Apparatus and method for transmitting a sequence of data
packets and decoder and apparatus for decoding a sequence
of data packets
Description
The present disclosure refers to data communication
applications and, in particular, to real-time communication
of audio signals via packet-oriented networks.
In real-time communication via packet-oriented networks,
like, for example, voice over IP (VoIP), normally it cannot
be ensured that all packets will arrive at the receiver in
the time required. The reason for this is that, when data
are transmitted in a packet-oriented manner, like, for
example, via the Internet, different packets take different
paths via the data network, and the paths taken via the
data network which the different packets may take from the
transmitter of the packets to the receiver of the packets
depend on the current network situation.
Even if packets are sent in the order in which they were
generated, the order of arrival at the receiver will differ
with high probability. A packet which has found a favorable
path may even "overtake" a packet which was sent at an
earlier time but has taken a longer path from the
transmitter to the receiver.
For applications which do not require real-time operation
or for real-time applications where relatively great delays
are allowed, this is not a problem since a receiver buffer
will simply buffer until all the packets of a sequence of
packets have arrived. Where a packet is positioned in a
sequence may exemplarily be determined by a packet number
or by a packet sequence indication at the packet and the

receiver will then sort the packets in the correct order
before reproducing them or passing them on.
However, the smaller this buffer is selected to be or the
smaller the allowable delay a packet may have when being
transmitted from the transmitter to the receiver is, the
greater the packet failure rates will be. Packet losses
will not only result when a packet has really been lost,
but packet losses will also result when a packet has
required too long a time to get from the transmitter to the
receiver. Another problematic situation is when a packet
has experienced data corruption on its way from the
transmitter to the receiver, i.e. really is faulty.
Such delay-critical applications are present in internet
telephony (Voice over IP) where the delay requirements
packets from the transmitter to the receiver have to meet
are relatively strict in order for the call not to be
interrupted. In particular when an audio encoder is
arranged on the transmitter side and a decoder is arranged
on the receiver and when there are no more data packets to
be encoded at the receiver, the result may be a call
interruption, namely when the decoder on the receiver side
"crashes" due to the lack of data present.
It is the object of the present invention to provide a
concept for communicating data packets which, despite
critical requirements, provides good communication quality.
This object is achieved by an apparatus for transmitting a
sequence of data packets of claim 1, a method for
transmitting a sequence of data packets of claim 19, a
decoder for decoding a sequence of data packets of claim
20, a method for decoding data packets of claim 24, or a
computer program of claim 25.
In order to avoid crashing on the decoder side and/or avoid
artefacts audible for a user on the decoder side, according

to the invention, a detection whether a packet of the
sequence of packets is missing or faulty is performed. If
such a packet loss or packet error has been determined, a
replacement packet will be provided, the replacement packet
being a valid packet with regard to the packet syntax,
wherein, however, audio contents of the replacement packet
have a certain contents characteristic. The apparatus for
transmitting the sequence of data packets according to the
invention outputs an undisturbed sequence of packets,
wherein, however, faulty packets or packets not received
have been replaced by replacement packets so that the
sequence output by the apparatus for transmitting comprises
at least one packet received and one replacement packet. In
one implementation, the contents and/or the contents
characteristic of a packet or frame in a replacement packet
is independent of the audio signal, i.e. does not depend on
the preceding or subsequent packet or frame. However, if
the packet is subjected to an error concealment measure,
the synthesis audio contents will depend on the preceding
or subsequent frame, i.e. will not be predetermined or
signal-independent.
Additionally, the replacement packet gives, in a payload
region, an indication for the fact that it is a replacement
packet, wherein the indication is ignorable or
interpretable by a basic decoder such that the replacement
packet will be decoded like a valid packet according to the
predetermined contents characteristic, and wherein the
indication is interpretable by an extension decoder which,
compared to the basic decoder, has an extended
functionality to perform an error concealment measure which
is generating contents for the replacement packet which has
a contents characteristic differing from the predetermined
contents characteristic. The replacement packet may be a
pure payload packet or may be a packet comprising a payload
portion and a header portion, the indication being present
not in the header portion, but preferably in the payload
portion.

A receiver of the sequence of data packets receives an
undisturbed stream of data packets all having a valid data
syntax. This receiver will be able to decode the sequence
of data packets easily. The receiver, when being a basic
receiver, will thus easily decode a replacement packet and
reproduce the predetermined audio contents. However, this
will result in a loss in quality since the audio contents
are predetermined, and thus will not fit the preceding
packet or frame and the subsequent packet or frame
perfectly. However, such a short temporary loss in quality
is unproblematic compared to a situation where the decoder
crashes completely due to a packet not being present and
thus the entire communication connection be interrupted.
An extension decoder in contrast is able to recognize,
using an indication in the replacement packet pointing out
that the packet is a replacement packet and not a normal
packet, this fact and not simply process the replacement
packet, but initiate an error concealment measure when a
replacement packet has been received.
Error concealment measures exemplarily include, apart from
frame/packet repetitions, extrapolations of previous
packets or subsequent packets already present in a memory
or interpolations between previous and subsequent packets.
Extrapolations or interpolations of this kind in particular
include band-wise energy measurements to then synthesize,
in an error concealment case, a short-time spectrum of an
audio signal the spectrum values of which have been
generated in a more or less random manner, wherein,
however, the band-wise energy of the randomly generated
audio signal depends on the energies of the previous and/or
subsequent correctly received packets or packets already
generated by error concealment measures.
In one embodiment, the predetermined contents
characteristic the replacement packet has is a zero

spectrum. The result here is that a basic decoder which
decodes the predetermined replacement packet performs
"muting". Alternatively, the contents characteristic may be
an audio signal the audio contents and/or the spectral
values of which correlate with the absolute listening
threshold and exemplarily are defined in a way that the
audio contents are smaller than two times the absolute
listening threshold so that a certain, but small amount of
noise is contained across all the bands, which under
certain circumstances may, subjectively, sound better than
simply "muting".
Embodiments of the present invention will be detailed
subsequently referring to the drawings, in which:
Fig. 1 shows a block circuit diagram of an
implementation of the apparatus for transmitting
a sequence of data packets;
Fig. 2 is a block circuit diagram of a decoder for
decoding a sequence of data packets;
Fig. 3 is a block circuit diagram of an audio encoder
for generating a sequence of data packets;
Fig. 4 shows, among other things, a more special
implementation of the audio rendering means of
Fig. 2;
Fig. 5A shows a sequence of data packets as are output by
an encoder;
Fig. 5B shows a sequence of data packets as are received
by a base station;
Fig. 5C shows a sequence of data packets as are output by
a base station and received by a mobile unit, a
replacement packet having been inserted;

Fig. 5D shows a sequence of audio contents as are
generated in audio rendering means;
Fig. 6A shows a syntax of extension_payload according to
ISO/IEC 14496-3:2005(E) MPEG4;
Fig. 6B shows a table for illustrating the values for the
extension_type field;
Fig. 7 shows an exemplary packet syntax for a transform-
based audio encoder/decoder;
Fig. 8 shows an example of a replacement packet
comprising a valid packet syntax and a
predetermined contents characteristic.
Fig. 1 shows an apparatus for transmitting a sequence of
data packets which represent an audio signal. The apparatus
for transmitting, which in Fig. 1 is exemplarily
implemented as a base station 10, includes means 11 for
receiving packets of the sequence, a packet syntax being
specified for the packets. The means 11 for receiving is
exemplarily connected to a packet-oriented transmission
network, like, for example, the Internet 12. Additionally,
the base station 10 includes means 13 for detecting whether
a packet of the sequence of packets is missing or faulty.
Furthermore, means 14 for providing a replacement packet is
provided in order to replace the faulty packet or missing
packet. The replacement packet is a valid packet with
regard to the packet syntax, however the audio contents of
the replacement packet have a predetermined contents
characteristic. In addition, the base station includes
means 15 for outputting the sequence of packets, the
sequence of packets comprising at least one received packet
and one replacement packet. Using the example of a base
station, the means 15 for outputting is an HF front end
which is coupled to an antenna 16 for transmitting in

accordance with a predetermined specification, like, for
example, the NG DECT specification, a sequence of data
packets to a mobile unit which will be discussed in greater
detail below referring to Fig. 2.
The means 11 of Fig. 1 is coupled to the means 15 for
outputting via a packet line 17 to transmit a normal, i.e.
not faulty, packet received on time to the means 15. In
addition, the means 14 for providing a replacement packet
is connected to the means 15 for outputting via a
replacement packet line 18. Preferably, the means 14 for
providing a replacement packet includes a memory where a
replacement packet is stored. Whenever a missing or faulty
packet has been recognized, the means 14 enables memory
access to recover the replacement packet from the memory
and feed same via the line 18 to the means 15 for
outputting.
In one embodiment, the means 13 for detecting is controlled
so as to detect packet loss and enable the means for
providing a replacement packet when no packet in a sequence
of packets has been received for a longer time than a
maximum delay. In one embodiment, this maximum delay is
exemplarily controllable via a control line 19. The means
13 for detecting may be fed, via the control line 19, a QoS
(quality of service) request which in a VoIP application
exemplarily includes a maximum delay. For other packet-
oriented applications, except for VoIP, there may be
different QoS requests which are fed to the means 13 for
detecting via the control line 19. Alternatively, the means
13 for detecting may also have a fixedly set criterion in
dependence on which a packet error or packet loss is
detected and the means 14 for providing the replacement
packet is enabled.
In one embodiment, the replacement packet which is provided
via the replacement packet line 18 to the means 15 for
outputting does not only have a predetermined contents

characteristic, but also an indication for the fact that
the packet is a replacement packet. In one embodiment, the
indication is such that a basic decoder receiving the
replacement packet ignores the indication and decodes the
packet like a valid packet according to the predetermined
contents characteristic and such that an extension decoder
which, compared to the basic decoder, has an extended
functionality, interprets the indication to perform an
error concealment measure which is generating contents for
the replacement packet differing from the predetermined
contents characteristic.
Fig. 2 shows a decoder for decoding a sequence of packets
which may exemplarily be located in a mobile unit 20. The
decoder includes a receiver 21 for receiving the sequence
of packets for which a packet syntax is specified, the
sequence comprising at least one data packet and at least
one replacement packet, the replacement packet being a
valid packet with regard to the packet syntax, and audio
contents of the replacement packet comprising a
predetermined contents characteristic. Additionally, the
replacement packet includes an indication for the fact that
it is a replacement packet. However, this indication can
only be interpreted by the extension decoder illustrated in
Fig. 2, whereas this indication cannot be interpreted by a
basic decoder.
The interpretation of the indication for a replacement
packet is done by a detector 22 which is implemented to
detect whether a packet contains the indication and thus is
a replacement packet. The extension decoder in Fig. 2
further includes error concealing means 23 for synthesizing
synthesis audio contents comprising a contents
characteristic differing from the predetermined contents
characteristic contained in the replacement packet. In
addition, the extension decoder includes audio rendering
means 24 for rendering the audio contents of a packet when
the packet is not a replacement packet and for rendering

the synthesis audio contents when the packet is a
replacement packet. The audio rendering means 24 is coupled
to an output unit 25 which may exemplarily include a D/A
converter, an amplifier and a speaker.
In particular, the audio rendering means is coupled to the
receiver 21 via a packet line 26 via which a regular packet
is transmitted. In addition, the audio rendering means 24
is connected to the error concealing means 23 via a
replacement packet line 27 via which synthesis audio
contents are transmitted from the error concealing means 23
to the audio rendering means 24.
The error concealing means 23 may perform error concealing
in different manners. Simple error concealing means is
simply repeating the previous frame and/or audio contents
of the previous packet or the audio contents of the
subsequent packet. This error concealing measure is
referred to as "frame repetition". Alternatively, the error
concealing means may be implemented to perform
extrapolation or interpolation. The extrapolation or
interpolation may be performed relative to spectral values
or bands. In the case of a spectral value extrapolation, a
spectral value for a replacement frame may be formed on the
basis of one or several spectral values having equal
frequencies of previous frames. Alternatively, error
concealing may also be performed with regard to bands in
that the spectral values in one band are exemplarily
generated by a buffer as a random number generator or
generated in a more or less determined way and that the
spectral values are then weighted such that the energy they
represent equals a target energy, this target energy having
been derived from one or several previous frames and/or one
or several subsequent frames. The previous and/or
subsequent frames may be valid received packets or may be
packet and/or frames having already been generated by error
concealing means when packet loss has occurred where not

only a single packet was lost, but several successive
packets were lost.
In an embodiment of the present invention, the audio
encoder generating the sequence of packets is a transform-
based audio encoder. Such a transform-based audio encoder
comprises a time-frequency converting stage 30 by means of
which a temporal audio signal is converted to a sequence of
short-time spectra. Every short-time spectrum is fed to a
quantizer 31 performing quantization which is controlled by
a psycho-acoustic model 32 such that quantization is
executed such that quantizing noise does not disturb the
subjective audio impression. Downstream of the quantizer is
an entropy encoder 33 which may exemplarily be a Huffman
encoder. The entropy encoder provides a sequence of bits
which in connection with side information which are
exemplarily provided by the quantizer 31 in the form of
scaling factors and exemplarily provided by the entropy
encoder 33 in the form of the encoding table used, form
data provided to a packet packer 34 outputting the sequence
of data packets on the output side. Apart from the packet
packer 34, the audio encoder shown in Fig. 2 represents a
typical transform-based encoder, as is exemplarily known
under the keyword MP3 (MPEG-1 Layer 3) or AAC (MPEG-4) or
AC-3 etc. It is to be pointed out that the packet packer
34, depending on requirements, is implemented to generate
one packet per audio frame, i.e. per short-time spectrum,
or generate more than one audio frame, i.e. bring several
encoded short-time spectra into a single packet.
Fig. 4 shows a more detailed representation of the audio
rendering means 24 of Fig. 2 and, in particular, the
cooperation of the audio rendering means 24 and the error
concealing means 23. On the input side, the audio rendering
means includes a packet unpacker 40 which unpacks a packet
so as to separate the side information from the "main
information".

The main information, i.e. the short-time spectra
represented by a sequence of bits, are fed to an entropy
decoder 41 which provides quantizing indices which are fed
to a de-quantizer 42 providing at its output quantized and
again de-quantized spectral values which are then used,
after a frequency-time conversion in block 43, to generate
the output audio signal. Both the entropy decoder 41 and
the de-quantizer 42 may be controlled by the side
information, the entropy decoder typically receiving a code
table index, whereas the de-quantizer 42 receives scaling
factors for performing correct de-quantization.
The packet unpacker 40 will then be able, when comprising a
detection characteristic as is present in the detector 22
of Fig. 2, to send the concealing means 22 a replacement
packet indication such that the concealing means 23 can
recognize that the current packet is not a packet generated
by the encoder of Fig. 3, but a replacement packet
generated by the base station. In this case, the concealing
means will with provide the entropy decoder either a
sequence of bits or provide the de-quantizer with a
sequence of quantizing indices or will provide the
frequency-time converting means with a sequence of spectral
values to feed the synthesis audio contents to the decoder
functional chain at any location. Synthesized spectral
audio contents are preferably fed at the end of the chain,
i.e. at the frequency-time converting stage 43.
These spectral contents preferably depend on the previous
spectra already received correctly or on the next spectra
which may already be present and may include error
concealing means relating to spectral values, bands or
both, i.e. relating to both spectral values and bands, to
synthesize the audio contents.
As has already been illustrated using Figs. 1 and 2, an
implementation of the invention is utilized in a base
station and a mobile unit, wherein it is assumed that the

base station can receive packets in an incorrect order,
like, for example, when the base station is coupled to the
Internet, whereas the mobile unit is dependent on receiving
a sequence of packets in the correct order. Such a
communication connection between a base station and a
mobile unit is, for example, defined by the DECT standard.
In real-time communication via packet-oriented networks
(e.g. VoIP), it cannot be ensured that all the packets will
reach the receiver in the time required. After a certain
(very limited) time, packets that have not yet arrived have
to be classified as lost. In the case of IP packet loss, a
replacement audio frame is generated and transmitted by the
base station.
Such a replacement will be performed by the base station of
a next generation (NG) DECT system. Instead of the original
(but not received) frames, the replaced frames are
transmitted from the base station to the mobile unit. The
NG DECT specification represents an extension of the
present DECT specification by, among other things, wide-
band and super-wide-band audio codecs and an IP terminal.
An NG DECT station includes one or several wireless
telephones receiving calls from an NG DECT base station.
Consequently, it becomes possible to make VoIP calls
directly via NG DECT phones. In the ideal case, the VoIP
voice packets can be transmitted, without being encoded
again in the base station, from the base station to the
mobile unit.
When, for example, audio codecs are used which in their
regular syntax offer no special way of signalizing a frame
loss, replacement packets are generated. Such a replacement
audio frame should be decodable by a decoder conforming to
standards, but, in one embodiment, should at the same time
offer an extended decoder a way of recognizing
unambiguously that this frame is a replacement frame so

that this extended decoder can enable corresponding
countermeasures, like, for example, error concealing.
Introducing a replacement frame should additionally be done
without greater calculating complexity and, in particular,
without special evaluation of previous audio data by a base
station such that a base station can operate as a pure
relay station, i.e. a station providing data transmission
without decoding and encoding again. Thus, a base station
should, if at all, perform only very little packet
unpacking which may exemplarily only be performed so as to
recover packet sequence information indicating where a data
packet is arranged in the sequence, so that correct
decoding of the audio contents represented by the sequence
of data packets may take place.
In one embodiment, special user-specified data regions are
used by transform-based encoders to provide signalization
for a decoder including extended functionality, i.e. an
extension decoder, indicating that the respective frame is
a replacement frame, even though the underlying bitstream
syntax standard has not provided for such signalization.
When a frame or packet carries a header and a payload
portion, the payload portion containing the useful data, it
is preferred to accommodate the indication in the payload
data, since, in subsequent audio processing when the header
is not considered anyway, the indication would drop out and
thus error concealing would no longer be possible.
When a packet only has a payload portion and no header,
only accommodating the indication in the payload portion
allows implementing the embodiment. Preferably, the
indication is accommodated in the audio data and/or audio
data portion.
In one implementation, the replacement audio frame and/or
replacement packet meets, in the case that no explicit way
of signalizing a frame loss is provided, a first criterion

which is that the replacement frame and/or replacement
packet be decodable by a decoder conforming to standards.
The second criterion is that the replacement frame should
offer an extended decoder a way of recognizing
unambiguously that this frame is a replacement frame so
that this extended decoder can enable error concealing.
A conventional decoder conforming to standards will be able
to decode the replacement frame if the replacement frame is
a valid frame according to the predetermined standardized
data stream syntax and/or packet syntax or frame syntax. In
one implementation, it is preferred to replace the missing
frame by a frame without audio contents, i.e. to perform
so-called muting. Muting means that the entire spectrum is
set to zero. In particular, using the AAC standard (MPEG-4-
Audio), that variation is preferred for generating the zero
spectrum where the highest scaling factor band for which
spectral values are transmitted is set to zero (max_sfb=0).
Alternatively, spectral values for scaling factor bands
which again are zero may also be transmitted. This may in
principle be realized using any available Huffman code book
of the different code books provided, wherein, when using
the code book "ZERO_HCB" (Zero Huffman Code Table), these
spectral lines do not explicitly have to be transmitted.
It is pointed out that the base station which is to detect
the faulty or missing frame may already perform error
concealing measures. However, according to the invention it
is preferred for the base station not to perform such an
error concealing measure since the error concealing measure
is complex with regard to calculation and additionally
requires evaluating the previous and, maybe, future audio
signals. In particular when a DECT base station "serves"
several mobile units, the result of this would be that a
base station, in order to be able to perform complete error
concealing, decodes and encodes again all the audio
contents continuously. Apart from the pertaining
requirement for high processor and memory resources, in

particular when using a lossy encoder, the result would be
the danger of additional quality deterioration due to
tandem encoding effects. Additionally, delay would increase
strongly.
Since, in the AAC standard, no explicit way of signalizing
a frame loss is provided, in one implementation a way of
signalization which is ignored by a regular decoder is
used. The data syntax and/or packet syntax is kept here. On
the other hand, simply setting the spectrum of a frame to
zero is not sufficient for providing a safe indication for
the fact that this is a replacement frame and/or
replacement packet, since frames of this kind may also be
present in regular operation when, for example, there is no
signal at the encoder input.
The indication for the fact that a frame is a replacement
frame and/or a replacement packet provides an extension
decoder with the information that the present frame is not
a frame where the spectrum is really zero, but that the
present frame is a frame having been introduced into a base
station due to transmission errors to avoid failure of the
decoder and/or failure of the voice link.
Audio encoding standards typically provide user-specific
data regions allowing additional payload transmission,
wherein this payload, however, is ignored by conventional
decoders, i.e. basic decoders with no extended
functionality. In the AAC standard, such a user-defined
payload is the so-called "extension_payload", as is defined
in Fig. 6A. Depending on the value of the "extension_type"
variable, as is shown in Fig. 6B, a different purpose is
provided for in the standard. Figs. 6A and 6B are extracts
from the standard ISO/IEC14496-3:2005(E) . The usage of the
fill element (FILL) described there is provided in the
standard for the following purpose. Fill elements have to
be added to the' bitstream when the overall number of bits
for all the audio data together with all the additional

data is smaller than the minimum number of bits in this
frame allowed to achieve the target bit rate. Dynamic range
control bits (DRC bits) are added when the encoder wants to
introduce such DRC information. Under normal conditions,
fill bits are, as the standard says, avoided and the free
bits are used to fill up the bit reservoir and/or bit
savings bank. Only if the bit reservoir is full will fill
bits be written. Any number of fill elements is allowed.
In one implementation of the invention, as is shown in Fig.
7 at 70, the "extension_type" is set to "0000" to write
- in contrast to the standard usage of fill bits - the
indication for a frame loss in the "other_bits" field.
The user-specific data region provided in the standard for
fill bits is used to provide a signalization of the
replacement frame, i.e. to accommodate the indication for a
replacement frame. Depending on the implementation,
however, other "extension_payload ()" may be used,
depending on a different setting of the "extension_type"
value. Since preferably a zero spectrum is transmitted
efficiently (by setting max_sfb to zero) or a Zero Huffman
code book is used, there are sufficient bits available for
various extension_payloads ().
It is to be pointed out that a typical MPEC-4-conforming
data stream and/or packet syntax and/or frame syntax for a
replacement packet is shown at 71, wherein, as is
illustrated at 72, the variation "max_sfb=0" is used.
Preferably, all the other data required for obtaining a
valid packet syntax are also set to zero. However, it is to
be pointed out that the data themselves are no reliable
indication for a replacement frame. Only the
extension_payload 70 will result in a reliable indication
since a normal encoder would not write such a frame with a
zero spectrum, but a special extension_payload.

Fig. 8 shows an exemplary replacement frame for new
generation DECT for a sampling rate of 48 kHz, a mono
signal and a bit rate of 64 kBit/s. It is pointed out that
the extension_payload, as is shown in Fig. 7 at 70, is not
to be seen easily in Fig. 8 since the extension_payload is
not "byte-aligned" in the bitstream.
Also, it is pointed out that the bit combination in the
"extension_payload" comprises a bit pattern which is also
referred to as "error pattern", which corresponds to the
ASCII code for "FRAME_LOSS". It is ensured by this that no
conflict is caused with other users of extension_payload,
since it is highly improbable that another user would use
the ASCII code for "FRAME_LOSS" to signal something which
has nothing to do with "FRAME_LOSS".
Different orders of packets or frames will be discussed
below referring to Fig. 5, as may occur at different points
in the transmission scenario from a transmitter of the
packet to a decoder and/or to audio rendering means in the
decoder.
Fig. 5A shows a sequence of packets of the number (i-1), i,
(i + 1), (i+2). Such a correct sequence of packets or frames
occurs at the input to a packet-oriented transmission
network, like, for example, the Internet.
Fig. 5B shows a sequence of packets at the input to the
base station, wherein it can be seen that the packet i, in
the time sector considered by Fig. 5B, has not yet arrived,
but that the order has become mixed up. The reason for this
is that the packet i has either got lost completely or has
obtained a very long path from the transmitter to the
receiver. On the other hand, the packet (i+2) has obtained
a very favorable path, so that this packet has "overtaken"
the packet (i+1) on its way from the transmitter, i.e. the
encoder, to the base station input.

The means 11 for receiving in Fig. 1 will sort the packets
as they have arrived in Fig. 5B again in the correct order.
Additionally, the means 13 for decoding will find out that
the packet having the number i is not present or is faulty.
For this reason, a replacement packet will be generated for
number i, as can be seen in Fig. 5C. Fig. 5C thus shows the
sequence of packets as is output by the means 15 for
outputting in Fig. 1. On the path from the base station
according to Fig. 1 to the mobile unit according to Fig. 2,
the order of the packets does not change. However, the
extension decoder, as is illustrated in Fig. 2, will
recognize that the packet of the number i is a replacement
packet. In contrast to the other packets shown in Fig. 5D,
for which normal rendering of the audio contents takes
place, synthesis audio contents going back to an error
concealing measure are generated for the replacement
packet.
Depending on the circumstances, the inventive method may be
implemented in either hardware or software. The
implementation may be on a digital storage medium, in
particular on a disc or CD having control signals which may
be read out electronically which can cooperate with a
programmable computer system such that the corresponding
method will be executed. Generally, the invention is also
in a computer program product comprising a program code
stored on a machine-readable carrier for performing the
method when the computer program product runs on a
computer. 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.

We Claim :
1. An apparatus for transmitting a sequence of data
packets representing an audio signal to a basic
decoder or to an extension decoder, wherein the
extension decoder comprises an extended functionality
as compared to the basic decoder, comprising:
means (11) for receiving packets of the sequence, a
packet syntax being specified for the packets;
means (13) for detecting whether a packet of the
sequence of packets is missing or faulty;
means (14) for providing a replacement packet to
replace the faulty packet or the missing packet, the
replacement packet being a valid packet with regard to
the packet syntax, audio contents of the replacement
packet being predetermined audio contents reproducible
by the basic decoder and having a predetermined
contents characteristic and the replacement packet
containing an indication for the fact that it is a
replacement packet, wherein the indication is
implemented such that the indication is ignorable or
interpretable by the basic decoder such that the
replacement packet is decoded like a valid packet
according to the predetermined contents
characteristic, and that the indication is
interpretable by the extension decoder to perform an
error concealing measure which is generating contents
for the replacement packet having a contents
characteristic differing from the predetermined
contents characteristic; and

means (15) for outputting the sequence of packets, the
sequence of packets comprising at least one received
packet and at least one replacement packet.
2. The apparatus of claim 1,
wherein a packet of the sequence receivable by the
means (11) for receiving comes from a transform-based
audio encoder and comprises a short-time spectrum of a
temporal sector of the audio signal, and
wherein the predetermined contents characteristic is
that all the spectral values equal zero or together
represent an energy smaller than two times the energy
represented by a psycho-acoustic resting listening
threshold.
3. The apparatus of claims 1 or 2, wherein the means (14)
for providing comprises:
a memory for storing the replacement packet;
a memory fetcher for fetching the replacement packet
from the memory whenever the means (13) for detecting
detects a faulty packet or a missing packet.
4. The apparatus of one of the preceding claims, wherein
the means (11) for receiving is an input interface of
a telephone base station and is connectable to a
network implemented for packet-based data
transmission.
5. The apparatus of one of the preceding claims, wherein
the replacement packet has a payload region where the
indication is located.
6. The apparatus of one of the preceding claims, wherein
the packet syntax is implemented to define an

extension payload field, and wherein the indication is
formed by data in the extension payload field.
7. The apparatus of claim 6, wherein the data are
generated in accordance with a letter code, the data
having a meaning indicating at a data loss.
8. The apparatus of claim 7, wherein the data represent
the expression "FRAME_LOSS" or "data loss".
9. The apparatus of claims 7 or 8, wherein the letter
code is the ASCII code.
10. The apparatus of claim 1, wherein the sequence of
packets is generated by an audio encoder according to
MPEG-1, layer 3, or MPEG-4 AAC, a bit reservoir
function being disabled.
11. The apparatus of one of the preceding claims, which is
implemented as a base station.
12. The apparatus of one of the preceding claims, wherein
the packet syntax comprises a fill data field which is
filled by a basic encoder when a minimum number of
bytes provided for the frame is not required for
encoding a portion of the audio signal, and
wherein the indication is represented by a
predetermined bit combination in the fill data field.
13. The apparatus of one of the preceding claims, wherein
the means (13) for detecting is implemented to wait
for a predetermined duration for a data packet having
a sequence position indication and to signalize, when
the predetermined duration has passed without
detecting the data packet, a packet loss to the means
(14) for providing.

14. The apparatus of claim 13, wherein the sequence
position information sought is set by a sequence
position indication of a preceding or subsequent valid
packet.
15. The apparatus of claims 13 or 14, wherein the
predetermined duration is settable and predeterminable
by a QoS request (19).
16. The apparatus of claim 15, wherein the predetermined
duration is smaller when the QoS request has a smaller
delay and is greater when the QoS request allows a
greater delay.
17. The apparatus of one of the preceding claims, wherein
the means (15) for outputting is implemented to output
the sequence of packets as an unbroken sequence of
packets and replacement packets representing a
complete continuous sequence.
18. The apparatus of one of the preceding claims, wherein
the means (11) for receiving is implemented to receive
packets but not, or at most partly, to decode the
packets received in order to detect sequence position
information in the packet.
19. A method for transmitting a sequence of data packets
representing an audio signalto a basic decoder or to
an extension decoder, wherein the extension decoder
comprises an extended functionality as compared to the
basic decoder comprising:
receiving (11) packets of the sequence, a packet
syntax being specified for the packets;
detecting (13) whether a packet of the sequence of
packets is missing or faulty;

providing (14) a replacement packet to replace the
faulty packet or missing packet, the replacement
packet being a valid packet with regard to the packet
syntax, audio contents of the replacement packet being
predetermined audio contents reproducible by the basic
decoder and having a predetermined contents
characteristic and the replacement packet containing
an indication for the fact that it is a replacement
packet, wherein the indication is implemented such
that the indication is ignorable or interpretable by
the basic decoder such that the replacement packet is
decoded like a valid packet according to the
predetermined contents characteristics, and that the
indication is interpretable by the extension decoder
to perform an error concealing measure which is
generating contents for the replacement packet having
a contents characteristic differing from the
predetermined contents characteristic; and
outputting (15) the sequence of packets, the sequence
of packets comprising at least one received packet and
at least one replacement packet.
20. An extension decoder for decoding a sequence of
packets, wherein the extension decoder comprises an
extended functionality as compared to the basic
decoder, comprising:
a receiver (21) for receiving the sequence of packets
for which a packet syntax is specified, the sequence
comprising at least one data packet and at least one
replacement packet, the replacement packet being a
valid packet with regard to the packet syntax, audio
contents of the replacement packet being predetermined
audio contents reproducible by the basic decoder and
having a predetermined contents characteristic and the
replacement packet containing an indication for the
fact that it is a replacement packet, wherein the

indication is such that the replacement packet may be
received by the basic decoder and that the replacement
packet may be decoded like a valid packet according to
the predetermined contents characteristics, and that
the extension decoder is implemented to interpret the
indication to perform an error concealing measure
which is generating contents for the replacement
packet having a contents characteristic differing from
the predetermined contents characteristic;
a detector (22) for detecting whether a packet
contains the indication and consequently is a
replacement packet;
error concealing means (23) for synthesizing synthesis
audio contents for the replacement packet the
characteristic of which differs from the predetermined
contents characteristic, using the error conceiling
means; and
audio rendering means (24) for rendering the audio
contents of a packet when the packet is not a
replacement packet, and for rendering the synthesis
audio contents when the packet is a replacement
packet.
21. The extension decoder of claim 20, wherein the
predetermined contents characteristic is a zero
spectrum,
wherein the received packets of the sequence of
packets have been generated by a transform-based audio
encoder, and
wherein the audio rendering means (24) comprises a
transform-based audio decoder (41, 42, 43).

22. The extension decoder of claim 21, wherein the error
concealing means (23) is implemented to generate
synthesis spectral values, and
wherein the audio rendering means is implemented to
convert the synthesis spectral values to a temporal
representation (43).
23. The extension decoder of claims 20, 21 or 22, wherein
the error concealing means (23) is implemented to
generate the synthesis audio contents by extrapolation
of audio contents from a previous or subsequent intact
or concealed audio packet, or
to generate the synthesis audio contents by
interpolation of audio contents of a previous intact
or concealed packet and a subsequent intact or
concealed packet.
24. A method for decoding a sequence of packets in an
extension decoder, wherein the extension decoder
comprises an extended functionality as compared to the
basic decoder, comprising:
receiving (21) the sequence of packets for which a
packet sequence is specified, the sequence comprising
at least one data packet and at least one replacement
packet, the replacement packet being a valid packet
with regard to the packet syntax, the audio contents
of the replacement packet being predetermined audio
contents reproducible by the basic decoder and
comprising a predetermined contents characteristic,
and the replacement packet containing an indication
for the fact that it is a replacement packet, wherein
the indication is such that the replacement packet may
be received by the basic decoder and that the
replacement packet can be decoded like a valid packet
according to the predetermined contents

characteristics, and that the extension decoder
interprets the indication to perform an error
concealing measure which is generating contents for
the replacement packet having a contents
characteristic differing from the predetermined
contents characteristic;
detecting (22) whether the packet contains the
indication and consequently is a replacement packet;
for the replacement packet, synthesizing (23)
synthesis audio contents the contents characteristic
of which differs from the predetermined contents
characteristic, by means of the error concealing
measure; and
rendering (24) the audio contents of a packet when the
packet is not a replacement packet, and rendering the
synthesis audio contents when the packet is a
replacement packet.
25. A computer program comprising a program code for
executing the method of claims 19 or 24 when the
method runs on a computer.

In the context of packet-oriented data transmission via a
network, an apparatus for transmitting a sequence of data
packets checks whether a packet is missing or faulty. In
this case, a replacement packet is provided which is a
valid packet with regard to a packet syntax which, however,
has a predetermined contents characteristic. On the decoder
side, a basic decoder recognizes the packet as a valid
packet and decodes same, whereas an extension decoder can
enable an error concealing measure on the basis of an
indication in the replacement packet to the fact that this
is a replacement packet, in order to provide a higher-
quality audio reproduction.