Background It is generally desirable to reduce the data rates for transmitting multimedia data. Attempts are thus also made to minimize the data volume of audio data which are to be transmitted, which are generated within the framework of MPEG audio methods, for example. As an example, the AAC (Advanced Audio Coding) method is considered here, which is used, inter alia, in MPEG-2 AAC, specified in ISO/IEC 13818 Part 7.

Invention The object of the present invention is to present methods and apparatuses for sending and for receiving multimedia data which ensure a reduced data transmission rate.

The invention achieves this object by means of a method for sending multimedia data which are transmitted in time blocks through selection of at least one first time block as a basis for an extrapolation, selection of at least one second time block of data which are to be estimated, sending of the data in the at least one first time block, and sending of representative data for the extrapolation, which have a reduced data volume as compared with the data which are to be estimated, during at least the second time block.

Correspondingly, an apparatus for sending multimedia data which can be transmitted in time blocks is obtained, having a transmission device for sending the data which are to be transmitted, having a selection device for selecting at least one first time block as a basis for an extrapolation and at least one second time block of data which are to be estimated, where the transmission device can transmit, during at least the second time block, representative data for the extrapolation which have a reduced data volume as compared with the data which are to be estimated.

At the reception end, the invention provides a method for receiving multimedia data which are transmitted in time blocks through reception of data in at least one first time block, reception of representative data in at least one second time block, extrapolation of the data in the at least one second time block on the basis of the data in the at least one first time block and the representative data.

In this regard, the invention finally provides an apparatus for receiving multimedia data which can be transmitted in time blocks, having a reception device for receiving data in at least one first time block and representative data in at least one second time block, and an extrapolation device for extrapolating the at least one second time block on the basis of the data in the at least one first time block and the representative data.

Preferably, the multimedia data comprise audio data and particularly those which are used in the AAC method.

The representative data may comprise a"null signal"or may be made up exclusively therefrom, with just a zero or a correspondingly different signal with minimum information content being transmitted. Transmission of

the null signal in selected time blocks reduces the overall data rate accordingly. The representative data may comprise additional information for the extrapolation. By way of example, it would merely be possible to transmit, instead of the data which are originally to be sent, power information and/or the envelope of the time signal for these data for the extrapolation. This allows the data to be extrapolated more exactly.

At the transmission end, coding and decoding allows an alias component which is obtained on the basis of the data reduction to be ascertained and transmitted. This is found to be advantageous particularly for coding methods in which the data to be transmitted are spread and hence overlaps arise.

In addition, the data for the at least one second time block may be extrapolated at the transmission end on the basis of the data in the at least one first time block and a corresponding signal for the difference between the extrapolated data and the data in the at least one second time block can be transmitted. This has the advantage that the receiver's signal acquisition is simulated at the transmission end and hence the signal to be sent can be optimized.

At the reception end, a signal for an alias component which has been obtained on the basis of the data reduction can be used for the extrapolation. This makes it possible to take into account the circumstance that the transmission end is using spreading methods for data transmission.

The extrapolation at the reception end may also take into account a signal for the difference between data extrapolated at the transmission end and the original data in the at least one second time block. As indicated above, the receiver simulation in the

transmitter allows the artefacts brought about by the receiver to be stopped to a very large extent.

Brief Description of the Drawings The present invention is now explained in more detail with reference to the appended drawings, in which: Figure 1 shows a synthesis window function for an MDCT; Figure 2 shows a block diagram of a transmission system based on the invention comprising a transmitter and a receiver ; Figure 3 shows a block diagram for a transmission system'based on the invention with alias compensation ; Figure 4 shows a block diagram for a transmission system based on the invention with difference signal coding ; and Figure 5 shows a data transmission system based on the invention with parametric coding for chosen frames.

Exemplary Embodiments The invention is based on the concept of one or more chosen time blocks (frames) not being transmitted, or being transmitted differently, e. g. using a null signal, during coding in order to reduce the data rate for the audio coding. A prerequisite for this is the use of a variable bit rate or the use of a buffer, e. g. bit cache in the case of MP3. If a null signal is transmitted in a time block, then in the case of simple coding the output signal in this time block is likewise set to zero. Accordingly, the output samples would be

set to zero upstream of the inverse transformation or the synthesis filter in the corresponding frame. For the audio coding using the AAC method, an MDCT (Modified Discrete Cosine Transformation) is used in order to transform the audio signal to the frequency domain. In this context, the synthesis filters overlap, as indicated in Figure 1. The drawing shows the time blocks N-2 to N+5. Under each block, a respective windowed signal is shown as an arc. The data in a time block are spread, in line with the MDCT, over two time blocks, which means that there is a 50% overlap with the signal in the next time block. In the present case, 1024 samples are spread over 2048 frequency lines, for example. If the intention is now to transmit a respective null signal in block N and block N+4, the decoder will produce a signal which is shown in bold in Figure 1. This means that the decoder also produces a signal during block N or N+4. This signal results from the preceding blocks N-1 and N+3 (alias components), which are"injected"into the null frames N and N+4. In the subsequent time block N+4 or N+5, the information to be transmitted there is gradually overlaid. Details in this regard can be found in the corresponding standards. The overlap in the windowed signals results in the alias errors in the region of the time blocks N and N+1 or N+4 and N+5 in which the null signals are actually decoded.

It is now an aim of the invention to smooth the decoder signal (bold line), which has intermittent dips on account of the transmission of null signals, by means of extrapolation.

As a result of the extrapolation, the chosen frames which have been set to zero are reconstructed on the basis of one or more preceding and/or one or more succeeding frames. To this end, it is possible to use the extrapolation methods known in the literature (concealment methods). One simple method would be block

repetition possibly with correct-phase transition, for example. A further extrapolation method which may be used here is described in the paper by Kauppinen et al.

"A Method for Long Extrapolation of Audiosignals", J. <BR> <BR> <P>Audio Eng. Soc. , Vol. 49, No. 12, December 2001, pages 1167 ff. In this context, all of the information for the extrapolation is taken from the signal itself, which means that no additional information about the signal content is required. Even signals which are subject to interference can be extrapolated, since the errors resulting from the interference are relatively small. The envelopes for the signal amplitudes may be polynomial or exponential functions.

A further extrapolation method is described in the <BR> <BR> paper by R. Sottek, "Kombination einer hochauflösenden Spektralschatzung mit einer Analyse der Einhullenden der Zeitfunktion"Combination of high-resolution spectral estimation with analysis of the envelope of the time function, DAGA 91-Bochum 1991, part B, pages 801 ff. In that case, the time signal is extrapolated appropriately through discrete Fourier transformation (FFT) of"correctly decoded"frames with subsequent estimation of the spectral lines (inverse convolution), subsequent convolution with a longer window and inverse, likewise longer transformation.

In line with the invention the extrapolation is thus carried out not only in cases of error, but rather is specifically controlled by an encoder 1 as shown in Figure 2. To this end, a selection unit 2 is used to select those frames or time blocks which can readily be extrapolated at the reception end. This selection information is transmitted directly or indirectly in the bit stream to a decoder 3. It can be transmitted, for example without additional bits, by a unique bit stream in which the scale factors are chosen equal to zero.

The decoder 3 decodes the data which are usually transmitted and makes them available as output signals (output). For those time blocks or frames in which the null signal has been transmitted, an extrapolation unit 4 produces an extrapolation signal and admixes it with the output signal. In this way, the dips in the decoder signal (bold line in Figure 1) can be compensated for by the extrapolation signal. This means that the extrapolation signal and/or the residual signal from the window overlap result in the new signal for frames N and N+1, for example.

The frame selection by the selection device 2 should take into account the decoder signal's output quality which is to be expected.

The selection information in the transmitted bit stream could be used to transmit further additional information, such as the power, the envelope of the time signal, which can be calculated using Hilbert transformation, for example. The decoder 3 or the extrapolation device 4 can then use this additional information to carry out level correction for the extrapolated data, for example.

The transmission system shown in Figure 2 can be extended in line with Figure 3 by alias compensation.

To this end, a decoder 5 at the transmission end simulates the decoding operation which is to be carried out in the receiver. Alias components can be ascertained from the data obtained from the decoder 5 and from the input data (input) for the coder 1. These alias components are coded in an alias encoder device 6 and are additionally transmitted to the receiver for the purpose of quality improvement for the output signal at the receiver end. An appropriate alias decoder device 7 decodes the alias components in the receiver. These are finally taken into account for the output signal from the receiver (output).

For the purpose of simulating the conditions in the receiver fully in the transmitter itself, an extrapolation device 8 may also be provided in the transmitter, in line with Figure 4. It is thus possible to produce a different signal from the original input signal and the decoding signal (complemented through extrapolation) from the decoder 5. This difference signal is coded in a difference signal encoder 9 at the transmission end and is transmitted to the receiver.

There, it is decoded in a difference signal decoder and is added to the decoding signal (complemented through extrapolation) from the decoder 3 in order to obtain the output signal from the receiver. Hence, a frame missing in the output signal from the decoder 3 is replaced using the extrapolation device 4 and the result is improved using the difference signal.

The extrapolation can be refined by transmitting the envelope of the original input signal. If Sottek's extrapolation is used, the estimation methods for the spectral lines could be optimized using new algorithms, which are used in the parametric coding in MPEG-4 (Parametric Audio Coding HILN, ISO/IEC 14496-3: 2001 (E) MPEG-4 Audio, Subpart 7).

A further embodiment of the inventive method or of the inventive apparatus is shown symbolically in Figure 5.

The receiver's encoder 1 estimates the frames which still exist and have been selected. Next, the harmonic and individual spectral lines and the envelope of the residual noise are transmitted as parameters. To this end, they are coded in a parameter encoder 11 at the transmitter end and are decoded in a parameter decoder 12 at the reception end. The output signal from the receiver is thus synthesized from the signal from the decoder 3 and from the parameter decoder 12. The components in the embodiments described above may be combined with one another as desired.