EP1815622B1 - Method for broadcasting a digital signal transmitted in the vicinity of an analog signal, and corresponding broadcasting device and digital signal. - Google Patents

Description

1. Field of the invention

The field of the invention is that of the diffusion of signals, in particular in systems of the DAB (Digital Audio Broadcasting) type, or DRM (for "Digital World Radio"), in frequency bands below 30 MHz.

More specifically, the invention relates to the diffusion, in the same propagation channel, of analog and digital signals, for example of the audio type.

The invention applies in particular, but not exclusively, to the simultaneous broadcasting, from the same transmission point, called "simulcast" broadcasting, of a digital signal of the DRM type, implementing an OFDM type modulation, and an analog signal carrying an audio signal modulating the amplitude of a carrier AM ("Amplitude Modulation" in English, "Amplitude Modulation" MA in French).

2. The prior art 2.1. Reminder on the principle of OFDM modulation

The OFDM modulation consists in distributing in the time / frequency space symbols of duration Tu (called useful symbol time) on a plurality of independently modulated carrier frequencies, for example in QPSK (of the English "Quadrature Phase Shift Keying" for "Quadrature phase shift keying") or QAM ("Quadrature Amplitude Modulation" for "Quadrature Amplitude Modulation") including 16, 64, 256, ..., states. The OFDM thus cuts the channel into cells along the axes of time 11 and frequencies 12, as shown in FIG. figure 1 . Each of the carriers is orthogonal to the previous one.

The channel, of predetermined width 13, is decomposed into a sequence of frequency sub-bands 14 and a sequence of time segments 15 (also called time slots).

Each time / frequency cell is assigned a dedicated carrier. We go therefore distribute the information to be transported on all of these carriers, each modulated at a low rate, for example by QPSK or QAM type modulation. An OFDM symbol comprises all the information carried by all the carriers at a time t.

• l'interférence intra symbole : addition d'un symbole avec lui-même légèrement déphasé ;
• l'interférence inter symbole : addition d'un symbole avec le suivant et/ou le précédent légèrement déphasé.

This modulation technique is particularly effective in situations where multipaths are encountered. So, as illustrated on the figure 2 which presents a set of OFDM symbols 21, the same sequence of symbols arriving at a receiver by two different paths is presented as the same information arriving at two different times and which add up. These echoes cause two types of defects:

• intra-symbol interference: addition of a symbol with itself slightly out of phase;
• inter-symbol interference: addition of a symbol with the next and / or the preceding slightly out of phase.

Between each transmitted symbol, a "dead" zone called guard interval 22 is inserted, the duration Δ of which is chosen to be sufficiently large relative to the spreading of the echoes. These precautions will limit inter-symbol interference (the latter being absorbed by the guard interval). Thus each OFDM symbol 21 includes a guard interval 22 and data 23.

At the reception, the carriers may have undergone either an attenuation (destructive echoes), an amplification (constructive echoes) and / or a phase rotation.

In order to calculate the transfer function of the channel, in order to equalize the signal before demodulation, synchronization pilot carriers (which are often of a higher amplitude than the useful data carriers), also called reference pilots, are also inserted. The value and location of these reference drivers in the time / frequency space are predefined and known to the receivers.

After interpolation in time and frequency, we obtain an estimate of the response of the channel, more or less relevant depending on the number reference pilots and their distribution in the time / frequency domain.

Thus, the reference pilots inserted into the multicarrier signal are used to estimate the propagation channel. The estimation of the propagation channel notably makes it possible to correct the received data, also called data pilots, at the receiver (equalization), and to obtain the impulse response of the propagation channel. The impulse response obtained can then be used to refine the temporal synchronization of the receiver (s).

2.2. Application in AM (DRM) bands

OFDM modulation is increasingly used in digital broadcasting because it is very well adapted to variations in the radio channel, which are mainly related to echoes and the Doppler effect. It has been selected for digital broadcasting in the AM (DRM) bands.

The figure 3 thus presents the OFDM structure in A mode of a set of DRM symbols, illustrating the distribution of the reference pilots 31 in the time / frequency space. This structure is described in particular in the ETSI ES 201 980 DRM standard.

2.3. The simulcast broadcast

Digital broadcasting, for example of the DRM type, gradually replaces the analog AM broadcast in the frequency bands below 30 MHz, which requires, during at least one transition period, the broadcasting (possibly simultaneous) of digital and analogue signals carrying a digital signal. similar information from the same site.

The diffusion of these two signals, carried by the same carrier frequency, in the same propagation channel, causes a degradation of the audio signal reproduced by the analogical receivers. Indeed, the digital signal disturbs the analog signal, and hinders its demodulation.

A first technique for broadcasting such signals has been proposed by D. Schill and J. Wildhagen, in the document EP 1 276 257 . According to this technique, a digital signal and an analog signal are combined in a so-called signal simulcast, the digital signal being modulated on a sideband of a carrier of the propagation channel and a correction signal being modulated on the other sideband.

According to this technique of the prior art, the correction signal is determined so that the envelope of the received signal corresponds to the envelope of a double band analog signal.

This technique thus makes it possible to broadcast in a same channel a digital signal and an analog signal, by modifying the analog part of the simulcast signal so that the overall envelope of this signal is seen as a dual band AM signal for a conventional analog receiver.

It is also known from the document EP 1 041 756 a technique for transmitting audio signals to avoid interference that may occur when two digital radio broadcast systems are locally close.

2.4. Disadvantages of the techniques of the prior art

According to the technique presented in the document EP 1 276 257 , the simulcast signal is obtained by successive iterations, so as to maintain the envelope modulation of the analog signal, that is to say so that the global simulcast signal is comparable to a dual-band AM signal in the time domain.

A receiver demodulating the signal envelope is thus only slightly disturbed by the digital part of the simulcast signal.

However, such an iterative technique is non-linear, and requires a fine-tuning of the parameters to obtain a satisfactory result, especially in terms of degradation for a demodulated audio signal. Therefore, this technique requires significant computing power for real-time processing.

Another major disadvantage of this technique is that it is not suitable for receivers using synchronous demodulation.

Moreover, it can be noted that this technique only modifies the analog part of the simulcast signal. It adds a signal compression analog (eg audio) to minimize the contribution of the digital signal, such a digital signal disturbing the desired signal by the analog receivers (that is to say, changing its "color").

3. Objectives of the invention

The invention particularly aims to overcome these disadvantages of the prior art.

More precisely, an object of the invention is to provide a technique for broadcasting, in the same propagation channel, a digital signal and an analogue signal, which is simple and effective.

Another objective of the invention is to propose such a technique making it possible to reduce, on reception, the influence of the digital signal on the analog signal.

In particular, an object of the invention is to provide such a technique for improving the reception performance of data at a receiver, and the audio signal restored.

Yet another object of the invention is to provide such a technique which is easy to implement, while remaining at a reasonable cost.

A further object of the invention is to provide such a technique having good performance with a receiver implementing a synchronous demodulation.

4. Main features of the invention

These objectives, as well as others that will appear later, are achieved by a method of broadcasting a digital signal transmitted in the vicinity of an analog signal carrying a source audio signal modulating the amplitude of an AM carrier (of the English "Amplitude Modulation", in French "Modulation Amplitude").

According to the invention, such a method comprises a step of amplitude modulation in the frequency domain of the digital signal by a masking signal representative of a psycho-acoustic masking curve of the source audio signal.

Thus, the invention proposes to calculate a masking level associated with a "conventional" audio signal carried by an analog signal and to apply this masking to the digital signal that it is desired to transmit in the same channel. propagation that the analog signal, so that the latter is not (or little) disturbed on large portions according to a psycho-acoustic criterion.

The analog signal received at a receiver is then undisturbed by the digital signal modulated by the masking signal, and can be easily demodulated.

It can thus be considered that the digital signal is "hidden" in the analog signal by its modulation by the masking signal.

The invention is particularly remarkable in that the digital and analog signals are transmitted in the same propagation channel.

Advantageously, the masking signal corresponds to the masking curve, after low-pass filtering in the frequency domain and / or in the time domain.

Such low-pass filtering makes it possible in particular to smooth the masking curve, and thus to improve (or at least not to degrade) the channel estimation implemented in reception of the digital signal.

Preferably, the masking signal always remains greater than a predetermined threshold.

In this way, the loss of information carried by the digital signal during silence in the source audio signal is avoided.

Indeed, the audio signals, in particular the speech signals, are generally not continuous in time.

During a silence in the audio signal, the corresponding masking curve being almost zero, the modulation of the digital signal by the masking signal would produce a virtually zero digital signal over longer or shorter time intervals. As a result, the information carried by the digital signal would be lost in the absence of a predetermined threshold below which said masking signal should not descend.

Advantageously, the masking signal is amplified according to a predetermined gain.

It is thus possible to amplify the masking signal to increase the power of the digital signal. By amplifying the masking signal, the reception quality of the digital signal is improved. However, if the digital signal strength is too great, the digital signal may be heard at an analog receiver. We must therefore find a compromise to improve the quality of reception of the digital signal without deteriorating the quality of reception of the analog signal.

The invention is also remarkable in that the digital signal and the analog signal are synchronized.

The synchronization between the digital signal, modulated by the masking signal, and the analog signal thus makes it possible to optimally reduce the contribution of the digital signal to the analogue signal on reception.

Indeed, during a perfect synchronization between the analog and digital signals, the modulated digital signal is entirely "hidden" in the analog signal, and the analog receivers are not (or very little) disturbed by the digital signal.

Preferably, the digital and analog signals carry the same source audio signal.

Thus, any receiver, whether of analog or digital type, can effectively receive and demodulate the source audio signal.

This feature is particularly advantageous for the transition period between the transition from analogue to digital broadcasting. Thus, users of digital receivers will be able to receive and demodulate information broadcast from a transmission site (source audio signal), and users of old analog receivers will also be able to receive the same information broadcast from the same transmission site, without have to change receiving device.

Advantageously, the digital signal is a multicarrier signal.

Such a signal, for example of the OFDM type, is well adapted to the variations of the radio channel, essentially related to multipaths. This modulation has In particular, it has been selected for digital broadcasting in the AM (DRM) bands. The invention can thus be applied to a digital signal of the DRM type.

• transformation mathématique du signal audio source du domaine temporel vers le domaine fréquentiel ;
• détermination d'un masque absolu, correspondant à un seuil d'audition pour chacune des fréquences correspondant aux composantes délivrées par la transformation mathématique ;
• identification des composantes tonales et des composantes non-tonales parmi les composantes ;
• décimation des composantes, par suppression des composantes inférieures au masque absolu ;
• calcul de masques individuels pour un nombre prédéterminé de composantes ;
• calcul du signal de masquage correspondant à une somme linéaire des masques individuels.

Preferably, the masking curve is obtained according to at least one of the following steps:

• mathematical transformation of the source audio signal from the time domain to the frequency domain;
• determining an absolute mask, corresponding to a hearing threshold for each of the frequencies corresponding to the components delivered by the mathematical transformation;
• identification of tonal components and non-tonal components among components;
• decimation of the components, by deleting the components lower than the absolute mask;
• calculating individual masks for a predetermined number of components;
• calculation of the masking signal corresponding to a linear sum of the individual masks.

According to the invention, the masking curve is obtained by adapting the conventional method to the bandwidth of the audio signal in the AM bands. A "classic" psychoacoustic masking curve that can be used is described in particular in the MPEG1 or MPEG2 specification.

The invention also relates to a device for broadcasting a corresponding digital signal.

Such a device comprises frequency modulation means in the frequency domain of the digital signal by a masking signal representative of a psycho-acoustic masking curve of the source audio signal.

The invention also relates to a digital signal intended to be transmitted in the vicinity of an analog signal carrying a source audio signal modulating the amplitude of a corresponding AM carrier.

The advantages of the broadcasting device and the corresponding signal are the same as those of the diffusion method. Therefore, they are not detailed further.

5. List of figures

• les figures 1 et 2, précédemment décrites en préambule, illustrent le principe général de la modulation OFDM à partir d'une représentation temps/fréquence d'un canal OFDM ;
• la figure 3, également décrite en préambule, présente un exemple de structure OFDM pour un ensemble de symboles DRM en mode A ;
• la figure 4 illustre une diffusion simulcast d'un signal numérique et d'un signal analogique portant un signal audio, modulés sur une même fréquence porteuse, selon l'invention ;
• la figure 5 présente l'algorithme de diffusion simulcast de la figure 4, à l'aide d'un schéma bloc ;
• les figures 6A à 6D illustrent, dans le domaine fréquentiel, les différentes sous-étapes du calcul de la courbe de masquage intervenant dans la modulation du signal numérique.

Other characteristics and advantages of the invention will appear more clearly on reading the following description of a preferred embodiment, given as a simple illustrative and nonlimiting example, and the appended drawings, among which:

• the Figures 1 and 2 , previously described in the preamble, illustrate the general principle of OFDM modulation from a time / frequency representation of an OFDM channel;
• the figure 3 , also described in the preamble, shows an exemplary OFDM structure for a set of mode A DRM symbols;
• the figure 4 illustrates a simulcast broadcast of a digital signal and an analog signal carrying an audio signal, modulated on the same carrier frequency, according to the invention;
• the figure 5 presents the simulcast diffusion algorithm of the figure 4 , using a block diagram;
• the Figures 6A to 6D illustrate, in the frequency domain, the different substeps of the calculation of the masking curve involved in the modulation of the digital signal.

6. Description of an embodiment of the invention

The general principle of the invention is based on the modulation of a digital signal by a masking signal from an analog signal carrying an audio signal, so that the digital signal does not disturb the reception of the analog signal, when the signal analog and digital signal are transmitted in the same channel.

In other words, the invention proposes to "hide" the digital signal in the analog signal so that the analog receivers can demodulate simple and effective way the audio signal, and without degrading the performance of receiving the digital signal.

We present, in relation to the figure 4 , a preferred embodiment of the invention, in which the digital signal, of the DRM type, and an analog signal carrying an audio signal modulating the amplitude of an AM carrier 43 are transmitted in the same channel of propagation of 9 or 10 kHz bandwidth, which corresponds to the bandwidth allocated by the ITU ("International Telecommunications Union", in English "International Telecommunications Union") for broadcasting AM (amplitude modulation) in the frequency bands below 30 MHz.

According to this preferred embodiment, the amplitude modulated digital signal 41 by the masking signal, and the analog signal 42 are simultaneously broadcast (simulcast mode), and are carried by the same carrier frequency 43, in the AM bands.

Thus, a first sideband 44 of the carrier 43 carries the analog signal, and the other sideband 45 carries the modulated digital signal.

Such a diffusion of the digital signal modulated by the masking signal 41 and the analog signal 42 on the same carrier frequency 43 is implemented, according to the invention, from a broadcasting method as shown in the block diagram. of the figure 5 .

• calcul (52) de la courbe de masquage, dans le domaine fréquentiel, à partir du signal analogique 42 portant un signal audio modulant l'amplitude de la porteuse AM (porteuse 43) ;
• filtrage (53) dans le domaine fréquentiel et/ou temporel de la courbe de masquage, délivrant le signal de masquage ;
• détermination d'un seuil minimal et ajout d'un gain(54) à appliquer au signal de masquage pour augmenter la puissance du signal numérique, de type DRM dans ce mode de réalisation préférentiel, ;
• modulation en amplitude, en fonction du signal de masquage, des porteuses de chaque symbole OFDM du signal numérique (55) ;
• synchronisation de la diffusion du signal analogique 42 et du signal DRM modulé 41 par le signal de masquage, dans le domaine temporel (57).

More specifically, the method of broadcasting a digital signal transmitted in the vicinity of an analog signal according to the invention implements the following steps:

• calculating (52) the masking curve, in the frequency domain, from the analog signal 42 carrying an audio signal modulating the amplitude of the carrier AM (carrier 43);
• filtering (53) in the frequency and / or time domain of the masking curve, delivering the masking signal;
• determining a minimum threshold and adding a gain (54) to be applied to the masking signal to increase the digital signal strength, of the DRM type in this preferred embodiment,;
• amplitude modulation, as a function of the masking signal, of the carriers of each OFDM symbol of the digital signal (55);
• synchronization of the diffusion of the analog signal 42 and the modulated DRM signal 41 by the masking signal, in the time domain (57).

6.1 Calculation of the masking curve (52)

Thus, a first step (52) of the invention is based on the calculation of the overall masking curve of the analog signal carrying the source audio signal modulating the amplitude of the AM carrier that is to be broadcast. Remember that the bandwidth of the audio signal is 4.5 or 5kHz.

We now present in relation with Figures 6A to 6D the different sub-steps of the calculation of the masking curve involved in the modulation of the digital signal.

This calculation requires an adaptation of the "classical" psycho-acoustic masking curve model resulting from the MPEG1 or MPEG2 specification, to adapt to the bandwidth of the audio signal in the AM bands.

• conversion temps-fréquence par transformée de Fourier;
• détermination du masque absolu ;
• mise en évidence des composantes tonales et non-tonales ;
• décimation des différentes composantes ;
• calcul des masques individuels ;
• calcul du masque global.

This step 52 of calculation of the masking curve implements the following substeps:

• time-frequency conversion by Fourier transform;
• determination of the absolute mask;
• highlighting tonal and non-tonal components;
• decimation of the different components;
• calculation of individual masks;
• calculation of the global mask.

6.1.1 Time-frequency conversion

The analog signal carrying a source audio signal modulating the amplitude of the AM carrier is first transposed into the frequency domain, from a Fourier Transform (FFT) type mathematical transformation (Fast Fourier Transform).

More precisely, the calculation of the masking curve requiring an estimation of the power spectral density for each of the frequencies of the audio signal, is implemented a Fourier transformation at 1024 samples for a sampling frequency of 48 kHz in order to obtain the different frequencies and their power spectral densities. The transform is calculated from the input analog signal to which a Hanning window is applied.

• Fenêtre de Hanning : $$h\left(i\right)=\frac{1}{2}\sqrt{{}^8{/}_3}\left[1-\mathrm{<figure-callout\; id="2"\; label="cos"\; filenames="imgf0003.png,imgf0004.png"\; state="\{\{state\}\}">cos</figure-callout>}\left(2\pi {/}_{\mathrm{<figure-callout\; id="0"\; label="N\; i"\; filenames="imgf0002.png"\; state="\{\{state\}\}">N</figure-callout>}}_i^{}\right)\right]\mathit{0}\le i\le N-\mathit{1};$$
  
• Densité spectrale de puissance : $$X\left(k\right)=\mathit{10}\log {\left|\frac{1}{N}{\displaystyle \sum _{l=0}^{N-\mathit{1}}}h\left(l\right)s\left(l\right)e^{-2\mathit{jk\pi }{/}_{\mathrm{<figure-callout\; id="2"\; label="N\; l"\; filenames="imgf0003.png,imgf0004.png"\; state="\{\{state\}\}">N</figure-callout>}}_l^{}}\right|}^2\mathit{dB\; k}=0,\dots ,{}^N{/}_2;$$
  

où N correspond à la largeur de la bande passante (ici, N est égal à 9 ou 10 kHz), et l'indice k correspond à la position de la composante par rapport à la fréquence centrale AM de la bande passante.The definition of a Hanning window and the expression of the power spectral density are described below:

• Hanning Window: $$h\left(i\right)=\frac{1}{2}\sqrt{{}^8{/}_3}\left[1-\mathrm{<figure-callout\; id="2"\; label="cos"\; filenames="imgf0003.png,imgf0004.png"\; state="\{\{state\}\}">cos</figure-callout>}\left(2\pi {}^i{/}_{\mathrm{NOT}}\right)\right]\mathit{0}\le i\le \mathrm{NOT}-\mathit{1};$$
  
• Spectral power density: $$X\left(k\right)=\mathit{10}\mathrm{<figure-callout\; id="1"\; label="log"\; filenames="imgf0003.png"\; state="\{\{state\}\}">log</figure-callout>}{\left|\frac{1}{\mathrm{NOT}}{\displaystyle \underset{l=0}{\overset{\mathrm{NOT}-\mathit{1}}{\Sigma }}}h\left(l\right)s\left(l\right)e^{-2\mathit{jk\pi }{/}_{\mathrm{<figure-callout\; id="2"\; label="NOT\; l"\; filenames="imgf0003.png,imgf0004.png"\; state="\{\{state\}\}">NOT\; </figure-callout>}}_l^{}}\right|}^2\mathit{dB\; k}=0,...,{}^{\mathrm{NOT}}{/}_2;$$
  

where N is the width of the bandwidth (here N is equal to 9 or 10 kHz), and the index k is the position of the component relative to the center frequency AM of the bandwidth.

6.1.2 Determination of the absolute mask

The second sub-step of the calculation of the masking curve of the analog signal requires the determination of an absolute mask, corresponding to the hearing threshold for each of the frequencies corresponding to the components delivered by the Fourier transform.

The hearing thresholds for the different frequencies, denoted LTq (k), come from the MPEG1 standard, and are presented in Appendix 1.

6.1.3 Highlighting the tonal and non-tonal components

The third substep implements a discrimination between the tonal components and the non-tonal components.

Indeed, the tone of a component has an influence on the contribution of the component in the calculation of the masking curve.

The separation of the tonal and non-tonal components takes place in several phases: determination of the local maxima, extraction of the tonal components, and calculation of the intensity of the non-tonal components of a critical band.

(i) Determination of local maxima

By definition, a component of the spectral density obtained by the Fourier transform will be called local maximum if: $$X\left(k\right)>X\left(k-1\right)\mathrm{and}X\left(k\right)\ge X\left(k+\mathit{1}\right)$$

According to the preferred embodiment of the invention, the frequency components are analyzed in packets of three, and if the central component has a higher level than the other two, it is considered as a local maximum.

The Figure 6A illustrates in particular the technique for determining local maxima, based on an example of a frequency spectrum 61 of the analog signal.

The first three frequency components (carriers 1, 2 and 3) of a portion 61 of the spectrum of the analog signal are first analyzed. Here, the central component (carrier 2) having a level, that is to say a spectral power density, higher than the other two (carriers 1 and 3), it is considered a local maximum. This analysis is then repeated for the other frequency components (carriers 2, 3 and 4, then carriers 3, 4 and 5, ...). Finally, in this example, the local maxima correspond to the carriers 2, 5, 7, 13, 15 and 17.

(ii) Extraction of tonal components

The tonal components correspond to the grouping of the lines contributing to the same harmonic component.

avec j tel que :

• si 2 < k < 63 , alors j ∈ {-2,2} ;
• si 62 < k < 127, alors j ∈ {-3,-2,2,3}.

By definition, a local maximum is considered to belong to the list of tonal components if the following conditions are met: $$X\left(k\right)-X\left(k+j\right)\ge 7\mathbf{dB}$$

with j such that:

• if 2 <k <63, then j ∈ {-2,2};
• if 62 < k < 127 , then j ∈ {-3, -2,2,3}.

In fact, to take account of the best frequency resolution of the ear in the low frequency region, the width of the analysis band is modified around the local maxima determined according to the position of the maxima in the spectrum, which implies that the value of j varies according to the values of k.

Thus, as defined in the MPEG1 specification, the analysis width is modified as follows: Δf (Hz) Frequency (kHz) MPEG1 125 0 <f ≤ 4 187.5 4 <f ≤ 8 375 8 <f ≤ 15

If, on examination of these different conditions, a local maximum corresponds to a tonal component, then its sound pressure level X _tm (k) is calculated, and all the components of the spectral density that have been used for the determination in the considered band are set at a level of -∞ dB.

The sound pressure level X _tm (k) is determined by the following relation: $$X_{\mathit{tm}}\left(k\right)=\mathit{10}\log \left[{\displaystyle \underset{\mathrm{not}=-\mathit{1}}{\overset{\mathit{1}}{\Sigma }}}{\mathit{10}}^{{}^{X\left(k+\mathrm{not}\right)}{/}_{\mathrm{NOT}}}\right]\mathbf{dB}$$

The Figure 6B illustrates in particular the determination of a tonal component, from the local maxima as determined previously.

In this example, if we consider the index k equal to 7, the spectrum components that were used for the determination of the tonal components are set to an infinitely low level: the index components 4, 5, 6, 8, 9 , 10 and 11 disappear.

(iii) Calculation of the intensity of the non-tonal components

The non-tonal components correspond to the grouping of lines that have no connection with harmonic components.

As illustrated in Figure 6C all remaining spectral density components after the calculation of the tonal components are used to determine the level of the non-tonal components in a critical band.

It is recalled that the critical bands are defined by E. Zwicker et al, "Audio Engineering and Psychoacoustics: Matching Signals to the Final Receiver, The Human Auditory System" (Journal of Audio Engineering Society, 39, No. 3, March 1991, pp. 115-126). ), and included for information purposes in Appendices 1 and 2.

Thus, in each critical band, a linear sum of the power of the remaining spectral components (after having put to infinite power the components of the spectrum that were used for the determination of the tonal components) in order to determine the sound pressure level the non-tonal component X _nm (k) corresponding to this critical band.

• où Bc est la bande critique considérée, et
• k correspond à l'indice de la composante la plus proche du centre géométrique de la bande critique.

The sound pressure level of the non-tonal component X _nm (k) is then defined by the relation: $$X_{\mathit{nm}}\left(k\right)=\mathit{10}\log \left({\displaystyle \underset{i\in \mathit{Bc}}{\Sigma }}{e}^{\left({}^{X\left(i\right)}/\mathit{10}\right)}\right);$$

• where Bc is the considered critical band, and
• k is the index of the component closest to the geometric center of the critical band.

Finally, the Figure 6D presents the portion of the spectrum remaining after extraction of the tonal and non-tonal components, and determination of their sound pressure levels.

6.1.4 Decimation of the different components

• les composantes tonales et non-tonales sont conservées si leurs niveaux (c'est-à-dire leurs densités spectrales de puissance) sont supérieurs au masque absolu (où le masque absolu correspond au seuil d'audition pour les différentes composantes, et est présenté en annexe 1) : $$X_{\mathit{tm}}\left(k\right)\ge {\mathit{LT}}_q\left(k\right){\mathrm{et}\mathit{X}}_{\mathit{nm}}\left(k\right)\ge {\mathit{LT}}_q\left(k\right)$$
  
• seule la composante tonale de puissance la plus élevée sera conservée dans une bande de 0,5 bark à l'intérieur d'une bande critique. Cette opération nécessite l'utilisation d'une fenêtre glissante de largeur 0,5 bark.

The fourth sub-step of the calculation of the masking curve of the analog signal implements a reduction in the number of components of the individual masks, eliminating certain tonal and non-tonal components of the remaining spectrum according to the following conditions:

• the tonal and non-tonal components are conserved if their levels (ie their spectral power densities) are greater than the absolute mask (where the absolute mask corresponds to the hearing threshold for the different components, and is presented in Annex 1): $$X_{\mathit{tm}}\left(k\right)\ge {\mathit{LT}}_q\left(k\right){\mathrm{and}\mathit{X}}_{\mathit{nm}}\left(k\right)\ge {\mathit{LT}}_q\left(k\right)$$
  
• only the highest power tonal component will be conserved in a band of 0.5 bark within a critical band. This operation requires the use of a sliding window of 0.5 bark width.

It is recalled that E. Zwicker et al, "Audio Engineering and Psychoacoustics: Matching Signals to the Final Receiver, The Human Auditory System" (Journal of Audio Engineering Society, 39, No. 3, March 1991, pp. 115-126). ), defined a scale to represent the width and position of the critical bands in the frequency spectrum. It is a non-linear scale of frequencies, ranging from 0 to 24 barks, where each of the 24 critical bands has a width of 1 bark.

6.1.5 Calculation of individual masks

Of all the samples computed during the Fourier transformation sub-step, only a certain number is retained for the calculation of the global mask (for example 72 or 75 samples for level 1 of the MPEG specification). According to the MPEG1 specification, all samples of the first six subbands are used, then one out of two for the next six subbands, and one out of four for the rest of the audio spectrum of 4.5 or 5kHz width.

We can notice that each of the tonal and non-tonal components, determined during the sub-step of extraction of the tonal and non-tonal components, must exist in the new selection (for the computation of the global mask) with a similar index as possible (or even identical) of the original index.

$${\mathit{LT}}_{\mathit{nm}}\left[z\left(j\right),z\left(i\right)\right]=X_{\mathit{nm}}\left[z\left(j\right)\right]+{\mathit{av}}_{\mathit{nm}}\left[z\left(j\right)\right]+\mathit{vf}\left[z\left(j\right),z\left(i\right)\right]\mathbf{dB}$$

où :

• LT_tm et LT_nm sont les masques individuels (composantes tonale et non-tonale) de la composante masquante z(j) en z(i) ;
• X_tm et X_nm sont les niveaux de pression acoustique des composantes tonales et non-tonales ;
• av l'index de masquage de la composante en z(j) ;
• vf la fonction de masquage de la composante en z(j).

After these selection operations, we obtain: $${\mathit{LT}}_{\mathit{tm}}\left[z\left(j\right),z\left(i\right)\right]=X_{\mathit{tm}}\left[z\left(j\right)\right]+{\mathit{BC}}_{\mathit{tm}}\left[z\left(j\right)\right]+\mathit{vF}\left[z\left(j\right),z\left(i\right)\right]\mathbf{dB}$$

$${\mathit{LT}}_{\mathit{nm}}\left[z\left(j\right),z\left(i\right)\right]=X_{\mathit{nm}}\left[z\left(j\right)\right]+{\mathit{BC}}_{\mathit{nm}}\left[z\left(j\right)\right]+\mathit{vF}\left[z\left(j\right),z\left(i\right)\right]\mathbf{dB}$$

or :

• LT _tm and LT _nm are the individual masks (tonal and non-tonal components) of the masking component z (j) in z (i);
• X _tm and X _nm are the sound pressure levels of the tonal and non-tonal components;
• av the masking index of the component in z (j);
• vf the function of masking the component in z (j).

The av index of masking of the component has a different value depending on whether it is a tonal component or a non-tonal component.

et dans le cas d'une composante non-tonale, on a : $${\mathit{av}}_{\mathit{nm}}=-2.025-0.175z\left(j\right)\mathbf{dB}\mathrm{.}$$

Thus, in the case of a tonal component, we have: $${\mathit{BC}}_{\mathit{tm}}=-\mathrm{6,025}-0275z\left(j\right)\mathbf{dB};$$

and in the case of a non-tonal component, we have: $${\mathit{BC}}_{\mathit{nm}}=-2025-0175z\left(j\right)\mathbf{dB}\mathrm{.}$$

The masking function vf , identical for the tonal and non-tonal components, is characterized by the distance, in the frequency domain, between the masking component (resulting from the analog signal) and the masked component (resulting from the digital signal): Δz = z (i) -z (j).

$$\mathit{vf}=\left(0.4X\left[z\left(j\right)\right]+6\right)\mathrm{\Delta }\mathit{z}\mathbf{dB\; pour}-1\le \mathrm{\Delta }\mathit{z}<0\mathbf{bark};$$

$$\mathit{vf}=-17\mathrm{\Delta }\mathit{z}\mathbf{dB\; pour}0\le \mathrm{\Delta }\mathit{z}<1\mathbf{bark};$$

$$\mathit{vf}=-\left(\mathrm{\Delta }\mathit{z}-1\right)\left(17-0.15X\left[z\left(j\right)\right]\right)-17\mathbf{dB\; pour}1\le \mathrm{\Delta }\mathit{z}<8\mathbf{bark}\mathrm{.}$$

Thus, the masking function is equal to: $$\mathit{vF}=17\left(\mathrm{\Delta }\mathit{z}+1\right)-\left(0.4X\left[z\left(j\right)\right]+6\right)\mathbf{dB\; for}-3\le \mathrm{\Delta }\mathit{z}<-1\mathbf{bark};$$

$$\mathit{vF}=\left(0.4X\left[z\left(j\right)\right]+6\right)\mathrm{\Delta }\mathit{z}\mathbf{dB\; for}-1\le \mathrm{\Delta }\mathit{z}<0\mathbf{bark};$$

$$\mathit{vF}=-17\mathrm{\Delta }\mathit{z}\mathbf{dB\; for}0\le \mathrm{\Delta }\mathit{z}<1\mathbf{bark};$$

$$\mathit{vF}=-\left(\mathrm{\Delta }\mathit{z}-1\right)\left(17-0.15X\left[z\left(j\right)\right]\right)-17\mathbf{dB\; for}1\le \mathrm{\Delta }\mathit{z}<8\mathbf{bark}\mathrm{.}$$

According to the MPEG specification, masking from components at values greater than 8 barks or less than -3 barks of the masked component is not used, for reasons of implementation complexity and irrelevance, these components being too far from the masking component to disrupt its power spectral density.

6.1.6 Computation of the global mask

The mask on the sample i , denoted by LT _g (i), is equal to the linear sum of the individual masks of the samples situated in the range [8; -3] barks and of the absolute mask: $${\mathit{LT}}_{\mathrm{boy\; Wut}}\left(i\right)=\mathit{10}\mathrm{<figure-callout\; id="10"\; label="log"\; filenames="imgf0002.png"\; state="\{\{state\}\}">log</figure-callout>}\left[{10}^{{}^{{\mathit{LT}}_q\left(i\right)}{/}_{10}}+{\displaystyle \underset{j=\mathit{1}}{\overset{\mathrm{<figure-callout\; id="10"\; label="m"\; filenames="imgf0002.png"\; state="\{\{state\}\}">m</figure-callout>}}{\Sigma }}}{10}^{{}^{{\mathit{LT}}_{\mathit{tm}}\left(i\right)}{/}_{10}}+{\displaystyle \underset{j=\mathit{1}}{\overset{j=\mathrm{not}}{\Sigma }}}{10}^{{}^{{\mathit{LT}}_{\mathit{nm}}\left[z\left(i\right),z\left(j\right)\right]}{/}_{10}}\right]$$

LT _g (i) represents the level of the masking curve for sample i (for the calculation of the overall mask, 72 or 75 samples are used, depending on the audio bandwidth, according to the MPEG1 specification).

In particular, the samples do not have the same frequency width as a function of their positions in the spectrum (from 62.5 to 250 Hz according to the MPEG1 specification).

6.2 Filtering (53) of the global mask

The second step of the invention, in its preferred embodiment, implements a filtering (53) in the frequency and / or time domain of the masking curve corresponding to the global mask, to deliver a masking signal.

Indeed, in reception, a conventional "envelope detection" type receiver demodulates the so-called "simulcast" signal (analog signal on a first sideband, and amplitude modulated digital signal by the masking signal on the second sideband). ), and delivers a base signal corresponding to the sum of the modulating signals on each of the two lateral bands of the carrier.

According to the techniques presented in the MPEG1 specification, the frequency masking makes it possible to render inaudible any signal that follows the amplitude variations of the masking curve.

However, according to the invention, if the masking is applied directly in frequency on the DRM type digital signal, the time and frequency variations of the masking curve will be too great for the DRM signal, because of its multicarrier structure ( OFDM), and it appears necessary to smooth the masking curve.

The invention thus proposes to limit these variations in time and in frequency by implementing a low-pass filtering of the masking curve in the time and frequency domains delivering the masking signal, according to this preferred embodiment of the invention. , in order to improve (or at least not to degrade) the channel estimation implemented in reception of the digital signal. The channel estimate thus follows the evolution of the masking signal.

6.3 Determination of a minimum threshold and a gain to be applied to the mask (54)

The invention also proposes, in its preferred embodiment, to determine a minimum threshold to be applied to the masking signal and to add a gain (that is to say to amplify it) during a third step. referenced 54.

Indeed, it can be seen that the audio signals are generally not continuous in time. This is particularly the case of speech signals, such as the human voice.

In the case of silences, if no improvement is proposed, the masking curve would produce a masking signal that is almost zero over longer or shorter time intervals. Thus, the data transported by the digital signal masked by a silence of the analog signal would be lost in reception.

To overcome this disadvantage and to ensure a continuity of the data reception (for example, a continuity of the DRM service), the invention proposes to define a minimum threshold below which the masking signal can not go down.

For example, it is possible to define a threshold of 40 dB below a peak-to-peak modulating signal of 96 dB, below which the masking signal can not descend.

It is also possible, in this preferred embodiment of the invention, to amplify the masking signal in order to increase the power of the digital signal, for example of the DRM type. However, this amplification of masking signal is to the detriment of the inaudibility of the digital signal in reception: by increasing the power of the digital signal, it improves its reception quality, but it also risks to disturb the reception of the audio signal carried by the signal analog by analog receivers.

Therefore, it is necessary to find a compromise between the masking of the digital signal and the digital and analog reception performance vis-à-vis the propagation channel.

6.4 Amplitude modulation of the digital signal by the filtered masking curve (55)

After determining a minimum threshold and a gain to be made to the filtered masking curve, i.e. the masking signal, a fourth step (55) is to amplitude modulate the digital signal by the masking signal.

The modulating signal, that is to say the masking signal, is then applied in the frequency domain to the carriers constituting an OFDM symbol of the DRM type digital signal.

According to a first variant, the amplitude modulation step in the frequency domain of the digital signal by the masking signal is implemented during the creation of the DRM signal on transmission, in the frequency domain.

According to a second variant, the signal DRM is processed at the end of the OFDM modulation chain, before transmission on an AM carrier, by going back to the frequency domain from a fast Fourier transform step 51, for example to 1024. points for a sampling frequency of 48kHz.

6.5 Synchronization of modulated digital signal and analogue signal before transmission (57)

During a fifth step 57, the modulated digital signal, again transposed in the time domain (56), is synchronized with the analog signal carrying the source audio signal modulating the amplitude of the AM carrier.

Indeed, to optimize the gain provided by this diffusion technique and to reduce the disturbances caused by the digital signal on the analog reception, a perfect synchronization between the signal DRM, modulated by the masking signal, and the analog signal is necessary before the broadcast of the so-called "simulcast" signal illustrated in figure 4 . The digital signal is then correctly "hidden" during the broadcast, and does not disturb (or very little) analog receivers, while maintaining a good reception quality for digital receivers.

For example, it may be interesting for the digital signal and the analog signal to carry the same audio signal, especially during the transition period between the transition from analogue to digital broadcasting.

The invention thus proposes a technique for broadcasting a digital signal and an analogue signal in the same channel, called "simulcat" broadcasting, based on the main stages of calculation of the masking level related to the analog signal, and of application of this masking to the digital signal.

It should be noted that this technique also has good performance with a receiver implementing a synchronous demodulation, for example single sideband.

Thus, the signal demodulated by a conventional receiver is not disturbed by the digital signal, if the digital signal follows the masking curve varying in time and frequency.

For good reception of the digital signal, however, it is necessary to limit the variations of the masking curve as a function of the structure of the digital signal (for example according to the OFDM symbols of a DRM signal). It is therefore necessary to find a compromise between the masking of the digital signal and the performance of the analogue or digital reception vis-à-vis the propagation channel.

Of course, the present invention is not limited to the details of the embodiments described here by way of example, but extends instead to modifications within the scope of those skilled in the art, without departing from the scope of the present invention. 'invention.

It will be noted moreover that the invention is not limited to a purely hardware implementation but that it can also be implemented in the form of a sequence of instructions of a computer program or any form mixing a hardware part and a software part. In the case where the invention is partially or totally implemented in software form, the corresponding instruction sequence can be stored in a removable storage means (such as for example a floppy disk, a CD-ROM or a DVD-ROM) or no, this storage means being partially or completely readable by a computer or a microprocessor.

NOTES

Appendix 1: Table of absolute masks according to the audio frequencies Frequency (Hz) Critical band rate (z) Absolute mask LT _q (dB) 62.50 0617 33.44 125.00 1232 19.20 187.50 1842 13.87 250.00 2445 11.01 312.50 3037 9.20 375.00 3618 7.94 437.50 4.185 7.00 500.00 4736 6.28 562.50 5.272 5.70 625.00 5789 5.21 687.50 6289 4.80 750.00 6770 4.45 812.50 7.233 4.14 875.00 7.677 3.86 937.50 8.103 3.61 1000.00 8511 3.37 1062.50 8901 3.15 1125.00 9275 2.93 1187.50 9632 2.73 1250.00 9974 2.53 1312.50 10.301 2.32 1375.00 10614 2.12 1437.50 10913 1.92 1500.00 11.199 1.71 1562.50 11474 1.49 1625.00 11736 1.27 1687.50 11988 1.04 1750.00 12,230 0.80 1812.50 12461 0.55 1875.00 12684 0.29 1937.50 12.898 0.02 2000.00 13.104 -0.25 2062.50 13302 -0.54 2125.00 13493 -0.83 2187.50 13.678 -1.12 2250.00 13855 -1.43 2312.50 14027 -1.73 2375.00 14.193 -2.04 2437.50 14354 -2.34 2500.00 14509 -2.64 2562.50 14660 -2.93 2625.00 14807 -3.22 2687.50 14949 -3.49 2750.00 15,087 -3.74 2812.50 15221 -3.98 2875.00 15351 -4.20 2937.50 15478 -4.40 3000.00 15602 -4.57 3125.00 15841 -4.82 3250.00 16069 -4.96 3375.00 16287 -4.97 3500.00 16496 -4.86 3625.00 16697 -4.63 3750.00 16,891 -4.29 3875.00 17078 -3.87 4000.00 17259 -3.39 4125.00 17434 -2.86 4250.00 17605 -2.31 4375.00 17770 -1.77 4500.00 17932 -1.24 4625.00 18089 -0.74 4750.00 18242 -0.29 4875.00 18392 0.12 5000.00 18539 0.48 Appendix 2: Table of Central Frequencies of Critical Bands Established by E. Zwicker Band number Central frequency (Hz) Width of the band (Hz) 1 50 80 2 150 100 3 250 100 4 350 100 5 450 100 6 570 120 7 700 140 8 840 150 9 1000 160 10 1170 190 11 1370 210 12 1600 240 13 1850 280 14 2150 320 15 2500 380 16 2900 450 17 3400 550 18 4000 700 19 4800 900 20 5800 1100 21 7000 1300 22 8500 1800 23 10500 2500 24 13500 3500

Claims (11)

1. Method for broadcasting a modulated digital signal and an analogue signal carrying a source audio signal modulating the amplitude of an AM carrier, comprising a step of modulating the amplitude (55), in the frequency domain, of a digital signal by a masking signal, delivering said modulated digital signal, and a step of broadcasting said modulated digital signal and analogue signal over one same carrying frequency, characterised in that the masking signal is representative of a psychoacoustic masking curve of said source audio signal.
2. Method for broadcasting according to claim 1, characterised in that said modulated digital signal and analogue signal are transmitted in one same propagation channel.
3. Method for broadcasting according to any one of claims 1 and 2, characterised in that said masking signal corresponds to said masking curve, after low-pass filtering (53) in the frequency domain and/or in the time domain.
4. Method for broadcasting according to any one of claims 1 to 3, characterised in that said masking signal always remains above a predetermined threshold.
5. Method for broadcasting according to any one of claims 1 to 4, characterised in that said masking signal is amplified (54) according to a predetermined gain.
6. Method for broadcasting according to any one of claims 1 to 5, characterised in that said modulated digital signal and analogue signal are synchronised (57).
7. Method for broadcasting according to any one of claims 1 to 6, characterised in that said modulated digital signal and analogue signal carry the same source audio signal.
8. Method for broadcasting according to any one of claims 1 to 7, characterised in that said modulated digital signal is a multicarrier signal.
9. Method for broadcasting according to any one of claims 1 to 8, characterised in that said masking curve is obtained according to at least one of the following steps:

   - mathematical transformation of said source audio signal from the time domain to the frequency domain;

   - determination of an absolute mask, corresponding to an auditory threshold for each one of the frequencies corresponding to the components delivered by said mathematical transformation;

   - identification of tonal components and non-tonal components among said components;

   - decimation of said components, by removing components below said absolute mask;

   - calculation of individual masks for a predetermined number of components;

   - calculation of said masking signal corresponding to a linear sum of said individual masks.
10. Device for broadcasting a modulated digital signal and an analogue signal carrying a source audio signal modulating the amplitude of an AM carrier, comprising means for modulating the amplitude (55) in the frequency domain, of a digital signal by a masking signal, delivering said modulated digital signal, and means for broadcasting said modulated digital signal and analogue signal over one same carrying frequency, characterised in that the masking signal is representative of a psychoacoustic masking curve of said source audio signal.
11. Modulated digital signal intended to be broadcast over one same carrying frequency as an analogue signal carrying a source audio signal modulating the amplitude of an AM carrier, obtained by modulating the amplitude of a digital signal, in the frequency domain, by a masking signal, characterised in that the masking signal is representative of a psychoacoustic masking curve of said source audio signal.

Images