Optimization process to reduce the number of channels in a stereo audio signal

Abstract

The present invention relates to a method for reducing the number of channels in a stereo signal in order to obtain a monaural signal, the method comprising: an application (610, 611) of two (T1 or T2, T3) channel reduction processing operations to the current frame of the signal, wherein one of the two processing operations uses re-phase filtering on at least one channel of the stereo signal, and the other processing operation does not use re-phase filtering; and a selection (612) of one of the processing operations to be applied to the current frame, the selection of which depends on the presence or absence of at least one transition in the stereo signal, or the quality level of filtering of the stereo signal used when performing the processing operation by using filtering of the stereo signal. The present invention also relates to a processing device for reducing channels to implement the method.

Description

Technical Field

[0001] The present invention relates to the general field of processing audio signals. In particular, the present invention relates to the downmixing of multi-channel audio signals. There is particular interest in the downmixing of stereo signals to monaural signals.

[0002] This type of processing is generally applicable in the field of audio technology, more specifically in the field of audio coding, regardless of whether it is in an encoding process or a decoding process.

Background Art

[0003] The essence of downmixing lies in inferring a signal y composed of a smaller number D of channels based on a combination of C channels of a multi-channel signal x. In practice, this is

Number

[0004] where D < C, and n represents the time index of the input and output signals.

[0005] These stereo signals can originate from a pair of stereo microphones or binaural capture, or even from the artistic mixing of audio tracks.

[0006] There are a certain number of pairs of microphones that enable the generation of stereo content. Among the most popular ones, an XY pair consisting of two cardioid microphones presenting an orientation angle difference of 90° to 135° can be referred to. The MS (mid-side) pair is another very broad pair composed of a cardioid directional microphone and a second microphone, called an "8-shaped" microphone oriented at 90°. By combining these two microphones (sum / difference), a left stereo content channel and a right stereo content channel can be generated. These pairs are called coincident pairs; in other words, the microphone capsules do not present a delay between them, and the spatial arrangement is perceived by the difference in acoustic intensity between the left channel and the right channel.

[0007] Another category of pairs called phase stereo pairs essentially uses microphones that are far from each other and thus generate a phase difference between the channels of a source located closer to one of the microphones. The most well-known one is the AB pair that utilizes two omnidirectional microphones separated by a few centimeters to a few meters. Another very broad pair is the ORTF pair that utilizes both a phase difference through a 17-cm spacing of the microphones and an amplitude difference through a cardioid directivity (the microphones present an angle of approximately 90°). The binaural pair essentially consists of placing omnidirectional microphones inside the ears of a human or an artificial head: this type of device enables the natural generation of binaural content that can be listened to via a headset.

[0008] Stereo content also includes speech signals and audio signals resulting from audio mixing or post-audio generation (e.g., channels stored on a CD or DVD or multicast over the Internet).

[0009] There are also other ways to generate stereo content that are not examined here.

[0010] The simplest way to generate a signal that is reduced by downmixing is called passive downmixing. Passive downmixing is

Number

Number

Number

Number

Number

Number

[0011] This extremely simple passive downmixing method works well when the right and left channels are in phase. However, when the microphones are not identical and exhibit a phase difference, such mixing produces comb filtering, which is revealed by the coloration of the original signal. This is due to the fact that, depending on the spacing of the microphones, the position of the source relative to the pair of microphones, and frequency, the left and right channels of the same source can be either in phase as depicted in Figure 1a or out of phase as depicted in Figure 1b. Therefore, when downmixing, the left and right signals either add (constructive interference) or cancel each other out (cancellative interference), thus producing variations in the level and effect of the coloration of the reduced signal m(t). The coloration arises from the effect of comb filtering: the effects of constructive and cancelative interference are evident in various frequency bands and thus modify the balance and, consequently, the tone of the signal resulting from the downmixing of the original signal.

[0012] To compensate for this flaw in intensity levels across frequencies, the e-AAC+ codec's downmixing performs level correction γ(k) to ensure that the reduced signal energy level remains equivalent to the original signal in each frequency band.

number

number

[0013] This technique allows for some degree of compensation for the drop in intensity caused by downmixing. However, this compensation can lead to over-amplification when the signal is close to out of phase, and in practice, the correction factor γ(k) is limited to its maximum value (e.g., an upper limit of 2).

[0014] Figure 7 illustrates how stereo channels are reduced to mono, which are integrated into a publicly available candidate IVAS codec available at address: https: / / forge.3gpp.org / rep / ivas-codec-pc / ivas-codec / - / tree / main.

[0015] A signal x(m) with two input channels (where m is an index of interleaved samples) is deinterleaved to discover both the left and right channels (block 701). Next, frequency analysis by windowing and Fourier transform (blocks 702 and 703) is performed to obtain the spectra of the two channels, and cross-correlation based on the phase spectra is estimated before determining the time difference between the channels (here, formally ICTDs but labeled ITDs between the two channels) by searching for peaks (block 705); block 705 also determines the correlation level R corresponding to the ITD. * To provide.

[0016] Next, the mixing coefficient g is determined as follows (block 706):

number

[0017] Next, the energies of the left channel x1(n) and right channel x2(n) of the mono signal m(n) are determined (blocks 708, 709, and 710). The left and right channels are separately reinjected (added) into the mono signal (which originates from block 706, depending on the energy compensation determined in block 711, based on the respective scale factors defined in blocks 712 and 713) in blocks 713 and 715.

[0018] These methods suffer from signal coloration resulting from downmixing (related to imperfectly compensated comb filtering), as well as lingering reverberation linked to the sum of signals when input channels are out of phase, and "click" artifacts when channels are reinjected in time within a single frame. Compensating for delays that are identical across all frequencies is impractical because this delay is related to the various sources that make up the content. In practice, this leads to additional coloration, as well as a potential loss of source clarity that defect level compensation cannot compensate for.

[0019] Other downmixing techniques attempt to avoid the aforementioned shortcomings. The principle of these methods is to re-phase the left and right signals before they are added together. In a non-patent document titled "A stereo to mono downmixing scheme for MPEG-4 parametric stereo encoder" (Non-patent document 1),

number

[0020] The ideal phase shift is referred to here as IPD (inter-channel phase difference),

number

[0021] This is calculated in the frequency domain: this allows for re-phase alignment of the spectral components of various sources, where a source with its own IPD may be dominant in frequency band k, while another source with a different IPD may be dominant at a different frequency k'.

[0022] In the case of the MPEG4 codec described in the Samsudin non-patent document cited above, the applicable IPD is estimated by averaging across the Bark bandwidth, rather than being calculated for each frequency band, the latter of which has been found to be particularly noisy and variable from one frame to the next. In addition, this method takes the left channel as the phase reference, and if the phase of this channel is poor, the downmixing will have degraded quality.

[0023] (Patent Document 1) proposes a downmixing method that combines the method proposed by Samsudin with passive downmixing. In particular, it proposes an ISD (inter-spectral distance) indicator that makes it possible to select which method is most suitable for re-phasering the signal for each frequency band:

number

[0024] This indicator makes it possible to measure whether the signals are in phase (ISD<1) (i.e., interval).

number

number

number

[0025] The passive downmixing proposed in low ISD situations (Patent Document 1) sometimes introduces increased reverberation.

[0026] Furthermore, the method proposed in this patent application requires the application of processing operations in the frequency domain, and the reduced signal is reconstructed based on the inverse short-time Fourier transform (STFT) of the signal M(t,k). This type of filtering generates cyclic convolutions that produce audible artifacts (echoes, pre-echoes, and static). To mask these artifacts, implementations with frame overlaps associated with suitable windowing (analysis and / or synthesis windows) can be used to ensure the reconstruction of the filtered signal. Generally, this overlap is incompatible with the operation of audio encoders working with adjacent frames (i.e., without overlap); in addition, this type of reconstruction results in a delay of generally 1 / 2 frame (conventional overlap) which is unacceptable for encoders with strong latency constraints.

[0027] Another overlap-add (OLA) implementation is possible: this method requires zero-padding the signal and filter to avoid cyclic convolution. However, several tests have shown that "with such filters whose phase changes very rapidly from one frame to the next, the OLA method does not allow for complete masking of artifacts in transitions between several frames, and artifacts remain audible."

[0028] Even in conventional methods where downmixing is achieved by switching (for example, in the frequency domain of Patent Document 1), different downmixing methods exist. In this case, it is important to ensure that the switching occurs "seamlessly" (i.e., without any difference in level between methods for avoiding discontinuities or artifacts). [Prior art documents] [Patent Documents]

[0029] [Patent Document 1] International Publication No. 2017103418 [Non-patent literature]

[0030] [Non-Patent Document 1] “A stereo to mono downmixing scheme for MPEG-4 parametric stereo encoder” by Samsudin,E.Kurniawati,N.Boon Poh,F.Sattar,S.George,in Proc.ICASSP,2006 [Overview of the project] [Problems that the invention aims to solve]

[0031] The present invention aims to improve upon the prior art. [Means for solving the problem]

[0032] To achieve this objective, the present invention targets a method for downmixing a stereo signal to obtain a mono signal, the method being described below: - Applying two downmixing operations to the current frame of the signal, wherein one downmixing operation uses phase realignment filtering on at least one channel of the stereo signal channels, and the other downmixing operation does not use phase realignment filtering. - Selecting one of the processing operations to be applied to the current frame, the following: -- Presence or absence of at least one transition in the stereo signal, or --The quality level of stereo signal filtering used when performing processing operations by using stereo signal filtering. The choice and the implementation depend on the choice and Includes.

[0033] Advantageously, the present invention provides two criteria that allow selection between two different downmixing operations, one of which uses filtering of the stereo signal and the other which does not.

[0034] In one embodiment, if the presence of at least one transition is detected in the stereo signal, or if it is not detected, a downmixing operation without filtering of the stereo signal or a downmixing operation with filtering of the stereo signal is selected, respectively.

[0035] This embodiment is as follows: -When one or more transitions are detected, a better rendering of the one or more transitions in the signal is obtained by selecting a downmixing operation that does not use filtering and, as a result, preserves the properties of the stereo signal. - If no transitions are detected, a downmixing operation is preferred that uses filtering better suited to the non-in-phase stereo signal components. This makes it possible.

[0036] In one embodiment, the at least one transition is detected before one of the two downmixing operations is applied to the stereo signal.

[0037] This embodiment allows for the prior inference that at least one transition is present or absent in the stereo signal, even before a downmixing operation is applied, without any additional processing operation being applied to the signal.

[0038] In one embodiment, the detection of the at least one transition involves the following steps: - The process of breaking down the current frame of a stereo signal into subblocks; - The process of calculating the energy of each subblock obtained from the subblocks; - Acquired energy and energy threshold (th 61 The process of comparing with ) and - A process in which the presence of a transition is not detected or is detected, respectively, if the energy is lower or higher than the energy threshold. Includes.

[0039] In one embodiment, the filtering quality level is compared to a filtering quality threshold, and if the filtering quality level is less than or greater than the filtering quality threshold, a downmixing operation without filtering of the stereo signal or a downmixing operation with filtering of the stereo signal is selected.

[0040] According to one embodiment, the filtering quality level is as follows: - Before filtering is applied, calculate the energy of the stereo signal, called the input energy. - After filtering is applied, calculate the energy of the monaural signal, called the output energy. - Calculate the ratio of input energy to output energy. It is measured by [method].

[0041] In one embodiment, an indicator is generated that indicates whether or not a transition has occurred.

[0042] In one embodiment, the two levels of the processed current frame are obtained after two downmixing operations have been applied to the current frame of the signal, one using stereo signal filtering and the other not. - Adjusting the level of the current frame to which the downmixing operation has been applied by using the filtering, such that the adjusted level is similar to or equal to the level of the current frame to which the downmixing operation has been applied without the filtering; - Comparing the processed current frame having the adjusted level with the processed current frame having the level obtained after a downmixing operation without filtering of the stereo signal is applied to the current frame of the signal, -Includes selecting one of the processing operations to be applied to the current frame based on the comparison.

[0043] Advantageously, the present invention makes it possible to maintain the level of the stereo signal as stably as possible without causing discontinuities in the perceived level of the signal when the downmixing operation with respect to the current frame of the stereo signal is changed.

[0044] The present invention targets a downmixing device that includes a processing circuit for carrying out the steps of the downmixing method described above.

[0045] The present invention also relates to a computer program that, when executed by a processor, includes instructions for performing a downmixing method according to the present invention according to any one of the particular embodiments described above.

[0046] Such instructions can be permanently stored in a non-temporary memory means of a downmixing device that performs a downmixing method according to the present invention.

[0047] This program may use any programming language and may be in the form of source code, object code, or intermediate code between source code and object code (e.g., special compilation format or any other preferred format).

[0048] The present invention also targets computer-readable storage media or information media containing computer program instructions as described above.

[0049] The storage medium can be any entity or device capable of storing a program. For example, this medium may include ROM (e.g., CD-ROM, or micro-electronic circuit ROM), or indeed, storage means such as magnetic storage means (e.g., mobile media, hard disk, or SSD).

[0050] Furthermore, the storage medium may be a transmittable medium such as an electrical or optical signal that can be delivered via an electrical cable or optical cable, wirelessly, or by other means, so that the computer program contained therein can be executed remotely. The program according to the present invention may, in particular, be downloaded from a network such as the Internet.

[0051] Alternatively, the storage medium may be an integrated circuit into which the program is taken up and which is adapted to be executed or used when the aforementioned downmixing method is performed.

[0052] In one embodiment, the technology is implemented by software components and / or hardware components. With this in mind, the terms “device” or “module” may, in this specification, be equivalent to a software component, a hardware component, or a set of hardware and software components.

[0053] Other features and advantages of the present invention will become clearer by reading the following description of the specific embodiments provided merely as illustrative and non-limiting examples and accompanying drawings. [Brief explanation of the drawing]

[0054] [Figure 1a] This shows the channels of the in-phase stereo signals described above. [Figure 1b] This shows the channels of the inverted-phase stereo signal described above. [Figure 2] The downmixing sequence in one embodiment of the present invention is shown in block diagram form. [Figure 3] This document presents one embodiment of selecting a downmixing operation, in addition to calculating the filter for downmixing. [Figure 4] This shows one embodiment of the adaptation stage for the downmixing filter. [Figure 5] This illustrates one embodiment of the transition of downmixing during the transition between two frames of a stereo signal. [Figure 6a] This document presents not only the filter calculation for downmixing but also another embodiment of the selection of the downmixing operation. [Figure 6b] This document presents not only the filter calculation for downmixing but also another embodiment of the selection of the downmixing operation. [Figure 7] The existing downmixing method described above is shown below. [Figure 8] An example of a structural embodiment of a downmixing device according to one embodiment of the present invention is shown. [Modes for carrying out the invention]

[0055] The present invention will now be described below with reference to Figures 2-8. To achieve this objective, the following symbols will be primarily used in the following specification.

[0056] index k frequency index l Temporal indicator of the signal (within the current frame) n is a temporal index of the signal. t Frame number

[0057] constant L Number of samples in the current frame N is the number of input channels. Number of output channels (M) P filter length Q: Crossfade length

[0058] signal x i (n) Input channel m(n) Downmixing (in the time domain) m(t,l) Downmixing within the current frame (in the time domain) M(t,k) Downmixing within the current frame (in the frequency domain)

[0059] Figure 2 shows an example sequence for processing a stereo audio signal (including processing operations such as downmixing).

[0060] At the input to this processing sequence, a stereo signal x consisting of two channels (x1 and x2) (also called the left channel and the right channel, respectively) is divided in the first step into frames of L samples (x1(t,l) and x2(t,l)) in 201, where t is the frame index and l is the sample index. Block 202 applies windowing and an FFT (Fast Fourier Transform) to obtain the signal in the frequency domain (X1(t,k) and X2(t,k)), where k is the frequency index. In 203, a downmixing filter to be applied to the signal frames is selected and determined. This step will be explained with reference to Figure 3. The filter thus determined is conditioned (or adapted) in 204 to optimize the filter in order to make it causal and to reduce processing complexity. In one embodiment, the windowing and FFT are similar to steps 702 and 703 in Figure 7.

[0061] This adaptation stage is explained with reference to Figure 4.

[0062] Once the filter is determined and adapted for the current frame, it is applied to this current frame at 205. To avoid audible artifacts caused by filter changes between two frames, a crossfade is applied to the filter of the previous frame.

number

[0063] This process will be explained with reference to Figure 5.

[0064] Figure 3 illustrates a detailed embodiment of block 203 in Figure 2. In this processing block, a first or second downmixing operation is selected and determined to be applied to the current frame of the stereo signal.

[0065] The first downmixing method T1 is defined here by simultaneously shifting the left and right channels of the stereo signal by an angle defined by half the phase difference (IPD) determined between the two channels of the stereo signal. Therefore, the following downmixed signal exists:

number

[0066] This method makes it possible to avoid the type of degradation caused by reverberation excess generated by conventional downmixing (averaging two signals) or the type of degradation caused by comb filtering in the case of a time shift between two channels.

[0067] This downmixing, defined in this way, can be considered as filtering in the frequency domain according to the following equation: M(t,k)=H1(t,k).X1(t,k)+H2(t,k).X2(t,k) Here,

number

[0068] The second downmixing method T2 is defined as the one proposed, for example, in the Samsudin non-patent document cited above.

[0069] In this second method,

number

[0070] The channels of the stereo signal in the frequency domain (X1(t,k) and X2(t,k)) are used, on the one hand, in 301 to calculate the ISD phase indicator (representing a measure of the degree of phase inversion between channels of the stereo signal) and on the other hand, in 303 to calculate the (IPD) phase difference between channels.

[0071] The phase indicator is defined by an ISD (interspectral distance) indicator, such as the one defined above.

[0072] In the embodiments described herein, the ISD is determined by the stereo signal frame so that a decision to select either a first downmix or a second downmix can be made on a frame-by-frame basis.

[0073] This frame-based determination, as in the patent application cited above, makes it possible to eliminate substantial phase shifts between frequencies and other audible artifacts that may arise when the determination shifts from one frequency to the next.

[0074] To apply frame-based decision-making, the ISD phase indicator decision criteria are defined in 302 according to the following formula:

number

[0075] This criterion measures the percentage of the frequency line of index k, and this criterion is favorable for downmixing by rephase adjustment (i.e., when ISD(t,k)>1.3).

number

number

[0076] Next, there is the following downmix selection decision between T1 and T2 (at 305):

number

[0077] In the 303, the phase difference IPD is [Number] which is obtained between two channels of a stereo signal, where ∠ indicates the phase of a complex operand.

[0078] However, the frame determination may sometimes be insufficient because the IPD calculated according to the above formula varies very rapidly over time. Also, the filter H i (t, k) can vary very rapidly from one frame to the next and can produce audible discontinuities when transitioning from one frame to the next. To limit this effect, a time-smoothed version of the IPD can be used to calculate the filter H i (t, k). One way to smooth the IPD is [Number] to apply first-order IIR low-pass filtering in 304 such that

[0079] The forgetting factor α(k) is selected adaptively. This makes it possible to maintain phase coherence in the lower part of the spectrum from one frame to the next, since the spectrum of the audio signal is in a steady state (especially in its lower part), and since a particular ear is sensitive to the phase within this part of the spectrum, it may be advantageous to select a high forgetting factor at lower frequencies (usually below 5 kHz). In contrast, it is advantageous to select a low forgetting factor at higher frequencies (i.e., at higher frequencies (usually above 5 kHz)), since the phase continuity of the audio signal is less prominent, and the ear is less sensitive to phase discontinuities within this part of the spectrum.

[0080] The following form of forgetting factor can be selected: [Number]

[0081] The value αmax , α min , k high and k low This is determined experimentally. For adjacent frames of length L=20ms, the following set of parameters corresponding to 5kHz may be selected: α max =0.94,α min =0.86,k low =0,k high .

[0082] In some variations, very low and very high frequencies are forced to remain untouched. Therefore, in some variations,

number

[0083] Figure 4 provides a detailed description of block 204 in Figure 2. For example, the filters H defined in the downmixing methods T1 and T2 defined above. i (t,k) is reconditioned or actually adapted to avoid cyclic convolution.

[0084] In the first step, at 401, filter H i Impulse response h (t,k) i (t,l) is h i (t,l)=FFT -1 {H i The calculation is performed such that (t,k)},i=1,2, and here the filter h i (t,l) is of size N, corresponding to the length of the FFT. Here, N = L.

[0085] In preferred embodiments, L=20ms or 320 samples at 16kHz, 640 samples at 32kHz, and 960 samples at 48kHz.

[0086] The impulse response defined in this way is conventionally acausal and corresponds to a finite impulse response filter. One conventional technique to make it causal is essentially to rotate L / 2 samples to center the response at L / 2. The drawback is that the filter thus reconstructed presents an L / 2 latency. To avoid this latency, a choice is made at 402 to asymmetrically truncate the filter, and the impulse response h i We retain only a portion of the first half of (t,l) as follows:

number

[0087] This filter offers the advantage of generally having a maximum value (absolute value) for its first sample that guarantees processing without additional delay. In terms of frequency response, this filter is an optimal re-phased filter h i This presents a phase approximately equal to (t,l).

[0088] The choice of P depends on several criteria: a P value close to N / 2 guarantees near-optimal rephase matching, while a lower value limits the computational power required for filtering. A low P also results in smoothing the filter's phase. This presents an advantage of the filter defined by the processing operation T2 as defined above, where the phase fluctuates very rapidly from one frequency to the next: these fluctuations can produce very high group delays perceptible to the human ear. A low P allows for the removal of all or some of these artifacts.

[0089] Filters that were discarded in this manner

number

Number

Number

[0090] This window can be the following triangular half-window or indeed a Hann half-window:

Number

[0091] Due to its truncation and windowing, the filter

Number

Number

[0092] According to the present invention, in 405, the quality level of the filtering performed is determined. To achieve this objective: -The energy E401 of the filter before truncation 402 is calculated; -The energy E403 of the time filter after windowing 403 is calculated; -E 401 and E 403 The ratio R4 is calculated.

[0093] Filtering quality also depends on the energy E of the time filter after normalization. 401 and energy E 404 This can be determined by comparing the two.

[0094] As a non-exhaustive alternative, the filtering quality level can be judged as follows: -Filter H i (t,k) directly, and / or -Normalization

number

number

[0095] In some variant forms, it is possible to calculate the following equation:

number

[0096] This criterion can be normalized by the following formula:

number

[0097] Another variation of quality may involve measuring the amount of energy lost during truncation in order to make the filter causal. In this variation, E 401 This can be calculated as the "anticausal" energy of the original filter:

number

[0098] Energy E 403 (or E 404 ) can be calculated as the energy of the composite filter before (or after) normalization:

number

number

[0099] To avoid bias linked to the window size, the energy of the anticausal part can also be calculated for the same number of samples, regardless of whether windowing is performed by window w(l), as follows:

number

number

[0100] In some variations, the frequency response of the filter obtained after truncation (at the output of 403)

number

number

[0101] This criterion can be normalized by the following formula:

number

[0102] Another variation of the quality judgment lies in comparing the phase coherence of filters, given that the objective is to re-phase the signal. Therefore, the coherence between the two filters

number

number

[0103] High coherence (i.e., close to 1) indicates that the synthesized filter is close to an ideal filter and therefore close to a high quality factor, while a low value (close to 0) conversely indicates a low quality criterion.

[0104] Figure 5 illustrates block 205 of Figure 2, where the OLS (overlap-save) filtering embodiment is performed in detail in the time domain or frequency domain. The choice of domain will depend on the complexity, which depends on the size of the filter relative to the size of the frame to be processed.

[0105] An OLS embodiment in the time domain having adjacent frames of size L is presented here.

[0106] In the first stage, at 501, the current frame is concatenated with P-1 samples from the previous frame (conservation principle):

number

[0107] Each channel is filtered by its phase shift filter in 503:

number

[0108] Next, the mono downmix is ​​generated by summing the two re-phased channels in the 505: m(t,l) = y1(t,l) + y2(t,l), 0 ≤ l <L

[0109] In reality, the filter can be very different between two consecutive frames (for example, when a new source appears). In this case, m t-1 The transition between (t,l) and m(t,l) can cause differences in the form of audible artifacts. To avoid these artifacts, a cross-fade process may be applied. To perform this cross-fading, a filter is used.

number

number

[0110] This is performed by block 502 in Figure 5.

[0111] The number of samples required for crossfading must be determined experimentally, depending on factors such as the filter size and sampling frequency.

[0112] Next, mono downmix m t-1 (t,l) is generated in 504 from the filter of the previous frame:m t-1 (t,l=y 1,t-1 (t,l)+y 2,t-1 (t,l), 0≦l

[0113] Downmix

number

number

number

[0114] It should be noted that, in integration with many other downmixing methods, the aforementioned downmixing methods T1 and T2 cannot necessarily be represented by a filtering operation. This is the case with method T3 in Figure 7, or any other method, whose gain depends on the instantaneous level of the signal.

[0115] According to the present invention, in cases where a selection of downmixing methods T1 and T2 or downmixing method T3 is made, the signal m will be described in the remainder of this specification. t-1 Steps 5040 and 5050 are performed to scale the respective signals (t,l) and m(t,l). To achieve this objective, the signal m to which downmix T1 or T2 is applied is then performed. t-1 The signal (n) acquired at the output of the downmix T3 in Figure 7, which has levels N3, N1, or N2 for each of the signals (t,l) and m(t,l), is adjusted to level N3 so that it is equal to or equivalent to level N3. t-1 Instead of applying this scaling to (t,l) and m(t,l) respectively, this scaling is also applied to the signal

number

[0116] Figure 6a now details a downmixing method according to the present invention, in which one of the downmixing methods T1 or T2 selected at the end of the selection method in Figure 3 is implemented at 610, and another downmixing method is implemented at 611, the gain of which depends on the instantaneous level of the signal, for example, method T3 in Figure 7.

[0117] In this case, one solution is to calculate the downmix of each method in the current frame of time index l and to perform a crossfading transition from one downmix to the next (in 612) to generate a monaural signal d(n). Specifically, in the case of a transition from downmix T1 (or T2) to downmix T3, the downmixing is performed on the signal acquired at the output of T1 (or T2).

number

[0118] According to the present invention, downmix T1 (or T2) or T3 is as follows: - Presence or absence of at least one transition in the stereo signal, - The quality level of filtering used for the stereo signal when downmixing T1 (or T2) is performed. It is selected depending on the circumstances.

[0119] In the case of detecting the presence or absence of at least one transition in a stereo signal 613, according to the first embodiment, transition detection is applied separately to each of the input signals x1 and x2 or to the current frames x1(t,l) and x2(t,l). Since transition detection is a conventional problem in audio coding, it has already been implemented in EVS codecs, and examples of modules described in 3GPP® standard TS 26.445 clause 5.1.8 can be cited. To achieve this objective, each of the input signals x1 and x2 or the current frames x1(t,l) and x2(t,l) is divided into subblocks, the energy of each subblock is calculated, and then smoothing is optionally performed. The obtained energy is then applied to an energy threshold th 61 This is compared to the energy obtained. 61 If it is less than th, the presence of a transition is not detected. The obtained energy is th 61 If the threshold is exceeded, the presence of a transition is considered to be detected. To achieve this objective, parameter Par1 is set to the following value (if T1 is selected at the end of the selection in Figure 3) or Par2 is set to the following value (if T2 is selected at the end of the selection in Figure 3): - A first value (e.g., 0) is used to indicate that no transition was detected. - A second value (e.g., 1) to indicate that the presence of a transition has been detected.

[0120] These parameters Par1 / Par2 are sent to the crossfading block 612 as criteria for judgment.

[0121] If downmix T1 (or T2) or downmix T3 is selected based on the quality level of filtering of the stereo signal used when downmix T1 (or T2) is performed, refer to Figure 4 and proceed as follows: -h i Filtering by (t,l)

number

number

[0122] E 61 and E 62 If the ratio is less than SE, the presence of a transition is likely to be detected. 61 and E 62 If the ratio exceeds SE, the presence of a transition is considered to be detected. To achieve this objective, parameter Par1 is set to the following value (if T1 is selected at the end of the selection in Figure 3) or Par2 is set to the following value (if T2 is selected at the end of the selection in Figure 3): -Set to a first value (e.g., 0) to indicate that no transition was detected. - This is set to a second value (e.g., 1) to indicate that the presence of a transition has been detected.

[0123] These parameters Par1 / Par2 are sent to the crossfading block 612 as criteria for judgment.

[0124] Therefore, according to the present invention, when a transition is detected in the current frame in one or both channels, downmix T3 is applied to the current frame rather than downmix T1 (or T2). Specifically, downmixes T1 and T2 use filtering by an impulse response which sometimes tends to broaden the signal envelope; that is, this problem is more evident in transients such as castanets, the sound of scissors clicking during binaural recording simulating a haircut in a hair salon, and so on.

[0125] To limit potential artifacts resulting from crossfading (block 612) between downmixing methods, the present invention may provide for applying signal scaling at the output of T1 (or T2) (in 614). According to the present invention, such scaling 614 involves the following steps: -The process of determining the energy E' of the downmixing signal at the output of T1 (or T2) by performing the same steps as steps 708, 709 and 710 in Figure 7 (preferably by smoothing), - Scale factor

number

[0126] Unlike the downmix in Figure 7, direct compensation of the downmix level at the output of T1 or T2 is applied here without reinjecting the input signal.

[0127] Figure 6b shows an alternative to the downmixing method described with reference to Figure 6a.

[0128] The embodiment shown in Figure 6b is distinguished from that in Figure 6a solely by the fact that the transition is detected according to the second embodiment. According to this second embodiment, such detection is performed during step 610 of downmixing T1 (or T2), instead of being applied separately to each of the input signals x1 and x2 or to the current frames x1(t,l) and x2(t,l). To achieve this, referring to Figure 5, the energy of each 2ms subblock is compared between the input and output signals of blocks 502 and 503. If the energy falls below a threshold (e.g., 3dB), the transition is considered to have been greatly attenuated, and in this case, downmix T3 is applied in the current frame. If the energy does not fall below a threshold (e.g., 3dB), the transition is considered not to have been attenuated, and in this case, downmix T1 (or T2) is applied in the current frame.

[0129] Figure 8 shows a downmixing device 800 within the scope of the present invention.

[0130] Device 800 is as follows: -Memory MEM1 for storing instruction data of a computer program within the scope of the present invention; - Stereo audio signal

number

number

[0131] Naturally, Figure 8 shows an example of a structural embodiment of a downmixing device within the scope of the present invention.

[0132] Figures 2-7, commented above, illustrate in detail the functional embodiments of this device.

[0133] Applications of this type of downmixing include, for example, audio encoding (when, for instance, a remote device does not have the capability to represent stereo sound on its terminal). In this case, there is no need to carry a stereo signal, thus saving bandwidth. This type of method can work during point-to-point conversations if one of the participants performs stereo or binaural recording. This type of method can also exist during multi-party calls: the conference bridge spatializes the scene by generating a stereo scene, but not all participants necessarily have the capability to render in stereo. The scene should then be downmixed in mono for these participants.

[0134] The present invention may also be applied to audio decoding: in this case, the user's renderer (rendering module) (e.g., a simple loudspeaker) will downmix stereo content to suit the user's reduced ability to hear.

Claims

1. A method for downmixing a stereo signal to obtain a mono signal, wherein the method is: - Applying two downmixing operations (T1 or T2, T3) to the current frame of the signal (610, 611), wherein one of the downmixing operations uses re-phase filtering for at least one channel of the stereo signal channels, and the other downmixing operation does not use re-phase filtering (610, 611). - Selecting one of the downmixing operations to be applied to the current frame (612), wherein the selection is as follows: -- Whether or not there is at least one transition in the stereo signal, --The quality level of the filtering of the stereo signal used when performing the downmixing operation by using the filtering of the stereo signal, The choice, which is carried out depending on A method that includes this.

2. The downmixing method according to claim 1, wherein if the presence of at least one transition is detected in the stereo signal (613) or if it is not detected (613), the downmixing operation without filtering the stereo signal or the downmixing operation with filtering the stereo signal is selected, respectively (612).

3. The downmixing method according to claim 1, wherein the at least one transition is detected before any of the two downmixing operations is applied to the stereo signal.

4. The detection of the at least one transition is as follows: - A step of decomposing the current frame of the stereo signal into subblocks; - A step of calculating the energy of each subblock obtained from the subblock; - The acquired energy and energy threshold (th ６１ The process of comparing with ) - If the energy is lower or higher than the energy threshold, the presence of the transition is not detected or is detected, respectively. A downmixing method according to any one of claims 1 to 3, including the following:

5. The downmixing method according to claim 1, wherein the filtering quality level is compared with a filtering quality threshold, and if the filtering quality level is less than or greater than the filtering quality threshold, a downmixing operation without filtering the stereo signal or a downmixing operation with filtering the stereo signal is selected.

6. The filtering quality levels are as follows: - Before the filtering is applied, calculate the energy of the stereo signal, called the input energy. - After the filtering is applied, the energy of the monaural signal, called the output energy, is calculated. - Calculate the ratio of the input energy to the output energy. The downmixing method according to claim 5, measured by [method].

7. The downmixing method according to claim 1, wherein an indicator (Par1 or Par2) indicating the presence or absence of the aforementioned transition is generated.

8. The two levels (E', E) of the processed current frame １ +E ２ A downmixing method according to any one of claims 1 to 6, wherein one uses filtering of the stereo signal and the other does not, and the two downmixing operations (T1 or T2, T3) are obtained after the current frame of the signal has been applied, - Adjusting the level (E') of the current frame to which the downmixing operation using the filtering is applied (614), wherein the adjusted level (E') is the same as the level (E) of the current frame to which the downmixing operation without the filtering is applied. １ +E ２ ) to be similar to or equal to, to adjust (614), - Comparing the processed current frame having the adjusted level with the processed current frame having the level obtained after the downmixing operation without filtering of the stereo signal is applied to the current frame of the signal, - A method for selecting one of the downmixing operations to be applied to the current frame based on the comparison described above.

9. A downmixing device including a processing circuit for carrying out the steps of the downmixing method according to any one of claims 1 to 7.

10. A storage medium for storing a computer program including instructions for performing the method according to any one of claims 1 to 7, which can be read by a processor.

Images