Time domain gain modeling in QMF domain
By calculating and applying broadband gain in the MFB domain, the problem of introducing delay in the time domain cross fade is solved, and efficient audio processing in the MFB domain is realized, reducing the delay and complexity of the audio processing chain.
Patent Information
- Application Number
- CN202380033669.8
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Priority Date
- 2023-03-16
- Filing Date
- 2023-04-13
- Publication Date
- 2024-11-22
AI Technical Summary
In audio encoding, time domain crossfade introduces additional processing delays and prevents audio processing of decoded signals in the QMF domain, resulting in the need to achieve efficient crossfade and broadband gain determination in the MFB domain.
By calculating broadband gain from the time domain target gain in the MFB domain, using MFB to analyze prototype filters and synthesize prototype filters, the broadband gain is optimized for application in the MFB domain, thereby simulating time domain gain and reducing latency and calculations of audio processing chains. Complexity.
Accurate and efficient broadband gain calculation in the MFB domain, reducing the overall delay and computing complexity of the audio processing chain, and supporting efficient audio processing on the decoder side.
Smart Images

Figure CN119013723A_ABST
Abstract
Description
[0001] CROSS-REFERENCE TO RELATED APPLICATIONS
[0002] This application claims priority to U.S. Provisional Patent Application No. 63 / 330,428 filed on April 13, 2022 and U.S. Provisional Patent Application No. 63 / 490,722 filed on March 16, 2023, each of which is incorporated herein by reference in its entirety. Technical Field
[0003] The present disclosure relates to audio processing techniques. In particular, the present disclosure describes time domain gain modeling in a modulated filter bank (MFB) domain (e.g., a quadrature mirror filter (QMF) domain), and cross-fading of an audio signal in the MFB domain when a target time domain cross-fade gain function should be achieved as accurately as possible. Background Art
[0004] In audio coding, the input to the encoder can be two or more audio channels that have been reversibly transformed, where the transformation varies with the time frame. When switching from the previous time frame to the current time frame, the transform parameters of the transform signal are smoothly cross-faded in the time domain before encoding. This cross-fading requires mirroring on the decoder side. Traditionally, cross-fading is performed in the time domain after parameter decoding in the QMF domain.
[0005] However, time domain cross-fading introduces additional processing delays and prevents audio processing of the decoded signal in the QMF domain.
[0006] Therefore, there is a need for techniques that enable efficient audio processing, especially at the decoder side. There is also a need for techniques that enable cross-fading of audio signals in the MFB domain (e.g., QMF domain), especially at the decoder side, and efficient determination of appropriate MFB domain wideband gains. Summary of the invention
[0007] In view of this need, the present disclosure provides a method and apparatus for processing audio, as well as a corresponding program and computer-readable storage medium, which have the features of the respective independent claims.
[0008] One aspect of the present disclosure relates to a method for processing audio. The method may include: determining a wideband gain in a modulation filter bank (MFB) domain for fading an audio signal according to a target gain in a time domain, such that applying the wideband gain in the MFB domain simulates applying the target gain in the time domain. Wherein, determining the wideband gain may include: calculating the wideband gain using the target gain, an MFB analysis prototype filter, and an MFB synthesis prototype filter. The target gain may be related to a target gain function as a function of time (or sample number or other suitable time index).
[0009] Configured according to the above definition, the proposed method can apply a broadband gain in the MFB domain, thereby achieving MFB domain processing of the audio signal after the gain is applied and reducing the overall delay and computational complexity of the audio processing chain.
[0010] In some embodiments, a respective wideband gain may be calculated for each of a plurality of MFB analysis time slots.
[0011] In some embodiments, calculating the wideband gain may include optimizing the wideband gain by calculating a least squares solution.
[0012] In some embodiments, determining the wideband gain may include: determining a corresponding MFB analysis signal for each of a plurality of MFB analysis time slots and each of a plurality of frequency bands based on an input training signal and the MFB analysis prototype filter. Determining the wideband gain may further include: determining a corresponding MFB synthesis signal for each of the plurality of MFB analysis time slots based on the MFB analysis signal in the corresponding MFB analysis time slot and the MFB synthesis prototype filter. Determining the wideband gain may further include: calculating the wideband gain across MFB analysis time slots based on the MFB synthesis signal and the target gain.
[0013] Thus, the optimal broadband gain can be calculated accurately and efficiently.
[0014] In some embodiments, calculating the wideband gain may include optimizing the wideband gain by calculating a least squares solution.
[0015] In some embodiments, the least squares solution may minimize the error between samples of the first audio signal and samples of the second audio signal. The first audio signal may be obtained by performing MFB analysis on the training signal, then performing MFB synthesis, overlap-adding, and applying the target gain. Alternatively, the first audio signal may be obtained by applying the target gain and delaying through the processing delay of the MFB analysis and the MFB synthesis. The second audio signal may be obtained by applying a corresponding (to be determined) wideband gain to a corresponding MFB synthesized signal in each MFB analysis time slot and summing the contributions of all MFB analysis time slots. The first audio signal and the second audio signal may be time domain audio signals.
[0016] In some embodiments, the least squares solution may be a solution based on an objective function depending on a transformation matrix T1 of the plurality of MFB synthesized signals and a target vector t1 depending on the target gain.
[0017] In some embodiments, the transformation matrix T1 can be T1=[w0(n), w1(n), ..., w K-1 (n)], where K is the number of MFB analysis time slots and n indicates the number of samples. The target vector t1 can be given by t1=x2(n)g(nD P ), where x2(n) is a time domain signal that can be obtained by performing MFB analysis on the training signal, followed by MFB synthesis and overlap-addition, and D P is a delay (eg, a processing delay). The least squares solution may be a solution to the equation T1G=t1, where G is a function of G=[G0, G1, ..., G K-1 ] T The broadband gain vector is given by T Indicates transposition. D P There may be a processing delay to perform MFB analysis followed by MFB synthesis and overlap-add.
[0018] In some embodiments, the least squares solution of the broadband gain vector G can be given by Given, where □ -1 Indicates inverse.
[0019] In some embodiments, the training signal may be a random signal or a DC signal. For example, the random signal may be a white noise signal.
[0020] In some embodiments, calculating the broadband gain may be performed iteratively, with each iteration after the first iteration being calculated using a corresponding modified training signal or a corresponding different training signal. The broadband gain in each iteration after the first iteration may be further calculated based on an average of the results of at least one previous iteration. For example, the final result of the broadband gain may be determined by averaging the results of all iterations.
[0021] By repeatedly / iteratively performing the wideband gain determination on different input training signals, better accuracy of the proposed method for a wide range of real-world audio signals can be achieved.
[0022] In some embodiments, determining the wideband gain may include: determining an MFB interpolation prototype filter based on the MFB analysis prototype filter and the MFB synthesis prototype filter. Determining the wideband gain may further include: calculating the wideband gain across MFB analysis time slots based on the MFB interpolation prototype filter and the target gain.
[0023] Thus, the optimal broadband gain can be calculated accurately and efficiently.
[0024] In some embodiments, the MFB interpolation prototype filter may be determined as a product of a mirror image (e.g., a time mirror image) and a shifted version of one of the MFB analysis prototype filter and the MFB synthesis prototype filter and the other of the MFB analysis prototype filter and the MFB synthesis prototype filter. For example, the shifted version may be shifted according to the effective length of the MFB interpolation prototype filter.
[0025] In some embodiments, calculating the wideband gain may include optimizing the wideband gain by calculating a least squares solution.
[0026] In some embodiments, the least squares solution may be a solution to an objective function based on the transformation matrix T2 and a target vector t2 that depends on the target gain.
[0027] In some embodiments, the least squares solution may be a solution based on an objective function that depends on a transformation matrix T2 of the MFB interpolation prototype filter and a target vector t2 that depends on the target gain.
[0028] In some embodiments, the transformation matrix T2 may be a matrix of shifted versions of the MFB interpolation prototype filter, each version being associated with a specific MFB analysis time slot.
[0029] In some embodiments, the transformation matrix T2 can be expressed as T2=[p i (n), p i (nS), ..., p i(n-(K-1)S)], where p i is the MFB interpolation prototype filter, K is the number of MFB analysis slots, n indicates the number of samples, and S is the slot length of the MFB analysis slot. The target vector t2 can be given by Given, where g is the target gain. The least squares solution may be a solution to the equation T2G=t2, where G is given by G=[G0, G1, ..., G K-1 ] T The broadband gain vector is given by T Indicates transpose.
[0030] In some embodiments, the least squares solution of the broadband gain vector G can be given by Given, where □ -1 Indicates inverse.
[0031] In some embodiments, the MFB interpolation prototype filter p i Can be p i (n) = p S (n)p A (Dn) is given by, where p A is the MFB analysis prototype filter, p S is the MFB synthesis prototype filter, and D+1 is the MFB interpolation prototype filter p i The effective length of
[0032] In some embodiments, the method may further comprise determining a set of (related) MFB analysis time slots by identifying a non-constant gain function portion of the target gain packed in time samples and determining associated time slots based on the non-constant gain function portion.
[0033] In some embodiments, the MFB domain may be a Quadrature Mirror Filter (QMF) domain.
[0034] In some embodiments, the method may include applying the determined wideband gain in the MFB domain.
[0035] In some embodiments, the method may include generating a time-domain wideband signal using the determined wideband gain.
[0036] In some embodiments, the method may include limiting (eg, mapping or shortening) the determined wideband gain to within a predetermined range, such as within a range of 0 to 1 (inclusive).
[0037] In some embodiments, the method may further include: decoding the transformed signal in the MFB domain using the wideband gain of each MFB analysis time slot, including fading the audio signal associated with the current parameter set and / or fading the audio signal associated with the previous parameter set. For example, this may correspond to cross-fading the aforementioned audio signal. Specifically, this may correspond to cross-fading the previous and current parameters (parameter sets) to achieve cross-fading of the signals associated with the previous and current parameters.
[0038] According to another aspect, a device for processing audio is provided. The device may include a processor and a memory, the memory being coupled to the processor and storing instructions for the processor. The processor may be configured to perform all steps of the method according to the aforementioned aspect and its embodiments.
[0039] According to another aspect, a computer program is described. The computer program may include executable instructions for performing the method or method steps outlined throughout the present disclosure when executed by a computing device.
[0040] According to a further aspect, a computer readable storage medium is described. The storage medium may store a computer program adapted to be executed on a processor and for performing the method or method steps outlined throughout the present disclosure when executed on the processor.
[0041] It should be noted that the methods and systems as outlined in this disclosure (including preferred embodiments thereof) can be used independently or in combination with other methods and systems disclosed in this document. In addition, all aspects of the methods and systems outlined in this disclosure can be combined arbitrarily. In particular, the features of the claims can be combined with each other in any manner.
[0042] It will be understood that device features and method steps can be interchanged in a variety of ways. In particular, as will be appreciated by the skilled person, details of the disclosed method(s) may be implemented by corresponding devices, and vice versa. Furthermore, any statements above regarding method(s) (and, for example, steps thereof) should be understood to apply equally to corresponding devices (and, for example, blocks, stages, units thereof), and vice versa. BRIEF DESCRIPTION OF THE DRAWINGS
[0043] The present invention is explained below in an exemplary manner with reference to the accompanying drawings, in which:
[0044] Figure 1 is a block diagram schematically illustrating an example of a basic operation of applying a gain in the time domain;
[0045] Figure 2is a block diagram schematically illustrating an example of an alternative operation of performing MFB analysis followed by MFB synthesis and applying a delayed time domain gain;
[0046] Figure 3 is a block diagram schematically illustrating an example of an operation of applying a gain in an MFB domain according to an embodiment of the present disclosure;
[0047] Figure 4 is a flowchart schematically illustrating a method for processing audio according to an embodiment of the present disclosure;
[0048] Figure 5 The present invention schematically illustrates an embodiment of the present invention. Figure 4 A flowchart of an example of an implementation of the steps of the method;
[0049] Figure 6 is a diagram showing an example of a transformation matrix based on a DC input training signal according to an embodiment of the present disclosure;
[0050] Figure 7 The present invention schematically illustrates an embodiment of the present invention. Figure 4 A flowchart of another example of an implementation of the steps of the method;
[0051] Figure 8 is a diagram showing an example of a transformation matrix based on an interpolation prototype filter according to an embodiment of the present disclosure;
[0052] Fig. 9 is a diagram showing an example of a time domain gain function to be modeled; and
[0053] Fig.10 is a block diagram of an example of an apparatus for executing a method according to an embodiment of the present disclosure. DETAILED DESCRIPTION
[0054] The present disclosure describes methods to compute MFB (eg, QMF) domain wideband gains in an optimal manner (eg, in a least squares error sense) to simulate any desired time domain gain.
[0055] Here, MFB refers to any real-valued or (preferably) complex-valued, sub-sampled, M-channel (or frequency band) modulated filter bank. Using this filter bank as an analysis filter bank, the time domain input signal is transformed into the MFB domain (e.g., QMF domain) via MFB analysis (e.g., QMF analysis) to produce an MFB domain signal (e.g., QMF domain signal). After optional signal processing in the MFB domain, the (possibly processed) MFB domain signal is transformed back into the time domain via MFB synthesis using an appropriate synthesis filter bank (e.g., inverse filter bank).
[0056] An example of an MFB is a QMF filterbank. The prototype filters of the analysis filterbank and the synthesis filterbank may be different and may be symmetrical or asymmetrical. The subsampling factor (or step) S is typically equal to M (critical sampling with respect to the amount of complex data) or less than M (oversampling). Typical values of M for perceptual audio coding and processing at a sampling rate of 48 kHz are, for example, 60, 64 or 77. The overall analysis-synthesis system delay D depends on the properties of the prototype filters. With subsampling (S>1), the processing delay D P Reduce to D P = D - S + 1. It is assumed that the MFB (eg, QMF filter bank) has nearly perfect reconstruction properties and can well suppress any aliasing errors due to filter bank domain processing.
[0057] Figure 1 An example of a basic operation of applying a gain (e.g., a cross-fade gain) in the time domain is schematically illustrated. This operation can be performed, for example, on the encoder side to cross-fade between different sets of parameters (e.g., transform parameters). n is understood to denote the number of samples or any other suitable time index. The time domain gain g(n) to be applied is 20 The input signal x(n) 10 is input to a multiplication block 160 where a temporal gain g(n) is applied, for example, in a sample-by-sample manner. The multiplication block 160 outputs an output signal y(n) 170 which is a faded (and delayed) version x(n) of the input signal faded by the temporal gain g(n).
[0058] Figure 2 Schematically illustrates an example of corresponding operations that can be performed, for example, at the decoder side. 10 is input to the MFB analysis block 40 (filter bank analysis, FB-A) for conversion to the MFB domain, thereby generating an MFB domain signal u(c, k) 45. Here, c∈0, ..., M-1 indicates a frequency band or channel, and k indicates an MFB time slot (e.g., a QMF time slot). Optionally, parameters (decoding) or other MFB domain processing may be applied to the MFB domain signal u(c, k). Subsequently, the MFB domain signal u(c, k) is input to the MFB synthesis block 50 (filter bank synthesis, FB-S) for conversion back to the time domain, thereby generating a time domain signal x2(n) 55. For the delayed time domain gain g(n -D P )25 application (which appropriately delays the processing delay of MFB analysis and synthesis D P ), the time domain signal x2(n) is input to a multiplication block 260 where a (delayed) time domain gain g(nD) is applied, for example, in a sample-by-sample manner. PThe multiplication block 260 outputs an output signal y2(n) 270 which is a faded (and delayed) version of the input signal x(n) faded by the time domain gain g(n).
[0059] As mentioned above, applying the (delayed) time domain gain after MFB analysis, potential MFB domain processing and MFB synthesis usually results in additional delay in the audio processing chain (e.g., decoding chain). That is, if additional processing of the cross-faded audio signal in the MFB domain is desired, another MFB analysis-synthesis stage is required after the time domain cross-fade, thereby delaying the further processing of the MFB analysis and synthesis by D. P to the overall delay. In addition, when crossfading two audio signals, Figure 2 The architecture of will need to run two instances of the MFB synthesis block 50 and then perform cross-fading in the time domain, which greatly increases the computational overhead. To address these issues, the present disclosure proposes applying the gain in the MFB domain. Figure 3 An example of the proposed operation of applying a wideband gain G per time slot k in the MFB domain (typically the filter bank domain; eg, the QMF domain) is illustrated in FIG.
[0060] Figure 3 The operations shown in can be performed, for example, on the decoder side. In the decoder, the transmitted / received audio signal can be decoded and subjected to MFB analysis (e.g., QMF analysis). For signal reconstruction, the transmitted parameters can be applied in the MFB domain and the processed MFB domain signals can be combined. The MFB wideband time slot gain can be applied to the parameters to perform a cross-fade from the previous frame parameters to the current frame parameters. For example, in a certain application, there may be 16 time slots per frame, and 16 wideband gains; however, a smaller number (e.g., only 3) of these gains can also be calculated in an optimal manner as described below to cover the desired cross-fade time range.
[0061] exist Figure 3In the embodiment of the present invention, the input signal x(n) 10 is input to the MFB analysis block 40 (filter bank analysis, FB-A) for conversion to the MFB domain, thereby producing the MFB domain signal u(c, k) 45. Then, the MFB domain wideband gain G(k) 30 is applied in a per-time slot manner. Optionally, parameters (decoding) or other MFB domain processing may be applied to the MFB domain signal u(c, k) 45 before or after the gain is applied. Subsequently, the MFB domain signal after the gain is applied is input to the MFB synthesis block 50 (filter bank synthesis, FB-S) for conversion back to the time domain, thereby producing the time domain signal y3(n) 370, which is a faded (and delayed) version of the input signal x(n) faded in and out by applying the MFB domain wideband gain G(k) 30. Here, the application of the MFB domain wideband gain G(k) 30 is intended to emulate (or simulate, implement) the application of a predefined time domain (target) gain g(n).
[0062] This disclosure proposes Figure 3 Two methods for calculating the optimal wideband gain G(k) used in the example processing chain.
[0063] As a general notation, K is used to denote the number of analysis slots of the correlation filter bank. Then, the column vector G containing the unknown (i.e., to be calculated) wideband gain G(k) can be defined as
[0064] G=[G(0)G(1)...G(K-1)] T , (1)
[0066] Among them, T Indicates transpose.
[0067] The general least squares solution to the problem of finding broadband gain can be given by
[0068] G=(T T T) -1 (T T t), (2)
[0070] Where T is a transformation matrix with K columns and t is a target column vector that depends on the target gain function (target gain in the time domain), which will be explained in more detail below for each proposed method.
[0071] The performance can be measured as the energy of the error e(n) = y3(n) - y2(n). The definition of the error intentionally excludes the filter bank reconstruction error and focuses on the modeling of the gain function. See Figure 2 , assuming that the studied MFB has nearly perfect reconstruction properties, the following relationship holds
[0072] x2(n)≈x(nDP ). (3)
[0074] Based on the above content, an example of the method 400 for processing audio according to an embodiment of the present disclosure is as follows: Figure 4 The method 400 includes steps S410 and S420, wherein step S420 is optional. Further, the method 400 mentions the MFB domain, an example of which is the QMF domain.
[0075] exist Step S410 , a modulation filter bank (MFB) domain wideband gain is determined based on a time domain target gain for fading the audio signal. Here, applying the wideband gain in the MFB domain should simulate (i.e., mimic or implement) applying the target gain in the time domain. In other words, the wideband gain can be determined to minimize the error e(n) as defined above, for example in a least squares sense. The (time domain) target gain can be related to a target gain function as a function of time (or sample number or any other suitable time index).
[0076] Determining the wideband gain may include calculating the wideband gain using the target gain, the MFB analysis prototype filter, and the MFB synthesis prototype filter. A corresponding wideband gain G(k) may be calculated for each of a plurality of MFB analysis time slots k. A detailed example will be described below.
[0077] Further, the determination (eg, calculation, operation) of the wideband gain may include optimizing the wideband gain by calculating a least squares solution.
[0078] In some embodiments, the determined wideband gain may be limited (e.g., mapped or shortened) to a predetermined target range before being applied to any audio signal. For example, the determined wideband gain may be limited to a range of 0 (0.0) to 1 (1.0), including the end value.
[0079] exist Step S420 , using the wideband gain of each MFB analysis time slot, decoding the transformed signal in the MFB domain, including fading the audio signal associated with the current parameter set and / or fading the audio signal associated with the previous parameter set. For example, the fading may correspond to cross-fading the audio signal. Specifically, the fading may correspond to cross-fading the previous and current parameters to achieve cross-fading of the signals associated with the previous and current parameters.
[0080] The decoding of this step may involve applying the determined wideband gain in the MFB domain. Additionally or alternatively, the decoding may involve generating a time domain wideband signal using the determined wideband gain.
[0081] Next, an example method of determining the wideband gain will be described in more detail.
[0082] Method 1: Training Signal
[0083] According to a first example approach, an optimum wideband gain may generally be calculated for a "training" signal and applied thereafter.
[0084] The MFB analysis signal (eg, QMF analysis signal) u at time slot k and frequency band c corresponding to the input training signal x(n) can be calculated as
[0085]
[0086] Among them, the real-valued analytical prototype filter p A A modulator m having a length L and a frequency band c c For example, given by the following formula
[0087]
[0088] The MFB synthesized signal (e.g., QMF synthesized signal) w is correlated with the analysis signal u at time slot k k It can be calculated as
[0089]
[0090] Among them, the synthetic prototype filter p S Has finite length.
[0091] The final synthesized signal x2 without filter bank processing (e.g., Figure 2 can be calculated as
[0092]
[0093] For example, the process of Eq. (7) can be called overlap-add.
[0094] The final output signal y3 (e.g., Figure 3 can be calculated as
[0095]
[0096] According to the above definition, the determination of the broadband gain in step S410 of method 400 can be performed by Figure 5 The exemplary method 500 shown in the flowchart of FIG. 500 includes steps S510 to S530.
[0097] exist Step S510 , based on the input training signal x(n) and the MFB analysis prototype filter p A, a corresponding MFB analysis signal u(c, k) is determined for each of a plurality of time slots (MFB analysis time slots) k and each of a plurality of frequency bands c.
[0098] exist Step S520 , based on the MFB analysis signal u(c, k) in the corresponding MFB analysis time slot k and the MFB synthesis prototype filter p S , for each of the plurality of MFB analysis time slots k, a corresponding MFB synthesis signal w is determined k .
[0099] Then, in Step S530 , based on the MFB synthesized signal w k and the target gain g(n), the wideband gain G(k) is calculated across MFB analysis time slot k.
[0100] An example of calculation details in step S530 , which involves optimizing the wideband gain by calculating a least squares solution, is described next.
[0101] With K as the number of relevant analysis time slots, the transformation matrix T1 can be defined as
[0102] T1=[w0(n)w1(n)...w K-1 (n)], (9)
[0104] Where n = 0, ..., (K-1)S+L-1, and the column vector of the broadband gain G can be defined as
[0105] G=[G0 G1 ... G K-1 ] T . (10)
[0107] Finally, the target column vector t1 can be defined as
[0108] t1=y2(n)=x2(n)g(nD P ) (11)
[0110] Here, similarly, n=0, ..., (K-1)S+L-1.
[0111] The broadband gain can then be determined by solving the following equation in a least squares sense,
[0112] T1G=t1. (12)
[0114] For example, the equation can be solved as follows
[0115]
[0116] According to the above example, the least squares solution calculated in step S530 of method 500 can minimize the error between samples of the first audio signal (e.g., the audio signal y2(n) defined in the above Eq. (11)) and samples of the second audio signal (e.g., the audio signal y3(n) defined in the above Eq. (8)). The first audio signal can be obtained by performing MFB analysis on the training signal x(n), then performing MFB synthesis, overlap-adding, and applying the (delayed) target gain g(n). Alternatively, the first audio signal can be obtained by applying the target gain and delaying through the processing delay of the MFB analysis and the MFB synthesis. The second audio signal can be obtained by applying the corresponding wideband gain G(k) to the corresponding MFB synthesized signal w in each MFB analysis time slot k. k , and summing the contributions of all MFB analysis time slots k to obtain. It can be understood that the first audio signal and the second audio signal can be time domain audio signals.
[0117] Further according to the above example, the least squares solution (e.g., the solution defined in Eq. (13) above) can be based on the multiple MFB composite signals w k The solution of the objective function (e.g., the objective function defined in Eq. (12) above) of the transformation matrix T1 and the target vector t1 that depends on the target gain g(n).
[0118] In a specific example, as described above, the transformation matrix T1 can be represented by T1=[w0(n), w1(n), ..., w K-1 (n)], where K is the number of MFB analysis time slots and n indicates the number of samples. Further, the target vector t1 can be given by t1=x2(n)g(nD P ), where x2(n) is a time domain signal that can be obtained by performing MFB analysis on the training signal x(n), followed by MFB synthesis and overlap-addition, and D P is a delay (e.g., a processing delay for performing MFB analysis, then performing MFB synthesis and overlap-add). The least squares solution may be a solution to the equation T1G=t1, where G is given by G=[G0, G1, ..., G K-1 ] T The broadband gain vector is given by T Indicates transposition. Specifically, the least squares solution of the broadband gain vector G can be obtained by Given, where □ -1 Indicates inverse.
[0119] As a variation of the above, the first audio signal may be obtained by applying the target gain and then delaying the MFB delay. Then, in addition to the actual gain target, the optimized wideband gain will also aim to minimize any MFB reconstruction error. In this case, the target vector t1 in Eq. (11) will be replaced by
[0120] t1=[x·g](nD P ). (14)
[0122] It can be seen that a white noise random input signal x(n) produces very good modeling results. However, other input training signals can also be used, such as a constant signal (e.g., a DC signal). The advantage of a constant signal is that the synthetic signal w k are shifted versions of each other. Figure 6 This is shown in the figure of , which shows an example transformation matrix T1 for a constant input S=60 and 3 time slots, and in which the horizontal axis indicates the time slot index and the vertical axis indicates the time sample. However, the error of the constant input may generally be higher than the error of the random input.
[0123] According to these findings, the input training signal used above can be a random signal (eg, white noise) or a DC signal.
[0124] The performance of the proposed method can be further improved by repeating the proposed method several times (eg, typically 2 or 3 times) and averaging the best broadband gains obtained.
[0125] Therefore, the calculation of the broadband gain through the steps of method 500 can be performed iteratively, wherein each iteration after the first iteration is calculated using the corresponding modified training signal or the corresponding different training signal. The broadband gain in each iteration after the first iteration can be further calculated based on the average value (e.g., weighted average value, arithmetic mean value, etc.) of the results of at least one previous iteration. For example, the final result of the broadband gain can be determined by averaging the results of all iterations.
[0126] Method 2: Interpolation Prototype
[0127] According to a second example method, an optimal wideband gain may be calculated using an MFB interpolation prototype filter (eg, QMF interpolation prototype filter) based on an MFB analysis prototype filter (eg, QMF analysis prototype filter) and an MFB synthesis prototype filter (eg, QMF synthesis prototype filter).
[0128] Therefore, the determination of the broadband gain in step S410 of method 400 can be performed by Figure 7 The example method 700 shown in the flowchart of FIG. 700 includes steps S710 and S720.
[0129] exist Step S710 , determining an MFB interpolation prototype filter based on the MFB analysis prototype filter and the MFB synthesis prototype filter.
[0130] Then, in Step S720 , calculating the wideband gain across MFB analysis time slots based on the MFB interpolation prototype filter and the target gain.
[0131] The advantage of this method is that no training data (training signal) is required, and the result depends only on the analysis prototype filter and the synthesis prototype filter. However, this method may ignore some aliasing products in the filter bank processing.
[0132] Next, an example of the computational details in steps S710 and S720 is described, which involves optimizing the wideband gain by computing the least squares solution. Specifically, it can be seen that the output x2(n) of the analysis-synthesis filter bank operation (e.g., Figure 2 can be approximated by the following equation,
[0133]
[0134] And the output signal x3(n) after applying the broadband gain G(k) (e.g., Figure 3 can be approximated as
[0135]
[0136] Among them, the interpolation prototype (interpolation prototype filter) p with an effective length D+1 i Defined as
[0137] p i (n) = P S (n)p A (Dn), (17) Among them, p A is the MFB analysis prototype filter, and p S It is an MFB synthetic prototype filter.
[0139] That is, the MFB interpolation prototype filter p i The MFB interpolation prototype filter may be determined (e.g., calculated) (e.g., in step S710) as the product of one of the MFB analysis prototype filter and the MFB synthesis prototype filter and a mirror image (e.g., time mirror image) and a shifted version of the other of the MFB analysis prototype filter and the MFB synthesis prototype filter. The mirror image and the shifted version may be shifted according to the effective length of the MFB interpolation prototype filter. For example, the MFB interpolation prototype filter pi (n) can be calculated as the MFB synthesis filter p S (n) and MFB analysis filter p A The mirrored and shifted version p of (n) A (Dn) product, calculated as p i (n) = p S (n)p A (Dn), where D+1 is the MFB interpolation prototype filter p i The effective length of
[0140] With K as the number of relevant analysis time slots, the transformation matrix T2 can be defined as
[0141] T2=[p i (n)p i (nS)...p i (n-KS)], (18)
[0143] Where n = 0, ..., (K-1)S+D, and the column vector of the broadband gain G can be defined as
[0144] G=[G0 G1 ... G K-1 ] T . (19)
[0146] Finally, the target column vector t2 can be defined as
[0147]
[0148] For n=0, ..., (K-1)S+L-1.
[0149] Then (e.g., at step S720), the following equation may be solved in a least squares sense,
[0150] T2G=t2. (twenty one) For example, the equation can be solved as follows
[0152]
[0153] According to the above example, the least squares solution calculated in step S720 of method 700 may be a solution of an objective function based on the transformation matrix T2 and the target vector t2 depending on the target gain g(n). Further, the least squares solution may be a solution based on the MFB interpolation prototype filter p iThe transformation matrix T2 of the MFB interpolation prototype filter p(n) and the solution of the objective function of the target vector t2 depending on the target gain g(n). Specifically, the transformation matrix T2 can be the MFB interpolation prototype filter p(n). i A matrix of shifted versions of , each version being associated with a specific MFB analysis time slot k.
[0154] For example, according to the above content, the transformation matrix T2 can be expressed as T2 = [p i (n), p i (nS), ..., p i (n-(K-1)S)], where p i is the MFB interpolation prototype filter, K is the number of MFB analysis slots, n indicates the number of samples, and S is the slot length of the MFB analysis slot. The target vector t2 can be given by Given, where g is the target gain. Then, the least squares solution can be the solution of the equation T2G=t2, where G is given by G=[G0, G1, ..., G K-1 ] T The broadband gain vector is given by T Indicates transposition. The least squares solution of the broadband gain vector G can be obtained by, for example Given, where □ -1 Indicates inverse.
[0155] Figure 8 is a diagram showing an example transformation matrix T2 based on an interpolation prototype for S=60 and 3 time slots. The horizontal axis indicates the time slot index and the vertical axis indicates the time sample.
[0156] Using any of the above methods, the MFB domain wideband gains may be calculated, for example, at decoder initialization. Alternatively, they may be calculated separately from the decoder, and the pre-calculated wideband gains may be stored at the decoder, for example in the form of one or more lookup tables. In yet another embodiment, the wideband gains may be calculated at the encoder side and sent to the decoder side, for example, together with the encoded audio signal.
[0157] Determine the relevant MFB time slot
[0158] One way to determine the relevant time slots for gain modeling may be to identify the portion of the non-constant gain function of the target gain encapsulated by time samples n0 and n1. Then, the associated time slots k0 and k1 (the total number of time slots K is given by K=k1-k0+1) may be calculated as
[0159]
[0160] The objective gain function g(nD P) and the transformation matrix can be restricted accordingly, for the first example method, the restriction is n=k0S...(k1-1)S+L-1, and for the second example method, the restriction is n=k0S...(k1)S+D.
[0161] According to the above content, the above method (e.g., method 400, method 500 or method 700) may further include: determining a set of (related) MFB analysis time slots by identifying a non-constant gain function portion of the target gain encapsulated (e.g., defined, limited) by time samples (e.g., n0 and n1) and determining associated time slots based on the non-constant gain function portion.
[0162] Fig. 9 An example of a gain function corresponding to a time domain target gain for modeling an MFB domain wideband gain (per MFB time slot) is shown in FIG, wherein the horizontal axis indicates time samples and the vertical axis indicates gain values, normalized to values between 0 and 1. In this figure, non-constant gain function portions can be identified as rising portions between constant gain function portions.
[0163] Device for implementing the method according to the present disclosure
[0164] Finally, the present disclosure is also directed to apparatuses (eg, computer-implemented apparatuses) for performing the methods and techniques described throughout this disclosure. Fig.10 An example of such an apparatus 1000 is shown. In particular, the apparatus 1000 includes a processor 1010 and a memory 1020 coupled to the processor 1010. The memory 1020 may store instructions for the processor 1010. Depending on the use case and / or implementation, the processor 1010 may also receive, among other things, suitable input data 1030 (e.g., audio input, time domain target gain, etc.). Depending on the use case and / or implementation, the processor 1010 may be adapted to perform the methods / techniques described throughout this disclosure (e.g., Figure 4 Method 400, Figure 5 Method 500 and / or Figure 7 The method 700 of FIG. 100a and generates corresponding output data 1040 (eg, MBF domain broadband gain or cross-fade audio signal).
[0165] The present disclosure also relates to a corresponding computer program, a computer program product, and a computer-readable storage medium storing such a computer program or computer program product.
[0166] Technical advantages
[0167] The techniques described herein can be applied to efficiently decode transform signals in the QMF domain, which may be necessary for subsequent processing. The input to the encoder may be two or more audio channels that have been reversibly transformed, where the transform varies with the time frame. When switching from the previous time frame to the current time frame, the transform parameters of the transform signal are smoothly cross-faded in the time domain before encoding.
[0168] At the decoder, the encoder process is reversed in the QMF domain by cross-fading the current and previous frame parameter sets using time-varying wideband gains every QMF analysis slot (transformed to the QMF domain, possibly combined with other processing). The present disclosure specifically describes how those QMF domain wideband gains can be advantageously calculated.
[0169] explain
[0170] Aspects of the systems described herein can be implemented in an appropriate computer-based sound processing network environment (e.g., a server or cloud environment) to process digital or digitized audio files. Portions of the adaptive audio system may include one or more networks comprising any desired number of independent machines including one or more routers (not shown) for buffering and routing data transmitted between computers. Such networks may be constructed on a variety of different network protocols and may be the Internet, a wide area network (WAN), a local area network (LAN), or any combination thereof.
[0171] One or more components, blocks, processes or other functional components may be implemented by a computer program executed by a processor-based computing device that controls the system. It should also be noted that any number of combinations of hardware, firmware and / or data and / or instructions embodied in various machine-readable or computer-readable media may be used to describe the various functions disclosed herein from the perspective of their behavior, register transfers, logic components and / or other features. Computer-readable media that may embody such formatted data and / or instructions include, but are not limited to, various forms of physical (non-transitory), non-volatile storage media, such as optical, magnetic or semiconductor storage media.
[0172] In particular, it should be understood that embodiments may include hardware, software, and electronic components or modules, and for discussion purposes, may be illustrated and described as if most components were implemented solely in hardware. However, one of ordinary skill in the art, and based on a reading of this detailed description, will recognize that in at least one embodiment, the electronic-based aspects may be implemented in software (e.g., stored on a non-transitory computer-readable medium) that is executable by one or more electronic processors, such as a microprocessor and / or an application specific integrated circuit (“ASIC”). Therefore, it should be noted that embodiments may be implemented using a plurality of hardware and software-based devices and a plurality of different structural components. For example, the above-mentioned Figure 1 , Figure 2 , Figure 3 The systems, encoders, decoders, or blocks described in the context of and / or Figure 19 may include one or more electronic processors, one or more computer-readable media modules, one or more input / output interfaces, and various connections (e.g., a system bus) connecting the various components.
[0173] Although one or more embodiments have been described by way of example and with respect to specific embodiments, it should be understood that one or more embodiments are not limited to the disclosed embodiments. On the contrary, it is intended to cover various modifications and similar arrangements that are obvious to those skilled in the art. Therefore, the scope of the appended claims should be given the broadest interpretation so as to cover all such modifications and similar arrangements.
[0174] Likewise, it should be understood that the phraseology and terminology used herein is for descriptive purposes and should not be regarded as limiting. The use of "including," "comprising," or "having" and variations thereof are meant to encompass the items listed thereafter and equivalents thereof as well as additional items. Unless otherwise specified or limited, the terms "mounted," "connected," "supported," and "coupled" and variations thereof are used broadly and encompass direct and indirect mountings, connections, supports, and couplings.
[0175] Enumerated example embodiments
[0176] Various aspects and embodiments of the present disclosure may also be understood from the following enumerated example embodiments (EEE), which are not claims.
[0177] EEE1. A method for processing audio, the method comprising: determining a broadband gain in an MFB domain of a modulation filter bank for fading in and out an audio signal according to a target gain in a time domain, so that applying the broadband gain in the MFB domain simulates applying the target gain in the time domain,
[0178] Wherein, determining the wideband gain includes calculating the wideband gain using the target gain, an MFB analysis prototype filter, and an MFB synthesis prototype filter.
[0179] EEE2. The method according to EEE1, wherein a corresponding wideband gain is calculated for each of a plurality of MFB analysis time slots.
[0180] EEE3. A method according to any one of the preceding EEEs, wherein calculating the broadband gain comprises optimizing the broadband gain by calculating a least squares solution.
[0181] EEE4. The method according to EEE1, wherein determining the broadband gain comprises:
[0182] determining a corresponding MFB analysis signal for each of a plurality of MFB analysis time slots and each of a plurality of frequency bands based on an input training signal and the MFB analysis prototype filter;
[0183] determining a corresponding MFB synthesis signal for each of the plurality of MFB analysis time slots based on the MFB analysis signal in the corresponding MFB analysis time slot and the MFB synthesis prototype filter; and
[0184] The wideband gain is calculated across MFB analysis time slots based on the MFB composite signal and the target gain.
[0185] EEE5. The method according to EEE4, wherein calculating the broadband gain includes optimizing the broadband gain by calculating a least squares solution.
[0186] EEE6. A method according to EEE5, wherein the least squares solution minimizes the error between samples of a first audio signal and samples of a second audio signal, the first audio signal can be obtained by performing MFB analysis on the training signal, then performing MFB synthesis, overlap-addition and applying the target gain, or by applying the target gain and delaying through processing delays of MFB analysis and MFB synthesis, and the second audio signal can be obtained by applying a corresponding wideband gain to a corresponding MFB synthesized signal in each MFB analysis time slot and summing the contributions of all MFB analysis time slots.
[0187] EEE7. A method according to EEE5 or EEE6, wherein the least squares solution is a solution based on an objective function depending on a transformation matrix T1 of the plurality of MFB synthesized signals and a target vector t1 depending on the target gain.
[0188] EEE8. The method according to EEE7, wherein the transformation matrix T1 is T1=[w0(n), w1(n), ..., wK-1 (n)], where K is the number of MFB analysis slots and n indicates the number of samples, and the target vector t1 is given by t1=x2(n)g(nD P ), where x2(n) is a time domain signal that can be obtained by performing MFB analysis on the training signal, followed by MFB synthesis and overlap-addition, and D P is delayed; and
[0189] The least squares solution is the solution of the equation T1G=t1, where G is G=[G0, G1, ..., G K-1 ] T The broadband gain vector is given by T Indicates transposition. EEE9. The method according to EEE8, wherein the least squares solution of the broadband gain vector G is given by Given, where □ -1 Indicates inverse.
[0190] EEE10. A method according to any one of EEE4 to EEE9, wherein the training signal is a random signal or a DC signal.
[0191] EEE11. A method according to any one of EEE4 to EEE10, wherein the calculation of the broadband gain is performed iteratively, and each iteration after the first iteration is calculated using a corresponding modified training signal or a corresponding different training signal.
[0192] EEE12. The method according to EEE1, wherein determining the broadband gain comprises:
[0193] Determine an MFB interpolation prototype filter based on the MFB analysis prototype filter and the MFB synthesis prototype filter; and
[0194] The wideband gain is calculated across MFB analysis time slots based on the MFB interpolation prototype filter and the target gain.
[0195] EEE13. The method according to EEE12, wherein the MFB interpolation prototype filter is determined as the product of one of the MFB analysis prototype filter and the MFB synthesis prototype filter and a mirror (e.g., time mirror) and shifted version of the other of the MFB analysis prototype filter and the MFB synthesis prototype filter. EEE14.
[0196] EEE14. A method according to EEE12 or EEE13, wherein calculating the broadband gain includes optimizing the broadband gain by calculating a least squares solution.
[0197] EEE15. A method according to EEE14, wherein the least squares solution is a solution to an objective function based on a transformation matrix T2 and a target vector t2 that depends on the target gain.
[0198] EEE16. A method according to EEE14, wherein the least squares solution is a solution based on an objective function that depends on a transformation matrix T2 of the MFB interpolation prototype filter and a target vector t2 that depends on the target gain. EEE16. A method according to EEE14, wherein the least squares solution is a solution based on an objective function that depends on a transformation matrix T2 of the MFB interpolation prototype filter and a target vector t2 of the target gain.
[0199] EEE17. A method according to EEE15 or EEE16, wherein the transformation matrix T2 is a matrix of shifted versions of the MFB interpolation prototype filter, each version being associated with a specific MFB analysis time slot. EEE17.
[0200] EEE18. A method according to any one of EEE15 to EEE17, wherein the transformation matrix T2 is T2 = [p i (n), p i (nS), ..., p i (n-(K-1)S)], where p i is the MFB interpolation prototype filter, K is the number of MFB analysis slots, n indicates the number of samples, and S is the slot length of the MFB analysis slot, and the target vector t is given by Given, where g is the target gain; and
[0201] The least squares solution is the solution of the equation T2G=t2, where G is G=[G0, G1, ..., G K-1 ] T The broadband gain vector is given by T Indicates transpose.
[0202] EEE19. The method according to EEE18, wherein the least squares solution of the broadband gain vector G is given by Given, where □ -1 Indicates inverse.
[0203] EEE20. The method according to EEE18 or EEE19, wherein the MFB interpolation prototype filter p i By p i (n) = p S (n)p A (Dn) is given by, where p A is the MFB analysis prototype filter, p S is the MFB synthesis prototype filter, and D+1 is the MFB interpolation prototype filter p i The effective length of
[0204] EEE21. The method according to any one of the preceding EEEs, further comprising determining a set of MFB analysis time slots by identifying a non-constant gain function portion of the target gain encapsulated in time samples and determining associated time slots based on the non-constant gain function portion. EEE22. The method according to any one of the preceding EEEs, further comprising determining a set of MFB analysis time slots by identifying a non-constant gain function portion of the target gain encapsulated in time samples and determining associated time slots based on the non-constant gain function portion.
[0205] EEE22. A method according to any one of the preceding EEEs, wherein the MFB domain is an orthogonal mirror filter QMF domain.
[0206] EEE23. A method according to any one of the preceding EEEs, comprising applying the determined wideband gain in the MFB domain.
[0207] EEE24. A method according to any one of the preceding EEEs, comprising generating a time-domain wideband signal using the determined wideband gain.
[0208] EEE25. A method according to any one of the preceding EEEs, comprising limiting the determined broadband gain to a range of 0 to 1 including end values.
[0209] EEE26. The method according to any one of the preceding EEEs, further comprising:
[0210] The transformed signal in the MFB domain is decoded using the wideband gain of each MFB analysis slot, including cross-fading an audio signal associated with a current parameter set and / or cross-fading an audio signal associated with a previous parameter set.
[0211] EEE27. An apparatus comprising a processor and a memory, the memory being coupled to the processor and storing instructions for the processor, wherein the processor is adapted to execute a method according to any one of EEE1 to EEE26.
[0212] EEE28. A program comprising instructions which, when executed by a processor, cause the processor to perform a method according to any one of EEE1 to EEE26.
[0213] EEE29. A computer-readable storage medium storing the program according to EEE28.
[0214] EEE30. A method for processing audio, the method comprising:
[0215] Calculating a wideband gain of the training signal, including calculating a modulation filter bank (MFB) analysis signal and an MFB synthesis signal related to the MFB analysis signal; and
[0216] The transformed signal is decoded in the MFB domain including cross-fading the current and previous parameter sets using a wideband gain per MFB analysis slot.
[0217] EEE31. The method according to EEE30, wherein the training signal is a random signal or a DC signal.
[0218] EEE32. The method according to EEE30, comprising determining the MFB analysis time slot by identifying a non-constant gain function portion of a time sample package and determining an associated time slot based on the non-constant gain function. EEE33.
[0219] EEE33. A method according to EEE30, wherein the calculation of the broadband gain is performed iteratively, and each iteration after the first iteration is calculated using the corresponding modified training signal or the average of the results of at least one previous iteration.
[0220] EEE34. A method for processing audio, the method comprising:
[0221] Calculate broadband gain, including:
[0222] Calculating an interpolated prototype filter based on one or more filter bank prototype filters; and
[0223] approximating wideband gains of a modulation filter bank (MFB) analysis signal and an MFB synthesis signal across time slots based on the interpolated prototype filter; and
[0224] The transformed signal is decoded in the MFB domain including cross-fading the current and previous parameter sets using a wideband gain per MFB analysis slot.
[0225] EEE35. A method according to any one of EEE30 to EEE34, comprising applying the broadband gain in a filter bank domain.
[0226] EEE36. A method according to any one of EEE30 to EEE34, comprising generating the target time-domain wideband signal using the wideband gain.
[0227] EEE37. A method according to any one of EEE30 to EEE36, wherein calculating the broadband gain includes optimizing a gain function by calculating a least squares solution.
[0228] EEE38. A method according to EEE37, wherein the least squares solution is a solution to an objective function based on a transformation matrix T and an objective vector t that depends on an objective gain function.
[0229] EEE39. A method according to EEE38, wherein the matrix T is a matrix of shifted versions of interpolated prototypes or filter combinations of data associated with a specific analysis time slot.
[0230] EEE40. A system comprising one or more processors configured to perform the operations described in any one of EEE30 to EEE39.
[0231] EEE41. A computer program product configured to cause one or more processors to perform the operations described in any one of EEE30 to EEE39.
Claims
1. A method for processing audio, the method comprising: Determine a broadband gain of a modulation filter bank MFB domain for fading in and out an audio signal according to a target gain in the time domain, so that applying the broadband gain in the MFB domain simulates applying the target gain in the time domain, Wherein, determining the wideband gain includes calculating the wideband gain using the target gain, an MFB analysis prototype filter, and an MFB synthesis prototype filter.
2. The method according to claim 1, wherein: A corresponding wideband gain is calculated for each of the plurality of MFB analysis time slots.
3. A method according to any one of the preceding claims, wherein: Calculating the wideband gain includes optimizing the wideband gain by calculating a least squares solution.
4. The method according to claim 1, wherein: Determining the broadband gain includes: determining a corresponding MFB analysis signal for each of a plurality of MFB analysis time slots and each of a plurality of frequency bands based on an input training signal and the MFB analysis prototype filter; determining a corresponding MFB synthesis signal for each of the plurality of MFB analysis time slots based on the MFB analysis signal in the corresponding MFB analysis time slot and the MFB synthesis prototype filter; and The wideband gain is calculated across MFB analysis time slots based on the MFB composite signal and the target gain.
5. The method according to claim 4, wherein: Calculating the wideband gain includes optimizing the wideband gain by calculating a least squares solution.
6. The method according to claim 5, wherein: The least squares solution minimizes an error between samples of a first audio signal and samples of a second audio signal, the first audio signal being obtainable by performing an MFB analysis on the training signal, then performing an MFB synthesis, overlap-adding, and applying the target gain, or by applying the target gain and delaying through a processing delay of the MFB analysis and the MFB synthesis, and the second audio signal being obtainable by applying a corresponding wideband gain to a corresponding MFB synthesized signal in each MFB analysis time slot and summing contributions of all MFB analysis time slots.
7. The method according to claim 5 or 6, wherein: The least square solution is a solution based on a target function that depends on a transformation matrix T1 of the plurality of MFB synthesized signals and a target vector t1 that depends on the target gain.
8. The method according to claim 7, wherein: The transformation matrix T1 is T1=[w0(n), w1(n), ..., w K-1 (n)], where K is the number of MFB analysis slots and n indicates the number of samples, and the target vector t1 is given by t1=x2(n)g(nD P ), where x2(n) is a time domain signal that can be obtained by performing MFB analysis on the training signal, then performing MFB synthesis and overlap addition, and D P is delayed; and The least squares solution is the solution of the equation T1G=t1, where G is G=[G0, G1, ..., G K-1 ] T The broadband gain vector is given by Indicates transpose.
9. The method according to claim 8, wherein: The least squares solution of the broadband gain vector G is given by Given, where Indicates inverse.
10. The method according to any one of claims 4 to 9, wherein: The training signal is a random signal or a constant signal.
11. The method according to any one of claims 4 to 10, wherein: The calculation of the wideband gain is performed iteratively, and each iteration after the first iteration is calculated using a corresponding modified training signal or a corresponding different training signal.
12. The method according to claim 1, wherein: Determining the broadband gain includes: Determine an MFB interpolation prototype filter based on the MFB analysis prototype filter and the MFB synthesis prototype filter; and The wideband gain is calculated across MFB analysis time slots based on the MFB interpolation prototype filter and the target gain.
13. The method according to claim 12, wherein: The MFB interpolation prototype filter is determined as a product of one of the MFB analysis prototype filter and the MFB synthesis prototype filter and a mirrored and shifted version of the other of the MFB analysis prototype filter and the MFB synthesis prototype filter.
14. The method according to claim 12 or 13, wherein: Calculating the wideband gain includes optimizing the wideband gain by calculating a least squares solution.
15. The method according to claim 14, wherein: The least squares solution is a solution to an objective function based on the transformation matrix T2 and a target vector t2 that depends on the target gain.
16. The method according to claim 14, wherein: The least squares solution is a solution based on an objective function that depends on a transformation matrix T2 of the MFB interpolation prototype filter and a target vector t2 that depends on the target gain.
17. The method according to claim 15 or 16, wherein: The transformation matrix T2 is a matrix of shifted versions of the MFB interpolation prototype filter, each version being associated with a specific MFB analysis time slot.
18. The method according to any one of claims 15 to 17, wherein: The transformation matrix T2 is given by T2=[p i (n), p i (nS), ..., p i (n-(K-1)S)] is given by Among them, p i is the MFB interpolation prototype filter, K is the number of MFB analysis slots, n indicates the number of samples, and S is the slot length of the MFB analysis slot, and the target vector t2 is given by Given, where g is the target gain; and The least squares solution is the solution of the equation T2G=t2, where G is G=[G0, G1, ..., G K-1 ] T The broadband gain vector is given by Indicates transpose.
19. The method according to claim 18, wherein: The least squares solution of the broadband gain vector G is given by Given, where Indicates inverse.
20. The method according to claim 18 or 19, wherein: The MFB interpolation prototype filter p i By p i (n) = p S (n)p A (Dn) is given by, where p A is the MFB analysis prototype filter, p S is the MFB synthesis prototype filter, and D+1 is the MFB interpolation prototype filter p i The effective length of 21. A method according to any preceding claim, further comprising determining a set of MFB analysis time slots by identifying a non-constant gain function portion of the target gain packed in time samples and determining associated time slots based on the non-constant gain function portion.
22. A method according to any one of the preceding claims, wherein: The MFB domain is the quadrature mirror filter QMF domain.
23. A method according to any preceding claim, comprising applying the determined wideband gain in the MFB domain.
24. A method according to any preceding claim, comprising generating a time domain wideband signal using the determined wideband gain.
25. A method according to any preceding claim, comprising constraining the determined wideband gain to be within the range of 0 to 1 inclusive.
26. The method according to any one of the preceding claims, further comprising: The transformed signal in the MFB domain is decoded using the wideband gain of each MFB analysis slot, including cross-fading an audio signal associated with a current parameter set and / or cross-fading an audio signal associated with a previous parameter set.
27. An apparatus comprising a processor and a memory, the memory being coupled to the processor and storing instructions for the processor, wherein: The processor is adapted to perform a method according to any one of claims 1 to 26.
28. A program comprising instructions which, when executed by a processor, cause the processor to perform the method according to any one of claims 1 to 26.
29. A computer-readable storage medium storing the program according to claim 28.
