Apparatus and Method for Handling Transient Sound Events in Audio Signals When
Changing the Replay Speed or Pitch
Description
The replay speed of audio signals can be changed while maintaining the pitch, for example
with the help of a phase vocoder (see for example J. L. Flanagan und R. M. Golden, "The Bell
System Technical Journal", November 1966, pages 1394 tol509; United States Patent No.
6,549,884 Laroche, J. & Dolson, M.: "Phase-vocoder pitch-shifting"; Jean Laroche and Mark
Dolson, "New Phase- Vocoder Techniques for Pitch- Shifting, Harmonizing And Other Exotic
Effects", Proc. 1999 IEEE Workshop on Applications of Signal Processing to Audio and
Acoustics, "New Paltz, New York, Oct. 17-20, 1999). n the same way, with such methods
transposition of the signal can be performed while maintaining the original replay duration.
The latter is obtained by replaying the stretched signal accelerated by the factor of time
stretching. In time discrete signal representation, this corresponds to down-sampling the
signal by the stretching factor while maintaining the sampling frequency. Conventionally, this
time stretching takes place in the time domain. Alternatively, the same can also take place
within a filter bank, such as a pseudo-quadrature mirror filterbank (pQMF). The pseudoquadrature
mirror filterbank (pQMF) is sometimes also called a QMF filterbank.
Specific challenges in stretching are transient events that are "blurred" in time during the
processing step of time stretching. This occurs because methods, such as the phase vocoder,
affect the so-called vertical coherence properties (with regard to a time frequency spectrogram
representation) of the signal.
Some current methods stretch the time more around the transients, in order to not have to
perform any or only little time stretching during the duration of the transient. This has been
described, for example, in:
- Laroche L., Dolson M.: Improved phase vocoder timescale modification of audio",
EEE Trans. Speech and Audio Processing, vol. 7, no. 3, pp. 32 —332
- Emmanuel Ravelli, Mark Sandler and Juan P. Bello: Fast implementation for non
linear time-scaling of stereo audio; Proc. of the 8th Int. Conference on Digital Audio
Effects (DAFx'05), Madrid, Spain, September 20-22, 2005
- Duxbury, C , M. Davies, and M. Sandler (2001, December). Separation of transient
information in musical audio using multi resolution analysis techniques. In
Proceedings of the COST G-6 Conference on Digital Audio Effects (DAFX-01),
Limerick, Ireland.
Another paper on the topic was written by Robel, A.: A NEW APPROACH TO TRANSIENT
PROCESSING IN THE PHASE VOCODER ; Proc. of the 6th Int. Conference on Digital
Audio Effects (DAFx-03), London, UK, September 8-1 1, 2003.
In time stretching of audio signals by phase vocoders, transient signal portions are "blurred"
by dispersions, since the so-called vertical coherency in spectrogram view of the signal is
affected. Methods operating with so-called overlap-add methods can generate spurious pre
echoes and post echoes of transient sound events. These problems can be handled by changing
time stretching in the environment of transients, no stretching during the actual transients and
stronger stretching in the surrounding. If, however, transposition is to take place, the
transposition factor will no longer be constant in the environment of the transients, i.e. the
pitch of superimposed (possibly tonal) signal portions changes in a spuriously audible
manner. When time stretching takes place within a filter bank, such as the pQMF, similar
problems occur.
The field of this application relates to a method for perceptually motivated handling of
transient sound events within such a process. n particular, transient sound events may be
removed during signal manipulation of time stretching. Subsequently, a precisely fitting
addition may be performed of the unprocessed transient signal portion to the changed
(stretched) signal under consideration of the stretching.
According to embodiments of the teachings disclosed in this document, an apparatus for
processing an audio signal, comprises a time manipulator for individually time manipulating a
plurality of subband signals of the audio signal. The time manipulator comprises an overlapadd
stage for overlapping and adding blocks of at least one of the plurality of subband signals
using an overlap-add-advance value being different from a block extraction advance value, a
transient detector for detecting a transient in the audio signal or a subband signal, and a
plurality of transient adders for adding a detected transient to a plurality of signals generated
by the overlap-add stage. The overlap-add stage is configured for reducing an influence of a
detected transient or for not using the detected transients when adding.
According to another embodiment, an apparatus for processing an audio signal, comprises an
analysis filterbank for generating subband signals; a time manipulator for individually time
manipulating a plurality of subband signals, the time manipulator comprising: an overlap-add
stage for overlapping and adding blocks of the subband signal using an overlap-add-advance
value being different from a block extraction advance value; a transient detector for detecting
a transient in the audio signal or a subband signal, wherein the overlap-adder stage is
configured for reducing an influence of a detected transient or for not using the detected
transients when adding; and a transient adder for adding a detected transient to a signal
generated by the overlap/add stage.
According to another embodiment, a method for processing an audio signal comprises:
Individually time manipulating a plurality of subband signals of the audio signal, the
time manipulating comprising:
Overlapping and adding blocks of a corresponding one of the plurality of subband
signals using an overlap-add advance value being different from a block extraction advance
value;
Detecting a transient in the audio signal or a subband signal;
Either reducing an influence of or discarding a detected transient when overlapping and
adding;
Adding a detected transient to a plurality of signals generated by the action of
overlapping and adding.
Another embodiment relates to a computer program for performing a method when the
computer program runs on a computer, the method comprising:
Individually time manipulating a plurality of subband signals of the audio signal, the
time manipulating comprising:
Overlapping and adding blocks of a corresponding one of the plurality of subband
signals using an overlap-add advance value being different from a block extraction advance
value;
Detecting a transient in the audio signal or a subband signal;
Either reducing an influence of or discarding a detected transient when overlapping and
adding;
Adding a detected transient to a plurality of signals generated by the action of
overlapping and adding.
According to related embodiments, the apparatus may further comprise a decimator for
decimating the audio signal or the plurality of audio signals. The time manipulator may be
configured for performing a time stretching of the plurality of subband signals.
According to a further embodiment, the transient detector may be configured to mark blocks
detected as comprising a transient; and in which the plurality of overlap-add stages is
configured to ignore the marked blocks.
According to a further embodiment, the plurality of overlap-add stages may be configured for
applying an overlap-add value being greater than a block extraction value for performing a
time stretching of the plurality of subband signals.
According to a further embodiment, the time manipulator may further comprise a block
extractor, a windower/phase adjustor, and a phase calculator for calculating a phase, based on
which the windower/phase adjustor performs the adjustment of an extracted block.
According to a further embodiment, the transient adder may be further configured to insert a
portion of the subband signal having the transient, wherein the length of the portion is
selected sufficiently long, such that a cross-fade from the signal output from the portion
having the transient to the output from the overlap-add-processing is possible.
According to a related embodiment, the transient adder may be configured for performing the
cross-fade operation.
According to a further embodiment, the transient detector may be configured for detecting
blocks extracted by a block extractor from the subband signal having a transient
characteristic. The overlap-add stage may be further configured for reducing an influence of
the detected blocks or for not using the detected blocks when adding.
According to a further embodiment, the transient detector may be configured for performing a
moving center of gravity calculation of energy across a predetermined time period of a signal
to be input into an analysis filterbank or a subband signal.
Exact determination of the position of the transient for the purpose of selecting an appropriate
section, can, for example, be performed with the help of a moving centroid calculation of the
energy across an appropriate time period. In particular, transient determination can be
performed in a frequency-selective manner within a filter bank. Additionally, the time period
of the section can be selected as a constant value or in a variable manner based on information
from the transient determination.
According to a further embodiment, the apparatus may further comprise an analysis filterbank
for generating the subband signals.
According to a further embodiment, the apparatus may further comprise a decimator arranged
at an input side or an output side of the analysis filter bank. The time manipulator may be
configured for performing a time stretching of the plurality of subband signals.
According to a further embodiment, the apparatus may further comprise a first analysis
filterbank, a second analysis filter bank, a resampler upstream of the second analysis filter
bank, and a plurality of phase vocoders for a second plurality of subband signals output by the
second analysis filterbank, the plurality of phase vocoders having a bandwidth extension
factor greater than one and a phase vocoder output being provided to the plurality of overlapadd
stages.
According to a further embodiment, the apparatus may further comprise a connecting stage
between the first analysis bank and the plurality of phase vocoders at an input side of the
connecting stage and the plurality of overlap-add stages at an output stage of the connecting
stage, the connecting stage being configured to control a provision of the blocks of the
corresponding one of the plurality of subband signals and phase-vocoder processed signal to
the overlap-add stage.
According to a further embodiment, the apparatus may further comprise: an amplitude
correction configured to compensate for amplitude affecting effects of different overlap
values.
The present application thus provides different aspects of apparatuses, methods or computer
programs for processing audio signals in the context of bandwidth extension and in the
context of other audio applications which are not related to bandwidth extension. The features
of the described and claimed individual aspects can be partly or fully combined, but can also
be used separately from each other, since the individual aspects already provide advantages
with respect to perceptual quality, computational complexity and processor/memory resources
when implemented in a computer system or micro processor.
According to the teachings disclosed herein, and in contrast to existing methods, a windowed
section including the transient may be removed from the signal to be manipulated. This may
be obtained by summing up only those time portions not including transients, block by block,
during the overlap-and-add (OLA) process. This results in a time stretched signal including no
transients. After terminating the time stretching, the unstretched transients that have been
removed from the original signal are added again.
Dispersion and echo effects hence no longer affect the subjective audio quality of the
transient.
By inserting the original signal portion, change of timbre or pitch will result when changing
the sampling rate. Generally, however, the transient psycho-acoustically masks this. If, in
particular, stretching by an integer factor takes place, the timbre will be changed only slightly,
since outside the environment of the transient, only every n-th (n = stretching factor)
harmonic is mapped.
The accompanying drawings are included to provide a further understanding of embodiments
and are incorporated in and constitute a part of this specification. The drawings illustrate
embodiments and, together with the description, serve to explain the principles of the
embodiments. Other embodiments and many of the intended advantages of embodiments will
be readily appreciated, as they become better understood with reference to the following
detailed description. Like reference numerals designate corresponding or similar parts.
Fig. 1 shows a signal waveform of an original signal consisting exemplarily of a
mixture of pitch pipe and castanets.
Fig. 2 shows a Discrete Fourier Transformation (DFT) spectrogram of the signal
waveform shown in Fig. 1.
Fig. 3 shows a QMF based spectogram based on a 64 band pQMF analysis filterbank
similar to the DFT spectrogram Fig. 2.
Fig. 4 shows a transient detection matrix.
Fig. 5 shows a signal waveform of a signal resulting from time stretching without
using the teachings disclosed herein.
Fig. 6 shows a signal waveform of a signal resulting from time stretching with using
the teachings disclosed herein.
Fig. 7 shows an FFT based spectrogram of a time stretched signal without transient
handling according to the teachings disclosed herein.
Fig. 8 shows an FFT based spectrogram of a time stretched signal with transient
handling according to the teachings disclosed herein.
Fig. 9 illustrates a schematic block diagram of an audio processing system comprising
an apparatus according to the teachings disclosed herein.
Fig. 10 illustrates a schematic block diagram of another audio processing system
comprising an apparatus according to the teachings disclosed herein.
Fig. A illustrates a schematic block diagram of a processing implementation for
processing a single subband signal.
Fig. 1IB illustrates a schematic block diagram of another processing implementation for
processing a single subband signal.
Figs. A to 12E illustrate the signal block processing according to the disclosed teachings.
Fig. 13 illustrates a schematic block diagram of an apparatus according to one
embodiment of the teachings disclosed herein.
Fig. 14 illustrates a schematic block diagram of an apparatus according to another
embodiment of the teachings disclosed herein.
Fig. 15 illustrates a schematic flow diagram of a method for processing an audio signal
according to the teachings disclosed herein.
Fig. 1 shows a time section of a signal waveform of an audio signal consisting exemplarily of
a mixture of pitch pipe and castanets. The depicted audio signal shall be used as an original
signal on which various time stretching actions are performed without or with applying the
teachings disclosed herein. The sound of the pitch pipe corresponds to a substantially periodic
signal having an amplitude of approximately 0.08 units in Fig. 1. Four castanet beats are
visible in Fig. 1 as four short impulses having an amplitude of approximately 0.45 units. The
pitch pipe produces a substantially tonal signal. The castanets, however, produce a highly
transient signal. In acoustics and audio, a transient is typically defined as a short-duration
signal that represents a non-harmonic attack phase of a musical sound or spoken word. It may
contain a high degree of non-periodic components and a higher magnitude of high frequencies
than the harmonic content of that sound. Transients typically do not directly depend on the
frequency of the tone they initiate.
Fig. 2 shows a Discrete Fourier Transform (DFT) spectrogram of the signal waveform of Fig.
1. Fig. 3 is similar to Fig. 2 and shows a 64 band pseudo-Quadrature Mirror interbank
(pQMF) spectrogram of the signal waveform of Fig. 1. As can be seen from the two
spectrogram illustrations in Figs. 2 and 3, the original audio signal includes a dense harmonic
partial sound structure (horizontal structures) and castanet beats (vertical structures).
Derived there from, Fig. 4 shows a binary transient detection matrix marking transient signal
portions in a frequency-selective manner. Detected transient signal portions are illustrated in
white. The same may be removed via vocoder for transposition and subsequently added again
based on the original signal. Alternatively, the detected transient signal portions may be
excluded from time stretching and replaced later with respective signal portions from the
original signal.
Figures 5 to 8 show the result of time stretching with and without the new transient handling
in the form of two time signals and the associated spectrograms. By comparing the time
signals and the spectrograms for the cases with transient handling and without transient
handling, it can be seen that undesired time blur of the transient signal portions is effectively
avoided by employing the teachings disclosed herein. For example, the time signal shown in
Fig. 5 and the corresponding spectrogram shown in Fig. 7 reveal that the castanet beats have
been widened, i.e. their duration is longer than in the original time signal shown in Fig. 1. In
contrast, the time signal shown in Fig. 6 and the corresponding spectrogram in Fig. 8, which
have been obtained by employing a transient handling according to the teachings disclosed
herein, demonstrate that the castanet beats have not undergone a substantial widening with
respect to their duration but are substantially preserved during the course of the signal
manipulation.
With the apparatus, method, and computer program according to the disclosed teachings,
artifacts (dispersions, pre and post echoes) resulting when processing transients by time
stretching and transposition methods, are effectively avoided. Above that, it is differentiated
in a frequency-selective manner whether stationary or transient portions in a subband
predominate, and the transient handling method is selected correspondingly. Additionally, the
time period of the signal portion to be inserted can be formed in a variable manner
considering parameters of transient determination for optimally adapting the time period of
the signal portion to the transient.
The method is suitable for all audio applications where the replay speed of audio signals or
their pitch is to be changed. Particularly suited are applications for bandwidth extension or in
the field of audio effects.
Fig. 9 illustrates an audio processing system which is in the field of audio bandwidth
extension. However, the invention can also be applied to other fields as well which do not
perform a bandwidth extension. A bitstream is input into a core decoder 100. The signal
output by the core decoder, i.e., a narrow bandwidth audio signal is input into respective
decimators 102a, 102b, 102c. The decimated signals which have a reduced time length
compared to the signal output by the core decoder 100 are input into corresponding pQMF
analysis stages 104a, 104b, 104c. The stages 104a, 104b, 104c can be implemented by any
other analysis filterbank which is not a pQMF filterbank. Plenty of different filterbank
implementations exist which all can be used for this purpose.
Each pQMF analysis stage 104a, 104b, 104c outputs a plurality of different subband signals in
different subband channels, where each subband signal has a reduced bandwidth and,
typically, a reduced sampling rate. In this case, the filterbank is a 2-times oversampled
filterbank which is preferred for the present invention. However, also a critically sampled
filterbank may be used.
The corresponding narrow band signal or subband signal output in a pQMF analysis channel
is input into a phase vocoder. Although Fig. 9 only illustrates three phase vocoders 106a,
106b, 106c, it is important to see that each individual pQMF analysis channel may have an
own phase vocoder. The phase vocoder algorithm can also be implemented by interpolation of
the base band or the first patch. The phase vocoders for different subband signals generated
by the same analysis filterbank, have a similar construction, and are different from the phase
vocoders for the subband signals from other filterbanks due to the bandwidth extension factor
illustrated in Fig. 9. The bandwidth extension factor is two in the phase vocoder 106a. In the
phase vocoder 106b, the bandwidth extension factor is three, and in the phase vocoder 106c,
the bandwidth extension factor is four. Note that it typically not necessary for the purposes of
the teachings disclosed herein to perform any bandwidth extension or even several different
bandwidth extensions. Thus, the decimators 102a, 102b, 102c may be omitted. The outputs
from the different phase vocoders are input into a pQMF synthesis filterbank 108. When the
analysis filterbanks in blocks 104a- 104c are implemented in a different technology, then the
synthesis filterbank 108 will also be implemented in a different technology, so that the
analysis filterbank technology and the synthesis filterbank technology match with each other.
An apparatus according to the teachings disclosed herein may be implemented in a distributed
manner in one or more of the QMF analysis stages 104a, 104b, 104c and the QMF synthesis
filterbank 108. In the same manner or a similar manner, a time manipulator which is a part of
the apparatus according to the disclosed teachings may be distributed aming the QMF analysis
stages 104a, 104b, 104c and the QMF synthesis filterbank 108. Accordingly, the one or more
of the QMF analysis stages 104a, 104b, 104c may omit blocks containing a transient from
time manipulation and forward the original blocks to the synthesis filterbank 108. The
synthesis filterbank 108 may provide the functionality of a transient adder by adding a
detected and typically unmodified transient to a signal generated by an overlap-add stage of
the synthesis filterbank 108. The schematic block diagram of Fig. 9 does not explicitly show
the transient detector. The transient detector could be part of the QMF analysis stages 104a,
104b, 104c. In the alternative, the transient detector could be a unit of its own.
Fig. 10 illustrates the different implementation, where the baseband signal on line 110 is input
into an analysis filterbank 112. Thus, the lowband signal is transformed into a plurality of
subband signals. Additionally, a switching stage or connecting stage 114 is provided, by
which different subband signals output by a phase vocoder 106a, 106b or output by the
baseband pQMF analysis 1 can be input into any arbitrarily selected synthesis band.
The individual phase vocoders are related to an individual pQMF band. In Fig. 10, the first
pQMF band and the last pQMF band of a first harmonic patch using the bandwidth extension
factor of two are illustrated as 106a. For the other harmonic patch using the bandwidth
extension factor of three, the first and the last pQMF band of this patch are illustrated as 106b.
The synthesized signal can be generated using an arbitrarily selected combination of phase
vocoder outputs and baseband pQMF analysis 11 outputs. It is to be noted that the switching
stage 4 can be a controlled switching stage which is controlled by an audio signal having a
certain side information, or which is controlled by a certain signal characteristic.
Alternatively, the stage 114 can be a simple connecting stage without any switching
capabilities. This is the case, when a certain distribution of output signals from elements 112
and 106a- 106b is fixedly set and fixedly programmed. In this case, the stage 114 will not
comprise any switches, but will comprise certain through-connections.
Fig. 1 A illustrates an embodiment of a processing implementation for processing a single
subband signal. The single subband signal may have been subjected to any kind of decimation
either before or after being filtered by an analysis filter bank not shown in Fig. 11A. In case a
decimation has been performed, the time length of the single subband signal is typically
shorter than the time length before forming the decimation. The single subband signal is input
into a block extractor 1800. The block extractor 1800 in Fig. 11A operates using a
sample/block advance value exemplarily called e. The sample/block advance value can be
variable or can be fixedly set and is illustrated in Fig. 11A as an arrow into block extractor
box 1800. At the output of the block extractor 1800, there exists a plurality of extracted
blocks. These blocks are highly overlapping, since the sample/block advance value e is
significantly smaller than the block length of the block extractor. An example is that the block
extractor extracts blocks of 12 samples. The first block comprises samples 0 to 1 , the second
block comprises samples 1 to 12, the third block comprises samples 2 to 13, and so on. In this
embodiment, the sample/block advance value e is equal to 1, and there is a 1 -fold
overlapping. The above example has values, which are provided by way of example and can
change from application to application.
The individual blocks are input into a windower 1802 for windowing the blocks using a
window function for each block. Additionally, a phase calculator 1804 is provided which
calculates a phase for each block. The phase calculator 1804 can either use the individual
block before windowing or subsequent to windowing. Then, a phase adjustment value p x k is
calculated and input into a phase adjuster 1806. The phase adjuster applies the adjustment
value to each sample in the block. Furthermore, the factor k is equal to the bandwidth
extension factor. When, for example, the bandwidth extension by a factor 2 is to be obtained,
then the phase p calculated for a block extracted by the block extractor 1800 is multiplied by
the factor 2 and the adjustment value applied to each sample of the block in the phase adjustor
1806 is p multiplied by 2. This is a value/rule provided by way of example. Alternatively, the
corrected phase for synthesis is k * p, p + (k-l)*p. So in this example the correction factor is
either 2, if multiplied or 1*p if added. Other values/rules can be applied for calculating the
phase correction value.
In an embodiment, the single subband signal is a complex subband signal, and the phase of a
block can be calculated by a plurality of different ways. One way is to take the sample in the
middle or around the middle of the block and to calculate the phase of this complex sample.
Although illustrated in Fig. 1A in the way that a phase adjustor operates subsequent to the
windower, these two blocks can also be interchanged, so that the phase adjustment is
performed to the blocks extracted by the block extractor and a subsequent windowing
operation is performed. Since both operations, i.e., windowing and phase adjustment are realvalued
or complex-valued multiplications, these two operations can be summarized into a
single operation using a complex multiplication factor which, itself, is the product of a phase
adjustment multiplication factor and a windowing factor.
The phase-adjusted blocks are input into an overlap/add and amplitude correction block 1808,
where the windowed and phase-adjusted blocks are overlap-added. Importantly, however, the
sample/block advance value in block 1808 is different from the value used in the block
extractor 1800. Particularly, the sample/block advance value in block 1808 is greater than the
value e used in block 1800, so that a time stretching of the signal output by block 1808 is
obtained. Thus, the processed subband signal output by block 1808 has a length which is
longer than the subband signal input into block 1800. When the bandwidth extension of two is
to be obtained, then the sample/block advance value is used which is two times the
corresponding value in blocks 1800. This results in a time stretching by a factor of two.
When, however, other time stretching factors are necessary, then other sample/block advance
values can be used so that the output of block 1808 has a required time length.
For addressing the overlap issue, an amplitude correction is preferably performed in order to
address the issue of different overlaps in block 1800 and 1808. This amplitude correction
could, however, be also introduced into the windower/phase adjustor multiplication factor, but
the amplitude correction can also be performed subsequent to the overlap/processing.
In the above example with a block length of 12 and a sample/block advance value in the block
extractor of one, the sample/block advance value for the overlap/add block 808 would be
equal to two, when a bandwidth extension by a factor of two is performed. This would still
result in an overlap of six blocks. When a bandwidth extension by a factor of three is to be
performed, then the sample/block advance value used by block 1808 would be equal to three,
and the overlap would drop to an overlap of four. When a four-fold bandwidth extension is to
be performed, then the overlap/add block 1808 would have to use a sample/block advance
value of four which would still result in an overlap of more than two blocks.
The phase vocoder for an individual subband signal illustrated in Fig. 11A preferably
comprises a transient detector 200 for performing a transient detection within the subband
signal indicated by connection 201a or for performing a transient detection of the signal
before the analysis filterbank processing as indicated by connection line 201b. As soon as the
transient detector 200 detects the transient, the overlap/add stage is controlled to not use the
blocks having the transient in the overlap/add processing as illustrated by control connection
203. In an embodiment, the signal on line 203 controls the overlap/add stage to remove all
blocks having the transient event. This will result in a signal at the output of this stage which
is stretched with respect to the signal before this stage, but which does not include any
transients.
The stretched signal without transients is input into the transient adder which is configured for
adding the transient to the stretched signal so that, at the output, there exists a stretched signal
having inserted transients, but these inserted transients have not been affected by a multiple
overlap/add processing.
In one embodiment, the transient portion is inserted from the subband signal itself as
illustrated by connection line 206 and line 201a. Alternatively, the signal can be taken out
from any other subband signal or from the signal before the subband analysis, since it is
characteristic for a transient that the transient occurs in a quite similar manner over the
individual subbands. On the other hand, however, using the transient event occurring in a
subband is preferable in some instances, since the sampling rate and other considerations are
as close as possible to a stretched signal.
Fig. 11B illustrates another possible embodiment of a processing implementation for
processing a single subband signal. Upstream of the block extractor 1800, a transient
suppression windower 1798 is inserted which acts on the single subband signal. The transient
suppression windower 1798 removes samples or blocks containing a transient. An evaluation
whether a sample contains a transient is performed by the transient detector 200. The single
subband signal is tapped at an input side of the transient suppression windower 1798 so that
the transient detector 200 receives the single subband signal as an input. Upon detection of a
transient, the transient detector 200 outputs a corresponding signal to the transient suppression
windower 1798 and the transient suppression windower 1798 reacts by suppressing the
sample(s) that has/have been indicated by the transient detector 200 as containing a transient.
Therefore, samples marked by the transient detector 200 as containing a sample do not enter
the block extractor 1800. The other, non-transient-containing samples are kept in the blocks
that are processed by the block extractor 1800, the windower 1802, the phase calculator 1804,
the phase adjuster 1806, and the overlap-add block 1808. The overlap-add block 1808 outputs
a stretched signal without transients.
The transient-containing samples are then added again to the stretched signal without
transients by the transient adder 204. The transient adder 204 receives a control signal from
the transient detector 200 and the original single subband signal as inputs. With this
information, the transient adder can identify the samples that have been suppressed by the
transient suppression windower 1798 and re-insert these samples in the stretched signal
without transients. At the output of the transient adder 204 the processed subband signal (long
time length) having inserted transients is obtained.
Figs. 12A to 12E illustrate how the audio signal or one of the plurality of subband signals may
be processed according to previously implemented methods and according to the teachings
disclosed herein. In Fig. 1 A, a sequence of audio samples 1202 is shown. The sequence 1202
may belong to one of the plurality of subband signals. The letter "T" marks a sample in which
a transient has been detected by a transient detector.
Beneath the sequence 1202 in Fig. 12A, a plurality of extracted blocks 1206 are represented.
The plurality of extracted blocks 1206 are each 1 samples long and comprise the sample with
the transient T. Above and beneath the plurality of extracted blocks 1 06 one preceding block
1204 and one subsequent block 1208 are shown that do not contain the transient T. As can be
seen in Fig. 12A, the entire plurality of extracted blocks 1204 extends over 23 blocks.
Fig. 12B illustrates how, in standard time manipulation methods, the preceding block 1204,
the blocks of the plurality of extracted blocks 1206, and the subsequent block 1208 are shifted
each by one block prior to overlapping and adding the individual blocks in order to perform a
time stretching of the audio signal. The shifted versions of the blocks or the plurality of blocks
are labeled 1204', 1206', and 1208'. The overlap-add-advance value is two in Fig. 12B,
whereas the block extraction advance value illustrated in Fig. 12A is one. As a result of the
shifting of the blocks, the samples comprising the transient T in the blocks of the plurality of
extracted blocks 1206 are not temporally aligned anymore, but are spread over a time span of
twelve samples.
Fig. 12C illustrates a removal of the blocks that contain the transient T in one or more of their
samples, in accordance with the teachings disclosed herein. The removed blocks belong to the
plurality of extracted blocks 1206' and are drawn in dashed line. The removal of the blocks
1206' leaves a gap which is 14 samples long. Furthermore, in a time span of 10 samples prior
to the gap and a time span of 10 samples subsequent to the gap, only a reduced number of
blocks instead of the usual six blocks are considered in the overlap-add process or by the
overlap-add stage of an apparatus for processing an audio signal. Note that Figs. 12B and 12C
are illustrative only and that the blocks of the plurality of extracted blocks 1206 of Fig. 12A
could be removed immediately after a transient has been detected, i.e. without performing the
time shifting action on these blocks of the plurality of extracted blocks 1206. In one possible
implementation of the teachings disclosed herein, the blocks of the plurality of extracted
blocks 1206 are rerouted to bypass the overlap-add stage and to be inserted downstream of the
overlap-add stage.
In Fig. 12D an insertion of the original transient section, i.e. the plurality of extracted blocks
1206, in the time manipulated audio signal. The original transient section is inserted in the gap
that has been left after the removal of the blocks containing the transient T. The original
transient section may be added to the time manipulated rest of the audio signal. At the
beginning or the left edge of the gap, the plurality of extracted blocks 1206 is superposed with
six regular blocks (three of which are shown in Fig. 12D, with a dot pattern). In the example
depicted in Fig. 12D, the regular blocks are processed with an overlap-add-advance value of
two. As can be seen in Fig. 12D, a residual gap remains between the end of the original
transient section and the subsequent block 1208'. It would be possible, to shift the plurality of
extracted blocks 1206 a few samples to the right, that is towards later time instants so that the
original transient section is more equally distributed and/or located within the gap between the
shifted preceding block 1204' and the shifted subsequent block 1208'.
The lower part of Fig. 12D shows how many blocks are superposed in each sample.
According to the regular time manipulation with a block length of twelve samples, a block
extraction value of one, and an overlap-add-advance value of two, six blocks are typically
considered during the overlap-add process for a particular sample of the time manipulated
audio signal. In other words, in one sample of the time manipulated signal there are
contributions from six different blocks extracted from the original audio signal. The curve in
Fig. 12D shows that during the processing of the original transient section initially six blocks
are considered. As the blocks of the plurality of extracted blocks 1206 are staggered with one
sample difference, the number of blocks to be superposed increases to reach the value twelve
for the sample were the transient T has been detected. Subsequently, the block count
decreases by one with every new sample to reach the value one at the end of the original
transient section. The block count may be used to correct an amplitude of the time
manipulated signal in section, in which the number of superposed blocks differs from the
regular value of six. To this end, the block count may be determined based on a detection of
the transient and fed to an amplitude correction. The amplitude correction may either act on
the blocks prior to overlapping, adding, and/or superposing, or on the resulting time
manipulated signal.
As mentioned above, a residual gap of two samples remains. When the regular blocks begin
again, starting with the subsequent block 1208',
Fig. 12E shows an optional implementation in which the gap has been shortened by two
samples so that no residual gap remains between the end of the original transient section and
the shifted subsequent block 1208'. Although this measure may lead to a slight corruption of
the resulting time manipulated signal (in particular, a slight shortening), the effect may be
negligible. As mentioned above, the original transient section could be inserted more centered
within the gap between the previous block 1204' and the subsequent block 1208'.
As an alternative to removing complete blocks that comprise one or more transient-containing
samples, as illustrated in Figs. 12A to 12E, the individual transient-containing sample(s) may
be removed within the block, while the remaining samples in the block are maintained. The
removal of the transient-containing samples may be implemented by setting a value of the
sample to zero. Hence, the transient-containing sample will not make a contribution to the
output of the overlap-add block 1808. An amplitude correction may be used in order to
increase a contribution of the other samples that are overlap-added with the zeroed sample.
The action of zeroing the transient-containing samples may be accompanied by fading-out
and fading-in the subband signal prior to the sample and subsequent to the sample,
respectively. For example, a few samples prior to the transient-containing sample and a few
samples subsequent to the transient containing sample the subband signal may be multiplied
with a fading factor signal in order to implement e.g. a triangular fading window around the
transient-containing sample(s).
Fig. 13 shows a schematic block diagram of a time manipulator that could be a part of an
apparatus for processing an audio signal according to the disclosed teachings. The time
manipulator receives a plurality of subband signals which together form the audio signal.
Within the time manipulator the plurality of subband signals may be temporarily stored by a
block extractor and buffer 0. The block extractor and buffer 8 0 extracts blocks from the
each one of the plurality of subband signals. The blocks have a specific block length L and are
extracted with a specific block extraction advance value e. For example, the block length L
may be twelve and the block extraction advance value e may be one. The block extractor and
buffer 1810 receives the block length L and the block extraction advance value e as input
parameters. In the alternative, the block length L and the block extraction advance value e
could be stored in a fixed manner in the block extractor and buffer 1810.
The block extractor and buffer 1810 outputs extracted blocks and provides them to an
overlap-add stage 1808 in which the extracted blocks are overlapped with an overlap-addadvance
value k * e different from the block extraction advance value e and added up to form
the time manipulated audio signal. The overlap-add stage 1808 may comprise a plurality of
overlap-add units, e.g. one overlap-add unit for a corresponding one of the plurality of
subband signals. Another option would be to use a single overlap-add stage or a few overlapadd
units in a time-sharing or multiplexed manner so that the subband signals are overlapadded
individually and successively.
The time manipulator further comprises a transient detector 200 which receives the plurality
of subband signals. The transient detector 200 may analyze the subband signals or the audio
signal with respect to e.g. a non-harmonic attack phase of a musical sound or spoken word or
a high degree of non-periodic components and/or a higher magnitude of high frequencies than
the harmonic content of that sound. An output of the transient detector 200 indicates whether
or not a transient has been identified in a current section of the audio signal and is provided to
the overlap-add stage 1808 and a transient adder 1812. In case the output of the transient
detector 200 indicates that a transient has been detected, the overlap-add stage 1808 is
controlled to ignore those blocks that contain the transient T when performing the overlap-add
action. The transient adder 1812, on its part, inserts the original transient section to the
otherwise time-manipulated audio signal upon reception of an indication from the transient
detector 200 that a transient has been detected. The time-manipulated signal with the added
transient forms an output of the time manipulator.
Fig. 14 shows a schematic block diagram of a time manipulator according to another
implementation according to the teachings disclosed herein. In addition to the elements of the
time manipulator shown in Fig. 13, the time manipulator of Fig. 14 comprises an amplitude
correction 1814. The amplitude correction 1814 receives the indication about a detected
transient from the transient detector 200. On the basis of this information, the amplitude
correction 1814 may modify the amplitude of signal blocks to account for a varying number
of blocks that are being used in the overlap-add process. The variation of the number of
blocks considered is due to the removal of the plurality of extracted blocks 1204 and possibly
due to the insertion of the original transient section. Typically, the time pattern how the
number of blocks varies is known and can be determined on the basis of the time instant of
the detected transient. Hence, it may be sufficient to provide a trigger signal to the amplitude
correction which then adjusts the amplitudes of subsequent blocks according to the time
pattern. A possible time pattern could be based on the waveform showing the evolution of the
number of blocks that are considered in the overlap-add process as illustrated in Figs. 12D
and 12E. An amplitude correction value could be, for example, a reciprocal of the block
count.
Fig. 15 shows a schematic flow diagram of a method for processing an audio signal according
to the teachings disclosed herein. After the start of the method, an action 1502 is performed in
which a plurality of subband signals of an audio signal are individually time-manipulated. The
action 1502 comprises sub-actions 1504 to 1510.
At 1504 the blocks of a corresponding subband signal of the plurality of subband signals are
overlapped and added. An overlap-add advance value is used that is different from a block
extraction advance value. The action 1504 represents the normal process flow in the absence
of transients and is performed continuously.
A transient detection action is performed at 1506 to detect a transient in the audio signal or in
a subband signal. The action 1506 may be performed concurrently with the action 1504 and
other actions shown in the flow diagram of Fig. 15.
An influence of a detected transient is either reduced, or the detected transient is discarded,
when performing the action 1504 of overlapping and adding.
A detected transient is then added, at action 1510, to a plurality of signals generated by the
action 1504 of overlapping and adding.
Although according to the teachings disclosed herein the transient section of the audio signal
has typically not undergone the same time manipulation as the rest of the audio signal, the
time-manipulated resulting signal typically renders the transient sections in a realistic manner.
This may be at least partly due to the fact that a transient is highly insensitive to many signal
manipulation methods, such as frequency shifting.
According to another aspect of the teachings disclosed herein, an apparatus for processing an
audio signal may comprise:
an analysis filterbank for generating subband signals;
a time manipulator for individually time manipulating a plurality of subband signals, the time
manipulator comprising:
an overlap-add stage for overlapping and adding blocks of the subband signal using an
overlap-add-advance value being different from a block extraction advance value;
a transient detector for detecting a transient in the audio signal or a subband signal,
wherein the overlap-adder stage is configured for reducing an influence of a detected transient
or for not using the detected transients when adding; and
a transient adder for adding a detected transient to a signal generated by the overlap/add stage.
According to another aspect of the teachings disclosed herein, an apparatus as previously
described, may further comprise a decimator arranged at an input side or an output side of the
analysis filterbank, wherein the time manipulator may be configured for performing a time
stretching of a subband signal.
According to another aspect of the teachings disclosed herein, in an apparatus as previously
described, the transient detector may be configured to mark blocks detected as comprising a
transient; and the overlap-adder-stage may be configured to ignore the marked blocks.
According to another aspect of the teachings disclosed herein, in an apparatus as previously
described, the overlap-add-stage may be configured for applying an overlap-add-advance
value being greater than a block-extraction-advance value for performing a time stretching of
the subband signal.
According to another aspect of the teachings disclosed herein, in an apparatus in accordance
with one of the preceding claims, the time manipulator may comprise: a block extractor; a
windower/phase adjustor; and a phase calculator for calculating a phase, based on which the
windower/phase adjuster performs the phase adjustment of an extracted block.
According to another aspect of the teachings disclosed herein, in an apparatus as previously
described, the transient detector may be configured to determine a length of a portion of the
subband signal containing the transient, the length matching the length of the signal to be
inserted by the transient adder.
According to another aspect of the teachings disclosed herein, in an apparatus as previously
described, the transient adder may be configured to insert a portion of the subband signal
having the transient, wherein the length of the portion may be selected sufficiently long, such
that a cross-fade from the signal output from the overlap-add-processing to the portion having
the transient or from the portion having the transient to the output from the overlap-addprocessing
is possible.
According to another aspect of the teachings disclosed herein, in an apparatus as previously
described, the transient adder may be configured for performing the cross-fade operation.
According to another aspect of the teachings disclosed herein, in an apparatus as previously
described, the transient detector may be configured for detecting blocks extracted by a block
extractor from the subband signal having a transient characteristic, and the overlap-add-stage
may be configured for reducing an influence of the detected blocks or for not using the
detected blocks when adding.
According to another aspect of the teachings disclosed herein, in an apparatus as previously
described, the transient detector may be configured for performing a moving center of gravity
calculation of an energy across a predetermined time period of a signal to be input into an
analysis filterbank or a subband signal.
Although some aspects have been described in the context of an apparatus, it is clear that
these aspects also represent a description of the corresponding method, where a block or
device corresponds to a method step or a feature of a method step. Analogously, aspects
described in the context of a method step also represent a description of a corresponding block
or item or feature of a corresponding apparatus.
The inventive encoded audio signal can be stored on a digital storage medium or can be
transmitted on a transmission medium such as a wireless transmission medium or a wired
transmission medium such as the Internet.
Depending on certain implementation requirements, embodiments of the invention can be
implemented in hardware or in software. The implementation can be performed using a digital
storage medium, for example a floppy disk, a DVD, a CD, a ROM, a PROM, an EPROM, an
EEPROM or a FLASH memory, having electronically readable control signals stored thereon
which cooperate (or are capable of cooperating) with a programmable computer system such
that the respective method is performed.
Some embodiments according to the invention comprise a data carrier having electronically
readable control signals which are capable of cooperating with a programmable computer
system, such that one of the methods described herein is performed.
Generally, embodiments of the present invention can be implemented as a computer program
product with a program code, the program code being operative for performing one of the
methods when the computer program product runs on a computer. The program code may for
example be stored on a machine readable carrier.
Other embodiments comprise the computer program for performing one of the methods
described herein, stored on a machine readable carrier.
In other words, an embodiment of the inventive method is, therefore, a computer program
having a program code for performing one of the methods described herein, when the
computer program runs on a computer.
A further embodiment of the inventive methods is, therefore, a data carrier (or a digital
storage medium, or a computer-readable medium) comprising, recorded thereon, the computer
program for performing one of the methods described herein.
A further embodiment of the inventive method is, therefore, a data stream or a sequence of
signals representing the computer program for performing one of the methods described
herein. The data stream or the sequence of signals may for example be configured to be
transferred via a data communication connection, for example via the Internet.
A further embodiment comprises a processing means, for example a computer, or a
programmable logic device, configured to or adapted to perform one of the methods described
herein.
A further embodiment comprises a computer having installed thereon the computer program
for performing one of the methods described herein.
In some embodiments, a programmable logic device (for example a field programmable gate
array) may be used to perform some or all of the functionalities of the methods described
herein. In some embodiments, a field programmable gate array may cooperate vsdth a
microprocessor in order to perform one of the methods described herein. Generally, the
methods are preferably performed by any hardware apparatus.
The above described embodiments are merely illustrative for the principles of the present
invention. It is understood that modifications and variations of the arrangements and the
details described herein will be apparent to others skilled in the art. It is the intent, therefore,
to be limited only by the scope of the impending patent claims and not by the specific details
presented by way of description and explanation of the embodiments herein.
Claims
1. Apparatus for processing an audio signal, comprising:
an overlap-add stage for overlapping and adding blocks of a at least one of a plurality of
subband signals representing the audio signal using an overlap-add-advance value;
a transient detector for detecting a transient in the audio signal or a subband signal of the
plurality of subband signals,
wherein the apparatus for processing is configured for reducing an influence of a
detected transient or for not using the detected transients when adding by the overlapadd
stage; and
a transient adder for adding a detected transient to a subband signal generated by the
overlap/add stage.
2. Apparatus in accordance with claim 1, further comprising a decimator for decimating
the audio signal or the plurality of subband signals,
wherein the time manipulator is configured for performing a time stretching of the
plurality of subband signals.
3. Apparatus in accordance with claim 1, in which the transient detector is configured to
mark blocks detected as comprising a transient; and
in which the overlap-add-stage is configured to ignore the marked blocks.
4. Apparatus in accordance with one of the preceding claims, in which the overlap-addadvance
value is different from a block-extraction-advance value used for extracting the
blocks from a subband signal of the plurality of subband signals.
5. Apparatus in accordance with one of the preceding claims, in which the overlap-addstage
is configured for applying an overlap-add-advance value being greater than a
block-extraction-advance value for performing a time stretching of the plurality of
subband signals.
6. Apparatus in accordance with one of the preceding claims, in which the time
manipulator comprises:
a block extractor;
a windower/phase adjustor; and
a phase calculator for calculating a phase, based on which the windower/phase adjuster
performs the phase adjustment of an extracted block.
7. Apparatus in accordance with one of the preceding claims, in which the transient
detector is configured to determine a length of a portion of the subband signal
containing the transient, the length matching the length of the signal to be inserted by
the transient adder.
8. Apparatus in accordance with one of the preceding claims, in which the transient adder
is configured to insert a portion of the subband signal having the transient, wherein the
length of the portion is selected sufficiently long, such that a cross-fade from the signal
output from the overlap-add-processing to the portion having the transient or from the
portion having the transient to the output from the overlap-add-processing is possible.
9. Apparatus in accordance with claim 7, in which the transient adder is configured for
performing the cross-fade operation.
10. Apparatus in accordance with one of the preceding claims, in which the transient
detector is configured for detecting blocks extracted by a block extractor from the
subband signal having a transient characteristic.
11. Apparatus in accordance with one of the preceding claims, in which the transient
detector is configured for performing a moving center of gravity calculation of an
energy across a predetermined time period of a signal to be input into an analysis
filterbarik or a subband signal.
12. Apparatus in accordance with one of the preceding claims, further comprising an
analysis filter bank for generating the plurality of subband signals.
\ . Apparatus accordance ciaim \1, farther comprising a decimator arranged at an
input side or an output side of the analysis filter bank,
wherein the time manipulator is configured for performing a time stretching of the
plurality of subband signals.
14. Apparatus in accordance with one of the claims 1 to 1, further comprising:
a first analysis filter bank;
a second analysis filter bank;
a resampler upstream of the second analysis filter bank; and
a plurality of phase vocoders for a second plurality of subband signals output by the
second analysis filter bank, the plurality of phase vocoders having a bandwidth
extension factor greater than one, wherein a phase vocoder output is provided to the
overlap-add stage.
15. Apparatus in accordance with claim 14, further comprising a connecting stage between
the first analysis filter bank and the plurality of vocoders at an input side of the
connecting stage and the overlap-add stage at an output side of the connecting stage, the
connecting stage being configured to control a provision of the blocks of the
corresponding one of the plurality of subband signals and phase-vocoder processed
blocks output by the plurality of phase vocoders to the overlap-add stage.
\ 6. Apparatus in accordance with one of the preceding ciaims, farther comprising.
an amplitude correction configured to compensate for amplitude affecting effects of
varying block counts in the context of the overlap-add stage.
17. Apparatus in accordance with one of the preceding claims, further comprising a timemanipulator
for individually time manipulating the plurality of subband signals of the audio
signal, wherein the time manipulator comprises the overlap-add stage, the transient detector,
and the transient adder.
1g. Method for processing an audio signal, comprising:
overlapping and adding blocks of a corresponding one of a plurality of subband
signals representing the audio signal using an overlap-add-advance value;
detecting a transient in the audio signal or a subband signal of the plurality of subband
signals;
either reducing an influence or discarding a detected transient when overlapping and
adding;
adding a detected transient to a subband signal generated by the action of overlapping
and adding.
1 . A computer program for performing a method for processing an audio signal when the
computer program runs on a computer, the method comprising:
overlapping and adding blocks of a corresponding one of a plurality of subband
signals representing the audio signal using an overlap-add-advance;
detecting a transient in the audio signal or a subband signal of the plurality of subband;
either reducing an influence or discarding a detected transient when overlapping and
adding;
adding a detected transient to a subband signal generated by the action of overlapping
and adding.

ABSTRACT

An apparatus for processing an audio signal comprises an overlap add stage for overlapping and adding blocks of a corresponding one of a plurality of subband signals using an overlap add advance value being different from a block extraction advance value. The apparatus further comprises a transient detector for detecting a transient in the audio signal or a subband signal of the plurality of subband signals. The overlap add stage is configured for reducing an influence of a detected transient or for not using the detected transients when adding. The apparatus further comprises a transient adder for adding a detected transient to a subband signal generated by the overlap/add stage. A related method for processing an audio signal comprises inter alia either reducing an influence or discarding a detected transient when overlapping and adding.