EP1159795A1 - Method for controlling the quality of a digital audio signal broadcast with an audio-visual programme - Google Patents

Abstract

The invention concerns a method for controlling the quality of a distributed digital audio signal. The digital audio signal (ADS) is subjected to a detection of spurious signals such as brief interruption (101), squeal (102), humming (103), left/right phase shift (104). A warning signal is produced (105, 106) if one at least of the spurious signals is present. The invention is applicable in radio and digital television broadcasting.

Description

METHOD FOR CONTROLLING THE QUALITY OF A BROADCASTED AUDIONUMERIC SIGNAL WITH AN AUDIOVISUAL PROGRAM

The invention relates to a method for controlling the quality of a digital audio signal distributed by broadcasting a video or audio program.

With the increase in the exchange and dissemination of information, the digital audio coding methods used by broadcasting services have made it possible to reduce the amount of information to be transmitted per program, and consequently to increase the number of programs broadcast. by television channels. On the other hand, this reduction can lead to an irremediable loss of the quality of the information, that is to say of the sound, compared to the source. The extent of the faults introduced depends on the bit rate allocated to the encoder, the complexity of the sound signal, as well as the problems associated with signal transmission.

For technical and / or accountability reasons relating to the broadcasting process, it is necessary, in order to meet user requirements, to assess the level of audio signal quality.

Currently, subjective equipment assessment methods, by human appraisal or monitoring, are used. These methods are cumbersome to implement, and unreliable. In addition, the implementation of such methods continuously is not very easy, because of the subjective implications vis-à-vis the personnel called upon to ensure their execution. Differential analysis methods have also been developed. These methods are based on a sys- human hearing aid of perception, bring into play a reference sound source and the sound source to be evaluated. However, such a solution turns out to be impractical since it is necessary to have the reference source.

Other operating modes, based on a priori knowledge of the defects generated by the chain of coding / transmission / decoding processes, can make it possible, by statistical methods, to assess the quality of the digital audio signal transmitted, by measuring the rate of appearance of these defects.

The object of the present invention is to remedy the drawbacks of the abovementioned methodologies by implementing a method for controlling the quality of a fully automated digital audio signal, in the absence of the use of a reference source.

Another object of the present invention is also the implementation of an automated method for controlling the quality of a digital audio signal capable of being implemented continuously or pseudocontinuously, the pseudo-continuous nature of this implementation. work being understood to be a periodic implementation with a sufficiently low repetition period to assure users the perception of a television broadcast with constant qualities during the television broadcast of one or more successive programs.

Another object of the present invention is also the implementation of a method for controlling the quality of a continuous digital audio signal making it possible, moreover, from separate quality control processes for this signal, since they relate to separate faults of this signal, to ensure overall quality control, resulting in unparalleled listening comfort to date of broadcast programs.

The method for controlling the quality of a digital audio signal, object of the present invention, is remarkable in that it consists, on a mono- or stereophonic digital audio signal, in detecting in this digital audio signal at least one of the signals parasites such as brief cut, hiss, buzzing and relative phase shift of the left and right channels of this digital audio signal, which makes it possible to generate an alarm signal in the presence of at least one of the parasitic signals.

According to one aspect of the method, object of the present invention, the step consisting in detecting in this digital audio signal a parasitic signal such as a brief cut-off consists in detecting on a series of successive samples of this digital signal a rapid decrease in the level energy of this digital audio signal to zero energy, thereby revealing an absence of reverberation of this digital audio signal.

According to another aspect of the method, object of the present invention, the step consisting in detecting in this digital audio signal a spurious signal such as a whistling sound consists in detecting in this digital signal a sudden and transient increase in the spectral energy of this digital audio signal in a frequency band whose low frequency is between 4.5 kHz and 6.5 kHz and whose high frequency can reach up to 20 kHz. According to another aspect of the method, object of the present invention, the step consisting in detecting, in this digital audio signal, a parasitic signal such as a humming consists in detecting in this parasitic signal a pink noise in a frequency band between 0 and 1100 Hz and of substantially constant level in this frequency band.

According to another aspect of the method, object of the present invention, the step consisting in detecting, in this digital audio signal, a spurious signal such as a phase shift between channels of the digital audio signal consists in calculating the value of phase shift between channels of the digital audio signal from the intercorrelation function of the digital audio signal present on each of the channels and to compare the calculated phase shift value with a threshold value. According to another aspect of the method which is the subject of the present invention, the step consisting in discriminating the mono- or stereophonic mode of the transmitted signal consists in detecting sudden and brief changes in context of the mono- or stereophonic mode of the transmitted or reciprocal signal. - ment, from a comparison of the energies of the right and left channels.

The method for controlling the quality of a digital audio signal which is the subject of the present invention finds application to any type of digital audio signal subjected to a coding, transmission and then decoding process, the coding / decoding operations being able to be assimilated to operations. compression / decompression, in particular to digital broadcasting signals such as DAB, or to signals of the audio frequency channel of a digital television signal for example. It will be better understood on reading the description and on observing the following drawings in which, in addition to FIG. 1 relating to prior techniques: - FIG. 2a represents, by way of nonlimiting example, a general flowchart illustrating the steps allowing the implementation of the method which is the subject of the present invention;

- Figure 2b shows, by way of non-limiting example, a general flowchart illustrating the steps allowing the implementation of the method object of the present invention, in a variant of Figure 2a in which a detection of the monophonic or stereophonic character of the transmitted digital audio signal is carried out;

- Figure 3a shows, by way of nonlimiting example, a general flowchart illustrating the steps allowing the implementation of the method object of the present invention, in a nonlimiting variant in which an overall quality of the signal is highlighted;

FIG. 3b represents, by way of nonlimiting example, a general flowchart illustrating the steps allowing the implementation of the method which is the subject of the present invention, in a variant of that illustrated in FIG. 3a and in which orders of separate priorities are introduced for the various spurious signals which may affect the digital audio signal;

FIG. 3c represents, by way of nonlimiting example, a process for the specific management of the orders of priority assigned to the spurious signals such as the hissing, phase shift and buzzing, which can be implemented in the context of the variant embodiment shown in FIG. 3a or 3b;

- Figure 4 shows, by way of nonlimiting example, a flowchart relating to the process of discrimination of the monophonic or stereophonic character of the transmitted digital audio signal;

- Figure 5a shows, by way of illustration, a process capable of being implemented to ensure the detection of a brief cut in the digital audio signal;

- Figure 5b shows, by way of illustrative example, a flow diagram relating to the steps allowing the implementation of the detection of a brief cut in the digital audio signal, in accordance with the process illustrated in Figure 5a;

- Figure 6a shows, by way of non-limiting example, a process capable of being implemented to ensure the detection of a whistling sound affecting the digital audio signal; - Figure 6b shows, by way of illustrative example, a flow diagram relating to the steps allowing the implementation of the detection of a whistling affecting the digital audio signal, in accordance with the process illustrated in Figure 6a; - Figure 7a shows, by way of nonlimiting example, a process capable of being implemented to ensure the detection of a hum affecting the digital audio signal;

FIG. 7b represents, by way of nonlimiting example, a flowchart relating to the steps allowing the implementation of the detection of an affected hum both the digital audio signal, according to the process illustrated in Figure 7a;

FIG. 8a represents, by way of illustration, a flowchart relating to the steps allowing the implementation of the detection of a phase shift affecting the right and left channels of the digital audio signal;

- Figure 8b shows, by way of non-limiting example, a detail of implementation of a step in the flow diagram of Figure 8a; - Figure 8c shows, purely by way of illustration, a calculation step implemented in the context of the method, object of the present invention, shown in Figure 8b;

- Figure 9 shows a device according to the object of the invention.

Prior to the actual description of the process which is the subject of the present invention, various elements relating to the conditions of transmission and distribution of digital audio signals will be described in conjunction with FIG. 1.

In general, with reference to the above-mentioned figure, it is indicated that the programs, digital audio program or audio channel of video program or digital TV, programs Pr _x , Pr ₂ , Pr ₃ , are subjected to a compression process and then , after MUX multiplexing, to a transmission proper by radio broadcasting, for example.

On reception, the digital audio signal is first demultiplexed DMUX then subjected to a decompression process which restores the programs, the programs subjected to successive operations of compression, mul- tiplexing, transmission, demultiplexing, decompression, that is to say in fact coded / decoded digital audio programs, denoted Pr,, Pr ₂ , Pr ₃ , are then subject to distribution to users. In the context of a DAB type transmission, for Digital Audio Broadcasting, the digital audio signal is subdivided into frames comprising for example, for MPEG1 LU type compression, 1152 samples and the length of the frame transmitted depends on the compression rate or coding rate used. The coded frame comprises synchronization bits S _yrιc , error correction code bits CRC and finally the digital audio data bits proper. In general, the aforementioned digital audio signal, for each program considered Pri to Pr _3, is either a monophonic signal or a stereophonic signal. When the signal is monophonic, this signal is present identically on the right and left channels of the digital audio channel, if necessary to the nearest phase shift value. When, on the contrary, the signal is stereophonic, each left and right channel of the digital audio channel transmits its own digital audio signal in order to restore the conditions of recording in the studio.

The method for controlling the quality of a digital audio signal in distribution, in accordance with the object of the present invention, is implemented on a coded / decoded digital audio signal, that is to say subject to all of the successive treatments described in connection with FIG. 1. It will now be described firstly in connection with FIGS. 2a and 2b. As shown in FIG. 2a, the method which is the subject of the present invention is applied to the signal previously mentioned coded / decoded digital ion, designated by

ADS, this signal corresponding to the programs Pr,, Pr ₂ , Pr ₃ , mentioned above.

According to a particularly advantageous characteristic of the method which is the subject of the present invention, it consists in detecting in this digital audio signal at least one of the spurious signals such as signal of short cut, hiss, buzzing, relative phase shift of the left and right channels of this digital audio signal. It is recalled that the parasitic signal of short cut is designated by "mute" in English language.

In FIG. 2a, there is shown, purely by way of illustration, the operations for detecting a short cut signal at step 101, for detecting a whistling sound at step 102, for detecting a hum at step 103 and detection of a phase shift in step 104, the phase shift being understood of course to a relative phase shift between the left and right channels G, D, of the audio-digital channel.

The detection of these spurious signals preferably means a detection independent of each of them, the detection operation being able to allow the establishment of a logic variable representative of the presence, respectively of the absence of this spurious signal. In FIG. 2a, purely by way of illustration, the corresponding logical variables are denoted Pj _. , P ₂ , P ₃ , P ₄ for the operations 101, 102, 103, 104 for detecting the aforementioned spurious signals, the complemented values of these logic variables representing the absence of a spurious signal for example. In addition, and in accordance with a particularly remarkable aspect of the method which is the subject of the present invention, as illustrated in FIG. 2a, the detection of at least one of the aforementioned signals makes it possible to generate an alarm signal, this operation being shown in steps 105, 106 of Figure 2a.

By way of nonlimiting example and in an illustrative mode, the signal A, control signal of a power alarm signal, can for example correspond as represented in step 105, in a logical combination OR of the 'all of the signals P _lr P ₂ , P ₃ , P ₄ , the actual power alarm signal, such as a sound, visual or other signal, being emitted in the following step 106.

Referring to FIG. 2a, it is indicated that, in accordance with a particularly remarkable aspect of the method, object of the present invention, the detection of one of the aforementioned signals allows for example the emission of the alarm signal considered.

It is understood in particular that during the transmission of this alarm signal, the operator of the broadcasting network is then made aware of the existence of a parasitic signal significantly disturbing the conditions of broadcasting and may therefore take any action to modify, for example, either the coding of the transmission channel, or any operation it deems necessary.

According to another particularly remarkable aspect of the method which is the subject of the present invention, as shown in FIG. 2b, it can also consist in discriminating the mode of mono or stereophonic transmission of the ADS digital audio signal. As shown in a purely illustrative manner in FIG. 2b, the method which is the subject of the present invention can then comprise a step 100 consisting in discriminating the mono or stereophonic character of the above-mentioned ADS signal. Preferably, but not limited to, this step of discriminating the mono or stereo character of the ADS signal is carried out before the implementation of steps 101, 102, 103, 104 of the spurious signals. Indeed, depending on the mono or stereophonic character of the digital audio signal transmitted ADS, different conditions can be applied for the implementation of the detection of the spurious signals, in particular in step 104 consisting in detecting the relative phase shift between the channels left and right of the digital audio channel. Under these conditions, and as represented in FIG. 2b, the step 105 of constitution of a logic alarm control signal can consist in establishing a signal verifying the relation Ai = Pi OR P ₂ OR P ₃ OR P ₃ when the signal is monophonic or A ₂ = Pi OR P ₂ OR P ₃ OR P ' ₄ when the signal is stereophonic.

It is understood, for example, that the signals P ₄ and P ′ ₄ can be logic signals subjected to different phase shift value conditions depending on whether the signal is mono- or stereophonic. The alarm control signal Ai, A ₂ then makes it possible to generate the power signal in step 106 as mentioned previously in the description.

Of course, the method which is the subject of the present invention must preferably allow the implementation of an overall quality control of the ADS digital audio signal. By overall quality of the audio signal meric is understood to mean a quality allowing the best listening comfort for users, taking into account the nature of the spurious signals and of course, where appropriate, the mono- or stereophonic nature of the digital ADS audio signal transmitted.

It is understood in particular that the aforementioned parasitic signals may, depending on the transmission conditions, not have the same importance relative to the degradation introduced on the listening conditions of the transmitted digital audio signal.

For this purpose, and with reference to FIG. 3a, it is indicated that the method, object of the present invention, in order to establish an overall quality control of the digital audio signal, can consist, over a sliding time window of determined length, c '' i.e. on a series of successive samples of the digital audio signal observed on this sliding window, to be carried out in combination with different operations aimed at weighting the relative value of each parasitic signal detected with respect to the degradation introduced under the conditions listening. As shown in FIG. 3a, the method which is the subject of the present invention can then consist in counting the number of occurrences N _M of spurious short-cut signals during a duration T _M of observation of this digital audio signal and to compare the number of occurrences N _M with a determined threshold value S _M. In FIG. 3a, the aforementioned operation has been represented as consisting, starting from step 101 previously described in connection with FIGS. 2a and 2b, of detection of a short cut and of a step 101a of initializing the number _M N of occurrences to zero, when the existence of a brief LYING pure in response to step 101, to cause the passage to the value N _M = N _M +1 in step 101b by incrementing a unit following this detection, this for the entire duration T _M of observation of the digital audio signal. At the end of this period, a comparison test of the final incremented value N _M is carried out in step 101c with respect to the aforementioned threshold value S _M. On a positive response to the comparison step 101c, a fault signal Pi * is then generated, this fault signal corresponding to a degradation of the overall quality of the digital audio signal.

In addition, as shown in FIG. 3a above, the method according to the invention also consists in counting the number of occurrences N _s of hissing parasitic signals for a duration T _s of observation of the digital audio signal ADS and in compare the number of occurrences N _s to a determined threshold value S _s . In FIG. 3a, and analogously to the process of detecting a short break, for the parasitic whistling signal step 102a designates the initialization of N _s to the value zero, 102b designates the incrementation of N _s to the value N _s +1 and 102c designates the comparison of N _s to the threshold value S _s . Upon positive comparison in step 102c, a logic signal P ₂ * corresponding to the degradation of the overall quality of the signal by a hissing noise is then generated.

The process which is the subject of the present invention consists in operating in the same way for the parasitic humming noise and the phase shift noise. Under these conditions, as shown in FIG. 3a, it consists in detecting for a duration τ _D , the phase shift value φ and the number of occurrences having to be at least equal to N _D of these phase shift values over a determined number D of phase shift calculations and then comparing the calculated phase shift value φ with a determined threshold value S _D.

It is the same with regard to the parasitic buzzing signals, the method which is the subject of the invention consisting in counting the number of occurrences N _B of parasitic buzzing signals for a duration τ _B of observation of this digital audio signal and comparing the number of occurrences N _B with a threshold value S _B.

In FIG. 3a, analogously to the parasitic short-cut and whistling signals, for the parasitic hum signal, 103a denotes the initialization of the value N _B to the value zero, 103b the incrementation of this value N _B a the value N _B +1, 103c designating the comparison of the number of occurrences of buzzing N _B with the determined threshold value S _B. On a positive response to test 103c, a signal P ₃ * of the presence of a buzzing signal affecting the overall quality of the digital audio signal is generated. Likewise, 104a designates the initialization of N _D to the value zero, 104b designates the incrementation of N to the value N _{D + 1} , 104c designates the comparison of the calculated phase shift value φ with the determined threshold value S _D. On a positive response to test 104c, a signal P ₄ * is generated, which represents the presence of a parasitic phase shift signal affecting the overall quality of the ADS signal.

Of course and without limitation, as shown in FIG. 3a, all of the signals Pi *, P ₂ *, P ₃ * and P ₄ * can, in step 105, be combined with the same way as in the case of FIG. 2a. The alarm control signal A then verifies the relationship: A = Pi * OR P ₂ * OR P ₃ * OR P ₄ *, the power alarm signal being emitted in step 106. In the setting mode implementation of the method which is the subject of the present invention, as shown in FIG. 3a, each signal resulting from the comparison tests 101c, 102c, 103c and 104c plays substantially the same role as regards the implementation of the control signal alarm A in step 105. However, it is of course possible to adapt the role of each of these signals, either from the window values or duration of observation of the ADS digital audio signal, or from each value of threshold against which the comparison is made in the above steps.

However, and in accordance with a particularly remarkable aspect of the method which is the subject of the present invention, in order to establish an overall quality control of the ADS signal, the step consisting in transmitting the alarm signal is preferably conditioned to an order of priority of the crossing of the aforementioned threshold values.

Such a preferred embodiment is now described in connection with FIG. 3b. In FIG. 3b, the same references naturally designate the same elements as in the case of FIG. 3a.

In general, and as shown in a nonlimiting manner in FIG. 3b, conditioning to an order of priority of the crossings of each of the aforementioned parasitic signals can for example be carried out by favoring one of the parasitic signals. Concerning its contribution to the alarm control signal A in step 105 and, ultimately, to the contribution of this parasitic signal to the degradation of the overall quality of the digital audio signal ADS. In FIG. 3b, by way of example, the spurious short-cut signal Pi * is considered more annoying than one or the other of the spurious signals P ₂ *, P ₃ * or P ₄ * relating to the hissing noises , buzzing, respectively phase shift. Under these conditions, the alarm control signal A can verify the logical relationship:

A = Pi * AND (P ₂ * OR P ₃ * OR P ₄ *). Of course, decreasing priority orders can be allocated to the aforementioned crossing by the numbers of occurrences of the spurious signals as a function of the decreasing relative importance of these signals relative to the degradation of the overall quality of listening to the digital audio signal. Such a nonlimiting embodiment will now be described in connection with FIG. 3c relatively for example to the spurious hiss, hum and phase shift signals, it being understood that the spurious short-cut signal is considered to be a priority for example. Under these conditions, the embodiment of FIG. 3c can correspond, without limitation, to a specific management of the parasitic signals of hissing P ₂ *, of buzzing P ₃ * or of phase shift P ₄ * in the mode of embodiment of FIG. 3b, the brief cut-off signal being considered to have maximum priority.

As shown in FIG. 3c, for a current series of samples of the digital audio signal ADS, the method can consist, in one step, 200 of observing ver the hiss noise in a sliding time window of duration T _s, then, at step 201, to carry out a detection at least S _s wheezing during the observation period. Of course, these operations 200 and 201 can correspond to operations 102, 102a, 102b, 102c in FIG. 3b. On a positive response to test 201, the parasitic signal P ₂ * is generated.

However, as shown in FIG. 3c, the detection of the phase shift in step 203 can be made conditional on the absence of a parasitic whistling signal, that is to say on the negative response to the test 201 cited above. Thus, in step 203, the phase shift is calculated D times over the duration T _D. On a positive response to the comparison step 203, a step 205 is performed in which a phase shift value φ appears N _D times on the determined calculation number D. On a positive response to the step 205, the comparison of the value of φ at the threshold value S _D is carried out in step 206. Steps 203, 205 and 206 can correspond to steps 104, 104a, 104b and 104c of FIG. 3b. On a positive response to the comparison step 206, the signal P ₄ * is then generated.

On the contrary, on a negative response in step 203, a step 204 is carried out, which consists in detecting the parasitic hum signal. Step 204 can correspond to steps 103, 103a, 103b and 103c of FIG. 3b. On a negative response to the above-mentioned step 204, the method is then updated for the next series of samples of the digital audio signal ADS.

On the contrary, on a positive response in step 204, the parasitic hum signal P ₃ * is generated, this signal capable of generating the alarm control signal A.

Upon observation of FIG. 3c, it can thus be seen that decreasing priority orders are successively assigned to the whistling, phase shift and buzzing signal.

Of course, the management of the block of the aforementioned parasitic signals with their corresponding relative priorities can be carried out either in accordance with step 105 represented in FIG. 3b, or, if necessary, in accordance with step 105 represented in FIG. 3a.

Finally, it is indicated that in FIGS. 3a and 3b, the step 100 for detecting the mono- or stereophonic character of the digital audio signal ADS has not been shown so as not to overload the drawing. It is understood that in the examples of implementation of the method which is the subject of the present invention according to FIGS. 3b and 3c, the detection of the mono- or stereophonic character of the signal can be carried out. Different modes of implementation of processes for detecting spurious signals and of the mono- or stereophonic character of the ADS digital audio signal previously mentioned will now be described in conjunction with FIG. 4 and the following figures. FIG. 4 relates to a specific procedure for detecting the mono- or stereophonic character of the aforementioned digital audio signal.

When an ADS digital audio signal is transmitted on two channel channels, right and left channels, of a digital audio channel, and transmitted in monophonic mode, the signals present on the right and left channels are identical to the value of a low value phase shift. The signal identity on the two channels is generally maintained for some time. An experimental observation of the nature of these signals on these two channels has been able to show that when there is a sudden and short change in the mode of transmission of the digital audio signal, such as for example passage from monophonic mode to stereophonic mode then return to monophonic mode , such changes indicate the presence of defects. Otherwise, the context of the transmission is not bound to change constantly over time. Consequently and with reference to FIG. 4, the step consisting in discriminating the mode of mono- or stereophonic transmission of the digital audio signal ADS consists in detecting sudden and brief changes of context of mono- or stereophonic mode of transmission, or vice versa, by comparing the energy variations of the two left and right channels.

In FIG. 4, a step consisting of a monophonic or stereophonic transmission context calculation is followed followed by a step of verifying this context making it possible to assign either the monophonic character or the stereophonic character to the digital audio signal transmitted. In a specific non-limiting particularly advantageous embodiment, as shown in FIG. 4, the discrimination of the mono- or stereophonic mode of transmission can comprise, for each successive series of samples of rank n, each series of samples being determined by a cut in sequences of N samples at a step 300 on the left and right channels of the digital audio signal ADS, a step 301 for calculating the respective energies of the right and left channels of the transmission signal, these energies being denoted E _n , _g for the left channel and E _n , _d for the right channel. The index n denotes the rank of the current series of samples considered.

The above-mentioned step 301 is then followed by a step 302 consisting in calculating the ratio M _n of the right and left energies with M _n = E _n , _g / E _n , _d .

This step 302 is itself followed by steps 303, 304 and 305, which allow the calculation of a binary variable C _n , context variable representative of the transmission context in mono- or stereophonic mode.

Step 303 consists in comparing the value of the ratio of the energies M _n to a first and a second threshold value, denoted Seuili respectively Threshold ₂ , according to the relation Seuili <M _n <Threshold ₂ . On a positive response to the aforementioned comparison test 303, the context variable C _n is assigned in step 304 the value 0 representative of a monophonic mode context, and the context variable C _n is on the contrary assigned the value 1 otherwise, in step 305. The value 1 of the context variable C _n is representative of a stereophonic mode context. Steps 300, 301, 302, 303, 304, 305 described in conjunction with FIG. 4 constitute the context calculation step previously mentioned in the description.

The context verification step previously mentioned is then constituted for example, as shown in FIG. 4, by a step 306 consisting in performing on a determined number of successive samples the calculation of a cumulative value of the binary variables of context C _n successive on a number C of sample sequences consecutive tillons. This cumulation is representative of the arithmetic sum of the context variables over the number C of series of samples. Step 306 is then followed by a step 307 consisting in comparing, by comparison of superiority, the value of this sum, that is to say the value of this accumulation, with a limit value, denoted Limi te, constituting a reference value. On positive response to the aforementioned superiority comparison in step 307, the mode of transmission of the digital audio signal ADS is assigned the stereophonic mode, while on negative response is assigned to this mode of transmission the monophonic mode. The operating mode of the process described in connection with FIG. 4 can be justified in the following manner.

When the signal is in stereophonic mode, the difference between the energies of the two left and right channels is not zero. Thus, to determine the mono- or stereophonic context, the distance between the energy of each right and left channel is calculated. The respective energies can then verify the relationship:

In these relationships, D _{n / 1} and G _n , _ι represent the value of the amplitude of each sample of rank i of the series of samples of rank n of the digital audio signal ADS.

In order not to consider a stereophonic signal passing through zero as a monophonic signal, the context variable C _n is then summed over a number C of successive series of samples during the context verification operation. The transmission mode is then reophonic if the sum of the context variables is at least equal to the limit value constituting the reference value.

An example of numerical value will now be given for an application to digital coding of the MPEGl LU type transmitted via a digital broadcasting channel. In such an application example, the number of each series of samples N can be between 256 and 1024 samples, the threshold values, Seuili and Threshold ₂ can be between:

Only [0, 95; 0, 9999] Threshold ₂ nd [1, 0001; 1, 05] C € [20; 60] and the limit value for comparison with test 307: Limi te e [15; 45].

A preferred process for detecting a spurious signal such as a short cut will now be described in conjunction with FIGS. 5a and 5b.

Firstly, it is recalled that a brief cut is a very brief interruption of the digital audio signal constituting digital silence, the digital audio signal ADS being in this case replaced by zeros or very low values due to a transmission problem, such as loss of synchronization for example, preventing the decoder, that is to say the decompression operator, from reconstructing the series of samples constituting the digital audio signal.

Consequently, and with reference to FIG. 5a, which represents the amplitude of the samples of the digital audio signal for a sequence of jazz music transmitted and collected at the edge of the service area, a brief cutoff, or Mute, is characterized by a total absence of sound for several milliseconds, 25 ms in FIG. 5a.

On the contrary, any sound intended to be listened to has a minimum of reverberation. The reverberation phenomenon is a phenomenon due to the multiple reflections of sound on the walls such as the walls and ceilings surrounding any sound source. This reverberation phenomenon still exists, even, but to a lesser extent, in free field.

In accordance with a particularly remarkable aspect of the method which is the subject of the present invention, the step consisting in detecting in the digital audio signal ADS a spurious signal such as a short cut, consists in detecting on a series of successive samples of this signal digital a rapid decrease in the energy level of this digital audio signal to zero energy. Such a signal form in fact reveals the absence of reverberation of this digital audio signal and, consequently, the introduction of a brief cutoff.

Thus, and in a specific, particularly advantageous, non-limiting implementation of the method which is the subject of the present invention, as shown in FIG. 5b, the step consisting in detecting in the digital audio signal that a spurious signal such as a brief cut includes a step consisting in determining separately on each left and right channel of the digital audio channel, for a plurality of sequences of N successive samples, the average energy E _n of the signal transported by this channel, n denoting the rank of each sequence of samples. Referring to FIG. 5b, it is indicated that the cutting of each channel on the digital audio signal ADS is carried out in a step 400, which is followed by a step 401 of calculating the time energy E _n of the series of samples considered.

The average energy E _n on each channel then verifies the relation:

In the previous relation, it is indicated that x [i] represents the temporal sample of rank i of the series of N samples, n denoting the rank of this series of samples. In this case, the time windows, that is to say the sequences of successive samples, are without overlap.

Step 401 of FIG. 5b is then followed by a step 402 consisting of comparing the evolution of the average energy for sequences of N successive samples, step 402 possibly corresponding, for example, to a verification of an inferiority relation of the value of the average energy E _n to the value Seuili.

The existence of a parasitic signal of short cut is revealed if at least one of the average energies is less than Seuili and if, as shown in step 403, one or more average energies close to this zero average energy are higher than a determined threshold value. Step 403 corresponds to a step of comparing the average energy of the series of samples of rank nk, denoted E _n - _k , with a threshold value Δ, Δ = 31 dB. E _n -) - designates the neighboring average energy (s).

If on the contrary, the response to the comparison step 402 is negative, the process is repeated by returning to step 401 for a series of samples of rank n + 1 by the return loop 404. The value of each energy mean E _n is then stored for a plurality of successive series of samples in order to allow the calculations to continue. An example of application to digital audio coding of MPEG1 LU type is now given below.

In this type of coding, the short cuts correspond to digital rests of length multiple of 1152 digital samples. In this case, the parameter N, number of samples constituting a series of samples, is equal to 256, which makes it possible to ensure the detection of the smallest area of the samples of low values, namely Seuili = 3, Δ = 31 dB with k = 2. In this case, the comparison in step 403 can be carried out by calculating the value:

ΔE = lO.LOGio [E _n _ ₂ ] -31. If the value ΔE is greater than zero, ΔE> 0, then there is a spurious short cut signal.

A more detailed description of a specific process for detecting a spurious signal such as a whistle will now be given in connection with FIGS. 6a and 6b.

More specifically, as shown in FIG. 6a, it is indicated that the presence of such a spurious signal corresponds to a sudden and transient increase in the spectral energy of the digital audio signal. only over a relatively wide frequency band not exceeding 15 to 16 kHz. Preferably, and according to a remarkable aspect of the method which is the subject of the present invention, the step consisting in detecting in the audio-digital signal ADS a spurious signal such as a whistling sound consists in detecting in this digital audio signal a sudden and transient increase in the spectral energy of this digital audio signal in a frequency band whose low frequency is between 4.5 and 6.5 kHz and whose high frequency can reach up to 20 kHz. This parasitic whistling signal is caused by errors in the error correcting code when reconstructing the series of digital audio samples. A specific embodiment of a process for detecting spurious hiss signals will now be described in connection with FIG. 6b, this process being based on the detection of variations in spectral energy of the digital audio signal.

As shown in FIG. 6b above, the digital audio signal ADS is firstly divided into successive sequences of N samples, each sequence being a sequence of given rank n subjected to a verification of the average energy of each sequence, as well as shown in step 500 of Figure 6b. For this purpose, step 500 is subdivided into a step 500a of cutting the signal into sequences of rank n each comprising N samples, the samples of each sequence being designated by e (i). Step 500a is followed by step 500b consisting in calculating the average time energy of each sequence sequence

N n, Emoy _n = Ye (i). The average energy of each sequence i = l of samples is compared by relation of superiority to a first threshold value, Seuili, relation:

Emoy _n > Seuili

in the next step 500c.

If the answer is negative in step 500c, the process is brought back to step 500a, the energy of at least one current sequence of rank n being considered as insufficient. Otherwise, on a positive response at step 500c, a step 500d is carried out, which consists in comparing according to a comparison of superiority the ratio of the energy of the current sequence to the energy of an immediately neighboring, non-adjacent sequence, of rank n-2, at a second threshold value, Threshold ₂ , according to the relationship:

Emoy _r

> Threshold ₂ Emoy _n .

On a negative response to the comparison of the test 500d, the process is brought back to the step of calculating the average temporal energy, the energy of the current sequence and of the neighboring sequence being similar.

On a positive response to the comparison of the test 500d, the average energy of two neighboring frames being increasing, and the ratio greater than the second threshold value, the whistling detection process is continued, step 500 having been satisfied. The aforementioned step 500 is then followed by a step consisting in calculating on a series of samples of the digital audio signal the spectral composition of this signal defined as the value of frequency components in frequency sub-bands central f _lr the value of the frequency components being designated by S _n (i), n denoting the rank of the series of samples considered. The values S _n (i) have a bandwidth δf. The aforementioned step is carried out and represented in FIG. 6b by the steps 501 consisting in fact of calculating the spectrum of the signal by Fourier transform, the components in sub-bands being denoted S _n (i). In step 502, the frequency value is limited in this spectrum to a value beyond the value F kHz, that is to say the value 4.5 kHz, and the interval 4.5 kHz is split up to '' at 20 kHz in frequency ranges of multiple width from δf kHz.

Step 502 is followed by a step 503 consisting in searching for the sub-band δf of maximum energy in each of the aforementioned ranges, defined previously in step 502. This sub-band δf is denoted S _n (imax ) and allows to center all the ranges defined in step 502 around the maximum energy sub-band considered.

Step 503 is itself followed by a step 504 consisting in calculating the average energy of each range centered around the maximum energy band for the series of samples of rank n considered.

In step 504, the energy of each range is noted E _n (sb) and verifies the relation:

E _n (sb) = ∑ (S _n (imax +/- δf / 2))

Step 504 is then followed by a step 505 consisting in calculating the ratio between the energy E _n (sb) of the ranges for the current series of samples and for a plurality of previous sequences, E _n - _S (sb) of successive samples. The aforementioned report is written:

E _n (sb)

R „(sb) =

E _n . _s (sb)

In the previous relation, s indicates the temporal past relating to s frequency spectra relating to a given number of series of corresponding samples.

The above-mentioned step 505 is itself followed by a step 506 consisting in calculating a hearing contrast value, denoted C _{n / S} b. The value of auditory contrast checks the relationship:

- ≠ -p, - (p-l), ... 0, ..., p-l, p

In this relation, R _n (sb + i) with i = -v and i ≠ -p, - (p-1), ..., 0, ..., p-1, p, p designating an arbitrary value , denotes the value of the ratio for neighboring ranges of the same series of samples of rank n and of the same spectrum S _n .

In step 506, the auditory contrast C _n , _sb is compared with a first whistle threshold value, denoted Si, by comparison of superiority. On a negative response to the aforementioned comparison, a return step 508 brings back to the series of samples of rank n + 1 following and in particular to step 501 of calculating the spectrum of the signal by Fourier transform. On the contrary, on positive response to the aforementioned comparison step of step 506, a step 507 is provided, which consists in calculating a proximity parameter denoted P _n , sb verifying the relation:

R- (sb) n, sb

^K i = l

Step 507 then includes a step of comparing the proximity parameter P _n , _Sb with a second whistle threshold value S ₂ , P _n , sb> S ₂ . On a negative response to the comparison of the abovementioned step 507, a return loop 509 makes it possible to return to step 501 of calculating the spectrum of the signal by Fourier transform for the following series of samples of rank n + 1. On the contrary, on a positive response to the comparison test of step 507, the presence of a hissing parasitic signal is revealed. It will of course be noted that the presence of the parasitic whistling signal is revealed in step 510 if the comparisons of superiority of the auditory contrast value C _n , _sb and of the proximity parameter Pn, sb with respect to the first threshold If respectively of the second threshold S ₂ are both verified.

Supporting evidence of the embodiment of whistling detection in the embodiment of Figure 6b will be given below.

The signal spectrum calculation in step 501 can be performed from fast Fourier transforms.

The values of the frequency components S _n (i) of the sub-band decomposition of the sample sequence tillons of rank n are observed on frequency ranges of width multiple of δf kHz and their temporal evolution is thus studied on a temporal past of s spectra. These operations are carried out in steps 502, 503, 504 and 505.

On each of the ranges resulting from the decomposition, the maximum of the energy of a sub-band of given rank is sought. The ranges are then refocused around the frequency, that is to say the rank i supporting this maximum, and the averages of the energy E _n (sb) are calculated for the series of samples of rank n considered. The evolution of the energy for each of these new ranges is observed from the ratio R _n (sb) and the auditory contrast criterion C _n , _Sb is then calculated by observing the behavior of the aforementioned ratio with respect to the neighborhood . The presence of a hiss is checked if the hearing contrast value is greater than the first threshold value and if the proximity parameter is greater than the second threshold value. In the previous relationships, v indicates the index of neighboring ranges of the same spectrum S _n relative to the same series of samples n, p denotes the number of sub-bands on either side of the maximum not taken into account in the calculation of the contrast value and k denotes the number of ranges.

For digital audio coding of MPEG1 LU type transmitted in digital broadcasting, the calculation of the Fourier transforms can be carried out over a length ranging from 256 to 4096 samples, the minimum overlap being from 25 to 75%. The ranges have a frequency width Δf e [500 Hz, 1500 Hz], the temporal past of observation tion est se [1,4] in number of successive spectra, that is to say of sequences of successive samples.

The neighborhood ve [1.6] and p € [0.4]. The threshold values are Si e [50,1000] and S ₂ € [2,20]. In addition and in a nonlimiting manner, to carry out the whistling detection process, the method which is the subject of the present invention can also consist in carrying out a step of filtering the spectral components into sub-bands not audible to the human ear. Under these conditions, this operating mode makes it possible to take into account the psycho-acoustic properties of the digital audio signal, the energy spectra being previously multiplied by the absolute hearing threshold, according to the formulation of the French standard ISO 226 of 1987, extended to -beyond 12.5 kHz.

A more detailed description of a process for detecting a spurious signal such as a hum will now be given in connection with FIGS. 7a and 7b. The buzzing fault introduced on digital audio signals is due to the difference in protection in the digital audio frame given to sensitive bits such as synchronization bits and error correction code bits, while the remaining bits of the frame are less well protected. Thus, when the frame is reconstructed after decompression, the digital audio data can still be erroneous due to this less good protection.

During the implementation of the process which is the subject of the present invention, investigations carried out in the presence of buzzing noises have shown that when this fault appears, a low frequency pink noise is added to the signal spectrum.

Consequently, and in accordance with a particularly advantageous aspect of the method which is the subject of the present invention, and with reference to FIG. 7a, the step consisting in detecting in the digital audio signal a parasitic signal such as a humming consists in detecting in this parasitic signal a pink noise in a frequency band between 0 and 1100 Hz and of substantially constant level in this frequency band. Referring to FIG. 7a, it is indicated that the substantially constant level of pink noise introduced during the appearance of this defect is of the order of 40 dB. FIG. 7a thus represents the spectrum of a digital audio signal before the appearance of pink noise, in dashed line, during the appearance of pink noise, in dotted line, and after the appearance of pink noise, in solid line . A specific mode for detecting the hum defect allowing the detection of the aforementioned pink noise will now be described by way of illustration, in conjunction with FIG. 7b.

The aforementioned detection process is implemented on at least one left or right channel of the digital audio channel. It consists in a step 700, as shown in the aforementioned figure, of cutting out the sequences of N samples of the digital audio signal ADS and then, in a step 701, of calculating the spectral composition of this signal on the series of samples considered. digital audio defined as the value S _n (i) of frequency component in sub-bands of central frequency fi, n designating the rank of the series of samples. It is understood in particular that step 701 represented in FIG. 7b can advantageously be carried out in the same way as step 501 in FIG. 6b, a single decomposition by Fourier transform then being carried out for all of the two detection processes. The previously mentioned step 701 is then followed by a step 702 consisting in calculating, for a determined number k of central frequency f- _. of the low frequency domain, that is to say between 0 and 1100 Hz, a first α _in and a second βι, _n ratio of the values of frequency components in sub-bands for the current sample suite and the following of previous samples, respectively for the current sample suite and the next sample suite. Thus, the first report αι, _n verifies the relation

α, ,, „n = - _Sn - _{i (i)} and the second relation βi _n checks the relation

' ^• "S _{n + I} (i)

In the preceding relations, S _n -ι (i) relates to the series of preceding samples and S _{n +} ι (i) relates to the series of following samples with respect to the current series of samples S _n (i). In step 702, the first and second ratios αι, _n and β _in are then compared with a first buzzing threshold value, denoted S'χ. On a negative response to the comparison of step 702, a return 703 is made to the implementation of step 702 for the component in sub-bands of rank i corresponding to the same series of samples of rank n.

On a positive response to the comparison test of step 702, a step 704 is carried out, this step consisting in subjecting the comparison of the first and second reports to a criterion of proportion p / k of the number p of verified comparisons with respect to all of the k comparisons made for the k center frequencies f- _. components in sub-bands considered. The p / k ratio can be expressed as values of P%. On a negative response from the comparison of the first and second reports to the aforementioned proportion criterion, a return by a loop 708 is carried out in step 701 of calculation of the signal spectrum for the following series of samples of rank n + 1 following.

On the contrary, on positive response to the comparison of step 704, a step 705 is carried out, which consists in discriminating, among the values in sub-bands S _n (i) of the frequency components in sub-bands, the value maximum S _n (i _max ) for the components in sub-bands of the values of frequency components relating to the current sequence of samples of rank n. Step 705 is then followed by a step 706 consisting in calculating the ratio of the aforementioned maximum value with the value of the frequency line situated at the same index i _m a _x of the spectrum of the sequence nl of samples, this ratio verifying the relationship :

^ n l'max) _n , i =

Sn-l max)

The above-mentioned ratio is then compared to a second buzzing threshold value, denoted S ′ ₂ . The comparison with the second hum threshold value S ' ₂ is a comparison of inferiority. On a negative response to the aforementioned comparison, a return loop 709 returns to step 701 for the series of samples of rank following n + 1.

Thus, steps 702, 704, 705 and 706, upon positive response to the comparisons with respect to the thresholds of buzzing S'i and S ' ₂ , comparison of superiority with respect to S'i and comparison with inferiority with respect to S' ₂ , of the ratios oi, _n and βi, n, respectively of the ratio M _n .i _f allow to conclude that a parasitic buzzing signal exists. Step 706, in this case, is then followed by a step 707 of statistical analysis consisting for example in determining the multiple occurrence of a parasitic buzzing signal possible over a given observation time τ _b of s seconds . To carry out step 707, this can consist of repeating the preceding operations of discrimination of the existence of a comparison of superiority of the first and second ratios to the first humming threshold value S'i and of existence of 'a comparison of the inferiority of the ratio M _n , i to the second threshold value S' ₂ . During the repetition of these operations in step 707, a binary variable for predetection of the existence of a parasitic hum signal is stored. This binary variable is assigned the value 1 when the criteria for comparing superiority and inferiority are satisfied and the value 0 otherwise.

In addition, a count in the duration τ _b = s seconds of the number of occurrences of the binary variable of pre-detection at the value 1 is carried out. This number, denoted NV _pd , is compared with a third buzzing threshold value S ' ₃ by comparison of superiority. On a negative response to this comparison, NV _pd > S ' ₃ , a return 710 by a loop is carried out in step 701 for the following series of samples of rank n + 1 following. On the contrary, on a positive response to the test against the third buzzing threshold S ' ₃ and carried out at in step 707, a parasitic hum signal is revealed when this comparison to this third hum threshold value is verified. The presence of the parasitic hum signal is obtained in step 711. A proof of the detection process of the parasitic hum signals will be given below.

Firstly, it is indicated that when the proportion of frequency lines for which the first ratios α _in and βi, _n are greater than the same hum threshold value S'i and greater than the aforementioned value P%, then it there may be a parasitic humming noise on the series of samples of rank n considered. While this condition appears necessary, it does not appear sufficient. For a parasitic hum signal to be detected, other conditions must be verified, in particular the comparison occurring on the most energetic line of the portion of frequency spectrum studied. This second condition concerns the comparison of the ratio M _ni to the second hum threshold value S ' ₂ -

Finally, the property of the signal linked to the stereophonic mode of the latter is implemented during the third comparison with the third threshold value. Thus, there can be detection of a parasitic hum signal only if the two comparisons relating to the hum thresholds S'i and S ' ₂ are satisfied on only one of the two channels of the digital audio signal for the current series of samples. of rank n. Statistical analysis is performed in step 707 and decisions are stored along a window temporal length τ _b = s seconds. If in this time window the number of detection all channels combined exceeds the threshold value S ' ₃ , then there is effectively a parasitic hum signal. The values of the previously described parameters are now given in the case of MPEG1 LU digital audio coding transmitted by a digital broadcasting channel.

In this case, the computation of the Fourier transforms can be carried out in the same manner as described previously in the description. The width of the spectrum observed can be from 500 Hz to 1.5 kHz at low frequencies, the number of stored spectra being equal to 3, that is to say for the preceding series of samples of rank nl, the current sequence of samples of rank n and the following sequence of samples of rank n + 1:

- the frequency band can be between [0 Hz, 1500 Hz]. the first hum threshold value

If [1.2] - P%> 25%

- the second hum value S ' ₂ e [1,2]

- τ _b = s, se [1s, 10s]

- the third hum threshold value

S ' ₃ e [1,10]. A more detailed description of a process allowing the detection of a spurious signal such as a phase shift between channels of the digital signal will now be described in conjunction with FIGS. 8a to 8c.

The presence of a spurious signal such as a shift between left and right channels, relative phase shift between the aforementioned channels of a digital audio channel is caused by the phase shift of the signal present on each of the channels. Such a defect and the corresponding spurious signal can only appear before the digital audio coding and compression of the signal, but it is nevertheless passed on all along the broadcasting chain.

When the ADS digital audio signal is monophonic, a certain phase shift between the right and left channels can be tolerated. Some program providers claim to simulate a stereophonic effect. However, beyond a certain phase shift value, the effect is no longer acceptable to listeners. When, on the contrary, the signal is stereophonic, the phase value not to be exceeded may be different.

Thus, with reference to FIG. 8a, the step consisting in detecting in the digital audio signal a spurious signal such as a phase shift between right and left channels of the digital signal can consist, in a step A, of calculating the value of phase shift between channels of the digital audio signal from the intercorrelation function of the digital audio signal present on each of the right and left channels of the digital audio channel. The aforementioned step A is followed by a step B consisting in comparing the calculated phase shift value with a threshold value. The relative phase shift between channels is noted φ and the threshold value is noted φ _ma χ- this value varying according to the mono or stereophonic transmission mode of the signal.

A specific non-limiting implementation of the detection process of the parasitic phase shift signal will now be given in connection with FIGS. 8b and 8c.

Referring to FIG. 8b, it is indicated that the method according to the object of the present invention consists, in a step 800, of cutting the digital audio signal following samples of N samples, each sequence comprising the rank n . This division is of course carried out on the left and right channels of the ADS digital audio signal. The above-mentioned step 800 is followed by a step consisting in calculating, on the aforementioned series of samples, given number N of samples, the intercorrelation function between the digital audio signal present on the left channel and on the right channel. The step of calculating the above-mentioned intercorrelation function can be carried out using a step 801 of calculating the complex spectra of the left and right channels by Fourier transform, value of the frequency component i of the series of rank samples not. This step 801 is followed by a step 802 of multiplication of a spectrum of a channel by the conjugate of the spectrum of the other channel, then of a step 803 of calculation proper of the inverse Fourier transform to obtain the intercorrelation function. The operations performed in steps 801, 802, 803 will not be described in detail since they correspond to conventional operations in processing the digital signal. The aforementioned step, following step 803, is followed by a step 804 consisting in determining the rank i of the sample of the cross-correlation function, sample denoted corr (i), corresponding to the maximum value corr ( i) of this intercorrelation function. This maximum search step can be carried out using a function of sorting on the value of the samples of the intercorrelation function.

Step 804 is then followed by a step 805 consisting, from a determined attenuation value A, in determining the attenuated rank i _ιnf , i _sup of the samples corr (i _ιnf ) and corr (i _sup ) of the intercorrelation function distributed on either side of rank i of the maximum sample corr (i) and corresponding to an attenuated value of the value A with respect to the maximum value of this intercorrelation function.

Step 805 also consists in calculating a first ratio of the maximum value to the corrected lower corrected value (i), this first report being written then a corr (i _mf ) second ratio of the maximum value to the upper attenuated value , this second report verifying the relation corr (i) corr (i _sup )

Step 805 finally consists in comparing the value of the above-mentioned first and second ratios with a first threshold value A, denoted S "ι. On comparison of equality or superiority to this threshold value, the left / right contrast of the digital audio signal between left channel and right channel is then considered to be significant. This contrast is significant because the indices of lower and upper value distributed on either side of the maximum of the intercorrelation function exist as well as the value of their ratio, these values can then be compared with the first phase shift threshold value S "ι. If these indices do not exist, in negative response to the comparison test carried out in step 805, a return loop 806 brings the process back to step 801 for the following series of samples of rank n + 1.

On a positive response to the comparison carried out in step 805, the lower and upper indices existing, this step 805 is followed by a step 807 consisting in searching for the rank noted j of the second relative maximum corr (j) of the function of intercorrelation.

With reference to FIG. 8c, on which the amplitude of the intercorrelation function as a function of time has been shown, that is to say successive samples representing this intercorrelation function. The first maximum corresponding to the sample i is represented, ie corr (i), the attenuated values of the value A and of index i _ιnf and i _sup corresponding, as well as the second maximum of index j. The value of the attenuation R corresponds to the difference between the maximum maximorum of the cross-correlation function and this second maximum. The index j of the second maximum of the cross- _correlation function is sought over the intervals [0; i _ιnf [and

Step 807 is then followed by a step 808 consisting in calculating a left / right contrast parameter C _g , _d ratio between the maximum value corr (i) and the value of second maximum corr (j). The left / right contrast value checks the relationship:

_ corr (i) corr (j)

Step 808 also includes a comparison of the value of the abovementioned contrast parameter C _{9 d} with a second phase shift threshold value, noted S " _2. On a negative response to the aforementioned comparison, a return by a return loop is carried out in step 801 for the following series of samples of rank n + 1. On the contrary, on a positive response to the comparison of the aforementioned step 808, the preceding operations successive to the comparison of the first and of the second ratio to the first maximum value of the cross-correlation function, that is to say the steps 805, 807 and 808, are repeated so as to determine in the successive rows the rank which has the most occurrences These operations are carried out, for example in step 809, where the result relating to the value i of the maximum of the function of intercorrelation is stored in a table, and in a step 810 where a statistical analysis is performed on the number of occurrences in this table. In step 810, if a value ia an occurrence greater than or equal to a third me phase shift threshold value S " ₃ , then, and in positive response to this comparison of superiority, the relative phase shift of the left and right channels of the digital audio channel is assigned a value corresponding to that of the rank which has the most occurrences, that is, the value of rank i. This allocation is carried out in step 811. On the contrary, if no occurrence is greater than or equal to the third phase shift threshold value S " ₃ , then a loop 812 leads to the following series of samples of rank n + 1. A proof of the operating process described in connection with FIGS. 8b and 8c will now be given below. As regards the calculation of the cross-correlation function, this can be estimated between the channels present on the left and right channels by a time average verifying the relationship:

In this relation, T (k) denotes the value of the cross-correlation function at point k, G (q) and D (q + k) denotes the sample of the left-hand lane respectively of rank q and corresponding q + k . In this relation, k varies from 0 to N-1.

The complexity in computation time of this estimator according to the aforementioned relation is proportional to N ² . The calculation of the circular convolution by complex FFT as described with steps 801, 802 and 803 makes it possible to reduce the computation complexity to the value (2N). Log ₂ (2N).

For MPEG1 LU type digital audio coding transmitted by a digital broadcasting channel, the observation and calculation lengths N and K can both be equal to 32768, a minimum value being equal to 1024.

The phase shift calculation threshold value Ξ "ι is between S" ι e [2,100].

The value of the second phase shift calculation threshold S " ₂ is between S" ₂ € [1,5]. The size of the table of results produced in step 810 for carrying out the statistical analysis of rank i corresponding to the maximum of the function d ¹ intercorrelation can be 10 successive values. The value of the third phase shift calculation threshold S " ₃ can for example be taken equal to 5.

A device for controlling the quality of a digital audio signal implementing the method which is the subject of the present invention previously described in the description will now be described in conjunction with FIG. 9.

Referring to the above-mentioned figure, it is indicated that the device for controlling the quality of a digital audio signal object of the present invention comprises at least one module 1 for converting the digital audio signal ADS into a digital signal of specialized format. By way of nonlimiting example, it is indicated that the module 1 for converting the digital audio signal into a digital signal of specialized format can be achieved by means of a professional quality IRD circuit receiving the digital audio signal ADS from a first input type BIS, for inter-sa telli te band, or from an input type MPEG2 TS. Of course, the implementation of such a module is not essential as such, this module can be replaced by a DAB receiver for example, for Digital Audio Broadcasting. The module 1 for converting the digital audio signal into a digital signal of specialized format delivers this signal in the EBU / AES format.

In addition, the device which is the subject of the present invention as shown in FIG. 9 comprises a portable computer type assembly comprising at least one module 2 for acquiring the left and right audio frequency components, this acquisition module 2 receiving the signal. digital format of specialized format delivered by the module 1 for converting the digital audio signal into digital signal of specialized format. The acquisition module 2 left and right audio frequency components then delivers a specialized digital audio signal for each of the left and right channels, denoted RL in FIG. 9.

The module 2 for acquiring the audio-frequency components is itself followed by a module 3 for detecting coding and transmission faults receiving the specialized digital audio signal for each of the left and right channels delivered by the aforementioned module 2. It makes it possible to detect at least one of the parasitic signals such as short cut, whistling, buzzing, relative phase shift of the left and right channels and thus to deliver a detection signal, in accordance with the method which is the subject of the present invention.

In addition, a management module 4 of the man-machine interface type receives the detection signal and makes it possible to generate an alarm signal in the presence of at least one of the above-mentioned interference signals.

As shown in FIG. 9, it is indicated that the device which is the subject of the present invention may include a module 5 for calculating and detecting additional parameters, this calculation module receiving the digital signal of specialized format delivered by the module 1 and delivering a signal representative of additional parameters such as mono or stereo mode, bit rate values of the digital audio signal. It is controlled by module 3 for detecting coding and transmission faults.

In general, it is indicated that all of the modules 2, 3, 4 and 5 can be produced by means of a microcomputer, which for this reason is represented by dotted lines in FIG. 9. In particular , the human interface platform management system machine allows remote control of the module 1 for converting the digital audio signal into a digital signal of specialized format.

Thus, the system constituted by the microcomputer performs the processing of the data, provides the results and orders them while allowing the management of the various signals to be processed by the module 1 for conversion to specialized format.

Thus, the module 2 for acquiring the components can be produced by a dedicated PCI type card interconnected with the format conversion module 1. The audio components of the left and right channels are thus acquired from the digital audio signal in the specialized EBU / AES format. The digital data supplied by module 2, and therefore by the PCI type card, are processed by module 3 for detecting faults, which of course makes it possible to implement in software form the various steps of the method which is the subject of the present invention such as described previously in the description. By way of nonlimiting example, all of the corresponding software elements can be installed in read-only memory, called in random access memory of the microcomputer and controlled from the management module 4 constituting the abovementioned human-machine interface HMI. All of the aforementioned software elements make it possible to detect in the digital audio signal at least one of the spurious signals such as a short cut, whistling, buzzing, relative phase shift of the left and right channels of this digital audio signal and to generate a signal. alarm in the presence of at least one of the spurious signals. Finally, module 5 provides additional results such as the detection of the mono or stereo transmission mode, the bit rate of the digital audio signal from the results delivered by module 3 as well as EBU / AES signals delivered by module 1.

Claims

1. A method for controlling the quality of a digital audio signal, characterized in that this method consists in detecting in this digital audio signal at least one of the spurious signals such as short cut, whistling, buzzing, relative phase shift of the left channels and right of this digital audio signal, which makes it possible to generate an alarm signal in the presence of at least one of said parasitic signals.

2. Method according to claim 1, characterized in that it further consists in discriminating the mode of mono- or stereophonic transmission of this signal.

3. Method according to claim 1 or 2, characterized in that, for the purpose of implementing an overall quality control of said digital audio signal, this consists, in combination, over a sliding time window allowing the observation of a series of successive samples of said digital audio signal:

- counting down the number of occurrences N _M of parasitic signals of short cut-off for a duration T _M of observation of this digital audio signal and comparing the number of occurrences N _M with a threshold value S _M ;

- counting down the number of occurrences N _s of hissing parasitic signals for a duration T _s of observation of this digital audio signal, and comparing the number of occurrences N _s with a determined threshold value

to detect, for a duration τ _D , the phase shift value φ and the number of occurrences N _D of these phase shift values over a determined number D of phase shift calculations phasing and comparing the phase shift value φ calculated with a determined threshold value S _D ;

- counting down the number of occurrences N _B of interference parasitic signals for a duration τ _B of observation of this digital audio signal and comparing the number of occurrences N _B with a threshold value S _B ;

- send an alarm signal of degradation of transmission quality of said digital audio signal on crossing of at least one of the threshold values determined by the number of corresponding occurrences counted down.

4. Method according to claim 3, characterized in that the step consisting in transmitting said alarm signal is conditioned to an order of priority of said crossings.

5. Method according to claim 4, characterized in that orders of decreasing priority are allocated to the crossings by the numbers of occurrences of the parasitic signals of short cut, of whistling, of displacement and of humming respectively.

6. Method according to one of claims 2 to 5, characterized in that said step consisting in discriminating the mono- or stereophonic mode of the transmitted signal consists in detecting sudden and brief changes of context in mono- or stereophonic mode, or vice versa , of the transmitted signal.

7. Method according to claim 6, characterized in that the step consisting in discriminating the mode of mono- or stereophonic transmission comprises, for a transmission of the digital audio signal in mono- or stereophonic, for each successive series of samples of rank n:

- a step of calculating the respective energies of the right and left channels of the transmission channel; a step of calculating the ratio M _n of the right and left energies;

a step of calculating a binary context variable C _n , the binary variable C _n being assigned the value 0 representative of a context of monophonic mode if the ratio M _n of the energies is between a first and a second value threshold, and a value 1 representative of a stereophonic mode context otherwise;

a step of verification of the context consisting, over a determined number of successive samples, in calculating an accumulation of the value of the successive binary variables C _n and in comparing the value of this accumulation with a reference value, and on comparison of superiority from this accumulation to this reference value, to be assigned to the transmission mode the stereophonic mode and the monophonic mode otherwise.

8. Method according to one of the preceding claims, characterized in that the step consisting in detecting, in this digital audio signal, a spurious signal such as a short cut, consists in detecting, on a series of successive samples of this digital signal, a rapid decrease in the energy level of this digital audio signal towards a substantially zero energy, revealing an absence of reverberation of this digital audio signal.

9. Method according to claim 8, characterized in that the step consisting in detecting, in this signal digital audio, a spurious signal such as a short cut includes:

a step consisting in determining separately on each stereophonic channel, for a plurality of sequences of N successive samples, the average energy E _n of the signal transported by this channel, n designating the rank of each series of samples;

a step consisting in comparing the evolution of the average energy for sequences of N successive samples, the existence of a parasitic short cut signal being revealed if at least one of the average energies is substantially zero and if one or more average energies close to this substantially zero average energy are greater than a determined threshold value.

10. Method according to one of the preceding claims, characterized in that the step consisting in detecting, in this digital audio signal, a spurious signal such as a whistling sound, consists in detecting in this digital audio signal a sudden and transient increase in the spectral energy of this digital audio signal in a frequency band whose low frequency is between 4.5 kHz and 6.5 kHz and whose high frequency can reach up to 20 kHz.

11. Method according to claim 10, characterized in that the step consisting in detecting, in this digital audio signal, a parasitic signal such as a whistling sound comprises at least:

a step consisting in calculating on a series of samples of the digital audio signal the spectral composition of this digital audio signal defined as the value S _n (i) of frequency components in sub-bands of center frequency fi and bandwidth δf, n denoting the rank of the series of samples; a step consisting in calculating the average value of the energy E _n (sb) of a range of said sub-bands for the series of samples of given rank n;

a step of calculating a hearing contrast value C _n , _sb from the value of the ratio

E (sb)

R (sb) = - between the energy E _n (sb) of this range for

E „_ _s (sb) the current sequence and for a plurality of previous sequences E _n _ _s (sb) of samples, the auditory contrast value Cn.s verifying the relation:

R _n (sb)

= Cn, sb

∑R _n (s + i)

2- (v-l) '1. = - V i ≠ -p, - (p-l), ... 0, ..., p-l, p

where R _n (sb + i), i = -v denote the value of the ratio for the neighboring sub-bands of the same series of samples of rank n and of the same spectrum S _n ;

a step of comparing the auditory contrast value C _n , sb with a first whistle threshold value Si, C _n , _S b>Si;

a step of calculating a proximity parameter P _n , sb verifying the relation:

a step of comparing the proximity parameter P _n , _Sb has a second whistling threshold value S ₂ , P _{n, sb} > S ₂ , the presence of a parasitic whistling signal being revealed if the comparisons of superiority of the the hearing contrast values C _n , _sb and the proximity parameter P _n , _sb are both verified.

12. Method according to claim 11, characterized in that it further comprises, prior to the step of calculating, on a series of samples of the digital audio signal, the spectral composition of this signal:

- a calculation step, on this series of samples e (i) of the mean time energy, n Emoy _n = ∑e (i); i = l

a first step of comparing the superiority of the calculated value of this average temporal energy with a first threshold value, Seuili, the process being reduced to the step of calculating the average temporal energy, the energy of at least a current sequence of rank n being considered as insufficient, and, on a positive response to said first comparison step,

- a second step of comparing the superiority of the ratio of the energy of the current sequence to the energy of an immediately adjacent, non-adjacent sequence, of rank n-2, to a second threshold value, Threshold ₂ , the process on negative response to this second comparison step being brought back to the step of calculating the average temporal energy, the energy of the current sequence and of the neighboring sequence being similar, the process, on positive response to this second step of comparison, being continued, the energy of two neighboring sequences being increasing.

13. The method of claim 11, characterized in that it further comprises a step of filtering the spectral components into sub-bands not audible to the human ear.

14. Method according to one of the preceding claims, characterized in that the step consisting in detecting, in this digital audio signal, a parasitic signal such as a humming consists in detecting in this parasitic signal a pink noise in a frequency band between 0 and 1100 Hz and of substantially constant level in this frequency band.

15. Method according to claim 14, characterized in that the step consisting in detecting, in this digital audio signal, a parasitic signal such as a buzzing comprises on at least one left or right channel of this signal:

a step consisting in calculating on a series of samples of the digital audio signal the spectral composition of this digital audio signal defined as the value S _n (i) of frequency components in sub-bands of central frequency fi, n denoting the rank of the suite of samples;

- to calculate, for a determined number k of central frequencies fi of the low frequency domain, a first and a second ratio of the values of frequency component in sub-bands for the series of samples

current and the preceding series of samples α; _n = -

'S _n -l (i) respectively the current series of samples and the

following series of samples βi _n = Sn (.

S _{n +} l (i) - comparing the value of the first and second reports with a first buzzing threshold value S'i; subjecting the comparison of the first and second reports to a criterion of proportion p / k of the number p of verified comparisons with respect to all of the k comparisons made for the k center frequencies fi and, if this criterion of proportion is verified;

- to discriminate, among the values S _n (i) of frequency components in sub-bands, the maximum value

S _n (i _a χ) of the values of frequency components relative to the current series of samples;

- calculating the ratio of said maximum value with the value corresponding to the index i _max of the spectrum

the previous sequence S _n -ι (i _ma χ), M _n; j = - -— ^ -, and to compa-

^ nl ^ max) rer the value of this ratio to a second humming threshold value S '₂;

- to discriminate, on at least one transmission channel in stereophonic mode of the digital audio signal, the existence of a comparison of superiority of the first and second ratios α.i, _n and βi, _n to the first hum threshold value S 'i and the existence of an inferiority comparison of the ratio of the maximum values M _n , i to the second humming threshold value S'₂; - to repeat the previous operations and to periodically memorize a predetermined binary variable for the existence of a buzzing parasitic signal over a fixed period, the value of the predetermined binary variable being assigned the value 1 when said criteria superiority and inferiority comparison res are satisfied, and the value 0 otherwise;

- counting down in time s determined the number NVp _d of the binary predetection variable to the value 1 and comparing this number to a third humming threshold value S ' ₃ , NV _pd >S' ₃ , the presence of a parasitic hum signal being revealed if the comparison of the superiority of the number NV _pd of the binary variable of predetection to this third hum threshold value is verified.

16. Method according to one of the preceding claims, characterized in that the step consisting in detecting, in this digital audio signal, a spurious signal such as a phase shift between channels of the digital signal consists in: - calculating the phase shift value between channels of the digital audio signal from the function of intercorrelation of the digital audio signal present on each of said channels;

- compare the calculated phase shift value with a threshold value.

17. Method according to claim 16, characterized in that the step consisting in detecting, in this digital audio signal, a spurious signal such as a relative phase shift of the left and right channels of the digital audio signal comprises the steps consisting in:

- to calculate on a given number of samples, the cross-correlation function between the digital audio signal present on the left channel and on the right channel;

- determining the rank i of the sample corr (i) corresponding to the maximum value corr (i) of this cross-correlation function; - to determine from an attenuation value A determined the attenuated rank (iinf), (isup) of the samples of the intercorrelation function distributed on either side of the rank of the maximum sample corr (i) and corresponding to an attenuated value corr (ii _nf ), corr (i _sup ) of the value A with respect to the maximum value of the cross-correlation function; corr (i)

- to calculate a first report and a corr (i) second ratio of the maximum value to the corrected (i _sup ) value lower respectively higher;

- comparing the value of the first and second ratios with a first phase shift threshold value S "ι and, on comparison of equality or inferiority to this threshold value, the contrast of the digital audio signal between left channel and channel on the right being significant,

- determining the rank (j) of the second relative maximum corr (j) of the intercorrelation function;

- calculating a contrast parameter C, ratio of the maximum value rank and the second maximum value rank of the function d ¹ intercorrelation;

- to compare the value of the contrast parameter C with a second phase shift threshold value S " ₂ , and on criteria of comparison of superiority to this second threshold value, - to repeat the preceding successive operations with the comparison of the first and the second ratio to the first maximum value of the intercorrelation function and to determine in the successive ranks the rank which has the most occurrence, at the relative phase shift left and right being assigned a value corresponding to that of the rank which has the most occurrence.

18. Device for controlling the quality of a digital audio signal, comprising at least: - a module for converting said digital audio signal into a digital signal of specialized format;

a module for acquiring the left and right audio frequency components receiving said digital signal of specialized format delivering a specialized digital audio signal for each of the left and right channels;

- a coding and transmission fault detection module receiving the specialized digital audio signal for each of the left and right channels and making it possible to detect at least one of the spurious signals such as short cut, whistling, buzzing, relative phase shift of the left channels and right, and to deliver a detection signal;

- a management module of the man-machine interface type receiving the detection signal and making it possible to generate an alarm signal in the presence of at least one of the parasitic signals.

19. Device according to claim 18, characterized in that it further comprises a module for calculating and detecting additional parameters, this module for calculating and detecting receiving said digital signal of specialized format and said detection signal and delivering a signal representative of additional parameters such as mono or stereo mode, bit rate value.

20. Software elements making it possible to detect in a digital audio signal at least one of the parasitic signals such as brief cut, whistling, buzzing, relative phase shift of the left and right channels of this alarm signal in the presence of at least one of the parasitic signals.