in summary in Fig. 5.

In Fig. 5, the system is represented for three services by way of example. The encoders
304-i transmit the necessary status data regarding the incoming information signals 301-i
to the predictor blocks 302-i, so that a data-rate estimation as was previously described
may be performed. On the basis of this data-rate estimation, the future delta-T values are
subsequently calculated. They are integrated into a bit stream by a DVB-H multiplexer 502
and are subsequently transferred to DVB-H terminals, for example. To this end, the
encoders 304-i may perform, via the time-slice structurer 308, short-term changes in the
DVB-H time-slice partitioning. The block 308 may be designed as a central unit, as is
depicted in Fig. 5, or it may be designed to be decentralized. I.e., the time-slice structurer
308 might be split up among the individual service providers, so that each service provider
(restructures its associated time slice - however, while taking into account the known past
or estimated restructuring of the respective other services that is still to be performed (see
equation (6)). Thus, in this case, at least linking of the decentralized restructuring units will
be required. In each case, boundary conditions and specifications as have been previously
described by way of example are observed in the redistribution.
Each redistribution is initiated by an individual encoder 304-i. In principle, an initiating
encoder may select between an extension and a shortening of a time slice. With the aid of
the payout function u[.], which was defined above, the encoder 304-i may determine the
cost and/or the benefit of a time-slice extension or shortening. In this manner, it has an
assessment tool available to it for controlling its own data-rate requirement.
The proposed inventive concept shortens a latency period or end-to-end delay TL, which
was already described at the outset with reference to Fig. 9, such that the moment when the
actual data-rate splitting must be known is moved closer to the reproduction time of the
actual payload data. This relationship is shown in Fig. 6.
Fig. 6 shows an information signal 301-i split up into various information signal portions
N-1, N, N+1, N+2,.... As for the information signal portion N, one may recognize that the
moment at which the data-rate analysis 602-N of the Nth information signal portion is
completed has moved closer to the reproduction time TN of the N,h time slice as compared
to Fig. 9 described at the outset. This is due to the fact that a data-rate analysis 602-(N+1)
is not needed for the subsequent information signal portion N+1 at the reproduction time
TN, but that, instead, an estimation 102, 104 of the delta-T values to be transmitted within
the time slice N is performed for the portion N+1. This estimation is less costly and does
not require the subsequent information signal portion N+1, but may be performed, as was

already described, by using past and current information signal portions (equations (1) and
(2)).
Finally, it shall be noted that the bit stream generated by the inventive concept may be
processed in a fully transparent manner by a terminal (e.g. DVB-H terminal) conforming to
standard.
The inventive concept introduced may be employed, for example, in a data-rate trading
system as will be described below. In such a system, data rates within a network, in
particular within a DVB-H network, are distributed among information signal and service
providers of the network via a trading system, similar to a market place. A trading platform
with data-rate acquisition devices or software agents and a data-rate allocation device or a
data-rate agent is used for controlling data rates of the individual information signal
providers. Participants in the trading platform are the software agents or the data-rate
acquisition devices. On behalf of the information signal providers, they take on trading
with data rates of the multiplex, and in this manner they acquire transmission capacities for
the information signal services assigned to them.
It shall be noted that the present invention is not limited to the respective components of
the devices or to the procedure explained, since said components and methods may vary.
The terms used here are merely intended to describe specific embodiments, and are not
used in a limiting sense. If quantities or indefinite articles are used in the description and in
the claims, they also refer to the plurality of said elements, unless clearly otherwise
indicated by the overall context, and vice versa.
Even though DVB-H networks are particularly suitable for the inventive concept since
DVB-H transmits any data streams in the form of IP datagrams, the inventive concept is
not limited to DVB-H networks.
In particular, it shall be noted that, depending on the circumstances, the inventive concept
may also be implemented in software. The implementation may be on a digital storage
medium, in particular a DVD, CD or disk with electronically readable control signals that
may cooperate with a programmable computer system and/or microcontroller such that the
corresponding method of transmitting the plurality of information signals in flexible
multiplexing is performed. Generally, the invention thus also consists in a computer
program product having a program code, stored on a machine-readable carrier, for
performing the inventive method, when the computer program product runs on a computer
and/or microcontroller. In other words, the invention may thus be realized as a computer

program having a program code for performing the method, when the computer program
runs on a computer and/or microcontroller.

We claim:
1. An apparatus (300) for transmitting a plurality of information signals (301-i) in
accordance with cyclic time-division multiplexing, comprising:
an estimator (302) configured to estimate, for each of the information signals (301-
i), a data rate (ri[n]; r[f(i;N+1)]) with which the information signal probably will be
encoded within a time slice (Bi[n+1]; B[f(i;n+1)]) which follows the current time
slice (Bi[n]; B[f(i;n)]);
a processor (306) configured to determine, for each of the information signals (301-
i) and on the basis of the estimated data rate (ri[n]; r[f(i;n+1)]), a relative waiting
time (delta-T) which indicates an estimated starting time (Ti[n + l]) of the
subsequent time slice (Bi[n+1]) of the information signal (301-i); and
a time-slice structurer (308) configured to assign an actual starting time (Ti[n+1];
T[f(i;n+1)) to each of the following time slices (Bj[n+1]; B[f(i;n+1)]) on the basis
of an actual data rate (reff,i[n]; reff[f(i;N+1)]) for the information signal (301-i), it
being possible for the actual data rate (reff,i[n]; reff[f(i;N+1)]) to deviate from a data
rate (ri[n]; r[f(i;N+1)]) estimated by the estimator (302),
each of the actual starting times (Ti[n+1]; T[f(i;n+1)) of the subsequent time slices
being larger than or equal to the respectively estimated starting time (Ti [n + 1]).
2. The apparatus as claimed in claim 1, wherein the processor (306) is configured to
determine the relative waiting time (delta-T) such that an estimated starting time
(Ti [n + 1]) and an actual starting time (Ti [n+1]; T[f(i;n+1)) of the subsequent time
slice (Bi[n+1]) perfectly coincide if the estimated data rate (ri[n]; r[f(i;N+1)]) and
the actual data rate (reff,i[n]; reff[f(i;N+1)]) of all of the information signals do not
differ from each other.
3. The apparatus as claimed in claims 1 or 2, wherein the processor (306) is
configured to integrate each of the determined relative waiting times (delta-Ti[n]) in
a current time slice (Bi[n]) of an associated information signal (301-i) so as to
transmit them to a receiver.
4. The apparatus as claimed in any of the previous claims, wherein the estimator (302)
is configured to estimate a data rate (reff[f(i;N+1)]) of an information signal (301-i)

for the following time slice (Bi[n+1]) at least on the basis of preceding and/or
current actual data rates (reff[f(i;n-1)]); reff[f(i;n)]) of the information signal.
5. The apparatus as claimed in any of the previous claims, wherein the estimator (302)
is configured to estimate a data rate (reff[f(i;N+1)]) of an information signal (301-i)
for the following time slice (Bi[n+1]) on the basis of the estimated data rate, the
actual data rate and a data rate of the information signal that is preferred for the
subsequent time slice.
6. The apparatus as claimed in any of the previous claims, wherein the time-slice
structurer (308) is configured to assign to each of the current time slices (B[f(i;n)])
an actual starting time and an actual data rate (reff[f(i;n)]) that may differ from a
previously estimated data rate (reff[f(i;n-1)]), and to change a duration of a current
time slice by a differential time duration (Δt[f(i;n)]) only when an estimated
duration of the subsequent time slice (B[f(i;n+1)]) is larger than the differential
time duration (Δt[f(i;n)]).
7. The apparatus as claimed in claim 6, wherein the time-slice structurer (308) is
configured to determine the differential time duration (Δt[f(i;n)]) as a function of a
differential time duration (Δtw[f(i;n)]) resulting from a preferred data rate for the
current time slice, of a differential time duration (Δtw[f(i;n-1)]) resulting from a data
rate that is preferred for the preceding time slice, and of a delay (ΔT[f(i;n)]) of the
actual starting time (T[f(i;n)]) as compared to a starting time (Ti[n]) estimated in
the preceding transmission cycle.
8. The apparatus as claimed in any of the previous claims, wherein the time-slice
structurer (308) is configured to assign to each of the subsequent time slices
(B[f(i;n+1)]) a differential time duration (Δt[f(i;n+1)]) in accordance with the actual
data rate (reff{f(i;N+1)]) of the information signal (301-i) so as to obtain the actual
starting times (T[f(i;n)]) in each case, the assigned differential time duration
(Δt[f(i;n+1)]) being dependent on a preferred differential time duration
(Atw[f(i;n+1)]) of the subsequent time slice and of a preferred differential time
duration (Δtw[f(i;n)]) of the current time slice (B[f(i;n)]) of the information signal.
9. The apparatus as claimed in claim 8, wherein the time-slice structurer (308) is
configured to select the differential time duration (Δt[f(i;n)]) of a time slice
(B[f(i;n)]) of an information signal to be larger than a negative accumulated overall

delay time (-ΔT[f(i;N+1)]) of temporally preceding time slices of a transmission
cycle.
10. The apparatus as claimed in claims 8 or 9, wherein the time-slice structurer (308) is
configured to select the differential time duration (Δt[f(i;n)]) of a time slice
(B[f(i;n)]) of an information signal to be smaller than a maximally possible delay
time (ATmax[f(i;n)]) of the time slice, the maximally possible delay time depending
on a size of a jitter buffer associated with the information signal.
11. The apparatus as claimed in any of the previous claims, wherein the time-slice
structurer (308) is configured to set each of the actual data rates (reff[f(i;N+1)]) in
the subsequent transmission cycle as a function of the information signal to be
transmitted in the subsequent transmission cycle and of data-rate cost, respectively.
12. The apparatus as claimed in claim 12, wherein the time-slice structurer (308) is
configured to credit each of the information signals (301-i) with a value (u[.])
dependent on the data-rate cost (c) when the time slice (B[f(i;n)]) of the information
signal (301-i) is shortened, and to subtract a value (u[.]) dependent on the data-rate
cost when the time slice (B[f(i;n)]) of the information signal is extended.
13. The apparatus as claimed in any of the previous claims, wherein the current and the
subsequent transmission cycles comprising the current and the subsequent time
slices are DVB-H transmission cycles in each case.
14. A method (100) for transmitting a plurality of information signals (30l-i) in
accordance with cyclic time-division multiplexing, comprising:
estimating (102), for each of the information signals (301-i), a data rate (ri[n];
r[f(i;n)]) with which the respective information signal probably will be encoded
within a time slice (Bi[n+1]; B[f(i;n+1)]) which follows the current time slice;
determining (104), for each of the information signals (301-i) and on the basis of
the estimated data rate (ri[n]; r[f(i;n)]), a relative waiting time (delta-T) which
indicates an estimated starting time (Ti[n + 1]) of the subsequent time slice
(Bi[n+1]; (B[f(i;n+1)]) of the information signal; and

assigning (106) an actual starting time (Ti[n+1]) and an actual data rate (reff,i[n];
reff[f(i;N+1)]) to each of the following time slices (Bi[n+1]; B[f(i;n+1)]), it being
possible for the actual data rate to deviate from an estimated data rate,
each of the actual starting times (Ti[n+1]) of the subsequent time slices being larger
than or equal to the respectively estimated starting time (Ti[n + 1]).
15. A computer program for performing the steps of the method as claimed in claim 14,
when the computer program runs on a computer and/or microcontroller.

ABSTRACT

Exact data-rate analysis of the information signal portions to be transmitted in a subsequent
transmission cycle per time-division multiplexing process shall be initially omitted.
Instead, on the basis of highly accurate estimated values for the subsequent data rates,
estimated values for relative waiting times (delta-T) are transmitted, in a current
transmission cycle, from the current time slice to the subsequent time slice of the same
service. In the subsequent transmission cycle, actual data rates may be set which may
deviate from the estimated data rates for the individual information signals, as a result of
which predicted time-slice boundaries for the subsequent transmission cycle may shift.
However, the potential shift in the time-slice boundaries is subject to several boundary
conditions. No time slice of the subsequent transmission cycle can start prior to its signaled
estimated starting time. In the event of constant data rates, the estimated time-slice
structure and the actual time-slice structure are identical, so that in this case, the concept
introduced exploits the advantages and the efficiency of time slicing.