EP2867893A1

EP2867893A1 - Effective pre-echo attenuation in a digital audio signal

Info

Publication number: EP2867893A1
Application number: EP13744654.8A
Authority: EP
Inventors: Balazs Kovesi; Stéphane RAGOT
Original assignee: Orange SA
Current assignee: Orange SA
Priority date: 2012-06-29
Filing date: 2013-06-28
Publication date: 2015-05-06
Anticipated expiration: 2033-06-28
Also published as: US20150170668A1; EP2867893B1; BR112014032587A2; RU2015102814A; US9489964B2; JP6271531B2; RU2607418C2; CN104395958B; CN104395958A; KR102082156B1; CA2874965C; WO2014001730A1; FR2992766A1; CA2874965A1; JP2015522847A; KR20150052812A; BR112014032587B1; ES2711132T3; MX2014015065A; MX349600B

Abstract

The invention relates to a method for processing pre-echo attenuation in a digital audio signal generated from a transform coding, wherein, at the decoding point, the method comprises steps of: detection (Detect.) of a position of attack in the decoded signal; determination (ZPE) of a pre-echo region preceding the position of attack detected in the decoded signal; calculation (F. Att.) of attenuation factors per sub-block of the pre-echo region, according to at least the frame wherein the attack has been detected and the preceding frame; and pre-echo attenuation (Att.) in the sub-blocks of the pre-echo region by the corresponding damping factors. The method also comprises the application of a filter (F) for the spectral shaping of the pre-echo region on the current frame up to the detected position of the attack. The invention also relates to a device implementing said method and to a decoder comprising such a device.

Description

Effective attenuation of pre-echoes in a digital audio signal

The invention relates to a pre-echo attenuation processing method and device for decoding a digital audio signal.

For the transport of digital audio signals on transmission networks, whether for example fixed or mobile networks, or for the storage of signals, compression processes (or source coding) using coding systems of the time coding type or frequency coding by transform.

The method and the device, which are the subject of the invention, thus have as their field of application the compression of sound signals, in particular frequency-coded digital audio signals.

FIG. 1 represents by way of illustration, a schematic diagram of the coding and decoding of a digital audio signal by transform including an addition / overlap synthesis analysis according to the prior art.

Certain musical sequences, such as percussion and certain segments of speech like the plosives (/ k /, lit, ...), are characterized by extremely sudden attacks which result in very fast transitions and a very strong variation of the dynamic signal within a few samples. An example of transition is given in Figure 1 from sample 410.

For the coding / decoding processing, the input signal is cut into blocks of samples of length L, represented in FIG. 1 by dotted vertical lines. The input signal is denoted x (ri), where n is the index of the sample. The division into successive blocks leads to defining the blocks ¾ («) = [x (NL) ... x (N.L + L1)] [x _N (0) ... x _N (L1)], where N is the index of the frame, L is the length of the frame. In Figure 1 we have L = 160 samples. In the case of the modulated modified cosine transform MDCT (for "Modified Discrete Cosine Transform" in English), two blocks X _N (n) and X _{N + 1} (n) are analyzed together to give a block of transformed coefficients associated with the frame of index N.

The division in blocks, also called frames, operated by the transform coding is totally independent of the sound signal and the transitions can therefore appear at any point in the analysis window. However, after transform decoding, the reconstructed signal is tainted by "noise" (or distortion) generated by the quantization (Q) - inverse quantization (Q ¹ ) operation. This coding noise is temporally distributed relatively uniformly over the entire temporal support of the transformed block, that is to say on any the length of the window of length 2L of samples (with overlap of L samples). The energy of the coding noise is generally proportional to the energy of the block and is a function of the coding / decoding rate.

For an attacking block (such as block 320-480 of FIG. 1) the signal energy is high, so the noise is also high.

In transform coding, the level of the coding noise is typically lower than that of the signal for the high energy segments that immediately follow the transition, but the level is higher than that of the signal for the lower energy segments, especially on the part preceding the transition (samples 160 - 410 of Figure 1). For the aforementioned part, the signal-to-noise ratio is negative and the resulting degradation can appear very troublesome to listen. Pre-echo is the coding noise prior to the transition and post-echo the noise after the transition.

It can be seen in Figure 1 that the pre-echo affects the frame before the transition and the frame where the transition occurs.

Psychoacoustic experiments have shown that the human ear performs a rather limited temporal pre-masking of sounds, of the order of a few milliseconds. The noise preceding the attack, or pre-echo, is audible when the duration of the pre-echo is greater than the duration of the pre-masking.

The human ear also performs a post-masking of a longer duration, from 5 to 60 milliseconds, during the passage of high energy sequences to low energy sequences. The rate or level of inconvenience acceptable for post-echoes is therefore greater than for pre-echoes.

The phenomenon of pre-echoes, more critical, is even more troublesome as the length of the blocks in number of samples is important. However, in transform coding, it is well known that for stationary signals plus the length of the transform increases, the higher the coding gain. At a fixed sampling rate and fixed rate, if we increase the number of points of the window (hence the length of the transform) we will have more bits per frame to code the frequency lines found useful by the psychoacoustic model, d where the advantage of using blocks of great length. MPEG AAC (Advanced Audio Coding) coding, for example, uses a long window that contains a fixed number of samples, 2048, over a period of 64 ms at a sample rate of 32 kHz; the problem of pre-echoes is managed by switching from these long windows to 8 short windows through intermediate windows (transition), which requires some delay in coding to detect the presence of a transition and adapt the windows. The length of these short windows is 8 ms. At low speed you can always have an audible pre-echo of a few ms. Switching windows allows you to attenuate the pre-echo but not to delete it. Transform encoders used for conversational applications such as ITU-T G.722.1, G.722.1C or G.719 often use a window of 40 ms duration at 16, 32 or 48 kHz (respectively) and a frame length of 20 ms. ms. It should be noted that the ITU-T G.719 encoder incorporates a window switch mechanism with transient detection, however the pre-echo is not completely reduced at low bit rate (typically at 32 kbit / s).

In order to reduce the aforementioned annoying effect of the pre-echo phenomenon, various solutions have been proposed at the encoder and / or the decoder.

Window switching has been mentioned previously. Another solution is to apply adaptive filtering. In the area preceding the attack, the reconstructed signal is seen as the sum of the original signal and the quantization noise.

A corresponding filtering technique has been described in the article High Quality Audio Transform Coding at 64 kbits, IEEE Trans. on Communications Vol 42, No. 11, November 1994, published by Y. Mahieux and J. P. Petit.

The implementation of such a filtering requires the knowledge of parameters some of which, like the prediction coefficients and the variance of the signal corrupted by the pre-echo, are estimated at the decoder from the noisy samples. On the other hand, information such as the energy of the original signal can only be known to the encoder and must therefore be transmitted. This requires additional information to be transmitted, which constrains the relative budget allocated to the transform coding. When the received block contains a sudden variation of dynamics, the filtering treatment is applied to it.

The aforementioned filtering process does not allow to find the original signal, but provides a strong reduction of pre-echo. However, it requires transmitting the additional parameters to the decoder.

Different pre-echo reduction techniques without specific transmission of information have been proposed. For example, a review of pre-echo reduction in the context of hierarchical coding is presented in article B. Kôvesi, S. Ragot, M. Gartner,

H. Taddei, "Pre-echo reduction in the ITU-T G.729.1 embedded coder," EUSIPCO,

Lausanne, Switzerland, August 2008.

A typical example of pre-echo attenuation method is described in French patent application FR 08 56248. In this example, factors are determined. sub-block attenuation, in the low energy sub-blocks preceding a sub-block in which a transition or attack has been detected.

The attenuation factor per sub-block g (k) is calculated for example as a function of the ratio R (k) between the energy of the sub-block of higher energy and the energy of the k-th sub-block in question. :

g {k) = f {R (k))

where / is a decreasing function with values between 0 and 1 and k is the number of the sub-block. Other definitions of the factor g (k) are possible, for example depending on the energy

In (k) in the current sub-block and energy En (k-1) in the previous sub-block.

If the variation of the energy with respect to the maximum energy is small, then no attenuation is necessary. The factor g (k) is then set to an attenuation-inhibiting attenuation value, i.e. 1. Otherwise, the attenuation factor is between 0 and 1.

In most cases, especially when pre-echo is annoying, the frame that precedes the pre-echo frame has a homogeneous energy that corresponds to the energy of a low energy segment (typically a background noise). According to experience, it is not useful or even desirable that after the pre-echo attenuation processing the signal energy becomes lower than the average energy per sub-block of the signal preceding the treatment zone.

(typically that of the previous frame In or that of the second half of the previous frame En ').

For the sub-block k to be processed, the limit value of the factor lim (k) can be calculated in order to obtain exactly the same energy as the average energy per sub-block of the segment preceding the sub-block to be processed. This value is of course limited to a maximum of 1 since it is of interest here to the attenuation values. More precisely :

where the average energy of the preceding segment is approximated by max ^ Era, En '.

The value lim _g (k) thus obtained serves as a lower limit in the final calculation of the attenuation factor of the sub-block: g (k) = max (g (k), lim _g (k))

The attenuation factors (or gains) g (k) determined by sub-blocks are then smoothed by an applied smoothing function sample by sample to avoid abrupt changes in the attenuation factor at the block boundaries.

For example, we can first define the gain per sample as a piecewise constant function:

S _pre {n) = g {k), n = kL ', - - -, (k + \) L' - \

where L 'represents the length of a sub-block.

The function is then smoothed according to the following equation:

g _pre {n): = a _gpn {n - 1) + {la) _gpre {n), n = 0, -, L - 1 with the convention that g _pre (-1) is the last attenuation factor obtained for the last sample of the previous sub-block, a is the smoothing coefficient, typically α = 0.85.

Other smoothing functions are also possible. Once the factors

S i ⁿ⁾ ^has i ^ns i calculated attenuation pre-echo is made on the reconstructed signal of the current frame, _rec x (n) by multiplying each sample by the factor:

x (n) = g (n) x (n), n = {), · · ·, L - \

where x _rec (n) is the signal decoded and post-processed by the pre-echo reduction.

Figures 2 and 3 illustrate the implementation of the attenuation method as described in the patent application of the state of the art, cited above, and summarized above.

In these examples the signal is sampled at 32 kHz, the length of the frame is

L = 640 samples and each frame is divided into 8 sub-blocks of K = 80 samples.

In part a) of Figure 2, a frame of an original signal sampled at 32 kHz is shown. An attack (or transition) in the signal is located in the sub-block beginning at the index 320. This signal has been coded by a low rate (24 kbit / s) MDCT type transform coder.

In part b) of Figure 2, the result of decoding without pre-echo processing is illustrated. Pre-echo can be observed from sample 160, in the preceding sub-blocks the one containing the attack.

Part c) shows the evolution of the pre-echo attenuation factor (solid line) obtained by the method described in the aforementioned prior art patent application. The dashed line represents the factor before smoothing. Note here that the position of the attack is estimated around the sample 380 (in the block delimited by the samples 320 and 400).

Part d) illustrates the result of the decoding after application of pre-echo processing (multiplication of signal b) with signal c)). We see that the pre-echo has been attenuated. Figure 2 also shows that the smoothed factor does not go back to 1 at the time of the attack, which implies a decrease in the amplitude of the attack. The noticeable impact of this decrease is very small but can nevertheless be avoided. FIG. 3 illustrates the same example as FIG. 2, in which, before smoothing, the attenuation factor value is forced to 1 for the few samples of the sub-block preceding the sub-block where the attack is located. Part (c) of Figure 3 gives an example of such a correction.

In this example, the value of factor 1 has been assigned to the last 16 samples of the sub-block preceding the attack, starting from the index 364. Thus, the smoothing function progressively increases the factor to have a value close to 1 at the moment. of the attack. The amplitude of the attack is then preserved, as illustrated in part d) of FIG. 3, but some pre-echo samples are not attenuated.

In the example of Figure 3 pre-echo attenuation reduction does not reduce the pre-echo to the level of the attack, because of the smoothing gain.

Another example with the same setting as that of Figure 3 is illustrated in Figure 4. This figure represents 2 frames to better show the nature of the signal before the attack. Here, the energy of the original signal before the attack is stronger (part a) than in the case illustrated in Figure 3, and the signal before the attack is audible (samples 0 - 850). On part b) we can observe the pre-echo on the decoded signal without pre-echo treatment in the 700-850 zone. According to the attenuation limitation procedure previously explained, the signal energy of the pre-echo zone is attenuated to the average energy of the signal preceding the treatment zone. It is observed in part c) that the attenuation factor calculated taking into account the energy limitation is close to 1 and that the pre-echo is always present on part d) after application of the pre-echo treatment ( multiplication of the signal b) with the signal c)), despite the good leveling of the signal in the pre-echo zone. We can indeed distinguish this pre-echo on the waveform where we notice that a high frequency component is superimposed on the signal in this area.

This high frequency component is well audible and annoying, and the attack is less clear (part d) figure 4). The explanation of this phenomenon is as follows: in the case of a very sudden, impulsive attack (as shown in Figure 4) the spectrum of the signal (in the frame containing the attack) is rather white and therefore also contains a lot high frequencies. Thus the quantization noise is also white and composed of high frequencies, which is not the case of the signal preceding the pre-echo zone. There is therefore a sudden change in the spectrum from one frame to the next, which results in an audible pre-echo despite the fact that the energy has been put to the right level.

This phenomenon is again represented in FIGS. 5a and 5b, which respectively show the spectrograms of the original signal at 5a, corresponding to the signal represented in part a) of FIG. 4, and the spectrogram of the signal with attenuation of preechos according to the state of the art, at 5b, corresponding to the signal represented in part d) of FIG. 4.

We notice a pre-echo still audible in the part framed in the figure

5b.

There is therefore a need for an improved attenuation technique for pre-decoding decoding, which also attenuates unwanted high frequencies or parasitic pre-echoes and without any auxiliary information being transmitted by the coder.

The present invention improves the state of the art.

For this purpose, the present invention relates to a pre-echo attenuation processing method in a digital audio signal generated from a transform coding, wherein, upon decoding, the method comprises the following steps:

detecting a driving position in the decoded signal;

determining a pre-echo zone preceding the detected driving position in the decoded signal;

calculating attenuation factors by sub-block of the pre-echo zone, as a function of at least the frame in which the attack has been detected and of the previous frame;

pre-echo attenuation in the sub-blocks of the pre-echo zone by the corresponding attenuation factors. The method is such that it further comprises:

the application of an adaptive spectral shaping filtering of the pre-echo zone on the current frame to the detected position of the attack.

Thus, the spectral shaping applied, improves the pre-echo attenuation. The treatment makes it possible to attenuate the pre-echo components that could remain during the implementation of the pre-echo attenuation as described in the state of the art. The filtering being applied to the detected position of the attack, it makes it possible to process the attenuation of the pre-echo up to the nearest attack. This therefore offsets the disadvantage of temporal attenuation echo control which is limited to a zone not going to the attack position (margin of 16 samples for example).

This filtering does not require information from the encoder.

This pre-echo attenuation processing technique may be implemented with or without knowledge of a signal derived from a time decoding and for the coding of a monophonic signal or a stereophonic signal.

The adaptation of the filtering makes it possible to adapt to the signal and to remove only the disturbing parasitic components.

The various particular embodiments mentioned below may be added independently or in combination with each other, to the steps of the method defined above.

In a particular embodiment, the method further comprises calculating at least one decision parameter on the filtering to be applied to the pre-echo zone and the adaptation of the filtering coefficients according to said at least one parameter of decision.

Thus, the treatment is then applied only when necessary to a suitable level of filtering.

In one embodiment, the at least one decision parameter is a measure of the strength of the detected attack.

The force of the attack determines the presence of audible high-frequency components in the pre-echo zone. When the attack is abrupt, the risk of having an annoying parasitic component in the pre-echo zone is large and the filtering to be implemented according to the invention is then to be expected.

In a possible calculation mode of this parameter, the measurement of the force of the detected attack is of the form:

P = max (EN (k), EN (k + 1) / min (EN (k1), EN (k-2)) with k, the number of the sub-block in which the attack was detected and EN ( k) the energy of the k ^th sub-block.

This calculation is of less complexity and makes it possible to define well the strength of the detected attack.

The said at least one decision parameter may also be the value of the attenuation factor in the sub-block preceding that containing the position of the attack.

Indeed, an attack can be considered abrupt if this attenuation is significant. In another embodiment, said at least one decision parameter is based on a spectral distribution analysis of the signal of the pre-echo zone and / or of the signal preceding the pre-echo zone.

This makes it possible, for example, to determine the importance of the high frequency components in the pre-echo signal and also to know if these high frequency components were already present in the signal before the pre-echo zone.

Thus, in the case where high-frequency components were already present before the pre-echo zone, it is then not necessary to perform a filtering to attenuate these high-frequency components, the adaptation of the filter coefficients is carried out then by setting 0 or a value close to 0 of the filter coefficients.

Thus, the adaptation of the filtering coefficients can be done in a discrete manner as a function of the comparison of at least one decision parameter with a predetermined threshold.

The filter coefficients can take predetermined values according to a set of values. The smallest set of values being one where only two values are possible, it is a saying for example the choice between a filtering and no filtering.

In an alternative embodiment, the adaptation of the filtering coefficients is carried out continuously according to said at least one decision parameter.

The adaptation is then more precise and more progressive.

In a particular embodiment, the filtering is a finite impulse response with a null phase transfer function:

c (n) z ^{~ l} + (l - 2c (n)) + c (n) z

with c (n) a coefficient between 0 and 0.25.

This type of filtering is of low complexity and allows more processing without delay (the processing stops before the end of the current frame). Thanks to its zero delay, the filtering can attenuate the high frequencies before the attack without modifying the attack itself.

This type of filtering makes it possible to avoid discontinuities and makes it possible to pass from an unfiltered signal to a filtered signal in a progressive manner.

According to one embodiment, the attenuation step is performed at the same time as the spectral shaping filtering by integrating the attenuation factors with the coefficients defining the filtering.

The present invention also relates to a pre-echo attenuation processing device in a digital audio signal generated from a transform coder, in which the device associated with a decoder comprises: a detection module for detecting a driving position in the decoded signal;

a determination module for determining a pre-echo zone preceding the detected driving position in the decoded signal;

a module for calculating attenuation factors per sub-block of the pre-echo zone, as a function of at least the frame in which the attack has been detected and of the previous frame;

an attenuation module for attenuating the pre-echoes in the sub-blocks of the pre-echo zone by the corresponding attenuation factors. The device is such that it further comprises:

an adaptive filtering module for effecting a spectral shaping of the pre-echo zone on the current frame to the detected position of the attack.

The invention relates to a decoder of a digital audio signal comprising a device as described above.

Finally, the invention is directed to a computer program comprising code instructions for implementing the steps of the attenuation processing method as described, when these instructions are executed by a processor.

Finally, the invention relates to a storage medium, readable by a processor, integrated or not to the processing device, possibly removable, storing a computer program implementing a method of treatment as described above.

Other features and advantages of the invention will appear more clearly on reading the following description, given solely by way of nonlimiting example, and with reference to the appended drawings, in which:

FIG. 1 previously described illustrates a state-of-the-art transform coding-decoding system;

- Figure 2 described above illustrates an example of a digital audio signal for which a mitigation method according to the state of the art is performed;

- Figure 3 described above illustrates another example of a digital audio signal for which a mitigation method according to the state of the art is performed;

- Figure 4 described above illustrates yet another example of a digital audio signal for which a mitigation method according to the state of the art is performed;

FIGS. 5a and 5b respectively show the spectrogram of the original signal and the spectrogram of the pre-echo attenuation signal according to the state of the art (corresponding to parts a) and d) respectively of FIG. 4); FIG. 6 illustrates a pre-echo attenuation processing device in a digital audio signal decoder, as well as the steps implemented by the processing method according to one embodiment of the invention;

FIG. 7 illustrates the frequency response of a spectral shaping filter implemented according to one embodiment of the invention, as a function of the parameter of the filter;

FIG. 8 illustrates an exemplary digital audio signal for which the processing according to the invention has been implemented;

FIG. 9 illustrates the spectrogram of the signal corresponding to the signal d) of FIG. 4, for which the treatment according to the invention is implemented;

FIG. 10 illustrates an exemplary signal having initially high frequency components for which a pre-echo mitigation method according to the state of the art is implemented;

FIG. 11 illustrates the same signal as FIG. 11, having originally high frequency components for which the processing according to the invention has been implemented without taking into account a decision criterion of the level of filtering. to apply;

FIG. 12 illustrates a hardware example of an attenuation processing device according to the invention.

With reference to FIG. 6, a preecho attenuation processing device 600 is described. In one embodiment, this device implements a pre-echo attenuation method in the decoded signal, for example that described in the patent application FR 08 56248. It also implements a shaping filtering. spectral of the pre-echo zone.

Thus, the device 600 comprises a detection module 601 able to implement a step of detecting (Detect.) The position of an attack in a decoded audio signal.

An onset (or onset) is a rapid transition and a sudden change in the dynamics (or amplitude) of the signal. This type of signal may be referred to by the more general term "transient". In the following and without loss of generality, we will use only the terms of attack or transition to designate also transients.

In one embodiment, each frame of L samples of the decoded signal ^x rec i ⁿ ) ^is divided into K sub-blocks of length L ', with for example L = 640 samples.

(20 ms) at 32 kHz, L '= 80 samples (2.5 ms) and K = 8. Special low-delay analysis-synthesis windows similar to those described in ITU-T G.718 are used for the analysis part and for the synthesis part of the MDCT transformation. Thus, the synthesis window MDCT contains only 415 non-zero samples, unlike the 640 samples in the case of using conventional sinusoidal windows. In a variation of this embodiment, other analysis / synthesis windows may be used, or switches between long and short windows may be used.

Furthermore, the MDCT x _MDCT (n) memory is used which gives a time-folding version ("folding") of the future signal. This memory is also divided into sub-blocks of length L 'and one retains - as a function of the window MDCT used - only the K' first sub-blocks, where K 'depends on the window used - for example K' = 4 for a sinusoidal window. Indeed, Figure 1 shows that the pre-echo influences the frame before the one where the attack is located, and it is desirable to detect an attack in the future frame which is partially contained in the MDCT memory.

The pre-echo reduction here depends on several parameters:

o The decoded signal in the current frame (which potentially contains pre-echoes) of length L,

o The memory of the inverse transformation MDCT which corresponds to the partially decoded signal in the following frame before addition-overlap.

o The average energy level in the previous frame (or half-frame). It can be noted that the signal contained in the MDCT memory includes time folding (which is compensated when the next frame is received). As explained below, the MDCT memory is used here essentially to estimate the energy by sub-blocks of the signal in the next (future) frame and it is considered that this estimate is sufficiently precise for the purposes of the detection and reduction of pre- echo when performed with the available MDCT memory at the current frame instead of the fully decoded signal at the future frame.

The current frame and the MDCT memory can be seen as concatenated signals forming a signal of length (K + K ') L' cut in (Κ + Κ ') consecutive sub-blocks. Under these conditions, the energy in the k-th sub-block is defined as:

(Fc + l), l

In {k) = Σ x _rec (n), k = 0, ..., K - l

n = kV when the k-th sub-block is in the current frame and, like:

(K + K-l) l-

In (k) = Σ x _MDCT (nf, k = K, ..., K + K '

n = (k-K) V

when the sub-block is in the MDCT memory (which represents the signal available for the future frame).

The average energy of the sub-blocks in the current frame is thus obtained as:

l K-l

In = - Ύ In (k)

K _k = o

The average energy of the sub-blocks in the second part of the current frame is also defined as:

K ^{~ l}

In '= -Σ In (k)

K k = K / 2

A transition associated with a pre-echo is detected if the ratio max (En {k))

k = 0, K + K

R (k) = exceeds a predefined threshold, in one of the sub-blocks considered.

In (k)

Other pre-echo detection criteria are possible without changing the nature of the invention.

In addition, we consider

where the limitation to L ensures that the MDCT memory is never changed. Other methods for more accurate estimation of the attack position are also possible.

In alternative embodiments with window switching, other methods giving the position of the attack can be used with a precision ranging from the scale of a sub-block to a position with the sample.

The device 600 also comprises a determination module 602 implementing a determination step (ZPE) of a pre-echo zone preceding the detected driving position.

The energies In [k) are concatenated in chronological order, with the time envelope of the decoded signal first, then the envelope of the signal of the next frame estimated from the memory of the MDCT transform. According to this envelope concatenated temporal and mean energies En and En 'of the previous frame, the presence of pre-echo is detected if the ratio R (k) is sufficiently strong.

The sub-blocks in which a pre-echo has been detected thus constitute a pre-echo zone, which in general covers the samples n = 0, · · ·, pos- 1, ie from the beginning of the current frame to the position of the attack (pos).

In alternative embodiments, the pre-echo zone does not necessarily start at the beginning of the frame, and may involve an estimate of the length of the pre-echo. If window switching is used, the pre-echo zone must be set to take into account the windows used.

A module 603 of the device 600 implements a sub-block attenuation factor calculation step of the determined pre-echo area, depending on the frame in which the attack was detected and the previous frame.

According to the description of the patent application FR 08 56248, the attenuations g (k) are estimated by sub-block. The attenuation factor by sub-block g (k) is calculated for example, as a function of the ratio R (k) between the energy of the sub-block of higher energy and the energy of the k-th sub-block in question

g {k) = f {R (k))

where / is a decreasing function with values between 0 and 1. Other definitions of the factor s {k) are possible, for example according to En {k) and En k-1).

If the variation of the energy with respect to the maximum energy is small, then no attenuation is necessary. The factor is then set to an attenuation-inhibiting attenuation value, i.e. 1. Otherwise, the attenuation factor is between 0 and 1.

These attenuations are limited according to the average energy of the previous frame.

For the sub-block to be processed, the limit value of the factor lim (k) can be calculated in order to obtain exactly the same energy as the average energy of the segment preceding the sub-block to be treated. This value is of course limited to a maximum of 1 since we are interested here in the attenuation values. More precisely :

The value lim _g (k) thus obtained serves as a lower limit in the final calculation of the attenuation factor of the sub-block:

g (k) = max (g (k), lim _g (k))

The attenuation factors g (k) determined by sub-blocks are then smoothed by an applied smoothing function sample by sample to avoid abrupt changes in the attenuation factor at the boundaries of the blocks.

The gain per sample is first defined as a piecewise constant function:

S _pre {n) = g {k), n = kL ', - - -, (k + \) L' - \

The smoothing function is for example defined by the following equations:

8 _pre {n): = g _pre (n -l) + (l-a) g _pre (n), n = 0, - - -, L -l with the convention that g _pre (- l) is the last the attenuation factor obtained for the last sample of the previous sub-block, a is the smoothing coefficient, typically α = 0.85.

Other smoothing functions are possible.

The module 604 of the device 600 of FIG. 6 implements the attenuation (Att.) In the sub-blocks of the pre-echo zone, by the attenuation factors obtained.

Thus, once the factors g _pre [n] computed, the pre-echo attenuation is made on the reconstructed signal of the current frame, x _rec [n), by multiplying each sample by the corresponding factor:

ec, _g ()) = 8 pre ()), = = 0, ^■■ , L - 1

where x _rec (n) is the decoded and post-processed signal for pre-echo reduction.

The device 600 comprises a filtering module 606 capable of performing the step (F) of applying a spectral shaping filtering of the pre-echo zone to the current frame of the decoded signal, to the position detected from the attack.

Typically, the spectral shaping filter used is a linear filter. Since the gain multiplication operation is also a linear operation, their order can be inverted: you can also do the formatting filtering first. Spectrum of the pre-echo area then the pre-echo attenuation by multiplying each sample of the pre-echo area by the corresponding factor.

In an exemplary embodiment, the filter used to attenuate the high frequencies in the pre-echo zone is a FIR filter (finite impulse response filter) with 3 coefficients and with a zero transfer function phase c (ri) z ^{~ l} + (l - 2c (n)) + c (n) z with c (n) a value between 0 and 0.25, where [c (n), l - 2c (n), c (n)] are the coefficients of spectral shaping filter; this filter is implemented with the difference equation:

^X _rec, fi ⁿ⁾ = ^{C n)} ^X rec, _g (n ^~ i) + {l ^~ 2 _C (t)) X _mg (tl) + C (n) X _{g rec} (t + 1)

with for example c (n) = 0.25 on the area n = 5, · · ·, pos- 5.

The frequency response of this filter is illustrated in FIG. 7, as a function of the coefficient c (n), for c (n) = 0.05, 0.1, 0.15, 0.2 and 0.25. The motivation for using this filter is its low complexity, its null phase and therefore its zero delay (possible because the processing stops before the end of the current frame) but also its frequency response which corresponds well to the desired low-pass characteristics for this filter.

The application of this filter can compensate for the fact that the temporal attenuation of the pre-echo is typically limited to a zone that does not go up to the position of the attack (with a margin of, for example, 16 samples), whereas the spectral shaping filtering as defined by the transfer function c (ri) z ^{~ l} + (l-2c (n)) + c (n) z can be applied up to the attack position, with possibly some interpolation samples of the filter coefficients.

To pass from an unfiltered signal to a filtered signal and to avoid discontinuities it is preferable to introduce the filtering in a progressive way. The proposed FIR filter makes it easy to smoothly move from the unfiltered domain to the filtered domain and vice versa, by interpolation or slow variation of its coefficients. For example, if the position of the attack is pos = 16, the filtering of the 16 samples in the pre-echo area n = 0, · · ·, pos - 1 can be performed as follows:

* (0) = ^ (0)

x _rec (1) = 0. \ x _rec (0) + 0.8x _rec (1) + 0. \ x _rec (2) x _rec (2) = 0. \ x _rec (1) + 0.8x _rec (2) ) + 0. \ x _rec (3) x _rec (3) = 0.15x _rec (2) + 0.7x _rec (3) + 0.15x _rec (4) x _rec (4) = 0.2x _rec (3) + 0.6x _rec (4) + 0.2x _rec (5) =

_rec x _f {n) = 0.25x _rec (n - 1) + 0.5x _rec (n) + 0.25x _rec (n + 1), n = 5, ^■■ ·, 11 x _recJ (12) _rec = 0.2x (11) + 0.6x _rec (12) + 0.2x _rec (13)

x _rec (13) = 0.15 x _rec (12) + 0 x _rec (13) + 0.15 x _rec (14)

x _rec (14) = 0. lx _rec (13) + 0.8x _rec (14) + 0. lx _rec (15)

x _recJ (15) = 0.05x _rec (14) + 0.9x _rec (15) + 0.05x _rec (16)

It can be observed that, thanks to its zero delay, the filter c (n) z ¹ + (l ^- 2c (n)) + c (n) z can attenuate the high frequencies before the attack without modifying the attack itself. .

An example of a digital audio signal, for which the processing as described here is carried out, is illustrated in part d) of FIG. 8. Parts a), b) and c) of this figure show the same signals as those described with reference in Figure 4 previously. Part d) differs by the implementation of filtering according to the invention. It can thus be noted that the disturbing high frequency component is greatly reduced, so that the signal decoded after filtering has a better quality than that described in part d) of FIG. 4.

The spectrogram representing this filtered signal is represented in FIG. 9. It is clearly observed with respect to FIG. 5b representing the same signal without shaping filtering, the attenuation of the disturbing high frequencies before the attack. The attack becomes sharper at decoding.

Of course, other types of spectral shaping filters may be considered to replace the filter c (nz ^{~ 1} + (l-2c (n)) + c (n) z ■ For example, it is possible to using a FIR filter of different order or with different coefficients Alternatively the spectral shaping filter can be infinite impulse response (IIR) In addition, the spectral shaping may be different from a pass filtering, by example a pass4 ande filter could be implemented.

A filter of order 1, of the form c (nz ^{~ 1} + (l ^- c (n)) can also be used in one embodiment of the invention.

In a particular embodiment, the filtering implemented according to the method described is an adaptive filtering. It can thus be adapted to the characteristics of the decoded audio signal. In this embodiment, a step of calculating a decision parameter (P) on the filtering to be applied to the pre-echo zone is implemented in the calculation module 605 of FIG. 6.

Indeed, there are cases like that illustrated for example in Figure 10 where it is preferable not to apply such filtering in the pre-echo area.

Indeed, in the case, rarer, illustrated in Figure 10, part a) the high frequencies are already present in the signal to be coded. In this case the attenuation of high frequencies could cause an audible degradation which must be avoided. In this signal example, it is observed that the attack is less abrupt than in the previous examples.

It is therefore advantageous to determine at least one parameter which makes it possible to decide whether to spectrally shape the zone of the signal containing a pre-echo, by attenuating (or not) the high frequencies.

In an exemplary embodiment, this decision parameter is representative of the presence of high frequency components in the pre-echo zone.

This parameter can be for example a measure of the strength of the attack (abrupt or not). If the attack is located in sub-block number k, the parameter can be calculated as:

max (In (k), In (k + 1))

~ min (In (k -i), In (k -2))

where k is the number of the sub-block and En (k) is the energy in the k-th sub-block.

According to an experimental setting, in this exemplary embodiment, P> = 32 indicates a sudden attack (very impulsive).

The force measurement of the attack can be completed taking into account also the attenuation determined for the sub-block preceding the attack g [k-1). An attack can be considered abrupt if this attenuation is significant, for example if g [k - l) <0.5. This shows that the energy in the pre-echo zone is considerably increased (more than doubled) because of the pre-echo, which also signals a sudden attack.

If P <32 and g (k - l)> 0.5, where k is the index of the sub-block containing the beginning of the attack, then filtering is not necessary. Indeed, if g (k-1)> 0.5, lim (fc)> 0.5, which means that the pre-echo zone has an energy comparable to that of the previous frame and as the attack that generates the pre-echo echo is not abrupt, the risk of having an annoying parasitic component is low. Thus, in this embodiment with the conditions (P <32 and gk - 1)> 0.5), no filtering will be done on the pre-echo area.

In the other cases (g (k - 1) <0.5 or P> 32), the spectral shaping filter is applied, according to the invention, from the beginning of the current frame to the position _p0 s of position of the 'attack.

In the exemplary embodiment described above, the spectral shaping of the pre-echo zone by filtering according to the invention is adaptive as a function of the parameter P and the attenuation values. Thus, the filtering is either applied with coefficients [0.25, 0.5, 0.25], or deactivated with coefficients [0, 1, 0].

The filter coefficients are then adapted in a discrete manner limited to a predefined set of values.

The adaptation of the filter coefficients (making it possible to adapt the attenuation level of the high frequencies) is thus determined by decision parameters which measure the force of the attack, such as the parameters P and g [k-1). is in this case one of adaptation of the coefficients of the filter in a discrete manner according to two sets of possible values ([0.25, 0.5, 0.25] or [0, 1, 0]). It can be noted that the set of coefficients [0, 1, 0] corresponds to a deactivation of the filtering.

A progressive transition between these two filters can be carried out using also for example the intermediate filters of coefficient [0.05, 0.9, 0.05], [0.1, 0.8, 0.1], [0.15, 0.7, 0.15] and [0.2, 0.6, 0.2 ].

In this case, it is an adaptation of the coefficients of the filter discretely according to several sets of possible values, if one takes into account the slow variation (or interpolation).

In alternative embodiments, other interpolation methods may be used.

For example, the filtering can be even more finely adaptive with c (n) = f (P) for example by using an intermediate filter with c (n) = [0.15, 0.7, 0.15] if 16 <P <32. c (n) can also be calculated continuously as a function of P, for example with

. , arctan (P / 10)

formula c (n) =

'2π

It is in this case an adaptation of the filter coefficients continuously according to possible values where c (n) is in the interval [0, 0.25]. Other decision parameters can also be used in the choice and adaptation of the filter, for example the zero crossing rate of the decoded signal of the pre-echo zone. the current frame and / or the previous frame. The rate of zero crossing can be calculated as follows if we consider the area n = 0, · · ·, L-1 as an example:

OR

Indeed, a high rate of zero crossing zc in the previous frame (so without pre-echo) signals the presence of high frequencies in the signal. In this case, for example when zc> L / 2 on the previous frame, it is preferable not to apply the filtering c (n) z ^{~ l} + (l - 2c (n)) + c (n) z.

In order to eliminate the bias of the DC component, a pre-filtering of the decoded signal is also possible before calculating the zero crossing rate, or the number of zero crossings of the estimated derivative x _{rec g} {n) - x _rec [ n- 1) can be used.

In a variant, a spectral analysis of the signal can also be made to assist the decision. For example, the spectral envelope in the MDCT domain resulting from the MDCT coding / decoding can be exploited in the choice of the filter to be used, however this variant assumes that the MDCT analysis / synthesis windows are sufficiently short for the local statistics of the MDCT to be used. signal before the attack remain stable over the length of a window.

Alternatively, it will be possible to filter the signal in the pre-echo zone and in the frame passed through a high-pass complementary filter as with for example c (n) = 0.25, and then the value of c (n) will be chosen so that the average energy of the filtered signals in the pre-echo zone and on the past frame are as close as possible; the choice of c (n) can be made on a limited set of possible values shown in FIG. 7 or on the basis of the energy ratio (or of an equivalent quantity such as the square root of the energy) of the signal after filtering high pass in the pre-echo area and in the past frame. Note that the high-pass filtering can also be implemented alternately by calculating the difference between the signal x _{rec g} (n) and the signal filtered by the low-pass filter c (n) z ^{~ l} + (l - 2c (n)) + c (n) z when c (n) = 0.25.

In another variant, when the shaping filtering is of type c (n) _Z - ^l + {\ - c (n)), we can set the value of c (n) according to the prediction coefficient -r (l) / r (0) from a linear prediction analysis (LPC for "Linear

Predictive Coding "in English) to the order 1 of the signal in the pre-echo area and the signal in the past frame.

In all these latter variants (zero crossing rate, MDCT spectral envelope, high pass filtering, LPC analysis), the decision parameter on the filtering to be applied to the pre-echo zone is based on a spectral distribution analysis of the signal the pre-echo zone and / or the preceding signal of the pre-echo zone; if the signal preceding the pre-echo zone already contains many high frequencies or if the amount of the high frequencies of the signal in the pre-echo zone and the signal preceding the pre-echo zone is substantially identical, the filtering according to the invention is not necessary and may even cause slight degradation. In these cases, the filtering according to the invention must be deactivated or attenuated by setting c (n) to 0 or to a low value close to 0.

In a variant of the invention the order between the attenuation and filtering step may be reversed.

It may indeed be that the filtering (F) of spectral shaping is done before the attenuation (Att.). Thus, after having performed the adaptive filtering of the samples of the pre-echo zone of the reconstructed signal of the current frame, these samples are then weighted by multiplying each sample by the corresponding attenuation factor calculated previously:

Xrec, f, _g { ⁿ ) ⁼ gpre ( ⁿ Kec, f { ⁿ )>"= 0, · · ·, L - 1 1

The attenuation of the amplitudes can also be combined (or integrated) by defining a set of "conjoint" filter coefficients, for example if for the sample n the filter has coefficients [c (ri), l-2c (") , c (ri)] and the attenuation factor is g (ri), we can directly use the filter [g _pm (n) c (n), g _pre (n) 2 g _pm (n) c (n) , g _pre (n) c (n)].

Figure 11 illustrates the advantage of making adaptive filtering. It uses the same signals parts a), b) and c) as Figure 10 and illustrates that the implementation of the nonadaptive filtering shown in part d), unnecessarily modifies the signal in the case where the high-frequency components are already present in the signal to be encoded. It is observed that from sample 640 the high frequencies are unnecessarily attenuated, which could lead to a slight deterioration in quality. The use of an adaptive filtering as described above makes it possible to inhibit or attenuate the filtering under these conditions, not to remove high frequencies already present in the signal to be coded and thus to avoid possible degradation due to the filtering.

Returning to FIG. 6, the attenuation processing device 600 as described is here included in a decoder comprising a reverse quantization module 610 (Q ¹ ) receiving a signal S, a reverse transformation module 620 (MDCT ¹ ) , a module 630 for reconstruction of the addition / recovery signal (add / rec) as described with reference to FIG. 1 and delivering a reconstructed signal to the attenuation processing device according to the invention.

At the output of the device 600, a processed signal Sa is provided in which a pre-echo attenuation has been performed. The processing performed improved the pre-echo attenuation by attenuating, if necessary, the high-frequency components in the pre-echo area.

An exemplary embodiment of an attenuation processing device according to the invention is now described with reference to FIG. 12.

Materially, this device 100 in the sense of the invention typically comprises a μΡ processor cooperating with a memory block BM including a storage and / or working memory, and a memory buffer MEM mentioned above as a means for storing all data. necessary to implement the attenuation processing method as described with reference to Figure 6. This device receives as input successive frames of the digital signal Se and delivers the reconstructed signal Sa with pre-echo attenuation and filtering of Spectral shaping, if any.

The memory block BM may comprise a computer program comprising the code instructions for implementing the steps of the method according to the invention when these instructions are executed by a processor μΡ of the device and in particular a step of detecting a position of etching in the decoded signal, of determining a pre-echo area preceding the detected attack position in the decoded signal, of calculating sub-block attenuation factors of the pre-echo area, as a function of the frame in which the attack has been detected and the previous frame, pre-echo attenuation in the sub-blocks of the pre-echo zone by the corresponding attenuation factors and further, a step of applying spectral shaping filtering of the pre-echo area on the current frame to the detected position of the attack. Figure 6 can illustrate the algorithm of such a computer program.

This attenuation device according to the invention can be independent or integrated into a digital signal decoder.

Claims

A method of pre-echo attenuation processing in a digital audio signal generated from a transform coding, wherein, upon decoding, the method comprises the following steps:

detection (Detect.) of a driving position in the decoded signal;

determination (ZPE) of a pre-echo zone preceding the detected attack position in the decoded signal;

- calculation (F. Att.) of attenuation factors per sub-block of the pre-echo area, based at least on the frame in which the attack was detected and the previous frame;

pre-echo attenuation (Att.) in the sub-blocks of the pre-echo zone by the corresponding attenuation factors;

the method being characterized in that it further comprises:

applying adaptive spectral shaping filtering (F) from the pre-echo area to the current frame to the detected position of the attack.

2. Method according to claim 1, characterized in that it further comprises the calculation of at least one decision parameter on the filtering to be applied to the pre-echo zone and the adaptation of the filtering coefficients according to said at least one decision parameter.

3. Method according to claim 2, characterized in that said at least one decision parameter is a measure of the strength of the detected attack.

4. Method according to claim 2, characterized in that said at least one decision parameter is the value of the attenuation factor in the sub-block preceding that containing the attack position.

5. Method according to claim 2, characterized in that said at least one decision parameter is based on a spectral distribution analysis of the signal of the pre-echo zone and / or of the signal preceding the pre-echo zone.

6. Method according to claim 3, characterized in that the measurement of the force of the detected attack is of the form: P = max (EN (k), EN (k + 1) / min (EN (k1), EN (k-2)) with k, the number of the sub-block in which the attack was detected and EN ( k) the energy of the k ^th sub-block.

7. Method according to claim 2, characterized in that the adaptation of the filtering coefficients is performed discretely as a function of the comparison of at least one decision parameter to a predetermined threshold.

8. Method according to claim 2, characterized in that the adaptation of the filtering coefficients is performed continuously according to said at least one decision parameter.

9. The method as claimed in claim 1, characterized in that the filtering has a finite impulse response with a zero transfer function phase:

c (n) z ^{~ l} + (l - 1c (n)) + c (n) z

with c (n) a coefficient between 0 and 0.25.

10. The method as claimed in claim 1, characterized in that the attenuation step is performed at the same time as the spectral shaping filtering by integrating the attenuation factors with the coefficients defining the filtering.

11. Pre-echo attenuation processing device in a digital audio signal generated from a transform coder, wherein the device associated with a decoder comprises:

a detection module (601) for detecting a driving position in the decoded signal;

a determination module (602) for determining a pre-echo area preceding the detected driving position in the decoded signal;

a module for calculating (603) attenuation factors per sub-block of the pre-echo zone, as a function of at least the frame in which the attack has been detected and of the previous frame;

an attenuation module (604) for attenuating the pre-echoes in the sub-blocks of the pre-echo zone by the corresponding attenuation factors;

the device further comprising: an adaptive filtering module (606) for spectrally shaping the pre-echo area on the current frame to the detected position of the attack.

12. Decoder of a digital audio signal comprising a device according to claim 11.

Computer program comprising code instructions for implementing the steps of the method according to one of claims 1 to 10, when these instructions are executed by a processor.