RU2523035C2 - Audio encoder and bandwidth extension decoder - Google Patents

Abstract

FIELD: physics, acoustics.

SUBSTANCE: invention relates to means of generating an equalised multichannel audio signal. An audio encoder for obtaining an output signal using an input audio signal comprises a patch generator, a comparator and an output interface. The patch generator generates at least one bandwidth extension signal, having a high-frequency band. The high-frequency band of the bandwidth extension signal is based on a low frequency band of the input audio signal. A comparator calculates a plurality of comparison parameters. A comparison parameter is calculated based on a comparison of the input audio signal and a generated bandwidth extension signal. Each comparison parameter of the plurality of comparison parameters is calculated based on a different offset frequency between the input audio signal and a generated high bandwidth signal. Further, the comparator determines a comparison parameter from the plurality of comparison parameters, wherein the determined comparison parameter satisfies a predefined criterion.

EFFECT: improved signal encoding quality at high bit rate.

17 cl, 22 dwg

Description

Embodiments of the invention relate to processing an audio signal, namely, an audio encoder, a method for producing an output signal, a decoder that provides an extension of the frequency range, and a method for producing an audio signal with an increased frequency range.

The encoding of audio signals taking into account their subsequent perception in order to reduce the amount of data for efficient storage and transmission of information has found application in many fields. Known encoding algorithms are, for example, MPEG 1/2 LAYER 3 "MP3" or MPEG 4 AAC. The use of these coding algorithms, especially at the lowest bitrates, leads to a decrease in sound quality. This is often because the encoder has a limited range of transmitted frequencies. The signal passed through the low-pass filter is encoded using the so-called central encoder, and the regions with higher frequencies are presented as parameters and then restored from the region passed through the low-pass filter.

According to WO 9857436, the audio signal undergoes a frequency range reduction procedure on the encoder side, and only the low frequency region of the audio signal is encoded using a high quality encoder. The high-frequency region is characterized very approximately using a set of parameters that allow you to restore the original spectrum of the high-frequency region. Then, on the decoder side, the high frequency band is synthesized. For this purpose, harmonic transposition is proposed when the low frequency band of the decoded audio signal is sent to the filter bank. The channels of the low-pass filter bank are connected to the channels of the high-pass filter bank, or “matched,” and each such signal passed through the pass-band filter is corrected. A synthesizing filter bank belonging to a special analyzing filter bank receives a filtered signal in the low frequency range and a corrected filtered signal in the low frequency range, which corresponds to the high frequency region. The signal at the output of the synthesizing filter bank is a signal expanded in accordance with the frequency range of the audio signal transmitted from the encoder to the decoder at a low data rate, because calculating the filter bank and the process of determining the correspondence between the high and low frequency ranges presents significant difficulties.

Less computationally expensive methods of expanding the frequency range of an audio signal with a limited frequency range use copying the low-frequency areas of the signal (LF) to the high-frequency areas (HF) in order to approximately fill in the missing information that is missing due to the range limitation. Similar methods are described in the following works: M. Dietz, L. Liljeryd, K. Kjörling and 0. Kunz, "Spectral Band Replication, a novel approach in audio coding," in 112th AES Convention, Munich, May 2002; S. Meltzer, R. Böhm and F. Henn, "SBR enhanced audio codecs for digital broadcasting such as" Digital Radio Mondiale "(DRM)," 112th AES Convention, Munich, May 2002; T. Ziegler, A. Ehret, P. Ekstrand and M. Lutzky, "Enhancing mp3 with SBR: Features and Capabilities of the new mp3 PRO Algorithm," in 112th AES Convention, Munich, May 2002; International Standard ISO / IEC 14496-3: 2001 / FPDAM 1, "Bandwidth Extension," ISO / IEC, 2002, or "Speech bandwidth extension method and apparatus", Vasu Iyengar et al. US Patent Nr. 5,455,888.

These methods do not use harmonic transposition. According to these methods, the low-frequency bandpass signal is artificially copied to an adjacent region of the high-frequency bandpass signal. This results in an inaccurate representation of the high frequency range. The rough approximation of the received signal to the original signal is then corrected in the next step by determining control parameters based on the original signal. For example, the MPEG-4 standard uses scale factors to adjust the spectral envelope, a combination of inverse filtering and the addition of a noise threshold to adjust tonality, and the insertion of sinusoidal parts of the signal to add tonal components.

In addition, there is the so-called "blind extension of the frequency range" described in the works of E. Larsen, R.M. Aarts, and M. Danessis, "Efficient high-frequency bandwidth extension of music and speech". However, this method was not mentioned at the 112th AES conference in Munich, Germany, May 2002. There is also the so-called “artificial bandwidth extension” method, which is discussed in K. Käyhkö, A Robust Wideband Enhancement for Narrowband Speech Signal; Research Report, Helsinki University of Technology, Laboratory of Acoustics and Audio signal Processing, 2001.

J. Makinen et al .: AMR-WB +: a new audio coding standard for 3 ^rd generation mobile audio services Broadcasts, IEEE, ICASSP ′05 describes a method for expanding the frequency range in which the operation of copying low-frequency components to the high-frequency range is performed using specular reflection obtained by increasing the sampling frequency of the low-pass filtered signal.

Alternatively, one-way range modulation can be applied, which is generally equivalent to the copy operation in the filter area. Ways that apply harmonic range extension? typically use the pitch tracking step, the nonlinear distortion step (see "U. Kornagel, Spectral widening of the excitation signal for telephone-band speech enhancement, in: Proceedings of the IWAENC, Darmstadt, Germany, September 2001, pp. 215-218 ") or phase voice encoders are used, as described in the patent application F.Nagel, S. Disch:" Apparatus and method of harmonic bandwidth extension in audio signals "" with the application number US 61/025129.

WO 02/41302 A1, for example, describes a method for improving the efficiency of coding systems that employ high frequency reconstruction. It shows how to increase the overall efficiency of such systems by gradually adapting the transition frequency between the low-frequency band encoded by the main encoder and the high-frequency band encoded by a system that applies high-frequency band reconstruction. According to this method, the main encoder must work with different transition frequencies on both the encoder side and the decoder side. Thus, the structure of the main encoder is complicated.

Ways to extend the frequency range are also described in the following works: R. M. Aarts, E. Larsen, and O. Oweltjes, A unified approach to low- and high-frequency bandwidth extension. In AES 115th Convention, New York, USA, October 2003 ", E. Larsen and R. M. Arts: Audio Bandwidth Extension - Application to psychoacoustics. Signal Processing and Loudspeaker Design. John Wiley & Sons, Ltd, 2004", E. Larsen, R. M. Arts, and M. Danessis: Efficient high-frequency bandwidth extension of music and speech. In AES 112th Convention, Munich, Germany, May 2002 "," J. Mahoul: Spectral Analysis of Speech by Linear Prediction. IEEE Transactions on Audio and Electroacoustics, AU-21 (3), June 1973 "," United States Patent Application 08 / 951.029, Ohmori et al .: Audio band width extending system and method "and" United States Patent 6895375, Malah, D & Cox, R. VS .: System for bandwidth extension of Narrow-band speech.

Methods of harmoniously expanding the frequency range are usually very complex, and when simplifying the methods of expanding the frequency range, quality deteriorates. In a particular case, when a low bitrate is combined with a small low-frequency band, interference may occur such as sudden transitions and timbre distortions.

The reason for this is that a portion of the high frequencies is often obtained approximately based on the copy operation, which does not determine the harmonic relationships between the parts of the tone signal. This applies both to the harmonic relationship between the low and high frequency ranges, and to the harmonic relationship between subsequent patches within the high frequency range. For example, in SBR, the adjacent arrangement of encoded components and copied components, which occurs at the boundary of the high and low frequency bands, may cause the sound to be perceived as uneven. This is shown in Fig. 18, where the tonal portions copied from the low frequency range to the high frequency range are spectrally close to the tonal portions of the low frequency range.

Figa illustrates the original spectrogram 1800A signal, consisting of three tones. Fig. 18B shows a diagram 1800B of a signal with an increased range that corresponds to the original signal in Fig. 18A. The abscissa indicates time, the ordinate indicates frequency. In particular, with regard to the last tone, here you can observe the potential problems 1810 (blurry lines 1810).

If the known methods consider the harmonic relationship, then this always happens on the basis of the F ₀ estimate. In this case, the success of these methods largely depends on the reliability of this assessment.

In general, well-known methods of increasing the frequency range allow you to get audio signals at low bitrate, with high bitrate the quality of the audio signal may not always be high.

An object of the present invention is to improve an audio signal coding scheme.

This problem is solved using an audio encoder according to claim 1, a decoder that increases the frequency range, according to claim 3 and claim 8 and the method according to claim 12, 13 and 14.

In an embodiment of the present invention, an audio encoder is used that generates an output signal based on an input audio signal. The audio encoder includes a patch generator, a comparator, and an output interface.

The patch generator generates at least one signal with an increased high-frequency range. A signal with an increased high-frequency range includes a high-frequency band, which is based on the low frequencies of the input audio signal.

If you created different signals with an increased range of high frequencies, then different signals will include different frequencies.

The comparator calculates many comparative parameters. The comparative parameter is calculated based on a comparison of the input audio signal and the generated signal with an increased high-frequency range. Each parameter from a plurality of comparison parameters is calculated based on a separate offset frequency between the input audio signal and the generated signal with an increased high-frequency range. Next, the comparator selects the comparison parameter from the set of comparison parameters so that the selected parameter corresponds to a predetermined criterion.

In other words, the comparator can be configured in such a way that it will determine the comparison parameter among the set of comparison parameters that best suits the given criterion.

The output interface provides an output signal for transmission or storage. The output signal includes an indication of the signal based on a bias frequency corresponding to a specific comparison parameter.

In other words, the output signal includes the selected comparison parameter, which indicates the optimal bias frequency.

A further embodiment of the invention includes a decoder that increases the frequency range to obtain an audio signal with an increased frequency range based on the input audio signal and the parametric signal. The parametric signal contains an indication of the bias frequency and the power density parameter. A decoder that increases the frequency range includes a patch generator, a combiner, and an output interface.

The patch generator creates a signal with an increased high-frequency range, which includes the high-frequency band. The high frequency band of a given signal is generated based on one or more frequency band offsets of the input audio signal. Frequency offsets are based on the offset frequency.

The patch generator can expand or decrease the high-frequency band of a signal with an increased high-frequency range by using a coefficient equal to the value of the power density parameter or the reciprocal of the power density parameter, respectively.

The combining unit connects the signal with the high frequency band and the input audio signal in order to obtain a signal with an increased high frequency range.

The output interface provides an output signal with an increased range.

A further embodiment of the present invention has a bandwidth increasing decoder for receiving an audio signal with an increased frequency range based on an input audio signal. The bandwidth increasing decoder includes a patch generator, a comparator, a combining unit, and an output interface.

The patch generator creates at least one signal with a high-frequency band, including a high-frequency band, obtained on the basis of the input audio signal. In this case, the lower limit frequency of the high frequency band of the generated signal is lower than the upper limit frequency of the input audio signal. If different signals with a high-frequency band are generated, then the received signals with a high-frequency band have different frequencies within their high-frequency bands.

The comparator calculates many comparison parameters. The comparison parameter is calculated based on a comparison of the input audio signal and the generated signal with the high frequency band. Each comparison parameter from a plurality of comparison parameters is calculated based on different bias frequencies between the input audio signal and the generated signal with a high frequency band. Next, the comparator selects the comparison parameter from the set of comparison parameters so that a certain parameter matches a predetermined criterion.

In other words, the comparator determines the comparison parameter among the set of parameters that meets the specified criterion as much as possible.

The combining unit combines the input audio signal and a signal with a high frequency band to obtain an audio signal with an increased frequency range. In this case, a signal with a high frequency band used to obtain a signal with an increased frequency range is based on the bias frequency corresponding to the selected comparison parameter.

The output interface provides an output of an audio signal with an increased frequency range.

Embodiments of the present invention are based on the basic idea that it is possible to create a signal with a high frequency band, also called a patch, based on the original input signal, as well as compare them. Using various bias frequencies of a signal with a high frequency band or several signals with a high frequency band with different bias frequencies, a plurality of comparison parameters corresponding to these bias frequencies can be calculated. The comparison parameters can be correlated with a number that is associated with the quality of the audio signal. Thus, when determining the comparison parameter, the correspondence of the signal with the high-frequency band and the input audio signal is guaranteed, thereby improving the sound quality.

The bitrate for transmitting and storing the encoded audio signal can be reduced by specifying a parameter based on an offset frequency corresponding to a specific comparison parameter for reconstructing the high frequency band of the original input audio signal. In this case, only a portion of the low frequencies of the input audio signal and an indication of the parameter are required for storage and transmission.

The terms “comparison parameter”, “transition frequency” and “parameter indication” will be defined later.

Some embodiments of the present invention include a comparator that uses cross-correlation when comparing the input audio signal and the generated signal with the high frequency band to calculate the comparison parameter.

Some embodiments of the present invention include a patch generator that generates a signal with a high frequency band in the time domain based on one-way modulation of the frequency band.

An advantage of the preferred forms of implementation of the present invention is an improved coding scheme for audio signals, which allows to increase the sound quality and / or reduce the bitrate for transmitting or storing the signal.

Embodiments of the present invention will be further described in detail in accordance with the following figures, which are given in the appendix:

Figure 1 is a block diagram of an audio encoder;

Figure 2 is a schematic illustration of the process of creating a signal with a high frequency band, comparing the input audio signal and the generated signal with a high frequency band, as well as the process of adapting the signal power with a high frequency band;

Figure 3 is a schematic illustration of the process of creating a signal with a high frequency band, comparing the input audio signal and the signal with the high frequency band, as well as the process of adapting the signal power with the high frequency band;

4 is a block diagram of an encoder increasing the frequency band;

5 is a block diagram of a decoder increasing the frequency band;

6 is a block diagram of a decoder increasing the frequency band;

7 is a diagram of a method of obtaining an output signal based on an input audio signal;

8 is a diagram of a method of obtaining a signal with an increased frequency range;

Fig.9 is a diagram of a method of obtaining an output signal based on an input audio signal;

10 is a diagram of a method for calculating a comparison parameter;

11 is a schematic illustration of an offset frequency interpolation;

12 is a block diagram of a decoder increasing a frequency band;

Fig. 13 is a diagram of a method for obtaining a signal with an increased frequency range;

Fig - block diagram of a method of obtaining a signal with an increased frequency range;

Fig - block diagram of an encoder that increases the frequency band;

Figa - spectrogram of three tones using a variable transition frequency;

Figv - spectrogram of the original audio signal, consisting of three tones;

17 is a power spectrum diagram of an original audio signal and an increased frequency range signal that use a constant transition frequency, and an increased frequency range signal, which uses a variable transition frequency;

Figa - spectrogram of three tones using a known method of increasing the frequency range;

Figv - spectrogram of the original audio signal, consisting of three tones.

In the following description, objects and functional modules having the same or similar functional properties will be indicated by the same serial numbers on the diagrams. Descriptions relating to any scheme will apply to other schemes in order to avoid repetitions in the description of variants of the invention.

Figure 1 shows a block diagram of an audio encoder 100 for receiving an output signal 132 in accordance with the present invention using an audio input signal 102. The output signal can be used to expand the frequency band on the decoder side. Thus, an audio encoder is also called an up-band encoder. The bandwidth increasing encoder 100 includes a patch generator 110, a comparator 120, and an output interface 130. A patch generator 110 is connected to a comparator 120, and a comparator 120 is connected to an output interface 130.

The patch generator 110 generates at least one signal with a high frequency band 112. The signal 112 has a high frequency band, which is created on the basis of the low frequency band of the input audio signal 102. In the event that various signals with high frequency bands 112 are generated, then signals 112 have different frequencies in the high frequency range.

Comparator 120 calculates a plurality of comparison parameters. The comparison parameter is calculated based on the comparison of the input audio signal 102 and the generated signal with the high frequency band 112. Each comparison parameter from a plurality of comparison parameters based on different offset frequencies between the input audio signal 102 and the created signal with the high frequency band 112. Next, the comparator determines the comparison parameter from the set of comparison parameters so that it meets a predetermined criterion.

The output interface 130 provides an output signal 132 for subsequent transmission or storage. The output signal 132 has a parameter indication based on the bias frequency in accordance with the determined comparison parameter.

By calculating a plurality of comparison parameters for different bias frequencies, a signal with a high-frequency band 122 that can best match the original input signal 102 can be found. This is achieved by generating a number of signals with a high-frequency band 112, each of which has a different bias frequency, or by create a single signal with a high-frequency band 112 and a further shift of the high-frequency band of the signal 112 using various offset frequencies. It is also possible to combine multiple signals with a high frequency band 112 with different bias frequencies and further shift the high frequency bands of these signals using other bias frequencies. For example, five different signals are generated with a high-frequency band 112, and each of them is shifted five times by a specific offset frequency.

FIG. 2 is a schematic illustration 200 of the process of creating a signal with a high frequency band, comparing a signal with a high frequency band with an input signal, and optionally adapting the signal power with a high frequency band if only one signal with a high frequency band is generated, which is shifted by the magnitude of the various offset frequencies.

The first diagram is “power vs. frequency "210 schematically shows the input audio signal 102. Based on this input signal 102, the patch generator 110 can create a signal with a high frequency band 112, for example, by shifting 222 the low frequency band of the input audio signal 102 towards higher frequencies (as indicated by ordinal number on the diagram). For example, the low frequency band is shifted by a frequency equal to the transition frequency of the main encoder, not shown in figure 1, which may be part of the encoder 100, increasing the frequency band, or to another predetermined frequency.

The generated signal with a high-frequency band 112 can then be shifted by different bias frequencies 232, for each bias frequency 232 (indicated by 230 in the diagram), a comparison parameter can be calculated by comparator 120. The offset frequency 232 may be determined, for example, with respect to the transition frequency of the main encoder, with respect to another predetermined frequency, or as an absolute frequency value.

Next, the comparator 120 determines a comparison parameter corresponding to a predetermined criterion. Thus, a signal with a high-frequency band 112 with a bias frequency of 242 corresponding to a specific comparison parameter (indicated by 240 in the diagram) can be determined.

Additionally, the power density parameter 252 can be determined (indicated on the diagram by number 250). The power density parameter 252 shows the ratio of the high frequency band of the generated signal to the bias frequency according to the determined comparison parameter to the corresponding frequency band of the input audio signal. For example, this ratio may be represented by a power density coefficient, a power factor, or another coefficient that relates to a power density of a frequency range.

In contrast to the previous schemes, FIG. 3 is a schematic illustration 300 of a process for creating a signal with a high frequency band, comparing the generated signals with an input audio signal and possibly adapting the power of the created signal with a high band if a plurality of signals with high frequency bands that have different offset frequencies.

In contrast to the sequence shown in FIG. 2, patch generator 110 generates a plurality of signals with high frequency bands 112 with different offset frequencies (indicated by 320). This occurs, as in the previous case, using the frequency shift 222 of the low frequency band of the input audio signal 102 in the high frequency range. The low frequency band of the input audio signal 102 may be shifted by a constant frequency plus an individual offset frequency 232 of each signal with a high frequency band 112. The constant frequency may be equal to the transition frequency of the main encoder or another specific frequency.

Then, the comparator 120 calculates a comparison parameter for each generated signal with a high frequency band 112, as well as a comparison parameter 240 corresponding to a predetermined criterion.

The power density parameter 250 is determined as described above.

The circuits shown in FIGS. 2 and 3 can also be combined.

Comparison of the input audio signal 102 and the generated signal with the high frequency band 112 may be based on the cross-correlation of both signals. In this case, the comparison parameter may be, for example, the result of cross-correlation between the input audio signal 102 and the generated signal with a high frequency band 112 for a certain offset frequency.

An indication of the parameter of the output signal 132 may be represented by the bias frequency itself, a quantized bias frequency, or another value based on the bias frequency.

When transmitting or storing only an indication of a parameter instead of the high frequency band of the input audio, the 102 bit rate for transmission or storage may be lowered. Choosing a parameter based on a bias frequency according to a comparison parameter corresponding to a given criterion leads to an improvement in sound quality compared to decoding an audio signal with a limited frequency band.

According to a predetermined criterion, a comparison parameter is determined from the set of comparison parameters, which indicates, for example, a signal with a high frequency band 112 with a necessary bias frequency, which corresponds to an input audio signal 102 better than the remaining 70% of signals with high frequency bands with other bias frequencies . The comparison parameter may also indicate a signal with a high-frequency band 112 with a desired bias frequency, which is one of the three best matches to an input audio signal 102, or a signal with a high-frequency band 112 with a desired bias frequency, which is most suitable for an input signal. This refers to the situation when a plurality of signals with a high frequency band 112 is generated with different bias frequencies, or the situation when a single signal with a high frequency band 112 is often generated that is shifted using different bias frequencies, or a combination of these circuits.

The comparison parameter may be the result of cross-correlation or another value that indicates how the maximum signal with a high-frequency band 112 with a certain offset frequency corresponds to the input audio signal 102.

The frequency range increasing encoder 100 may include a main encoder for encoding a low frequency band of an input audio signal 102. The main encoder can operate with a transition frequency that corresponds to an upper limit frequency of a coded low frequency band of an input audio signal 102. The transition frequency of the main encoder can be constant or variable. The use of a variable transition frequency complicates the operation of the main encoder, but can also increase the flexibility of the encoding process.

The process shown in FIG. 2 and / or FIG. 3 may be repeated for bands with higher frequencies or patches. For example, the low frequency band of the input audio signal 102 has an upper limit frequency of 4 kHz. Thus, if the lower frequency band of the input audio signal 102 is shifted by an amount equal to the upper limit frequency of the low frequency band, in order to obtain a signal with a high frequency band often 112, then the signal 112 has a high frequency band with a low limit frequency of 4 kHz and upper limiting frequency of 8 kHz. The process can be repeated if you shift the low frequency band of the input audio signal 102 by an amount equal to the upper limit frequency of the low frequency band, doubled. So, the created signal with a high-frequency band 112 will have a high-frequency band with a lower limit frequency of 8 kHz and an upper limit frequency of 12 kHz. This process can be repeated until the required high frequency is reached.

Alternatively, this procedure can be accomplished by creating a single signal with multiple high frequency bands.

As shown in the example, the low-frequency bandwidth of the input audio signal and the high-frequency bandwidth of the generated signal are the same. Alternatively, the low frequency band of the input audio signal may be stretched or shifted in order to obtain a signal with a high frequency band.

When determining a signal with a high-frequency band 112 with an offset frequency 232 corresponding to a specific comparison parameter, a gap may remain between the low-frequency band of the input audio signal 102 and the high-frequency band of the generated signal 112 depending on the offset frequency 242. This gap can be filled by creating frequency portions corresponding to a space and containing, for example, noise limited in a strip. The space can also be left blank because sound quality will not suffer significantly.

Figure 4 shows a block diagram of an encoder that increases the frequency range, 400, which creates an output signal 132 in the process of processing the input audio signal 102 in accordance with one embodiment of the present invention. The frequency range increasing encoder 400 includes a patch generator 110, a comparator 120, an output interface 130, a main encoder 410, a bandpass filter 420 and a parameter allocation unit 430. A main encoder 410 is connected to an output interface 130 and a patch generator 110, a patch generator 110 is connected to a comparator 120, a comparator 120 is connected to a parameter extraction unit 430, a parameter extraction unit 430 is connected to an output interface 130, and a bandpass filter 420 is connected to a comparator 120.

The patch generator 110 can be implemented as a modulator to create a signal with a high frequency band 112 based on the input signal 102. The comparator 120 compares the input audio signal 102 filtered by the band pass filter 420 and the generated signal with the high frequency band 112 based on their mutual correlations. The determination of a comparison parameter corresponding to a predetermined criterion may also be called lag calculation.

The output interface may also have the function of a bitstream generating unit and include a combining unit for connecting the low-frequency signal provided by the main encoder 410 and a parametric signal including a parameter determined based on the offset frequency provided by the parameter allocation unit 430. Further, the output interface 130 may include an entropy encoder or differential encoder to reduce the bitrate of the output signal 132. The combining unit and the entropy or differential encoder may be part of the output interface 130, as shown in this example, or be separate units.

The audio signal 102 can be divided into two parts: low frequencies and high frequencies. Separation is carried out using a low-pass filter of the main encoder 410 or a band-pass filter 420. The low-pass filter can be part of the main encoder 410 or be an independent filter connected to the main encoder 410.

The low-frequency part is processed by the main encoder 410, similar to MPEG 1/2 Layer 3 "MP3" or MPEG 4 AAC standard or voice encoder.

The low frequency band is shifted by a fixed amount by converting the sideband or fast Fourier transform (FFT) in the frequency domain. Thus, it is located above the original low-frequency region in the corresponding patch. In addition, the low frequency band can be obtained directly from the input signal 102 using an independent low-pass filter connected to the patch generator 110.

In the case of constant time intervals, a cross-correlation between the amplitude-frequency characteristics of the windows of the original high-frequency band (the original audio signal) and the resulting high-frequency band (the generated signal with the high-frequency band) can be calculated. In this way, a delay (offset frequency) can be determined for maximum correlation. This delay means a correction factor in terms of the original one-way frequency range conversion, i.e. one-way frequency range conversion can be further adjusted by delay in order to maximize cross-correlation. In other words, a bias frequency, also called a delay, is determined, corresponding to the comparison parameter, which, in turn, meets a predetermined criterion. Thus, the comparison parameter corresponds to cross-correlation, and the task of a predetermined criterion is to find the maximum correlation.

Additionally, the relationship between the absolute values of the amplitude-frequency characteristics is determined. This calculates the coefficient by which the received high-frequency signal decreases or increases. In other words, a power density parameter can be determined, which denotes a power factor, power densities, absolute values of amplitude-frequency characteristics or other values correlated with the ratio of the power density between the high frequency band of the generated signal 112 and the corresponding frequency band of the original audio input signal 102. The procedure is carried out by a power density comparator, which can be part of a parameter extraction unit 430, as shown in this example, or an independent unit. To determine the power density parameter, a signal with a high-frequency band 112, which was created by shifting the low-frequency band of the input audio signal 102 by a constant frequency or a signal with a high-frequency band 112 corresponding to a specific comparison parameter, or another created signal with a high-frequency band, can be used 112. The corresponding frequency band in this case denotes a frequency band with the same frequency range. For example, if the high frequency band of the generated signal includes frequencies from 4 kHz to 8 kHz, then the corresponding frequency band of the input audio signal also has a range from 4 kHz to 8 kHz.

The obtained correction factors (bias frequency, power density parameter) corresponding to the delay and the absolute value of the amplitude can be interpolated in time. In other words, it is possible to interpolate the parameter obtained for one window of the signal (for the time frame) for each time period of the signal.

This modulation (control) of the signal (parametric signal) or its parametric representation can be stored or transmitted to the decoder. In other words, the parametric signal 432 can be combined with the low frequency band of the input audio signal 102 processed by the main encoder 410 in order to obtain an output signal 132, which can be stored or transmitted to the decoder.

In addition, parameters for adaptation, for example, noise level and / or tonality, can be defined. The calculations are performed by the parameter extraction unit 430. Additional parameters are added to the parametric signal 432.

The example shown in FIG. 4 illustrates the computation of time-varying modulations on the encoder side. Variables in modulation time in this case refer to signals with a high-frequency band 112 with different bias frequencies. The bias frequency corresponding to a specific comparison parameter that meets a predetermined criterion may vary over time.

FIG. 5 shows a block diagram of a bandwidth increasing decoder 500 for obtaining a signal with an increased frequency range 532 based on an input audio signal 502 and a parametric signal 504 in accordance with one embodiment of the present invention. The parametric signal 504 includes a bias frequency metric and a power density parameter. The bandwidth increasing decoder 500 includes a patch generator 510, a combining unit 520 and an output interface 530. A patch generator 510 is connected to a combining unit 520, and a combining unit 520 is connected to an output interface 530.

The patch generator 510 generates a signal with a high frequency band 512, which has a high frequency band created on the basis of the input audio signal 502. The high frequency band of the generated signal 512 is generated based on the frequency shift of the frequency band of the input audio signal 502, while the frequency shift is based on offset frequencies.

Next, the patch generator 510 increases or decreases the high frequency band of the generated signal 512 by a factor equal to the value of the power density parameter or inverse to the value of the power density parameter.

The combiner 520 connects the signal to the high frequency band 512 and the input audio signal 502 in order to receive an audio signal with an increased frequency range 532, and the output interface 530 outputs a signal with an increased frequency range 532.

Creating a signal with a high-frequency band 112 based on the bias frequency allows you to expand the high frequency range of the input audio signal, if, for example, the bias frequency is determined as previously indicated. This improves the sound quality of the signal with an increased frequency range of 532.

In addition, the power density when expanding the high-frequency band of the input audio signal 502 is set in an effective way - by increasing or decreasing the high-frequency band of the generated signal 112 by the power density coefficient. In this case, no adjustment is necessary.

The patch generator 510 generates a signal with a high frequency band 512 by shifting the frequency band of the input audio signal 512 by a constant frequency, as well as by an offset frequency. If the bias frequency indicates a frequency shift to lower frequencies, then the combining unit may ignore part of the high frequency band of the generated signal 512, which includes frequencies below the upper frequency limit of the input audio signal 502.

The patch generator 510 generates a signal with a high frequency band 512 in the time or frequency domain. In the time domain, patch generator 510 can generate a signal with a high frequency band 512 based on one-way frequency range conversion.

In addition, the output interface can amplify the output signal.

Figure 6 shows a block diagram of a decoder that increases the frequency band, 600 to obtain a signal with an increased frequency range 532 based on the input audio signal 502 and the parametric signal 504 according to one embodiment of the present invention. The bandwidth increasing decoder includes a patch generator 510, a combiner 520, an output interface 530, a main decoder 610 and a parameter extractor 620. A main decoder 610 is connected to a patch generator 510 and a combiner 520, a parameter extractor 620 is connected to a patch generator 510 and an output interface 530, a patch generator 510 is connected to a combining unit 520, and a combining unit 520 is connected to an output interface 530.

The main decoder 610 decodes the received bitstream 602 and directs the input audio signal 502 to the patch generator 510 and combining unit 520. The input audio signal 502 may have an upper limit frequency equal to the transition frequency of the main decoder 610. The transition frequency may be constant or variable over time. A time variable means that it changes for different time intervals or time frames, but remains constant for one time interval or time frame.

Parameter extraction unit 620 can separate the parametric signal 504 from the received bitstream 602 and send it to the patch generator 510. In addition, the parametric signal 504 or the selected noise and / or tonality parameter can be directed to the output interface 530.

The patch generator 510 can convert the input audio signal 502 based on the offset frequency in order to obtain a signal with a high-frequency band 512, as well as increase or decrease the signal with a high-frequency band 512 based on the parameter of the power density, which is in the parametric signal 504. The signal with a high frequency band 512 is sent to the combining unit 530. In other words, the patch generator 510 converts the input audio signal 502 using the offset frequency and the power density parameter in order to obtain a high A signal. The process can occur, for example, in the time domain using one-way conversion of the frequency band 634 together with interpolation and / or filtering 632 for each time interval.

The combining unit 520 connects the input audio signal 502 and the generated signal to the high frequency band 512 in order to obtain a signal with an extended frequency range 532.

The output interface 530 outputs a signal with an extended frequency range 532 and may further include a correction unit. The correction block performs tonality and / or noise correction based on the parameters provided by the parameter extraction unit 620. The correction block may be part of the output interface 530, as shown in FIG. 6, or a separate block. A correction block may also be located between the patch generator 510 and combining block 520. In this case, the correction block corrects the tonality and / or noise of the created signal with a high frequency band 512 only. The correction of the tonality and noise of the input audio signal 502 may not be taken into account, i.e. to. the input audio signal 502 corresponds to the original audio signal.

Summarizing, it can be said that a band-increasing decoder 600 can synthesize and shape a high-frequency signal spectrum based on the output signal of an audio decoder or a main decoder (input audio signal using a transmitted conversion function. A transmitted conversion function denotes a frequency-based conversion function displacement or power density parameter. Then it is possible to connect signals with high frequency and low frequency, then it is possible to use parameters to adapt noise and tonal nnosti.

7 shows a diagram of a method 700 for obtaining an output signal based on an audio input signal in accordance with an embodiment of the present invention. The method includes creating 710 at least one signal with a high frequency band, computing 720 a plurality of comparison parameters, extracting a 730 comparison parameter from the plurality of comparison parameters and providing 740 an output signal for subsequent transmission or storage.

The created signal with an extended frequency band has a high frequency band. The high frequency band of the generated signal is based on the low frequency band of the input audio signal. Different signals with extended frequency bands have different frequencies within the high frequency band.

The comparison parameter is calculated based on a comparison of the input audio signal and the generated signal with the high frequency band. Each comparison parameter from a plurality of parameters is calculated based on different bias frequencies between the input audio signal and the generated signal with a high frequency band.

The highlighted comparison parameter matches a predefined criterion.

The output signal includes an indication of a parameter based on a bias frequency that corresponds to a specific comparison parameter.

On Fig shows a diagram of a method 800 for receiving an audio signal with an increased frequency range based on the input audio signal and a parametric signal according to one embodiment of the present invention. The parametric signal indicates the offset frequency and the power density parameter. The method includes creating an 810 signal with a high frequency band, increasing 820 or decreasing the high frequency band of a signal with an extended frequency band, combining 830 a signal with a high frequency band and an input audio signal to obtain an extended frequency band signal, and outputting an 840 wide band signal.

An extended band signal has a high frequency band. The high frequency band of the extended band signal is created 810 based on the band shift of the input audio signal. The frequency shift occurs using the offset frequency.

The high frequency band of a signal with an extended frequency band increases 820 or decreases by a factor equal to the value of the power density parameter or inverse to the value of the power density parameter.

FIG. 9 shows a diagram of a method 900 for receiving and outputting a signal based on an input audio signal according to one embodiment of the present invention. The diagram illustrates the possibility of applying the algorithm on the encoder side. The sequence can be described mathematically as follows. Real-time signals are indicated in uppercase Latin letters, signals that have passed the Hilbert transformation stage are denoted by the corresponding letters of the Greek alphabet, signals that have passed the Fourier transformation stage are indicated in capital letters of the Latin or Greek alphabet.

указывает на трансформацию Фурье, j обозначает воображаемое число, трансформация Гилберта Н(.) определяется обычным образом:The input signal is denoted as f (n), the output signal is o (n).

indicates the Fourier transform, j denotes an imaginary number, the Hilbert transform H (.) is defined in the usual way:

φ (m): = H (f (n)) = F ^-1 (-j sgn (ω) F (jω)),

Where

F (jω): = F (f (n))

xOver denotes the limiting frequency of the main encoder, n∈N denotes time. k _max > k∈N denotes ke extension or patch. α _k denotes the border of the band of perceived bands related to xOver in accordance, for example, with the Bark scale or ERB. Alternatively, α _k may increase linearly, i.e. α _{k + 1} -α _{k is a} constant. Hilbert transform can be effectively calculated by filtering the signal with a modulated low-pass filter.

при временном инкременте

(Fs указывает частоту выборки). Математически это можно описать с помощью следующих формул:First, an analytic function of modulator 902 can be created with modulation frequencies α _k and resulting phase increments

with time increment

(Fs indicates the sampling rate). Mathematically, this can be described using the following formulas:

The sum can be replaced only by n if γ _k is independent of n.

The input audio signal 102 or the real audio signal f is subjected to a filtering process using a band-pass filter to the band α _{k + 1} -α _k , which can be expressed as follows:

f _LF = f * filt _LF

In this case, each patch will have the same bandwidth.

Alternatively, the input signal f 102 can be filtered using a band-pass filter to the bands α _k with different widths, which is expressed by the formula:

Then, the areas of the original signal that should be reconstructed using the method under consideration are determined. Bounded areas are designated as:

and are located in the intervals (α _k , α _{k + 1} ).

The conversion of input signals filtered by a low-pass filter, 904 may occur in the frequency or time domain.

In the frequency domain, the input signals are first divided into windows, which is expressed as follows:

where NFFT is the number of fast Fourier transforms (for example, with a sampling frequency of 512), ξ is the number of windows, and win (.) is the window function. Windows or time frames can be partially aligned in terms of time. For example, the formula above represents a time overlay equal to half the window. N∈N are the time intervals of the original signal and the associated set of amplitude spectral characteristics F _ξ (ω) for ξ≤N as the absolute values of the Fourier transform

represent the measure of the boundary of the strip k in the Fourier transform.

Then, the signal is converted in the frequency domain by shifting the digital results of the FFT transforms (fast Fourier transforms). There is no need for an implicit Hilbert transform here, but it allows us to describe the next steps in an equivalent way:

for ω≥0 and

Φ _ξ (ω): = Ψ _ξ (ω): ≡0∀ω <0

In the time domain, the transformation of Hilbert 906 of the input audio signal f 102 to create an analytical signal 908 is performed primarily.

φ: = f + jH (f)

and

подвергается преобразованию при помощи односторонней модуляции полосы частот 710 посредством модулятора µ(n) 902:Then analytical signal

undergoes conversion using one-way modulation of the frequency band 710 by means of a modulator µ (n) 902:

or

ψ (n): = φ _LF (n)

In this case, an extended bandwidth signal is created, which is also called modulated signal 910.

Next, the window is divided (with possible overlap) of the input signal 912 and the signal with an extended frequency range 914, as well as the Fourier transform:

and

where NFFT is the number of digital results (bins) of the fast Fourier transform (for example, 256, 512, 1024 or another number ranging from 2 ⁴ to 2 ³² ), ξ is the number of windows, win (.) is the window function. The time intervals N∈N are created on the basis of the initial signal and the set of amplitude-spectral characteristics Ф _ξ (ω), Ψ _ξ (ω) associated with it at ξ≤N as absolute values of the Fourier transform 916

represent the measure of the boundary of the strip k in the Fourier transform.

The process in the time domain is shown in Fig.9.

The next step is the calculation of 720 cross-correlation R _{ξ, k} (the comparison parameter may be equal to the result of cross-correlation) of the amplitude-spectral characteristics of the original signal and the signal with an extended frequency range, which is mathematically expressed as follows:

at

Φ _ξ (ω): ≡Ψ _ξ (ω): ≡0∀ω <0; v≤Λ

с обоих сторон спектра определенной полосы (патча).δ denotes the maximum delay (maximum offset frequency) for which cross-correlation is calculated. If cross-correlation should be calculated with an offset, i.e. preference is given to small delays and, accordingly, large overlays, it is necessary to choose β = 0. On the contrary, if it is necessary to compensate for the fact that a smaller number of digital results of fast Fourier transforms (FFt-bins) overlap each other with large delays, then it is necessary to choose β = 1. In general, 0≤β∈P can be arbitrarily chosen. Additionally or alternatively, 2 <δ∈N; mod (δ, 2) = 0 in order to select the cross-correlation region slightly larger than the created frequency band (patch). Moreover, the region that is considered in cross-correlation can be expanded with

on both sides of the spectrum of a certain band (patch).

Based on the results of cross-correlation, the maximum cross-correlation is determined 730

and lag d _{ξ, k} cross-correlation

R _{ξ, k} (d _{ξ, k} ) = m _{ξ, k}

Additionally, energy or power factors in the patch are determined using power density spectra:

If it is impossible to determine a clear maximum of 924, the delay is set back to 0 (the diagram shows number 922). Otherwise, the estimated delay 918 may be a delay corresponding to the maximum cross-correlation. For this, a suitable threshold criterion d _{ξ, k} > τ is determined, where it is necessary to choose τ. In addition, a bend or even part of the cross-correlation spectrum (SFN) may be observed, for example:

or

at

Delays d _{ξ, k} and power density parameters ζ _{ξ, k} can be interpolated 926 in order to obtain a value for each time interval:

ζ _k (n) = interp (c _{ξ, k} ); λ _k (n) = interp (d _{ξ, k} ).

Further, after modifications, amplitude transformations and frequency shift, the general modulation function is determined:

A common modulation function or parameters of a common modulation function are provided 740 along with an output signal for further storage or transmission.

Additionally, parameters can be determined to adjust noise and / or tonality.

Modulation on the side of the decoder is carried out according to the formula:

and also an additional number of k partial modulations (if more than one patch is present). For this, the common modulation function µ _k (n) or µ (n) or the parameters ζ _k (n) and λ _k (n) or c _{ξ, k} and d _{ξ, k of the} general modulation function are encoded in a suitable way, for example, by quantization. If necessary, the sampling rate can be reduced and hysteresis introduced.

The delay calculation can be omitted if there is no tone, such as a period of silence, a transition moment, noise. In these cases, the delay can be set to 0.

10, an example 1000 is examined in detail where a delay is determined.

For a time frame or window ξ = i 1010, the delay v is set to minus λ as the initial value. Then the cross-correlation R _{ξ, k} (v) 720 is calculated. If v is less than Λ 1030, v increases 1032 and 720 the next comparison parameter is calculated within the cross-correlation. If v is equal to or greater than Λ 1030, a delay 730 corresponding to the maximum calculated cross-correlation can be determined. If the maximum is clearly defined 924, then a certain delay is used as the parameter d _{ξ, k} 918. Otherwise, the delay is set to 0 and used as the parameter d _{ξ, k} = 0922.

Next, the whole process is repeated 1040 for the next time frame ξ = ξ + 1 1050. Certain delays are interpolated 926 in order to obtain a parameter for each time interval N.

The calculation of many comparison parameters, for example, cross-correlation results, can be carried out in parallel if the required number of comparators is used. Processing of various time frames can also be carried out in parallel if the necessary hardware is used several times. The cross-correlation calculation cycle starts at + Λ and can be reduced at each cycle to v≤Λ.

11 shows a schematic illustration of interpolation 926 of the offset frequencies of different time frames, time intervals, or windows. 11 a shows interpolation 1100 if time frames do not intersect. The delay d _{ξ, k} is determined for the entire time frame 1110. The easiest way to interpolate the parameter for each time interval 1120 is to set the parameters of the time frame 1110 equal to the corresponding delay d _{ξ, k} for all time intervals 1120. A delay can be selected at the borders of the time frame previous or next time frame. For example, the parameters from λ _k (n) to λ _k (n + 3) are equal to d _{ξ, k} , and the parameters from λ _k (n + 4) to λ _k (n + 7) are equal to d _{ξ + 1, k} .

Alternatively, the delays of time frames 1110 can be interpolated linearly between time frames, for example:

λ _k (n + 2) = d _{ξ, k}

11B shows an example 1150 for intersecting time frames 1110. In this case, one time interval 1120 refers to more than one time frame 1110. Thus, more than one specific delay refers to one time interval 1120. Thus, certain delays can be interpolated 926 in order to obtain one parameter for each time interval 1120. Certain delays corresponding to one time interval 1120 can be interpolated linearly. For example, the following interpolation is possible:

λ _k (n) = d _{ξ-1, k}

λ _k (n + 2) = d _{ξ, k}

Alternatively, the interpolation can be carried out, for example, using medial filtering.

Interpolation can be carried out using special tools designed for interpolation. These tools can be part of a parameter retrieval unit or output interface, as well as a separate unit.

On the side of the decoder, the extension of the frequency range is carried out according to the formula:

и φ_LF(N) как выходного сигнала основного кодера. Дополнительно

может быть адаптирован с помощью параметров, полученных заранее от исходного сигнала, по уровню тональности и/или шума.After decoding

and φ _LF (N) as the output of the main encoder. Additionally

can be adapted using parameters obtained in advance from the original signal, according to the level of tonality and / or noise.

The calculation of the total modulation function on the decoder side is performed according to one of the two formulas indicated below:

and

ψ (n) = φ _LF (n) μ (n) + noise (n)

The imaginary part of the signal can be omitted:

o (n) = Re (ψ (n))

Further, as indicated above, a tonality adjustment may be followed by, for example, reverse filtering.

12 is a block diagram of a bandwidth increasing decoder 1200 for receiving an extended band audio signal 532 based on an input audio signal 502 according to an embodiment of the present invention. The bandwidth increasing decoder 1200 includes a patch generator 1210, a comparator 1220, a combining unit 1230 and an output interface 1240. A patch generator 1210 is connected to a comparator 1220, a comparator 1220 is connected to a combining unit 1230, a combining unit 1230 is connected to an output interface 1240.

The patch generator 1210 generates at least one signal with a high-frequency band 1212, which has a high-frequency band based on the input signal 502, while the lower limit frequency of the high frequency band of the generated signal 1212 is lower than the upper limit frequency of the input audio signal 502. Various signals with high-frequency bands 1212 have different frequencies within the high-frequency band.

Comparator 1220 computes a plurality of comparison parameters. The comparison parameter is calculated based on the comparison of the input audio signal 502 and the generated signal with the high frequency band 1212. Each comparison parameter from the set of comparison parameters is calculated based on different offset frequencies between the audio input signal 502 and the created signal with the high frequency band 1212. Next, the comparator selects the parameter comparisons from the set of comparison parameters so that it matches a predetermined criterion.

The combining unit 1230 connects the input audio signal 502 and the generated signal to the high frequency band 1212 in order to obtain a signal with an extended frequency range 532, while the created signal with the high frequency band is based on the offset frequency corresponding to the selected comparison parameter.

The output interface 1240 provides an output signal with an extended frequency range 532.

Compared to the decoder shown in FIG. 5, the decoder 1200 in question itself determines the offset frequency. Thus, there is no need to obtain this parameter with the input audio signal 502. In this case, the bit rate for transmitting or storing audio signals can be reduced.

As indicated by referring to FIG. 1, patch generator 1210 can generate multiple signals with a high frequency band with different bias frequencies, or a single signal with a high frequency band that is biased using different bias frequencies. It is also possible to use a combination of these two possibilities.

On Fig shows a diagram of a method 1300 for receiving a signal with an extended frequency range according to one embodiment of the present invention. Method 1300 includes creating 1310 at least one signal with a high frequency band, calculating 1320 the set of comparison parameters, extracting 1330 one parameter from the set of comparison parameters, combining 1340 input audio signal and the created signal with high frequency band, and outputting 1350 signal with extended frequency range.

A signal with an extended frequency range has a high frequency band created based on the input audio signal. The lower limit frequency of the high frequency band of the signal with the extended frequency range is lower than the upper limit frequency of the input audio signal. If different signals are generated, then signals with high frequency bands have different frequencies within the high frequency band.

The comparison parameter is calculated based on a comparison of the input audio signal and the generated signal with the high frequency band. Each comparison parameter from a plurality of comparison parameters is calculated based on different bias frequencies between the input audio signal and the generated signal with a high frequency band.

The highlighted comparison parameter matches a predefined criterion.

The generated signal with a high frequency band, which is connected to the input audio signal in order to obtain a signal with an extended frequency range, is based on the offset frequency corresponding to the selected comparison parameter.

On Fig shows a diagram of a method 1400 for receiving an audio signal with an extended frequency range according to one embodiment of the present invention.

After receiving 1402 bit stream containing the input audio signal, the main decoder decodes 1410 the input audio signal. Based on the input audio signal, a 1310 signal with a high frequency band is generated and 1320 a plurality of comparison parameters are calculated in terms of mutual correlation between the input audio signal and the generated signal with a high frequency band. Then, a comparison parameter corresponding to a predetermined criterion is determined 1330; The comparison parameter is also called the estimated delay.

Based on a bias frequency that corresponds to a specific comparison parameter, the modulator modulates 1420 the input audio signal. Additionally, a parameter 1430 may be extracted from the received bitstream 1402 to adapt, for example, the power density of the modulated signal. The modulated signal is then connected 1340 to the input audio signal. In addition, the 1440 tone and noise of a signal with an extended frequency range can be adjusted. This step may be carried out before combining with the input signal. Then, audio data, namely an audio signal with an extended frequency range, is sent 1350, for example, for audio playback.

In this case, the calculation of the modulation, which is variable over time, is carried out on the side of the decoder.

Unlike a modulator that modulates the input audio signal 1420 and creates a patch, for example, a previously created signal with a high frequency band can be used, or a patch generator can create a signal with a high frequency band (patch) based on an offset frequency that corresponds to a certain parameter comparisons.

In other words, if a low data rate is more important than the complex structure of the decoder, then the determination of the frequency modulation that the modulators perform can also be done on the side of the decoder. To this end, on the decoder side, with some changes, the algorithm shown in Fig. 9 is executed. Since on the decoder side the original signal is not available for calculating cross-correlation, the correlation is calculated between the original signal (audio input signal) and the original signal (audio input signal), which has undergone a shift in the intersection area. For example, the signal shifts between zero and α _k , for example, α _k is divided by 2, α _k is divided by 3 or α _k is divided by 4. α _k denotes the boundary of the strip k, for example, α ₁ denotes the transition frequency of the main encoder.

Calculations can occur in the same way on the encoder side as well as on the decoder side. On the decoder side, parameters are extracted to form the spectrum, adjust noise and / or tonality, which are then transmitted to the decoder.

On Fig shows a block diagram of an encoder that increases the frequency band, 1500 to obtain an output signal based on the input signal according to one embodiment of the present invention. Encoder 1500 corresponds to the encoder shown in FIG. However, the encoder 1500 itself does not provide an output signal 132 with an indication of the parameter based on the offset frequency. It can only determine the power density parameter and optional parameters for correcting tonality and noise and include the indication of these parameters in the output signal 132. The power density parameter (like other parameters in the case of their calculation) are determined based on the offset frequency corresponding to the selected comparison parameter.

For example, the power density parameter may indicate the relationship between the input audio signal 102 and the generated signal with a high frequency band with an offset frequency corresponding to a specific comparison parameter. Thus, the indication of the parameter, which correlates with the parameter of the power density and optionally with the parameters for adjusting the tonality and / or noise, is based on the offset frequency corresponding to a certain comparison parameter.

The next difference between the encoder 1500 and the encoder shown in FIG. 4 is that the patch generator 110 generates a signal with a high frequency band in the same way as the patch generator of the decoder 1400. In this case, the encoder 1500 and the decoder can receive the same frequencies offsets, so the parameters extracted by encoder 1500 are also valid for patches created by the decoder.

Some variants of the present invention are implemented by means of devices and a method of expanding the frequency band of audio signals in the time domain using modulators that vary in time. In other words, a patch is generated with a changing limit frequency, for example, for each time interval, each time frame, part of a time frame, or groups of time frames.

The described method of expanding the frequency band of an audio signal can be applied on the encoder side and on the decoder side, and also only on the decoder side. Unlike other known methods, the described new method can perform the so-called harmonic extension of the frequency band without requiring accurate information about the fundamental frequency of the audio signal. In contrast to the so-called harmonic bandwidth extensions, considered, for example, in the patent application in the United States "F.Nagel, S. Disch:" Apparatus and method of harmonic bandwidth extension in audio signals "" under the number US 61/025129, which are produced using phase voice encoders, in this case, the spectrum is not subject to expansion, so the density remains unchanged. To ensure harmony, a correlation is used between the extended and base frequency bands. This correlation is calculated both on the side of the encoder and on the side of the decoder, depending on the possibility of the calculations, the amount of memory and the data rate.

For example, bandwidth expansion can be performed using amplitude modulation (AM) or frequency shift through one-way bandwidth modulation (SSB). In the course of the subsequent processing, in accordance with additional parameters, an attempt is made to bring the spectral envelope, noise level and other parameters to the original signals.

The new signal conversion method avoids the problems that arise during simple copying or during the mirroring operation, as in this case, a harmonious correct continuation of the spectrum is created with the help of a time-varying transition frequency limit between the low frequency range (LF) and the high frequency range (HF), as well as between subsequent high frequency bands, the so-called patches. Limit frequencies are selected in such a way that the created patches match the actual harmonic raster as much as possible.

On Fig shows a modulator with three time-varying amplitudes and limit frequencies, with which three patches can be created by one-way modulation of the base bands. On figa shows a diagram 1600A of the spectrum of the signal with an extended frequency range when using time-varying limit frequencies 1610. On figv shows a diagram 1600B of the spectrum of the audio signal, consisting of three tones. In contrast to the spectrogram shown in FIG. 18B, lines 1620 are significantly less blurred.

17 shows a diagram 1700. The power density spectrum of the three tones of the original audio signal is shown at 1710, the signal with a constant limit frequency at 1720, the signal with a constant limit frequency at 1730. In contrast to using a constant limit frequency, harmonic the structure is preserved using a variable limiting frequency of 1730.

With harmonious continuation of the spectrum, problems can be avoided at the transition point between the base band (main encoder) and the extended band, as well as between subsequent patches. Without an estimate of F ₀ , which is a necessary condition for the functioning of the system, a harmonic continuation of random signals without audible interference is possible, while harmony and transient sound moments are not violated.

Embodiments of the present invention relate to a method that is suitable for all audio applications where the full frequency range is not available.

For example, the method can be used to broadcast audio information using digital radio, the Internet stream or applications that use audio communication.

Other embodiments of the present invention relate to a bandwidth increasing decoder that receives a signal with an extended frequency range based on an input audio signal and a parametric signal, wherein the parametric signal indicates a shift frequency and a power density parameter. The bandwidth increasing decoder includes a patch generator, a combiner, and an output interface. The patch generator creates a signal with a high frequency band, and the high frequency band of the generated signal is generated based on the frequency shift of the frequency band of the input audio signal. Frequency offset is based on the offset frequency. The patch generator increases or decreases the high frequency band of the generated signal by using a coefficient equal to the value of the power density parameter or equal to the value inverse to the value of the power density parameter. The combining unit connects the generated signal with a high frequency band and an input audio signal to obtain an audio signal with an extended frequency range. The output interface provides a signal output with an extended frequency range.

Other embodiments of the present invention relate to a decoder increasing the frequency band as described above, wherein the patch generator increases or decreases the high frequency band of the signal with the high frequency band by a factor equal to the value of the power density parameter or equal to the reciprocal of the density parameter power. The power density parameter in this case is indicated in the input audio signal.

Despite the fact that the present invention has been described in several embodiments, other changes and functioning options are possible within the scope of this invention. It should be noted that there are many alternative ways to implement the methods, as well as the ways of composition of the present invention. The following formula is intended to include all such options and changes as they do not go beyond the scope and essence of the present invention.

It should be noted that, depending on the conditions, the scheme of the invention can be implemented as software. The implementation of the invention can be made on digital storage media. For example, on a floppy disk or CD, which are readable and have the ability to work with a programmable computer system in order to perform the corresponding method. In general, the invention consists of a computer program product with program code that is stored on a readable medium for implementing the inventive method when the program product is executed on a computer. In other words, the invention can be implemented as a computer program with program code for implementing the method when the computer program product is executed on a computer.

Claims (17)

1. Audio encoder (100), (400), (1500) to obtain the output signal (132) using the input signal (101), including a patch generator (110), designed to generate at least one high-frequency signal with an extended band frequencies (112), which contains a high frequency band based on the low frequency band of the input audio signal (102), while if different high-frequency signals with an extended frequency band (112) are generated, they contain different frequencies in their bands;
a comparator (120) calculates a plurality of comparison parameters, wherein the comparison parameter is calculated based on a comparison of the input audio signal (102) and the generated signal with the high frequency band (112); each comparison parameter from a plurality of comparison parameters is calculated based on a different offset frequency between the input audio signal (102) and the generated signal with a high frequency band (112); the comparator (120) determines one comparison parameter from the set of comparison parameters so that the selected comparison parameter corresponds to a predetermined criterion;
the output interface (130) provides an output signal (132) for subsequent transmission or storage, while the output signal (132) includes a parameter based on a bias frequency that corresponds to a specific comparison parameter. 2. The audio encoder according to claim 1, including a power density comparator (430), which compares the parameters based on the power density of the high frequency band of the generated signal with the high frequency band (112) and the corresponding band of the input audio signal (102) to obtain a power density parameter ; the power density parameter indicates the ratio of the power density of the high-frequency band of the generated signal with the high-frequency band (112) and the corresponding frequency band of the input audio signal (102); the indication of the output audio signal parameter (132) is based on the power density parameter. 3. The audio encoder according to claim 1, in which the comparator compares the input audio signal and the generated signal with the high frequency band by calculating the result of the mutual correlation of the input audio signal and the generated signal with the high frequency band; the calculated comparison parameter is based on the result of cross-correlation; the cross-correlation parameter is the frequency of the signal offset with the high-frequency band and, thus, correlates with the calculated comparison parameter. 4. The bandwidth expansion decoder (500), (600) for receiving an audio signal with an extended frequency range (532) based on the input audio signal (502) and the parametric signal (504), while the parametric signal (504) includes an indication of the frequency bias and power density parameter; a bandwidth increasing decoder including a patch generator (510) that generates a signal with a high frequency band (512), wherein the high frequency band of the signal with the high frequency band (512) is created based on the frequency shift of the frequency band of the input audio signal (502) ; frequency offset is based on the offset frequency; the patch generator (510) increases or decreases the high frequency band of the signal with the high frequency band (512) by a factor equal to the value of the power density parameter or, correspondingly, equal to the inverse of the power density parameter;
a combining unit (529) that connects the signal to the high frequency band (512) and the input audio signal (502) to obtain a signal with an extended frequency range (532);
however, the combining unit provides the ability to neglect part of the high frequency band of the RF signal with an extended band (512), where the neglected part of the high frequency band of the RF signal with an extended band (512) contains frequencies lower than the upper cut-off frequency of the input audio signal (502) ; and
the output interface (530) provides a signal output with an extended frequency range (532). 5. The decoder according to claim 4, including the main decoder (610), which generates an input audio signal (502) based on the encoded input audio signal (602), while the main decoder (610) generates an input signal (502) with a constant upper limit frequency; a patch generator (510) generates a high frequency band of a signal with a high frequency band (512) by shifting the frequency band of the input audio signal (502) by a frequency equal to the upper limit frequency of the input audio signal (502) together with the shift frequency. 6. The decoder according to claim 4, in which the patch generator (510) generates a signal with a high frequency band (512) in the time domain. 7. The decoder according to claim 6, in which the patch generator (510) generates a signal with a high frequency band (512) based on one-way modulation of the frequency band. 8. A bandwidth expansion decoder (1200) for receiving a signal with an extended frequency range (532) based on an input audio signal (502), including a patch generator (1210) that generates at least one signal with a high frequency band (1212) having a high frequency band based on an input signal (502), wherein the lower limit frequency of the high frequency band of the generated signal with the high frequency band (1212) is lower than the upper limit frequency of the input audio signal (502); if different signals with high frequency bands (1212) are generated, then signals (1212) have different frequencies within the high frequency bands;
a comparator (1220) calculates a plurality of comparison parameters, wherein the comparison parameter is calculated based on a comparison of the input audio signal (502) and the generated signal with the high frequency band (1212); each comparison parameter from a plurality of comparison parameters is calculated based on a different offset frequency between the input audio signal (502) and the generated signal with a high frequency band (1212);
the comparator (1220) determines one comparison parameter from the set of comparison parameters so that the selected comparison parameter matches a predetermined criterion;
the combining unit (1230) connects the input audio signal (502) and the signal with a high frequency band to obtain a signal with an extended frequency range (532), while the generated signal with a high frequency band that is used to obtain a signal with an extended frequency range (532) based on a bias frequency corresponding to a specific comparison parameter;
the output interface (1240) provides a signal output with an extended frequency range (532). 9. The decoder of claim 8, in which the patch generator (1210) increases or decreases the high frequency band of the signal with the high frequency band (1212) by a factor equal to the value of the power density parameter or, correspondingly, equal to the inverse of the power density parameter, The indication of the power density parameter is contained in the input audio signal (502). 10. The decoder according to claim 4, which has interpolation means, wherein the time frame includes a plurality of time intervals; each time frame has a corresponding offset frequency; by means of interpolation, the offset frequency of the time frame or a plurality of offset frequencies of different time frames are interpolated for each time interval of the time frame in order to obtain an interpolated offset frequency for each time interval. 11. The bandwidth expansion decoder of claim 8, wherein the comparator compares the input audio signal and the generated signal with the high frequency band by calculating the result of cross-correlation of the input audio signal and the generated signal with the high frequency band; the calculated comparison parameter is based on the result of cross-correlation; the cross-correlation parameter is the frequency of the signal offset with the high-frequency band and, thus, correlates with the calculated comparison parameter. 12. A method (700) for obtaining an output signal using an input audio signal, including generating (710) at least one signal with an extended frequency band, wherein the generated signal has a high frequency band; the high frequency band of the generated signal is based on the low frequency band of the input audio signal; if different signals with high frequency bands are generated, then the created signals have different frequencies within the high frequency bands;
calculating (720) a plurality of comparison parameters, wherein the comparison parameter is calculated based on comparing the input audio signal and the generated signal with the high frequency band; each comparison parameter from a plurality of comparison parameters is calculated based on a different offset frequency between the input audio signal and the generated signal with a high frequency band;
determining (730) a comparison parameter from a plurality of comparison parameters, wherein the highlighted comparison parameter corresponds to a predetermined criterion;
providing (740) an output signal for subsequent transmission or storage, wherein the output signal includes an indication of a parameter based on an offset frequency that corresponds to a specific comparison parameter. 13. A method (800) for generating an audio signal with an extended frequency band based on an input audio signal and a parametric signal containing an indication of the offset frequency and the power density parameter, including:
generating (810) an RF signal with an extended frequency band, wherein the high frequency band of the generated signal is generated based on the frequency shift of the frequency band of the input audio signal; the frequency shift, in turn, is based on the bias frequency;
increase (820) or decrease the frequency band of the signal with the high-frequency band by a factor equal to the value of the power density parameter or equal to the inverse of the power density parameter;
combining (830) a signal with a high frequency band and an input audio signal to produce an audio signal with an extended frequency range;
while part of the high-frequency band of the RF signal with the extended frequency band (512) can be neglected, and the neglected part of the high-frequency band of the RF signal with the expanded frequency band (512) contains frequencies lower than the upper cut-off frequency of the input audio signal (502) ; and
output (840) of an audio signal with an extended frequency band. 14. A method (1300) for generating a signal with an extended frequency range based on an input audio signal, including generating (1310) at least one signal with a high frequency band based on the input audio signal, wherein the lower limit frequency of the high frequency band of the generated signal is lower than the upper limit frequency of the input audio signal;
if different signals with high frequency bands are generated, then the created signals have different frequencies within the high frequency bands;
calculating (1320) a plurality of comparison parameters, wherein the comparison parameter is calculated based on comparing the input audio signal and the generated signal with the high frequency band; each comparison parameter from a plurality of comparison parameters is calculated based on a different offset frequency between the input audio signal and the generated signal with a high frequency band;
determining (1330) the comparison parameter from the plurality of comparison parameters, wherein the highlighted comparison parameter corresponds to a predetermined criterion;
combining (1340) an input audio signal and a signal with a high frequency band to obtain an audio signal with an extended frequency range; wherein the signal with an extended frequency range is based on a bias frequency that corresponds to a specific compensation parameter;
output (1350) of an audio signal with an extended frequency range. 15. Machine-readable medium storing a computer program with program code for implementing the method according to item 12, provided that the computer program is executed using a computer or microcontroller. 16. A computer-readable medium storing a computer program with a program code for implementing the method according to item 13, provided that the computer program is executed using a computer or microcontroller. 17. A computer-readable medium storing a computer program having a program code for implementing the method according to 14, provided that the computer program is executed using a computer or microcontroller.

Images