EP3403260B1 - Method and apparatus for conditioning an audio signal subjected to lossy compression - Google Patents

Description

The invention relates to a method for processing a lossy compressed audio signal.

The data compression of audio signals or information, such as. B. music files is known in and of itself. The purpose of data compression is to reduce the data size of corresponding audio signals. In principle, data compression can be lossy or not lossy. In addition, the lossy data compression should be considered in particular, which can be implemented, for example, by data-based rejection of frequency components lying on the edge of the human hearing range. The subjective hearing perception by a listener should hardly be impaired in this way.

Because of the reduced sound quality in comparison of lossy compressed audio signals, it is sometimes desirable to prepare lossy compressed audio signals, i. H. to at least partially restore correspondingly rejected frequency components or to replace them with comparable frequency components.

To date, different technical approaches have been known for processing lossy compressed audio signals. These known approaches are regularly comparatively (arithmetically) complex and not very efficiently designed. There is therefore a need for further development of improved methods for processing a lossy compressed audio signal.

The document is in the prior art EP 1501190 A1 is known which discloses a method and a device for equalizing an audio signal with an external interference signal to increase the intelligibility of the audio signal.

The invention is therefore based on the object of specifying an improved method for processing a lossy compressed audio signal.

The object is achieved by a method according to claim 1. The dependent claims relate to advantageous embodiments of the method. The object is further achieved by the device according to claim 14 and by the audio device according to claim 15.

The method described herein is generally used to process a lossy compressed audio signal. With a processable or processed audio signal, it can e.g. B. a lossy compressed audio file or part of such. Specifically, it can be e.g. B. a lossy compressed audio file using an mp3 algorithm, d. H. is an mp3-encoded audio file or mp3 file.

The audio file or parts of it can already be decoded. For the above example In an mp3-coded audio file, for example, suitable decoding algorithms can be used, by means of which an at least partial decoding of the mp3-coded audio file was carried out. The same applies, of course, to audio data that was not coded using an mp3 algorithm, but rather using other algorithms.

In all cases, the audio file can e.g. B. audio signals e.g. B. include a piece of music.

Processing is generally an at least partial restoration that is missing, i. H. z. B. discarded as part of data compression, frequency components or an at least partial replacement missing, d. H. z. B. in the context of data compression discarded to understand frequency components by comparable frequency components. As can be seen further below, for the processing according to the method, lossy compressed audio signals in particular require an at least partial replacement of missing, ie. H. z. B. discarded in the context of data compression, frequency components relevant.

The individual steps of the method described here are explained in more detail below:
In a first step of the method, a lossy compressed audio signal to be processed is provided. A corresponding audio signal can in principle be provided via any physical or non-physical audio source, that is to say, for example, from an audio device for processing and outputting audio signals.

In a second step of the method, the audio signal is transmitted into a frequency spectrum. In the frequency spectrum, energies of the audio signal are correlated with frequencies of the audio signal. In other words, the content of the audio signal is reduced to its energy, i.e. H. Amplitude or frequency components are examined and the individual energy components of the audio signal are transmitted or converted in terms of data into a frequency-dependent representation. For this purpose, the audio signal is typically divided into individual, possibly overlapping, time intervals, which are individually transmitted or converted into the frequency spectrum. The transmission or conversion of the audio signal into the frequency spectrum takes place by means of suitable algorithms, i. H. z. B. using (faster) Fourier transform algorithms. The length of the algorithms is basically variable. The examination of the content of the audio signal for its energy components can include a classification and grouping of the energy components and an estimate of the energy components of the audio signal.

In a third step of the method, frequencies of local amplitude maxima are determined in the frequency spectrum. In other words, the frequency spectrum is examined for local amplitude maxima and the frequencies associated with the respective amplitude maxima are determined. Below a local amplitude maximum is an amplitude maximum value to be understood in a defined frequency range. Local amplitude maxima are determined using suitable analysis algorithms.

In a fourth step of the method, a first selection criterion is defined. On the basis of the first selection criterion, the frequencies of two immediately following (local) amplitude maxima are preselected, which frequencies meet the first selection criterion. In the fourth step, the frequencies of pairs of immediately consecutive amplitude maxima are examined with regard to the first selection criterion. In the fourth step, the frequencies of immediately successive amplitude maxima are then examined in pairs to determine whether the frequencies associated with the respective amplitude maxima meet the first selection criterion. In the further steps of the method, only the frequencies that meet the first selection criterion are typically considered. In the fourth step, the frequencies to be considered below and the associated amplitude maxima are preselected.

The first selection criterion typically describes a specific limit frequency value (range) (“threshold”). Frequencies of immediately successive amplitude maxima satisfy the first selection criterion if their frequency difference exceeds the limit frequency value (range) described by the first selection criterion, cf. the relationship represented by the formula I shown below: $$\mathrm{\Delta }{f}_i>\left|\mathrm{\Delta }{f}_T\right|$$

In this case, Δf _i applies: frequency difference between two immediately following amplitude maxima; Δf _T : cutoff frequency value (range).

The limit frequency value (range) can be determined by transferring the preselected frequencies to a Bark scale. As is known, frequencies can generally be transferred to a Bark scale. The preselected frequencies are transferred to a Bark scale on the basis of the relationship represented by Formula II below: $$\mathrm{e.g.}=13\cdot \arctan \left(0.00076\cdot f\right)+3.5\cdot \arctan {\left(\frac{f}{7500}\right)}^2$$

Z: Bark; f: frequency value to be transferred to the Bark scale.

Both the preselected frequencies and the limit frequency value described by the first selection criterion can be transferred to the Bark scale using the relationship represented by Formula II.

The limit frequency value can fundamentally correspond to a bark or a bark adjusted by an adaptation factor or multiplied by an adaptation factor. The adjustment factor is typically between 0.7 and 1.1, in particular 0.9. The limit frequency value thus typically corresponds to 0.7 to 1.1, in particular 0.9, Bark. In other words, the frequency difference of the respective frequencies should correspond to a bark or almost a bark in order to meet the first selection criterion. A certain variability of the limit frequency value is given by the adjustment factor.

In a fifth step of the process, a second selection criterion is defined. On the basis of the second selection criterion, (based on the first selection criterion) preselected frequencies of two immediately successive local amplitude maxima are selected which satisfy the second selection criterion. In the fifth step, preselected frequencies are considered with regard to the second selection criterion. In the fifth step, preselected frequencies are then examined to determine whether (additionally) they meet the second selection criterion.

The second selection criterion can describe a limit energy value (range). Respective preselected frequencies meet the second selection criterion if the energy content between them falls below the limit energy value (range) described by the second selection criterion (“threshold”).

The limit energy value (range) can be defined by a specified limit energy content. Respective preselected frequencies meet the second selection criterion if they fall below the limit energy content described by the second selection criterion, cf. the relationship represented by the formula III shown below: $${\displaystyle \underset{f1}{\overset{f2}{\int }}{\left|\mathrm{<figure-callout\; id="2"\; label="S\; f"\; filenames="imgf0001.png,imgf0002.png"\; state="\{\{state\}\}">S\; </figure-callout>}\left(f\right)\left(\right)\right|}^2\mathit{df}<T}$$

The following applies: S (f): the area described by the frequencies or frequency values f ₁ , f _{2 of} the two immediately successive amplitude maxima (energy content between the frequencies or frequency values f ₁ , f _{2 of} the two immediately successive amplitude maxima); T: Limit energy content.

The limit energy value (range) can alternatively also be determined by using a first energy curve starting from the preselected frequency ("lower frequency") which belongs to the lower (lower frequency) amplitude maximum and a frequency ("upper frequency") which immediately follows the upper (frequency higher) maximum amplitude is associated, outgoing second energy curve is generated and the two energy curves are transferred to the frequency spectrum. The limit energy value is then defined by the respective energy profiles. The first energy curve runs from the frequency of the (frequency-wise) lower amplitude maximum of the two immediately following amplitude maxima in the direction of the frequency of the (frequency-wise) upper (higher) amplitude maximum of the two immediately successive amplitude maxima. The second energy curve starts from the frequency of the (in terms of frequency) upper amplitude maximum of the two immediately following amplitude maxima in the direction of the frequency of (in terms of frequency) lower (lower) amplitude maximum of the two immediately following amplitude maxima. The energy profiles generated can be transferred into the frequency spectrum in terms of data. A closed area or area is defined by the actual frequency curve between the frequencies and the energy curves. The range is defined in terms of frequency by the frequencies of the two immediately adjacent amplitude maxima and in terms of energy by the actual frequency profile between the amplitude maxima and the energy profiles running between them. The range typically only contains energy values ≥ zero. If one looks at the area geometrically in relation to the frequency spectrum, the area corresponds to the area geometrically defined by the two immediately adjacent amplitude maxima, the energy or frequency profiles running between these and the frequency axis (x-axis).

The energy profiles are typically generated on the basis of a psychoacoustic model. A psychoacoustic model is therefore typically used to generate the energy profiles, or the energy profiles are derived from a psychoacoustic model. The psychoacoustic model generally describes those frequency components of a certain sound which are heard by the human ear in a certain noise environment, i.e. H. if necessary in the presence of other noises. A preferred used psychoacoustic model is the model of spectral masking or masking, by which it is described that the human hearing ability cannot perceive certain frequency components of a certain noise or only with reduced sensitivity. These masking or masking effects are essentially based on the anatomical or mechanical conditions of the human inner ear and, for example, mean that low-energy or quiet tones in the middle frequency range are not perceivable with simultaneous reproduction of high-energy or loud tones in the low frequency range; the tones in the low frequency range mask the tones in the medium frequency range.

The energy profiles are derived in particular from the hearing thresholds of human hearing given by the respective psychoacoustic model at the respective preselected frequencies. This means that the psychoacoustic model is applied to the frequencies of the two immediately following amplitude maxima.

The first energy curve corresponds to the part of the hearing threshold derived from the psychoacoustic model for the frequency of the lower amplitude maximum, which extends in the direction of increasing frequencies. The second energy curve corresponds to the part of the hearing threshold derived from the psychoacoustic model for the frequency of the upper amplitude maximum, which extends in the direction of falling frequencies.

It is essential for the method that frequency ranges between the respective frequencies of two immediately following amplitude maxima, which frequencies satisfy both the first and the second selection criterion, are processed. The steps of the method described so far relate to the determination of frequency ranges to be processed within the audio signal to be processed.

In a sixth step of the method, an audio fill signal is generated or generated. The audio fill signal is typically generated in a targeted manner with regard to the previously determined frequency ranges to be processed within the audio signal to be processed. The audio fill signal is therefore typically generated in a targeted manner with regard to the frequency range defined by immediately successive frequencies that satisfy both the first and the second selection criterion in order to fill it and to fill the "energy valley" between the frequencies, at least in sections, in particular completely. The generated audio fill signal therefore expediently has a frequency range lying between the frequencies of respective immediately successive amplitude maxima. The generation of the audio fill signal takes place e.g. B. by means of a suitable signal generator.

In a seventh step of the method, the actual processing of the audio signal is carried out by introducing the audio fill signal into respective frequency ranges between respective frequencies satisfying the first and second selection criteria, so that a respective frequency range is filled with the audio fill signal at least in sections, in particular completely.

In other words, according to the method, corresponding "energy troughs" resulting from the data compression of the audio signal are determined and specifically filled with a certain data content in the form of the audio fill signal generated with regard to the determined "energy troughs", as a result of which the audio signal is processed. It follows from this that the processing of the audio signal according to the method, as mentioned above, in particular by an at least partial replacement of missing, ie. H. z. B. in the context of data compression discarded, frequency components of the audio signal is realized.

The steps of the method described result in an improved method for processing a lossy compressed audio signal, in particular with regard to the efficiency of the processing and the quality of the processed audio signal.

Of course, in an optional eighth step of the method it is possible to process the correspondingly prepared audio signal via at least one z. B. trained as a speaker device or at least one such comprehensive signal output device. An optional eighth step of the method can then provide for the output of a processed audio signal via at least one signal output device. Alternatively or in addition, in the eighth step of the method it is possible to store the correspondingly processed audio signal in a memory device, ie. H. z. B. a hard disk space to store (between). A correspondingly prepared, stored audio signal can be output at a later point in time via at least one corresponding signal output device and / or transmitted to at least one communication partner via a suitable, in particular wireless, communication network. An optional eighth step of the method can therefore (also) provide for storing a processed audio signal in at least one storage device and / or transmitting a processed audio signal to at least one communication partner. The processed audio signal can be subjected to an inverse Fourier transformation before it is output and / or stored and / or transmitted.

It is possible that, before processing the audio signal, by introducing the audio fill signal into the frequency range between the frequencies which satisfy the second selection criterion, a frequency which is based on the selected frequency (“lower frequency”) and which is associated with the lower (lower in frequency) amplitude maximum, if necessary third, energy curve and a fourth energy curve, possibly a fourth, are generated from the selected frequency ("upper frequency"), which is associated with the (higher frequency) amplitude maximum, and these two energy curves are transmitted into the frequency spectrum. The third energy profile, if any, proceeds from the frequency of the (in terms of frequency) lower amplitude maximum of the two immediately following amplitude maxima in the direction of the frequency of (in terms of frequency) the upper amplitude maximum of the two immediately following amplitude maxima. The fourth energy curve, if any, starts from the frequency of the (frequency-wise) upper (higher) amplitude maximum of the two immediately successive amplitude maxima in the direction of the frequency of the (frequency-wise) lower (lower) amplitude maximum of the two immediately successive amplitude maxima. The energy profiles generated can in turn be transmitted in terms of data to the frequency spectrum. A closed area or area is also defined by the frequencies and the energy profiles. In terms of frequency, the range is in turn defined by the frequencies of the two immediately following amplitude maxima and in terms of energy by the energy profiles running between them. The range typically only contains energy values ≥ zero. If the area is viewed geometrically with respect to the frequency spectrum, the area in turn corresponds to the area defined geometrically by the two immediately adjacent amplitude maxima, the energy or frequency courses running between these and the frequency axis (x-axis) Area.

The generation of the third and fourth energy profiles, if appropriate, is typically also carried out on the basis of a psychoacoustic model. A psychoacoustic model is therefore typically also used to generate the energy profiles, or the energy profiles are derived from a psychoacoustic model. The explanations in connection with the first two energy profiles apply analogously.

The third and fourth energy profiles, if any, are likewise derived in particular from the hearing thresholds of human hearing given by the respective psychoacoustic model at the respective preselected frequencies. This means that the psychoacoustic model is applied to the frequencies of the two immediately following amplitude maxima. The possibly third energy profile corresponds to the part of the hearing threshold derived from the psychoacoustic model for the frequency of the lower amplitude maximum, which extends in the direction of increasing frequencies. The possibly fourth energy curve corresponds to the part of the hearing threshold derived from the psychoacoustic model for the frequency of the upper amplitude maximum, which extends in the direction of falling frequencies.

If, as explained above, corresponding energy profiles should also be generated in connection with the limit energy value described by the second selection criterion and transferred to the frequency spectrum, these (first two) energy profiles can differ from the (third and fourth) energy profiles mentioned in the previous paragraph differentiate.

The audio fill signal is subsequently introduced at least in sections, in particular completely, into the region of the frequency spectrum defined by the two preselected frequencies and the respective energy profiles. The audio signal is processed here by introducing the audio fill signal into the frequency range of the frequency spectrum defined by the frequencies of the two immediately adjacent amplitude maxima and the respective energy profiles, so that the range of the frequency spectrum defined by the frequencies of the two immediately adjacent amplitude maxima and the respective energy profiles Frequency spectrum at least in sections, in particular completely, is or will be filled with the audio fill signal.

In all cases it applies that the audio fill signal can be generated as a function of or independently of acoustic parameters of the audio signal to be processed, in particular with regard to the respective energy and frequency components of the audio signal. However, the audio fill signal is expediently generated independently of acoustic parameters of the audio signal, that is to say solely with regard to the at least partial filling of the range of the frequency spectrum defined by the frequencies of the two immediately adjacent amplitude maxima, since the computing effort for generating the audio fill signal may be such can be significantly reduced.

If the audio fill signal is generated as a function of acoustic parameters of the audio signal, the filling or filling of the range of the frequency spectrum defined by the frequencies of the two immediately consecutive amplitude maxima can be carried out depending on certain acoustic parameters of the audio signal, in particular the amplitude and / or frequency curve. or certain acoustic parameters of a further audio signal to be processed, in particular of the amplitude and / or frequency curve. In this way, a more natural perception of the prepared audio signal for the human ear can be realized.

Basically, a Bark scale can be used as the frequency spectrum in which the audio signal is transmitted according to the method. The 24 individual barks or bands of the Bark scale are known to correspond to the 24 individual frequency groups of human hearing, i.e. H. those frequency ranges that are jointly evaluated by human hearing. The individual barks or bands of the Bark scale contain different frequencies or frequency ranges or bandwidths. Possible frequency bands of the frequency spectrum can correspond to the 24 barks or bands of the Bark scale.

• Übertragung eines Audiosignals in ein Frequenzspektrum, in welchem Energien des Audiosignals mit Frequenzen des Audiosignals korreliert werden,
• Ermittlung von Frequenzen lokaler Amplitudenmaxima in dem Frequenzspektrum,
• Festlegung eines ersten Auswahlkriteriums und Vorauswahl der Frequenzen von zwei unmittelbar aufeinander folgenden lokalen Amplitudenmaxima, welche Frequenzen dem ersten Auswahlkriterium genügen,
• Festlegung eines zweiten Auswahlkriteriums und Auswahl von vorausgewählten dem ersten Auswahlkriterium genügenden Frequenzen von zwei unmittelbar aufeinander folgenden Amplitudenmaxima, welche zusätzlich dem zweiten Auswahlkriterium genügen,
• Erzeugung eines Audiofüllsignals und
• Aufbereitung des Audiosignals durch Einbringen des Audiofüllsignals in einen Bereich zwischen den dem zweiten Auswahlkriterium genügenden Frequenzen, sodass der Bereich zumindest abschnittsweise, insbesondere vollständig, mit dem Audiofüllsignal befüllt ist, eingerichtet ist.

In addition to the described method, the invention further relates to a device for processing a lossy compressed audio signal according to the method described above. The device comprises at least one hardware and / or software implemented control device, which is characterized in that it is used for

• Transfer of an audio signal into a frequency spectrum in which energies of the audio signal are correlated with frequencies of the audio signal,
• Determination of frequencies of local amplitude maxima in the frequency spectrum,
• Definition of a first selection criterion and preselection of the frequencies of two immediately following local amplitude maxima, which frequencies meet the first selection criterion,
• Definition of a second selection criterion and selection of preselected frequencies of two immediately successive amplitude maxima which satisfy the first selection criterion and which additionally satisfy the second selection criterion,
• Generation of an audio fill signal and
• Processing of the audio signal by introducing the audio fill signal into a range between the frequencies satisfying the second selection criterion, so that the range is at least partially, in particular completely, filled with the audio fill signal.

Of course, individual, several or all of the steps carried out according to the method can also be implemented in separate hardware and / or software Devices of the control device are made. In this case, the device comprises a control device equipped or communicating with appropriate devices. As follows in the following, the device can be part of an audio device or an audio system for a motor vehicle.

The invention further relates to an audio device or an audio system for a motor vehicle. The audio device can be part of a motor vehicle-side multimedia device for outputting multimedia content, in particular audio and / or video content, to occupants of a motor vehicle. The audio device comprises at least one signal output device, i. H. z. B. a loudspeaker device, which is set up for acoustic output of prepared audio signals in an at least part of a passenger compartment of an interior of a motor vehicle. The audio device is characterized in that it has at least one device for processing lossy compressed audio signals as described for the preparation of lossy compressed audio signals.

All statements in connection with the described method apply analogously both to the device for processing a lossy compressed audio signal and to the audio device.

Fig. 1

eine Prinzipdarstellung einer Vorrichtung zur Durchführung eines Verfahrens gemäß einem Ausführungsbeispiel;

Fig. 2

ein Blockdiagramm eines Verfahrens gemäß einem Ausführungsbeispiel;

Fig. 3, 4

jeweils eine Prinzipdarstellung eines psychoakustischen Modells gemäß einem Ausführungsbeispiel; und

Fig. 5 - 8

jeweils eine Prinzipdarstellung eines Frequenzspektrums, in welchem Energien eines Audiosignals mit Frequenzen des Audiosignals korreliert werden, gemäß einem Ausführungsbeispiel.

Exemplary embodiments of the invention are explained in more detail below with reference to the drawing figures. Show:

Fig. 1

a schematic diagram of an apparatus for performing a method according to an embodiment;

Fig. 2

a block diagram of a method according to an embodiment;

3, 4

a schematic representation of a psychoacoustic model according to an embodiment; and

5 - 8

a schematic representation of a frequency spectrum, in which energies of an audio signal are correlated with frequencies of the audio signal, according to an embodiment.

Fig. 1 shows a schematic diagram of a device 1 for processing a lossy compressed audio signal 2. The audio signal 2 may, for. B. is a lossy compressed audio file. Specifically, it can be e.g. B. can be a lossy compressed mp3-encoded audio file ("mp3 file") by means of an mp3 algorithm. The audio file can already be at least partially decoded. The audio file can e.g. B. include a piece of music.

The device 1 shown in the exemplary embodiment forms part of an audio device 3 or an audio system of a motor vehicle 4. The audio device 3 can be part of a motor vehicle-side multimedia device (not shown) for outputting multimedia content, in particular audio and / or video content, to occupants of the motor vehicle 4. The audio device 3 comprises at least one z. B. designed as a loudspeaker device or at least one such comprehensive signal output device 5, which is set up for acoustic output of prepared audio signals 6 in an at least part of the passenger compartment interior 7 of the motor vehicle 4.

The device 1 comprises a central hardware and / or software implemented control device 8, which is set up in the following with reference to Fig. 2 to implement the described method for processing lossy compressed audio signals 2.

Individual, multiple, or all of those discussed below with reference to Fig. 2 Steps S1-S7 (S8) carried out according to the method can be carried out in separate hardware and / or software implemented devices (not shown) of the control device 8. In this case, the device 1 comprises a control device 8 equipped with corresponding devices.

Fig. 2 shows a block diagram of an embodiment of a method for processing lossy compressed audio signals 2. The method can be carried out with the device 1 described above.

In the first step S1 of the method, the lossy compressed audio signal 2 to be processed is provided. The provision of the audio signal 2 can in principle be via any physical or non-physical audio source, i. H. z. B. from the audio device 3. Specifically, the audio signal 2 z. B. from a data memory (not shown) of the audio device 3.

In the second step S2 of the method, the audio signal 2 is transmitted into a frequency spectrum. In the frequency spectrum, energies of the audio signal 2 are correlated with frequencies of the audio signal 2. For this purpose, the content of the audio signal 2 is examined for its energy, that is to say amplitude and frequency components, and the individual energy components of the audio signal 2 are transmitted in terms of data into a frequency-dependent representation by means of suitable algorithms, for example by means of (faster) Fourier transformation algorithms. A corresponding frequency spectrum is included in Fig. 5 shown in a schematic diagram ..

In the third step S3 of the method, frequencies f _{i of} local amplitude maxima are determined in the frequency spectrum; the frequency spectrum is therefore examined for local amplitude maxima and the frequencies f _i associated with the respective amplitude maxima are determined. Under one in the 5 - 8 by a dot graphically highlighted local amplitude maximum is to be understood as an amplitude maximum value in a defined frequency environment range.

In the fourth step S4 of the method, a first selection criterion is defined. On the basis of the first selection criterion, the frequencies f _{i of} two immediately following (local) amplitude maxima are preselected, which frequencies meet the first selection criterion. In the fourth step S4, the frequencies f _i of pairs of immediately consecutive amplitude maxima are examined with regard to the first selection criterion to determine whether the frequencies f _i meet the first selection criterion. In the further steps S5-S7 of the method, only the frequencies f _i which meet the first selection criterion are considered. The fourth step S4 then preselects the frequencies f _i to be considered below.

The first selection criterion describes a certain limit frequency value Δf _T. Frequencies f _{i of} immediately successive amplitude maxima satisfy the first selection criterion if their frequency difference Δf _i exceeds the limit frequency value Δf _T described by the first selection criterion, cf. the relationship represented by the following formula: $$\mathrm{\Delta }{f}_i>\left|\mathrm{\Delta }{f}_T\right|$$

In this case, Δf _i applies: frequency difference between two immediately following amplitude maxima; Δf _T : cutoff frequency value.

The limit frequency value Δf _T is determined by transferring the preselected frequencies f _i to a Bark scale. The preselected frequencies f _{i are transferred} to a Bark scale on the basis of the relationship represented by the following formula: $$\mathrm{e.g.}=13\cdot \arctan \left(0.00076\cdot f\right)+3.5\cdot \arctan {\left(\frac{f}{7500}\right)}^2$$

Z: Bark; f: frequency value to be transferred to the Bark scale.

Both the preselected frequencies f _i and the limit frequency values Δf _T described by the first selection criterion can be transferred to the Bark scale using the relationship represented by the above formula.

The limit frequency value Δf _T can correspond to a bark or to a bark adjusted by an adaptation factor or multiplied by an adaptation factor. The adjustment factor is typically between 0.7 and 1.1, in particular 0.9. The limit frequency value thus typically corresponds to 0.7 to 1.1, in particular 0.9, Bark.

A second selection criterion is defined in the fifth step S5 of the method. On the basis of the second selection criterion, (based on the first selection criterion) preselected frequencies f _{i are} selected which (additionally) meet the second selection criterion. In the fifth step S5, preselected frequencies f _{i are} then examined to determine whether they (additionally) meet the second selection criterion. The frequencies f _i (additionally) satisfying the second selection criterion can in turn be transferred to a Bark scale.

The second selection criterion can describe a limit energy value. Respective preselected frequencies f _i satisfy the second selection criterion if the amount of energy between them falls below the limit energy value described by the second selection criterion.

The limit energy value can be defined by a defined limit energy content T. Respective preselected frequencies f _i meet the second selection criterion if they fall below the limit energy content T described by the second selection criterion, cf. the relationship represented by the following formula: $${\displaystyle \underset{f1}{\overset{f2}{\int }}{\left|\mathrm{<figure-callout\; id="2"\; label="S\; f"\; filenames="imgf0001.png,imgf0002.png"\; state="\{\{state\}\}">S\; </figure-callout>}\left(f\right)\left(\right)\right|}^2\mathit{df}<T}$$

The following applies: S (f): the area described by the frequencies f ₁ , f _{2 of} the two immediately consecutive amplitudes (energy content between the frequencies f ₁ , f _{2 of} the two immediately consecutive amplitude maxima); T: Limit energy content.

In this context to the in Fig. 6 Schematic representation of a frequency spectrum containing two preselected frequencies f ₁ , f ₂ , which is also a section of another, namely the one in Fig. 5 shown frequency spectrum is to refer. Out Fig. 6 the area described by the frequencies f ₁ , f _{2 of} the two immediately successive amplitude maxima (hatched) and the limit energy content T represented by a horizontal line are illustrated. The hatched area corresponds to the integral represented by the above formula.

The limit energy value can alternatively also be determined by a first energy curve EV1 starting from the preselected frequency f ₁ ("lower frequency"), which belongs to the lower (lower in frequency) amplitude maximum, and a first energy curve EV1 from the preselected frequency f ₂ ("upper frequency ), which is associated with the upper (higher frequency) amplitude maximum, the outgoing second energy curve EV2 is generated and the two energy curves EV1, EV2 are transmitted into the frequency spectrum Energy profiles EV1, EV2 defined.

Based on Fig. 7 it can be seen that the generated energy profiles EV1, EV2 can be transmitted in terms of data into the frequency spectrum. The first energy curve EV1 proceeds from the lower frequency f ₁ in the direction of the upper frequency f ₂ . The second energy curve EV2 proceeds from the upper frequency f ₂ in the direction of the lower frequency f ₁ .

A closed area or area is defined by the actual frequency curve between the frequencies f _{1, 2} and the energy curves EV1, EV2. The range is defined in terms of frequency share by the two frequencies f _{1, 2} and in terms of energy share by the actual frequency curve and the energy curves EV1, EV2 running between them. The range typically only contains energy values ≥ zero. If one looks at the area geometrically with respect to the frequency spectrum, the area corresponds to the geometrically defined amplitude or maxima defined by the frequencies f _{1, 2 of} the two immediately adjacent amplitude maxima, and the frequency axis (x-axis) between them Fig. 7 hatched area.

The energy profiles EV1, EV2 are generated on the basis of a psychoacoustic model. A preferred used psychoacoustic model is the spectral masking or masking model. Based on Fig. 3 it can be seen that the energy profiles EV1, EV2 are derived from the hearing thresholds of the human ear given the respective preselected frequencies f _{1, 2} by the respective psychoacoustic model. This means that the psychoacoustic model used is applied to the two frequencies f _{1, 2} . The first energy curve EV1 corresponds to the part of the hearing threshold derived from the psychoacoustic model for the lower frequency f ₁ , which extends in the direction of increasing frequencies (see left curly bracket in Fig. 3 ). The second energy curve EV2 corresponds to the part of the hearing threshold derived from the psychoacoustic model for the upper frequency f ₂ , which extends in the direction of falling frequencies (cf. right-hand curly bracket in Fig. 3 ). Of course, it is contrary to the representation in Fig. 3 it is also possible that the energy profiles EV1, EV2 intersect or intersect in a value range above the x-axis.

It is essential for the method that frequency ranges between the respective frequencies f _i or f _{1, 2 of} the two immediately following amplitude maxima, which frequencies satisfy both the first and the second selection criterion, are processed. The steps S1-S5 of the method described so far relate to the determination of frequency ranges to be processed according to the method within the audio signal 2 to be processed.

In a sixth step S6 of the method, a suitable signal generator is used generates an audio fill signal AFS. The audio fill signal AFS is generated in a targeted manner with regard to the previously determined frequency ranges to be processed within the audio signal 2 to be processed. The audio fill signal AFS is thus generated in a targeted manner with regard to the frequency range defined by the frequencies f _i and f _{1, 2 of} the two immediately successive amplitude maxima, which satisfy both the first and the second selection criterion, in order to fill this and that between the frequencies f _i fill given "energy valley". The generated audio fill signal AFS therefore has a frequency range lying between the frequencies f _{i of the} respective directly successive amplitude maxima.

The audio fill signal AFS can be generated as a function of or independently of acoustic parameters of the audio signal 2, in particular with regard to the respective energy and frequency components of the audio signal 2. In the exemplary embodiment described, the audio filling signal AFS becomes independent of acoustic parameters of the audio signal 2, ie solely with regard to the filling of the frequency component with frequencies f _{1, 2} and energy component with the actual frequency profile and the range defined between these energy profiles EV3, EV4. generated.

In a seventh step S7 of the method, the actual processing of the audio signal 2 takes place by introducing the audio fill signal AFS into respective frequency ranges between respective frequencies f _i satisfying the first and second selection criteria, so that a respective frequency range is filled with the audio fill signal AFS.

Before the audio signal 2 is processed by introducing the audio fill signal AFS, a further or third energy curve EV3 starting from the selected lower (lower) frequency f ₁ , which is associated with the lower (lower in frequency) amplitude, and one from the selected upper (higher ) Frequency f ₂ , which is associated with the upper (higher frequency) amplitude maximum, generates outgoing further or fourth energy curve EV4.

Based on Fig. 8 it can be seen that the generated energy profiles EV3, EV4 - analogously to the energy profiles EV1, EV2 - are transmitted to the frequency spectrum in terms of data. The third energy curve EV3 proceeds from the lower frequency f ₁ in the direction of the upper frequency f ₂ . The fourth energy curve EV4 proceeds from the upper frequency f ₂ in the direction of the lower frequency f ₁ .

A closed area or area is defined by the actual frequency curve between the frequencies f _{1, 2} and the energy curves EV3, EV4. The range is defined in terms of frequency share by the frequencies f _{1, 2 of} the amplitude maxima and in terms of energy share by the actual frequency curve and the energy curves EV3, EV4 running between them. The range typically only contains energy values ≥ zero. Looking at the area geometrically with respect to the frequency spectrum, corresponds to the range of the geometrically defined energy or frequency profiles and the frequency axis (x-axis) defined by the frequencies f _{1, 2 of} the two immediately adjacent amplitude maxima, in Fig. 8 hatched area.

The energy profiles EV3, EV4 are also generated on the basis of a psychoacoustic model. A preferred used psychoacoustic model is the model of spectral masking or masking (cf. Fig. 4 ). Based on Fig. 4 it can be seen that the energy profiles EV3, EV4 are derived from the hearing thresholds of the human ear given the respective preselected frequencies f _{1, 2} by the respective psychoacoustic model. Here too, this means that the psychoacoustic model used is applied to the two immediately successive frequencies f _{1, 2} . The third energy curve EV3 corresponds to the part of the hearing threshold derived from the psychoacoustic model for the lower frequency f ₁ , which extends in the direction of increasing frequencies (see left brace in Fig. 4 ). The fourth energy curve EV4 corresponds to the part of the hearing threshold derived from the psychoacoustic model for the upper frequency f ₂ , which extends in the direction of falling frequencies (see right curly bracket in Fig. 4 ). Of course, it is contrary to the representation in Fig. 4 it is also possible here that the energy profiles EV3, EV4 intersect or intersect in a range of values above the x-axis.

In general, the (first two) energy profiles EV1, EV2 can differ from the third and fourth energy profiles Ev3, EV4.

Overall, according to the method, "energy valleys" resulting from the data compression of the audio signal 2 are determined and specifically filled with a certain data content in the form of the audio fill signal AFS generated with regard to the determined "energy valleys", whereby the audio signal 2 is processed. It follows from this that the processing of the audio signal 2 according to the method by an at least partial replacement of missing, ie. H. z. B. in the context of data compression discarded, frequency components of the audio signal 2 is realized.

An optional eighth step S8 of the method can output a processed audio signal 2 via at least one signal output device 5 and / or store a processed audio signal 2 in at least one storage device (not shown) and / or transmit a processed audio signal 2 to at least one communication partner ( not shown). The processed audio signal 2 can be subjected to an inverse Fourier transformation before it is output and / or stored and / or transmitted.

The described steps S1-S7 (S8) of the method result in an improved method for processing a lossy compressed audio signal 2, in particular with regard to the efficiency of the processing and the quality of the processed audio signal 6.

REFERENCE SIGN LIST

1

contraption

2

Audio signal (compressed)

3

Audio device

4

Motor vehicle

5

Signal output device

6

Audio signal (processed)

7

inner space

8th

Control device

AFS

Audio fill signal

EV1 - EV4

Energy flow

f _i

frequency

Δf _T

Cutoff frequency value

T

Limit energy content

S1 - S8

Procedural step

Claims (15)

1. Method for conditioning an audio signal (2) subjected to lossy compression, characterised by the following steps:

   - provision of an audio signal (2) subjected to lossy compression, which is an already decoded audio file subjected to lossy compression,

   - transfer of the audio signal (2) to a frequency spectrum in which energies of the audio signal (2) are correlated with frequencies of the audio signal (2),

   - determination of frequencies (f_i) of local amplitude maxima in the frequency spectrum,

   - definition of a first selection criterion and preselection of the frequencies (f_i) of two local amplitude maxima directly following one another, which frequencies satisfy the first selection criterion,

   - definition of a second selection criterion and selection of preselected frequencies (f_i) satisfying the first selection criterion of two local amplitude maxima directly following one another, which frequencies additionally satisfy the second selection criterion,

   - generation of an audio filler signal (AFS) and

   - conditioning of the audio signal (2) by introduction of the audio filler signal (AFS) into a frequency range between the frequencies (f_i) satisfying the second selection criterion, so that the frequency range is filled at least in sections, in particular completely, with the audio filler signal (AFS).
2. Method according to claim 1, characterised in that the frequencies (f_i) satisfy the first selection criterion if their frequency difference exceeds in amount a limit frequency value (Δf_i) described by the first selection criterion.
3. Method according to claim 2, characterised in that the limit frequency value (Δf_i) is defined by transferring the frequencies (f_i) to a Bark scale, wherein the limit frequency value (Δf_i) corresponds to a Bark or a Bark adjusted by way of an adjustment factor.
4. Method according to claim 3, characterised in that the adjustment factor used corresponds to a value between 0.7 and 1.1 Bark, in particular 0.9 Bark.
5. Method according to any of the preceding claims, characterised in that the frequencies (f_i) satisfy the second selection criterion if the energy content between the frequencies (f_i) falls below a limit energy value in amount.
6. Method according to claim 5, characterised in that the limit energy value is defined by a fixed limit energy content (T).
7. Method according to claim 5, characterised in that the limit energy value is defined in that a first energy curve (EV1) starting out from the selected lower frequency (f₁) and a second energy curve (EV2) starting out from the selected upper frequency (f₂) is generated and the two energy curves (EV1, EV2) are transferred to the frequency spectrum, wherein the limit energy value is defined by the respective energy curves (EV1, EV2).
8. Method according to claim 7, characterised in that the generation of the first and second energy curves (EV1, EV2) takes place on the basis of a psychoacoustic model.
9. Method according to any one of the preceding claims, characterised in that before the conditioning of the audio signal (2) by introducing the audio filler signal (AFS) into the frequency range between the frequencies (f_i) satisfying the second selection criterion, so that the frequency range is filled at least in sections, in particular completely, by the audio filler signal (AFS),
   an, if applicable third, energy curve (EV3) starting out from the selected lower frequency (f₁) and an, if applicable fourth, energy curve (EV4) starting out from the selected upper frequency (f₂) is generated and the two energy curves (EV3, EV4) are transferred to the frequency spectrum.
10. Method according to claim 9, characterised in that the audio filler signal (AFS) is introduced at least in sections, in particular completely, into a range of the frequency spectrum defined by the two selected frequencies (f₁, f₂) and the respective energy curves (EV3, EV4).
11. Method according to claim 9 or 10, characterised in that the generation of the energy curves (EV3, EV4) takes place on the basis of a psychoacoustic model.
12. Method according to any one of the preceding claims, characterised in that the audio filler signal (AFS) is generated depending on or independently of acoustic parameters of the audio signal (2).
13. Method according to claim 12, characterised in that the audio filler signal (AFS) is generated depending on acoustic parameters of the audio signal (2), wherein the filling of the range (A) takes place as a function of certain acoustic parameters of the audio signal (2) or of another audio signal (2) to be conditioned.
14. Apparatus (1) for conditioning an audio signal (2) subjected to lossy compression according to a method according to any one of the preceding claims, characterised by at least one control device (8), which is adapted to

    - provide an audio signal (2) subjected to lossy compression,

    - transfer the audio signal (2) to a frequency spectrum in which energies of the audio signal (2) are correlated with frequencies of the audio signal (2),

    - determine frequencies (f_i) of local amplitude maxima in the frequency spectrum,

    - define a first selection criterion and preselect the frequencies (f_i) of two local amplitude maxima directly following one another, which frequencies satisfy the first selection criterion,

    - define a second selection criterion and select preselected frequencies (f_i) satisfying the first selection criterion of two local amplitude maxima directly following one another, which frequencies additionally satisfy the second selection criterion,

    - generate an audio filler signal (AFS) and

    - condition the audio signal (2) by introduction of the audio filler signal (AFS) into a range between the frequencies (f_i) satisfying the second selection criterion, so that the range is filled at least in sections, in particular completely, with the audio filler signal (AFS).
15. Audio device (3) for a motor vehicle (4), comprising at least one signal output device (5), which is adapted for the acoustic output of conditioned audio signals (6) into an interior (7) of a motor vehicle (4) forming at least part of a passenger compartment, characterised in that for conditioning audio signals (2) subjected to lossy compression the audio device has at least one apparatus (1) according to claim 14.

Images