TWI420918B - Low-complexity audio matrix decoder - Google Patents

Description

Low complexity audio matrix decoder Field of invention

The present invention relates to audio signal processing. In particular, the present invention relates to low complexity adaptive audio matrix decoders or decoding programs, all of which are suitable for decoding of encoded or uncoded input signals. Although it can be used as an independent decoder or decoder, the decoder or decoder can be advantageously used in conjunction with a "virtualizer" or "virtualization" program to make the decoder or decoder to virtual or virtual. The program provides multi-channel input. The invention also relates to a computer program stored on a computer readable medium for causing a computer to perform a decoding process or a decoding and virtualization program in accordance with the teachings of the present invention.

Combined reference

Each of the patents, publications, and references cited herein are hereby incorporated by reference.

Background technique

"Virtual Headset" and "Virtual Speaker" audio processors ("virtualizers") typically encode each multi-channel audio signal associated with a direction into two encoded channels such that when the encoded channel is applied to When a transducer such as a pair of headphones or a pair of speakers is paired, the listener appropriately positioned for the transducers senses that the audio signals appear to be from a different location than the transducer, and these directions are best Is associated with the direction of the multi-channel audio signal. The headset virtualizer is typically formed as a result of the listener feeling that the sound is "out of the head" rather than inside the head. Virtual headset and virtual speaker processor are involved And the use of a head related transfer function (HRTF) of the multi-channel audio signal applied thereto. In this technical field, virtual earphones and virtual speaker processors are well known and similar to each other (virtual speaker processors may include, for example, a "crosstalk canceller" to be different from a virtual earphone processor).

Examples of headphone and speaker virtualizers include virtual machines sold under the trademarks "Dolby Headphone" and "Dolby Virtual Speaker". "Dolby", "Dolby Headphone" and "Dolby Virtual Speaker" are trademarks of Dolby Laboratories Licensing Corporation. The patents and applications of the Dolby Headphone and the Dolby Virtual Speaker include U.S. Patent Nos. 6,370,256, 6,574,649 and 6,741,706, and the entire disclosure of the entire disclosure of the entire disclosure of the entire disclosures of Other "virtualizers" include those disclosed in, for example, U.S. Patent No. 6,449,368, the disclosure of which is incorporated herein by reference.

The Dolby Headphone and Dolby Virtual Speaker offer the feeling of using a pair of standard headphones or a pair of standard speakers for multi-channel surround sound. Recently, low-complexity versions of the Dolby Headphone and Dolby Virtual Speaker have been introduced, for example, in a wide variety of new low-cost products such as multimedia mobile phones, portable media players, portable game consoles and Low cost TV. However, such low cost products are typically two-channel stereo sound (stereo) devices, while virtual devices require multi-channel surround sound input.

Although, for example, Dolby Pro Logic II and its predecessor, Pro Logic's existing matrix decoder, have a multi-channel surround sound input that combines the two-channel stereo audio output of a low-cost device with the Dolby Headphone virtual device. Useful, but in the use of some low-cost devices, existing matrix decoders may generally be more complex and resource-intensive than desired. "Dolby Pro Logic" and "Dolby Pro Logic II" are trademarks of Dolby Laboratories Licensing Corporation. The idea of Dolby Pro Logic II is set forth in U.S. Patent Nos. 6,920,223 and 6,970,567, the entire disclosure of which is incorporated herein by reference. The views of Dolby Pro Logic are set forth in U.S. Patent Nos. 4,799,260, 4,941,177 and 5,046,098.

Thus, the need for low complexity matrix decoders exists, especially for matrix decoders that are used on virtual machines and for use on virtual machines, especially for virtual applications such as Dolby Headphone and Dolby Virtual Speaker. The device. Ideally, such a matrix decoder should minimize the complexity of each program stage to achieve performance similar to the Dolby Pro Logic II decoder.

Summary of the invention

The invention relates to a method for processing an audio signal, comprising: (1) deriving n audio output signals from m audio input signals, where m and n are positive integers, and the n audio output signals Deriving using an adaptive matrix or matrixing procedure responsive to one or more control signals, the matrix or matrixing process generating n audio signals in response to m audio signals; and (2) by the m The audio input signal derives a plurality of time varying control signals, wherein the control signals are derived from the m audio input signals using: (a) in response to the m audio signals a processor or program for generating a plurality of directional dominant signals under the input signal, wherein at least one directional dominant signal is associated with a first directional axis and at least one other directional dominant signal is associated with a second directional axis; And (b) generating a processor or program of the control signals in response to the directional dominant signals.

The adaptive matrix or matrixing program may include: (1) a passive matrix or matrixing process that generates n audio signals in response to the m audio signals; (2) amplitude scale adjuster or amplitude ratio adjustment And each of the control signals generated by the passive matrix or the matrixing process is amplitude-scale adjusted in response to the one-time variable amplitude scaling factor control signal to generate the n audio output signals, wherein the plurality of The time varying control signals are n time varying amplitude scaling adjustment factor control signals, which are used to adjust the amplitude ratio of each audio signal generated by the passive matrix or matrixing program.

The m value can be 2, and the value of n can be 4 or 5.

A processor or program for generating a directional dominant signal may use: (1) a passive processor or program that generates a plurality of pairs of signals in response to the m audio input signals, the first pair of signals representing a first a signal strength of the opposite direction of the direction axis, and the second pair of signals represents signal strengths in opposite directions along a second direction axis; and (2) generating the plurality of directions in response to the two pairs of signals A processor or program of one of the dominant dominant signals associated with one of the first and second directional axes.

The processor or program that generates multiple directional dominant signals can be used A linear amplitude domain subtractor or subtraction procedure that produces a positive or negative difference between the magnitudes of each pair of signals, amplifies each of the equalization amplifiers or amplification procedures, and substantially limits each amplified difference to one A chopper or a chopping program at the positive chopping level and a negative chopping level, and a smoother or smoothing procedure for time-averaging the difference between each of the amplified and limited.

The processor or program that generates the plurality of directional dominant signals may use a linear amplitude domain subtractor or subtraction procedure that obtains a positive or negative difference between the magnitudes of each pair of signals, substantially the difference between each amplified Limiting the chopping or clipping process at a positive chopping level to a negative chopping level, amplifying each of the equalizing amplifiers or amplification procedures, and time-averaging the difference between each amplified and limited Motion smoother or smoothing program.

The amplification factor of the amplifier or amplification program and the relationship between the chopping or clipping program function used to limit the clipping level of the amplification difference may constitute positive and negative threshold values of the magnitude, below which the limit value is limited and amplified. The difference may have an amplitude between 0 and substantially the chopping level, and the difference between the limited and amplified values above the threshold may have an amplitude that is substantially the chopping level for this.

In the absence of an associated audio input signal, the directional dominant signal may approximate a directional dominant signal based on the ratio of the more pairs of signals, and the directional dominant signal may tend to be negative or related to the associated audio input signal. Positive chopping level.

The difference between the limit and the amplification is substantially linear between the threshold values for the difference transfer function.

A difference above a positive threshold may indicate a positive dominance along a direction axis, and a difference below a negative threshold may indicate a negative dominance along a direction axis, and The difference between positive and negative thresholds can indicate non-domination along one direction axis.

The processor or program that generates the plurality of directional dominant signals may also modify the amplified and limited difference before or after the smoothing action such that the derived directional dominant signal is biased on the axis associated with the directional dominant signal .

The processor or program that produces the plurality of directional dominant signals can also modify the difference between the amplification and the limit differentially when there is non-domination and positive or negative dominance along the direction axis.

A processor or program that generates the control signals in response to the plurality of directional dominant signals may apply at least one left and right balance function to each of the plurality of directional dominant signals.

In another aspect, the present invention can derive p audio signals from the n audio output signals, where p is 2 and the p audio signals are derived from the n audio signals using a virtualizer or virtual program. Such that when the p audio signals are applied to a pair of transducers, one of the listeners is appropriately positioned for the transducers to feel that the n audio signals appear to be from a different location than the transducers . The virtualizer or virtual program can include applying one or more head related conversion functions to the plurality of the n audio output signals. The transducers can be a pair of headphones or a pair of speakers.

While the present invention is also applicable to other types of matrix decoders, in an exemplary embodiment, a fixed matrix variable gain approach is employed for reasons that are comparable to variable matrix practices. Low complexity. The superior isolation of a single source produced using a variable gain decoder is acceptable in game sounds that are often monophonic, if there is no better choice.

When working with a virtualizer, you want to minimize inter-channel leakage because of the interaction and credit between the head-related transfer functions (HRTF) of different channels. The variable gain approach allows some channels to be completely turned off while maintaining minimal inter-channel leakage.

Furthermore, when using a variable gain decoder, the "pumping" side effects that occur in certain signal conditions are not as annoying as when used with a virtual machine. This is due to the nature of the output of the two channels produced by each input channel of the virtual machine. Although the variable gain matrix decoder will cause some speakers to be completely turned off, as long as at least one of the inputs is turned on, the output of one of the virtual devices will not be turned off.

As explained further below, optimization may be done for another known disadvantage of the program variable gain approach - missing non-dominated signals, resulting in a decoder that combines the best of both worlds.

Meanwhile, since one of the matrix decoders according to the viewpoint of the present invention is to derive multi-channel content for the virtual device, the number of outputs can be limited to four: left, right, left surround, and right surround. In fact, the main goal of the virtual machine is to convey all the good directionality around the listener, which can be achieved by omitting the center channel using only four channels. The inclusion of the center channel can significantly increase the processing execution time. It only marginally enhances the perception of directionality.

Since the head related conversion functions (HRTFs) are added together to cause destructive interference, it is preferable to avoid correlation between channels. In other words, the virtualizer will perform better when the sound source is manipulated as much as possible toward one horn. However, the achievement of this result should be balanced in the compromise of the overall sound class.

Schematic description

1 is a schematic functional block diagram showing an example of a processor or program for deriving a plurality of pairs of intermediate control signals from a plurality of audio input signals in accordance with the teachings of the present invention, wherein the plurality of pairs of intermediate control signals are representative along The signal strength of the opposite direction of a direction axis. In this example, it can be designated as the "first stage", which has two audio input signals Lin and Rin, and has two pairs of intermediate control signals L-R and F-B.

2 is a schematic functional block diagram showing an example of a processor or program for deriving a plurality of directional dominant signals in accordance with the teachings of the present invention, at least one such signal being derived for each pair of intermediate control signals. In this example, it can be designated as the "second stage", which has two pairs of intermediate control signals L-R and F-B and two directional dominant signals LR and LB.

Figure 3 shows an example of a conceptual or theoretical directional dominating vector in a two-dimensional plane based on orthogonal LR and FB axes.

Figure 4 is an idealized representation of the signal amplitude versus time, which shows the absolute values L and R of a two-channel stereo signal, respectively, where the left input channel (Lin) has a peak amplitude of 0.4 before taking the absolute value. 50Hz sine wave, and the right input channel (Rin) has (50*) before taking the absolute value A sine wave with a frequency of Hz and a peak amplitude of 1.0. The frequencies of the sinusoids are irrelevant, and the level of the left channel is 0.4 times the level of the right channel.

Figure 5 is an idealized representation of the signal amplitude versus time, showing the result of subtracting L from R and multiplying the difference and then chopping at -1.0 and 1.0 to provide a quasi-rectangular wave.

Figure 6 is an idealized representation of the signal amplitude versus time, shown by The quasi-rectangular wave feeds through the smoothed LR intermediate control signal caused by a smoother filter, which illustrates that for substantially uncorrelated signal input, the directional dominant signal will approach the control along the middle of the LR The value of the ratio comparison of the signal strengths of the signal-related direction axes.

Figure 7 is a schematic functional block diagram showing the modification of the processor or program in accordance with the teachings of the present invention shown in Figure 2. In this example, this can also be designated as the "second phase", and the BF difference of the amplified and chopped is limited to a value less than 0 to bias the FB dominating signal backward.

Figure 8 is an idealized representation of the gain versus upper angle in arcs, showing the common left-right balance rule (pan-law) between the left (L) and right (R) audio channels, ie the sine/cosine left and right balance rule Where L = cos(x) * input, and R = sin(x) * input, and x varies between 0 and π/2.

Figure 9a shows an idealized representation of the gain versus the directional dominant signal level for panL and panR when the same sine/cosine left and right balance rule is applied to the LR axis in Figure 8, respectively. Represents the gain distribution from left and right.

Figure 9b shows an idealized representation of the gain versus the directional dominant signal level for panB and panF when the same sine/cosine left and right balance rule is applied to the FB axis in Figure 8, respectively. Represents the gain distribution from the back and the front.

Figure 10 is an idealized diagram showing the quasi-three-dimensional representation of the LGain formula (its three axes are the normalized gain and the values of FB and LR).

Figure 11 is an idealized diagram showing the quasi-three-dimensional representation of the LGain, RGain, LsGain and RsGain formulas (the three axes are the normalized gain and FB). With the value of LR).

Figure 12 is an idealized diagram showing a second-order polynomial approximation of a cosine wave with a cosine between 0 and π/2, which is shown in the range of 0 < x < 1, y = (1-x ² ) The approximation is reasonably close to y=cos(x* π/2). The lower curve in the figure is the approximation.

Figure 13 is an idealized diagram showing the quasi-three-dimensional representation of the LGain, RGain, LsGain and RsGain formulas (the three axes are the normalized gain and the values of FB and LR). When calculating LGain and RGain, the left and right balance components of LR are not used.

Figure 14 is a schematic functional block diagram showing an example of a processor or program for deriving a plurality of control signals from a plurality of directional dominant signals in accordance with the teachings of the present invention. In this example, which can be designated as the "third stage", the four control signals LGain, RGain, LsGain and RsGain are derived from the two directional dominant signals LR and FB.

Figure 15 is a schematic functional block diagram showing an example of an adaptive matrix or matrixing procedure for deriving a plurality of audio output signals from an audio input signal and a plurality of control signals in accordance with the teachings of the present invention. In this example, which can be designated as the "fourth stage", a pair of audio input signals Lin and Rin are applied to a passive matrix, and the level of each matrix output is controlled by four control signals LGain, RGain, LsGain and One of the RsGains is controlled separately to generate four audio output signals LOut, POut, LsOut and RsOut.

Figure 16 is a schematic block diagram showing an overview of all four phases of this example and indicating their relationship.

Preferred embodiment of the invention

The aspects of the present invention are best understood in conjunction with the exemplary embodiments, which are divided into four "stages" for ease of description. The four stages receive m audio input signals, in this case two signals of Lin and Rin, and output n audio signals, in this case LOut (left output), ROut (right output), LsOut The overall relationship in the context of the adaptive audio decoder or decoding program for the four signals (left surround output) and RsOut (right surround output) is shown in Figure 16. The decoder or decoder has a control path that includes first, second, third stages, and a signal path including an adaptive matrix or matrixing procedure in the fourth stage. In this example, a plurality of time varying control signals, the control path produces four control signals and the signals are applied to the adaptive matrix or matrixing procedure.

The first stage

Looking at the first stage shown in Figure 1, in this example, m audio input signals Lin and Rin are applied to a processor or program that derives multiple pairs in response to the m audio output signals a signal, i.e., a first pair of signals L and R representing the relative signal strength along a first direction axis (in this case LR or left-right axis), and a representative along a second direction axis ( In this example, the second pair of signals F and B of the relative signal strength of the FB or the front-rear axis). Although this example utilizes two orthogonal direction axes, there may be more than two direction axes (and thus there may be more than two pairs of signals representing the relative signal strength along the respective additional direction axes), and The equiaxions are not necessarily orthogonal (see, for example, U.S. Patent No. 6,970,567). The first stage processor or program can be considered a passive matrix or matrix program. In this case, a simple matrix calculates left and right, The sum and difference signals, and their absolute values are used as intermediate control signals L, R, F and B. More specifically, the passive matrix or passive matrixing procedure in this example can be characterized by the following formula: L=|Lin|

R=|Rin|

F=|(0.5*Lin)+(0.5*Rin)|

B=|(0.5*Lin)-(0.5*Rin)|

second stage

Looking now at the second stage shown in Figure 2, multiple pairs of signals (each pair of signals representing the relative signal strength along a direction axis) are applied to one of the processors that generate multiple directional dominant signals or program. In this example, two pairs of signals L-R and F-B are applied to the second stage, and two directional dominant signals are generated by the second stage. In principle, as mentioned above, there may be more than two directional axes (and thus more than two pairs of signals and more than two directional dominant signals). It is also possible to generate more directional dominant signals than these pairs of signals and related axes. It is possible to achieve this by processing a pair of applied signals in more than one way to generate multiple directional dominant signals in response to a particular pair of applied signals. Before looking at the details of the second phase of the example, it is useful to explain the principle of operation of the second phase.

After the signal strength measurement data of each of the four directions (L, R, F, B) has been obtained, the intensity in one direction and the intensity in the opposite direction (L ratio R, and F ratio B) are compared. ) to provide measurement data that governs the axis along this direction. Since the four directions in this example provide two directions of 90° to each other (orthogonal axis), so the dominance can be interpreted as a single one of the two-dimensional LR/FB planes. Such a conceptual or theoretical dominance vector can be shown as an example in FIG. Although such dominating vectors are built into the operation of a matrix or matrixing program in accordance with one aspect of the present invention, such dominating vectors need not be explicitly calculated.

A negative value along the LR axis may indicate a dominance toward the left, while a positive value along the LR axis may indicate a dominance toward the right. Similarly, a negative FB value may indicate a dominance toward the rear, while a positive FB value may indicate a dominance toward the front. By interpreting the two dominant values as components of the 2D vector, one can visualize the dominance of a signal to be anywhere on the LR/FB plane.

In most modern matrix decoders (including Dolby Pro Logic and Dolby Pro Logic II), the LR direction is calculated using the ratio of L to R, and the FB direction is dominated by the ratio of F to B. Calculation. Since the ratio is independent of the magnitude of the second signal being compared, it provides a stable dominant direction throughout the natural amplitude variation of the real audio signal. Unfortunately, if implemented using a computer program that controls a digital signal processor (DSP), this requires a statement in the program to select the numerator and denominator, and a sign to give the dominant value. More importantly, common methods such as deriving a ratio of division or subtraction in a logarithmic domain require a large amount of computational resources. The simpler way of subtracting these two numbers in a linear amplitude domain (such as a non-logarithmic domain) is computationally efficient, but the dominant signal produced by such subtraction will quickly vary with the natural amplitude of the signal amplitude. Change in place.

To reduce implementation complexity, the present invention retains most of the amplitude independence of the ratio comparison, but requires much less computation.

The second stage processor or program uses a linear amplitude domain subtractor or subtraction procedure to generate a plurality of directional dominant signals that are positive or between the magnitudes of each pair of applied signals. Negative difference. This subtraction can be implemented with very low computational resources. The result of each subtraction is amplified by an amplifier or amplification procedure, and the amplified difference is applied to a chopper or interceptor that substantially limits each amplified difference to a positive chopping level. With a negative chopping level. Alternatively, the order of the amplifier or amplification procedure and the chopper or interceptor can be reversed using an appropriate chopping level to produce an equivalent result. The smoother or smoothing program can perform a time averaging action on each of the difference between the amplification and the limit to provide a directional dominant signal.

The relationship between the amplification factor of the amplifier or amplification procedure and the chopping level at which the chopper or chopping program is located constitutes the positive and negative thresholds of the magnitude, and the difference between the limits and amplifications below this threshold may be substantial. There is an amplitude between 0 and the chopping level, and the difference between the limited and the amplified values above the threshold may have an amplitude that is substantially the chopping level for this. Although this particular transfer function is not critical and may take many forms, the difference between such limited and amplified differences is substantially linear with respect to the difference between the critical values. Calculate the requirements and be appropriate.

The second stage processor or program may include modifications to the amplified and limited differences before or after the smoothing action at the time of processing such that the derived directional dominant signal is along an axis associated with the directional dominant signal Being "offset". This offset can be fixed or adaptive. For example, a difference signal after amplification and chopping may be scaled in amplitude, and/or shifted (ie, offset) in amplitude, and/or fixed in a fixed manner on amplitude or sign (eg, As a function of the amplitude, the sign, or the amplitude and sign of the difference signal that is amplified and limited. The result may, for example, include applying less bias to the non-dominated signal than to the dominant signal (dominance and non-domination will be further explained below). An example of applying a "bias" to a directional control will be described below in conjunction with Figure 7.

In the example of the second phase of Figure 2, two pairs of signals L-R and F-B are applied to produce bidirectional dominating signals LR and FB. As described above, if there are four intermediate directional signals (L, R, F, B), one would derive the two dominating signal components LR and FB by comparing the directivity along each axis. In accordance with the teachings of the present invention, this is accomplished by subtracting R and F minus B (or vice versa in each case) to provide an amplitude difference signal along each axis. The heavy gain is applied to the equal difference signal, and the amplified difference is chopped (hard limited) to -1.0 and +1.0. The chopped difference signal is then applied to a time smoothing filter.

By applying the weight gain and the chopping to the difference signal, the dominance of any magnitude in one direction is essentially considered to be absolutely dominant in one of the directions. In the case of a signal whose instantaneous directivity is changed from one polarity to another, the result of this operation is similar to a rectangular wave having a varying frequency and duty cycle. The temporal smoothing filter averages the most rectangular waves to provide a continuous wave that approximates the ratio of the original directional signal pairs to each other. Although the actual filter used is a design choice, the filter can also be effectively applied, for example, to a first order digital IIR low pass filter having a time constant of about 40 ms.

In addition to detecting the formulation of each axis, it may be advantageous to present "non-dominated". For example, an input signal that is manipulated purely to the left should exhibit strong dominance on the left-right axis, but should not be dominant along the front and rear axes. Another example is The extremely low level signal of background noise, people's preference does not cause any manipulation effect. In accordance with the teachings of the present invention, such a general practice is to select a threshold value, and specify a magnitude greater than -1.0 or 1.0 for the difference (depending on the sign of the difference), and specify a difference between the two values for the difference. A certain amount of value less than this threshold. There is a possibility to specify a value of 0.0 for all differences below the threshold. For this reason, some situational statements and numerical comparisons are needed in a program-controlled DSP. A preferred approach from a low complexity point of view is to amplify the difference with a large gain such that the output below the threshold value follows a linear function from -1.0 to +1.0. This gain is the reciprocal of the critical value. This is very efficient - both the gain and the chopping phase can be set in the program-controlled DSP with the "saturation logic" of the DSP (such as setting one of the DSP control registers/bits, so that when the ALU overflows When the result is set to the maximum positive or minimum negative value exhibited by the platform according to its sign, it is applied as an arithmetic left shift (in terms of a gain of 2). A gain with a power of 2 can be applied with only a slight increase in processing complexity.

The dominance signal (positive dominance, negative dominance, and non-domination) of a three region allows the resolution between dominant and non-dominated along a direction axis before smoothing. The discriminating between dominance and non-domination promotes the "offset" of adaptive application of directional dominant signals as described above, an example of which is given below in conjunction with Figure 7. As shown below, in the perspective of the present invention, it is useful to resolve the unique left steering signal from the left surround steering signal and the unique right steering signal from the right surround steering signal prior to the smoothing action.

In a practical embodiment of the invention, it is necessary to determine the discrimination of the side (left or right) steering signals from the surround (left or right surround) steering signals. The minimum gain is decoded by the music material encoded by the Dolby Pro Logic II matrix encoder. The average (FB) difference signal is measured for the left or right surround manipulation input and is used as an estimate of one of the maximum resolutions (minimum gain) that maintains a clear resolution between left and left surrounds (or right and right surrounds). . In a practical embodiment of a decoder in accordance with the teachings of the present invention, a gain factor of 1024 is used which corresponds to a threshold of about 0.001 that is normalized to a signal of [-1, +1]. A threshold value of less than 0.001 produces edge hearing improvement, while a larger threshold value reduces the isolation between the sides (left and right) and the surround (left and right surround) to an unacceptable level. In general, the critical level is not critical.

To illustrate this technique, consider a two-channel stereo signal in which the left input channel (Lin) is a 50 Hz sine wave with a 0.4 peak amplitude and the right input channel (Rin) is (50*). A sine wave with a frequency of Hz and a peak amplitude of 1.0. These signals are shown in Figure 4. The frequency of the sine wave is irrelevant, while the level of the left channel is 0.4 times the level of the right channel. Using a ratio comparison as described above, this provides a dominance of 0.6 in the right direction (defined herein as positive). As shown in the first stage, the L and R intermediate signals are the magnitudes of the input signals Lin and Rin.

After L is subtracted from R, the difference is amplified by, for example, 1024 (implemented with an arithmetic left shift of 10 bits), then truncated at -1.0 and +1.0 to provide a quasi-rectangular wave. Figure 5 shows the difference signal before and after the chop.

The quasi-rectangular wave is fed through a smoothing filter that provides one of the LR directional dominant signals. In the case where the input signal has a fixed level, the directional dominant signal eventually reaches a value of about 0.65 as shown in Fig. 6, which is close to The dominant value calculated using the ratio comparison is oscillated in the vicinity thereof. The smoothness of the oscillation is a function of the order and characteristics of the smoothing filter.

This example represents an audio material that has a large amount of uncorrelated signals at each input, such as an uncoded two-channel stereo signal, where the polarity of the amplified difference of the cut-off is reversed very frequently. Under these input conditions, the control signal derived from the subtraction/amplification/clash produces a result that is close to the ratio comparison.

However, just as for materials having a common (i.e., correlated) signal in the two channels of a monophonic sound source included in the matrix encoded content, the chopped difference signal does not contain many zero crossings. In such cases, even if the polarity of the difference signal eventually reverses, the smoothed control signal tends to "lock" to the dipole value by smoothing the transition to another extreme value (ie + One of 1.0 and -1.0). The "lock" of such a component can be imagined as pulling a two-dimensional dominating vector along the LR/FB plane. When both components are "locked", the dominating vector is pulled to one of the four corners of the LR/FB plane. In accordance with the teachings of the present invention, such hard left and right balance improves spatial imaging of matrix encoded content by providing a relatively discrete, single input channel to the virtualizer.

Front and rear dominance bias

A disadvantage of the variable gain approach is that non-dominated signals may be lost in the decoded output. This is quite evident in music sources where a large number of sound sources are mixed together at many different levels and phase differences. It is often the case that there are a few main instruments and vocals that are equally mixed in the left and right, although there are many other weaker tones and are added to the overall space and the sound of the sound field. sound. Since the decoder uses only the direction of the most dominant acoustic component, the traditional variable gain approach to this material results in an output of almost no phase material from the rear decoder output (in this case, left surround and The result of the right surround output).

In accordance with one aspect of the present invention, this problem is mitigated by biasing the FB dominating signal toward the back to ensure that the out-of-phase material is not removed from the surrounding output. The way to accomplish this is to limit the FB signal to a negative value before smoothing the filter. This is shown in the example of Figure 7. For a purely rectangular wave between -1.0 and 1.0, this is equivalent to scaling down one-half the output of a smoothing filter with a fixed offset of -0.5. Thus, this modification can be applied before or after smoother filtering. However, the difference signal after clipping may not be a pure rectangular wave. Rather, when the difference signal falls below a threshold and indicates non-domination along a particular axis, it will contain a value between them. When the magnitude of the chopped difference signal is less than 1.0, the process of limiting the FB to a negative value results in less than a negligible effective offset after smoothing. It is thus possible to resolve the non-dominated and positive or negative dominance before smoothing in this way, allowing the left and right steering signals to maintain a high degree of separation from the surround, while giving a significant bias to the other signals backwards. .

The third stage

The third stage processor or program generates control signals for applying one or more left and right balance functions in response to the plurality of directional dominant signals (the left and right balance functions are representative of the channels) The left and right balance "characteristic transfer function" to each directional control signal to control the adaptive matrix or matrixing program. One or more left and right balance functions are available Use one or more of the following items: a trigonometric conversion function (such as a sine or cosine conversion function), a logarithmic conversion function, a linear conversion function, and. A mathematically simplified approximation of a trigonometric transfer function.

In this example, the goal of the third stage is to obtain the LR and FB dominating signals that were calculated in the previous stage, and to derive the output that is applied to the passive matrix to produce the decoded output.

A general approach to a matrix decoder or decoding procedure in accordance with the teachings of the present invention is to emphasize the output channel closest to the dominant position after the dominant directionality of the input has been detected, and to remove the most dominant position. Output. Between the two outputs closest to the dominant position, the problem can be reduced to a pairwise left and right balance, which can be expressed as a left and right balance function.

Sine/cosine left and right balance rule

The most common left-right balance rule between two channels is the sine/cosine rule, where L = cos(x)* input, and R = sin(x)* input, and x varies between 0 and π/2. See Figure 8.

Two-dimensional sine/cosine left and right balance rule

The gain of each decoder output channel must be expressed in LR and FB functions: LGain=f _L (LR, FB)

RGain=f _R (LR, FB)

LsGain=f _Ls (LR, FB)

RsGain=f _Ls (LR, FB)

The same sine/cosine left and right balance rule can be applied to the LR and FB axes, and the left and right balance curves shown in the figures 9a and 9b are obtained, wherein panL, panR, panB and panF represent the left, right, back, and front respectively. Gain distribution.

After understanding the sine function as the cosine function after phase shifting, you can use only the cosine function to obtain the following left and right balance formula: panL=cos((LR+1)/2* π/2)

panR=sin((LR+1)/2* π/2)=cos((LR-1)/2* π/2)

panB=cos((FB+1)* π/2)

panF=sin((FR+1)* π/2)=cos(FR* π/2)

With the nature of the left channel position on the LR/FB plane (see Figure 3), LGain should be maximum only when both panL and panF are maximum, and should be spaced away from the two axes or one of the axes. And decrease. This can be achieved by multiplying panL by panF. The same principle can be applied to RGain, LsGain and RsGain, and the final formula of the gain becomes: LGain=panL*panF

RGain=panR*panF

LsGain=panL*panB

RsGain=panR*panB

The use of multiplication can also be seen as a mutual scaling of the two sine/cosine amplitude left and right balance functions, where the minimum of the two components becomes the maximum value at which the overall gain can be reached.

Figure 10 shows the 3D rendering of the LGain formula, and Figure 11 shows the 3D rendering of all four gains.

Polynomial approximation of cosine function

As shown in Fig. 8, the left and right balance rule is synthesized by the cos(x) and sin(x) two curves (the sin function can be replaced by the cos function with appropriate phase shift). To avoid complicated calculations or to use large checklists, a second order polynomial approximation of the cosine curve between 0 and π/2 can alternatively be used in accordance with the teachings of the present invention. The formula y=(1-x ² ) is reasonably close to y=(x* π/2) in the range of 0<x<1. (See Figure 12, where the lower curve is an approximate curve). The results obtained with this approximation may only have little to no auditory differences.

Front left and right balance adjustment

Since the expected audio input source is stereo that has been mixed to naturally balance the two channels between L and R, one aspect of the present invention is that the left and right balance components of LR are not considered in the calculation of LGain and RGain. The extra left-right balance in the variable gain does not significantly improve the separation in this case because L and R have been separated very well. In addition to saving some calculations, it also allows a more stable sound field in front by avoiding unnecessary gain loading. After removing the LR component, these formulas are obtained: LGain=panF

RGain=panF

LsGain=panL*panB

RsGain=panR*panB

The three-dimensional rendering of these new formulas is shown in Figure 13.

Note that a similar simplification can be applied to the Ls gain and Rs gain equations without using an additional LR left and right balance, and using the natural left and right balance within the source signal to create separation between the two surround signals. In this case, however, Ls The separation from Rs is limited by the passive decoding that occurs in the fourth phase. A passive decoding matrix or matrixing procedure such as forming part of the present invention can only achieve a 3 dB separation between Ls and Rs, so such simplification is unacceptable from a channel separation standpoint. In order to maintain a high degree of separation, the LR component in the formula of LsGain and RsGain is retained.

Final gain formula

Substituting the cosine of each left and right balance term into a polynomial approximation, the final formula can be derived for each gain factor: LGain=1-FB ²

When FB=0→LGain=1

When FB=-1→LGain=0

RGain=1-FB ²

When FB=0→RGain=1

When FB=-1→RGain=0

LsGain=[1-((LR+1)/2) ² ]*[1-(FB+1) ² ]

When FB=0→LsGain=0

When FB=-1 and LR=-1→LsGain=1

When FB=-1 and LR=1→LsGain=0

RsGain=[1-((LR-1)/2) ² ]*[1-(FB+1) ² ]

When FB=0→RsGain=0

When FB=-1 and LR=-1→RsGain=0

When FB=-1 and LR=1→RsGain=1

Referring to Figure 14, the control signals LGain, RGain, LsGain, and RsGain are applied by applying a left and right balance function to a directional dominant signal, and/or applying one The left and right balance functions are derived from the product of a directional control signal and a directional control signal applied to the left and right balance functions, wherein each of the left and right balance functions may be different from all or part of the other left and right balance functions. These left and right balance functions are not the balance functions inherent in the n input audio signals. In this example, one of the directional axes is a left and right axis, and the left and right balance functions are left and right balance functions that do not include left and right balance components. The following items can be applied to this embodiment: the LR directional dominant signal is applied to a panL left and right balance function and a panR function; the FB directional dominant signal (as in Figure 2, there is no offset or as in Figure 7) The offset is applied to a panF balance function and a panB balance function; the result of applying the panF function to the FB directional control signal is applied to the passive decoder or decoding program of the fourth stage as generally as LGain and Rgain; The result of applying the panB function to the FB dominating signal is multiplied by the result of applying the panL function to the LR dominating signal, and is generally applied to the fourth stage passive decoder or decoding program as LsGain; the result of applying the panR function to the LR dominating signal is Multiply the result of applying the panB function to the FB dominating signal and is generally applied to the fourth stage passive decoder or decoding procedure as RsGain.

Fourth stage

Figure 15 shows a passive matrix or matrixing procedure for generating n audio signals in response to m audio signals, and an amplitude scaler or amplitude scaling procedure, each in response to a time varying amplitude scale adjustment factor control signal One of the audio signals generated by the passive matrix or the matrixing process is amplitude-scaled to generate n audio output signals, wherein the plurality of time-varying control signals are n time-varying amplitude-proportional adjustment factor control signals No. is used to adjust the amplitude ratio of each audio signal generated by the passive matrix or matrixing program. In the example of Fig. 14, there are two audio output signals Lin and Rin, four audio output signals LOut, ROut, LsOut, RsOut, and four proportional adjustment factor control signals LGain, RGain, LsGain and RsGain (from the first Three stages).

In the example of Figure 15, the four matrix or matrixing procedures can be characterized by the following formula: LOut=LGain*(a*Lin+b*Rin)

ROut=RGain*(c*Lin+d*Rin)

LsOut=LsGain*(e*Lin+f*Rin)

RsOut=RsGain*(g*Lin+h*Lin)

Here, a to h shown in Fig. 15 are matrix coefficients. These coefficients a to h can be selected to match the user in the Dolby Pro Logic II encoding/decoding system, where: a=1.0, b=0.0, c=0.0, d=1.0, e=0.8710,f =-0.4898, g=0.4898, h=0.8710

This provides the following final formula: LOut=LGain*Lin

ROut=RGain*Rin

LsOut=LsGain*(0.8710*Lin-0.4898*Rin)

RsOut=RsGain*(0.8710*Rin-0.4898*Lin)

Figure 16 shows an overview of all four phases of this example and points out Interrelationship.

Implementation

The invention may be implemented in hardware or software or a combination of both, such as a programmable logic array. Unless specifically stated otherwise, the rules included as part of the invention are not fixedly related to any particular computer or other device. In particular, a variety of general purpose machines, such as digital signal processors, can be used in conjunction with programs written in accordance with the teachings herein, or more convenient to construct more specialized devices (e.g., integrated circuits) to perform the required method steps. Thus, the present invention can be implemented in one or more computer programs executing one or more stylized computer systems, each programmable computer system including at least one processor, at least one data storage system (including power and Non-electrical memory and/or storage element), at least one input device or input port, and at least one output device or output port. The code is applied to the input data to perform the functions as described herein and to generate output information. The output information is applied to one or more output devices in a conventional manner.

Each of these programs can communicate with a computer system in any desired computer language (including machine language, combination language, or higher level programming language, logical language, or target oriented programming language). In any case, the language can be a compiled and interpreted language.

Each such computer program is preferably stored or downloaded to a general purpose or special purpose programmable computer readable storage medium or device (such as solid state memory or media, or magnetic or optical media) for The storage media or device is assembled and operated by a computer system to perform the procedures described herein. It is also conceivable to use the computer of the present invention in a computer program. The brain is readable by a storage medium, wherein the storage medium so configured causes the computer system to operate in a specific and predefined manner to perform the functions described herein.

The practical embodiment of the present invention implemented in a computer program suitable for controlling a digital signal processor has been implemented using 30 lines of C language code, operating at approximately 3 MIPS, and memory is not used virtually. This is probably 15% of the MIPS used by the Dolby Pro Logic II decoder. The program can be left entirely in the time domain and executed on a benchmark basis (no block program). In order to minimize the execution time of each sample, the application can avoid the use of branches and mathematical functions such as square root, sine, cosine and division. The implementation also avoids the use of checklists and advance delays, which increase memory requirements and increase execution time. The perspective of the present invention can be implemented with a very simple computer program and a very basic digital signal processor. In particular, from the standpoint of simplicity, the idea of the present invention can also be implemented using analog circuits.

Various embodiments of the invention have been described. However, it will be appreciated that various modifications may be made without departing from the spirit and scope of the invention. For example, some of the steps described herein can be independent in sequence and thus can be performed in a different order than described herein.

Lin‧‧‧ left input signal

Rin‧‧‧Right input signal

FB, LR‧‧‧ directional dominant signal

L, R, F, B‧‧‧ intermediate control signals

panL‧‧‧Left gain allocation

panR‧‧‧right gain distribution

panB‧‧‧ post gain allocation

panF‧‧‧ front gain distribution

Lgain, Rgain, LsGain, RsGain‧‧‧ control signals

a~h‧‧‧matrix coefficient

Lout‧‧‧left output signal

Rout‧‧‧Right output signal

LsOut‧‧‧ left surround output signal

RsOut‧‧‧Right surround output signal

1 is a schematic functional block diagram showing an example of a processor or program for deriving a plurality of pairs of intermediate control signals from a plurality of audio input signals in accordance with the teachings of the present invention.

Figure 2 is a schematic functional block diagram showing an example of a processor or program for deriving a plurality of directional dominant signals in accordance with the teachings of the present invention.

Figure 3 shows an example of a conceptual or theoretical directional dominating vector in a two-dimensional plane based on orthogonal LR and FB axes.

Figure 4 is an idealized representation of the signal amplitude versus time, which shows the absolute values L and R of a two-channel stereo signal, respectively.

Figure 5 is an idealized representation of the signal amplitude versus time, showing the result of subtracting L from R and multiplying the difference and then truncating at -1.0 and 1.0 to provide a quasi-rectangular wave.

Figure 6 is an idealized representation of the signal amplitude versus time, showing the smoothed LR intermediate control signal resulting from feeding the quasi-rectangular wave through a smoother filter.

Figure 7 is a schematic functional block diagram showing the modification of the processor or program in accordance with the teachings of the present invention shown in Figure 2.

Figure 8 is an idealized representation of the gain versus upper angle in arcs, showing the common left-right balance rule (pan-law) between the left (L) and right (R) audio channels, ie the sine/cosine left and right balance rule .

Figure 9a shows an idealized representation of the gain versus the directional dominant signal level for panL and panR, respectively, when the same sine/cosine left and right balance rule is applied to the LR axis in Figure 8.

Figure 9b shows an idealized representation of the gain versus the directional dominant signal level for panB and panF, respectively, when the same sine/cosine left and right balance law is applied to the FB axis in Figure 8.

Figure 10 is an idealized diagram showing the quasi-three-dimensional representation of the LGain formula (its three axes are the normalized gain and the values of FB and LR).

Figure 11 is an idealized diagram showing LGain, RGain, LsGain Quasi-three-dimensional representation with the RsGain formula (its three axes are the normalized gain and the values of FB and LR).

Figure 12 is an idealized diagram showing a cosine wave approximation of a second order polynomial with a cosine between 0 and π/2.

Figure 13 is an idealized diagram showing the quasi-three-dimensional representation of the LGain, RGain, LsGain and RsGain formulas (the three axes are the normalized gain and the values of FB and LR).

Figure 14 is a schematic functional block diagram showing an example of a processor or program for deriving a plurality of control signals from a plurality of directional dominant signals in accordance with the teachings of the present invention.

Figure 15 is a schematic functional block diagram showing an example of an adaptive matrix or matrixing procedure for deriving a plurality of audio output signals from an audio input signal and a plurality of control signals in accordance with the teachings of the present invention.

Figure 16 is a schematic block diagram showing an overview of all four phases of this example and indicating their relationship.

Lin‧‧‧ left input signal

Rin‧‧‧Right input signal

FB, LR‧‧‧ directional dominant signal

L, R, F, B‧‧‧ intermediate control signals

Claims (24)

A method for processing an audio signal, comprising the steps of: deriving n audio output signals from m audio input signals, wherein m and n are positive integers, and the n audio output signals are used in response to n Deriving an adaptive matrix or matrixing procedure of the variable control signal, the matrix or matrixing process generating n audio signals in response to the m audio signals; deriving the n control signals from the m audio input signals The deriving step uses: a passive matrix or matrixing program that generates pairs of signals in response to the m audio input signals, wherein the relative directions along a first direction axis are present a first pair of signals of direction signal strength, and a second pair of signals indicative of relative direction signal strength along a second direction axis; a first processor or program responsive to the inputs The signal pair produces a plurality of directional dominant signals, wherein at least one directional dominant signal is associated with a first directional axis and at least one other directional branch The signal system is associated with a second direction axis; and a second processor or program that generates the control signals in response to the directional dominant signals, characterized by: generating a plurality of directivities The first processor or program that governs the signal: a positive or negative difference between the magnitudes of each pair of signals is obtained a plurality of linear amplitude domain subtractors or subtraction programs; each of the equalities is substantially limited to a chopper or a chopping program at a positive chopping level and a negative chopping level; 1) amplifying each of the equalities such that the chopper or chopping program limits each of the amplified differences, or (2) amplifying each of the limited differences, an amplifier or amplification procedure; Each smoothing or smoothing procedure that is amplifying and limiting the difference or the difference between the limited and amplified times as a time-averaging action. The method of claim 1, wherein for the unrelated plurality of audio input signals, each directional dominant signal approximates a directional dominant signal according to a ratio of the signal pairs, and for the related number For an audio input signal, the directional dominant signal tends to a negative or positive chopping level. The method of claim 2, wherein a difference above the positive chopping level indicates a positive dominance along an axis of direction, and a difference below the negative chopping level indicates a A negative control of one of the directional axes, and a difference between the positive chopping level and the negative chopping level indicates a non-dominance along one direction axis. The method of claim 3, wherein the processor or program that generates the plurality of directional dominant signals is modified when it is non-dominated along one direction axis, and is modified differently when it is dominant or negatively dominated. The difference that is amplified and limited or the difference that is limited and amplified. The method of claim 1, wherein the processor or program that generates the plurality of directional dominant signals is also limited to the output of one of the chopper or the chopping program before a smoother or smoothing procedure. Measured value or Negative value. The method of claim 5, wherein the processor or program that generates the plurality of directional dominant signals limits the at least one of the choppers or interceptors prior to the smoother or smoothing procedure The positive value of the output. The method of claim 6, wherein the first direction axis is a front/rear axis, and the processor or program that generates the plurality of directional dominant signals limits processing of a front/rear axis directional dominant signal The positive value of the output of the chopper or chopping program. The method of claim 1, wherein the second processor or program that generates the control signals in response to the plurality of directional dominant signals applies at least each of the plurality of directional dominant signals A left and right balance function. The method of claim 8, wherein one or more of the left and right balance functions employ a triangular transfer function. The method of claim 8, wherein one or more of the left and right balance functions employ a logarithmic transfer function. The method of claim 8, wherein one or more of the left and right balance functions employ a linear transfer function. The method of claim 8, wherein the one or more of the left and right balance functions employ a mathematically simplified approximation of a triangular transfer function. The method of claim 8, wherein the control signals are derived by the following items: Application of the application of a left and right balance function to a directional dominant signal, and/or application of a left and right balance function to a directional dominant signal and application of a left and right balance function to another directional dominant signal The product of the results, wherein each of the left and right balance functions may be different from all other left and right balance functions or different from some of the other left and right balance functions. The method of claim 8, wherein the left and right balance functions are not left and right balance functions inherent in the n audio input signals. The method of claim 14, wherein one of the direction axes is a left/right axis, and the left and right balance functions are left and right balance functions that do not include a left/right left and right balance component. The method of claim 8, wherein at least some of the n time-varying proportional adjustment factor signals are derived from a result of application of a single left-right balance function on a directional dominant signal, and The other signals in the n time-varying proportional adjustment factor signals are the product of the application result of a single left-right balance function on one directional dominant signal and the application result of another single left-right balance function on the other directional dominant signal. To export. The method of claim 16, wherein the direction axis of one of the directional dominant signals is a left/right axis, and the other of the directional signals governs the signal The direction axis is a front/rear axis, at least some of the n time-varying proportional adjustments The factorial signal is derived from the application of a single left and right balance function on the anterior/posterior directional dominant signal, and at least some of the n time varying proportional adjustment factor signals are left/right by a single left and right balance function The result of the application on the directional dominant signal is derived from the product of the application of another single left and right balance function on the anterior/posterior directional dominating signal. The method of claim 1, further comprising the step of deriving p audio signals from the n audio output signals, wherein p is two, and the p audio signals are using a virtual device or virtual The program is derived from the n audio signals such that when the p audio signals are applied to a pair of transducers, a listener positioned appropriately with respect to the transducers senses the n The audio signals appear to be from a different location than the transducers. The method of claim 18, wherein the virtualizer or virtualization program includes the application of one or more head related conversion functions on some of the n audio output signals. The method of claim 18, wherein the pair of transducers is a pair of earphones. The method of claim 18, wherein the pair of transducers are a pair of speakers. An apparatus adapted to perform the method of any one of claims 1 to 21. A computer program stored on a computer readable medium for causing a computer to perform the method of any one of claims 1 to 21 method. The method of any one of claims 1 to 21, wherein m is 2 and n is 4 or 5.