JP2016504622A - Method for calculating at least two individual signals from at least two output signals - Google Patents

Abstract

The present invention relates to a method of calculating at least two individual signals from at least two output signals formed from at least two source signals, identifying a mixing ratio on the output signal of at least two source signals, and mixing ratio By multiplying the output signal by the factor formed from the output signal and calculating at least two subtraction signals from the output signal, one source signal is eliminated in each subtraction signal, and this subtraction signal is Fourier transformed, thereby converting the subtraction signal Then, a remainder signal is determined from the converted signal, at least two individual signals are calculated based on the converted signal and the remainder signal, and the individual signal is subjected to inverse Fourier transform.

Description

  The invention relates to a method for calculating at least two individual signals according to the preamble of claim 1.

  It is known that a stereo signal can be generated by recording two or more sound sources. For this, a level difference or arrival time difference is used. By using a stereo signal, a three-dimensional sound is born.

  Known stereo techniques are intensity stereo in the form of AB-method, XY-method and MS-method, as well as a compromise technique such as ORTF-method and OSS-method.

  In recent years, the number of acoustic systems with more than two speakers has gradually increased in the market. In particular, such an acoustic system has already existed, for example, in cars for many years, and such cars have at least two speakers in the side door and two in the car. Has a speaker.

  This is why several methods have been developed to allow more than two separate signals, as is the case with stereo. In this case, there are generally two different approaches.

  First, there is a method capable of calculating each source signal from a large number of output signals. In this case, each instrument, each singing voice and speaking voice is interpreted as a source signal. In the case of chorus, a plurality of singers can also be regarded as one source signal. For example, even if there are 20 female singers, “soprano” can be recognized as one singing voice.

  Overall, the problem is to extract an arbitrary number of source signals from two output signals, such as present in a stereo signal, or from more output signals.

  Such a method for signal source separation is carried out under the so-called “blind separation” catchphrase. There are two main types of this method. On the one hand, the method is performed using statistical methods such as independent component analysis, principal component analysis, maximum posterior probability method, maximum likelihood method and the like. On the other hand, it is also known to use recursive optimization methods. The known methods then allow any source signal to be calculated from a smaller number of output signals at a high computational cost, or more than or equal to the number of source signals for real-time calculations. Requires an output signal.

  It is therefore an object of the present invention to develop a method having the features of the premise of claim 1 and to enable signal source separation in real time even in the case of fewer output signals than the number of source signals. .

  This problem is solved by the characterizing part of claim 1. Advantageous developments of the invention are given in the subclaims.

  The core of the present invention is that a subtraction signal is calculated from the output signal, the subtraction signal is Fourier-transformed, and an individual signal is calculated in the converted signal thus obtained. This eliminates the need for statistical and recursive methods and achieves the calculation of individual signals within real time. Here, real time means that the signal to be played can be provided without interruption while the incoming data stream has a marginal time lag that occurs when calculating the first block of the signal. Means that. Thus, for example, a specific singing voice and, for example, a piano can be extracted from a recording and assigned to a specific speaker at a mixing ratio to be selected, so that the output signal does not have to be processed in advance. . Thus, for example, you should be able to set a CD in a car, decide the song and, in some cases, the distribution of the signal to be extracted and the signal to be extracted, and start playing the song or CD without noticing the delay. is there.

  The mixing ratio of the source signal on the output signal can be calculated by specifying the direction. For this purpose, the output signal is Fourier transformed and an amplitude value is calculated. These amplitude values are collected in one histogram. For this purpose, point pairs are created from the Fourier-transformed output signal, and the point pairs in the histogram can be combined according to the obtained angles. At this time, the angle of the point pair takes a value between 0 ° and 90 °, and is obtained as follows.

Here, X ~ 1, org is the first output signal X 1 which is the Fourier transform, X ~ 2, org, the second shows the output signal X 2, which is the Fourier transform, a x and a y are , A vector, which contains the magnitude. L of spelling, the output signal X 1, X 2 or Fourier transformed output signals X ~ 1, org, a term index number of X ~ 2, org, point pairs are ascertained on the basis of this. Although the first value of the Fourier transformed output signals X ~ 1, first numerical Fourier transformed output signals X ~ 2, org the org is nothing but the vector, these first number, thus l Form a first pair of points = 1. Of course, the term index l may start from 0, in which case the term index continues to L-1. Here, L indicates the number of points (for example, 4096 points) obtained by Fourier transform of the output signals X 1 and X 2 . θ is an angle between 0 ° and 90 °, and ψ l is the size of the point pair with index l.

The vector Ψ _l contains the sum of the magnitudes of the vectors a _x and a _y .

  The Fourier transform is performed in blocks. As the input signals, powers of 2, that is, powers of 2 are used. The 10th, 11th, or 12th power, ie, 1024, 2048, or 4096 data points have been found to be particularly efficient. Particularly preferred are 4096 data points because the calculation time is optimal considering the calculation cost.

  Histogram segmentation is preferably performed in units of 1 °. That is, the histogram includes 90 intervals. The value of the angle θ is then rounded to an integer value to delimit the numerical range of the histogram.

  Finally, the histogram is flattened so that the maximum (maximums) is clear. The following function is used as the flattening function.

  The parameter T is an integer and gives how many adjacent points are incorporated into the flattening. Thus, an average over multiple data points is performed. At the edge of the histogram where there are no S adjacent points, a value of 0 is used for the corresponding position to be supplemented. When calculating the first frequency (Zahlenwert) of the histogram, therefore, the value of 0 should be applied 8 times on one side, while the existing frequency is used on the other side. In the flattened histogram, all local extrema are identified and sorted by frequency, ie frequency. Each location in the histogram corresponds to one angle as described above, and as a result, one angle corresponds to each extreme value. The angle corresponding to the found maximum (s) is identified and a preset number of maxima is used, or the maxima of all that exceed a preset threshold are used.

  The flattened histogram-vector thus becomes:

  Here, i = 0,..., 90. Here, a possible problem is that the source signal does not appear over the entire period of the output signal. For example, a specific instrument or singing voice is paused. After determining two or more histograms and identifying the corresponding angles, this pause will give the next histogram a weight, for example a low pass, so that this pause does not cause an error in identifying the angles. You can multiply functions. The low pass can be expressed by the following equation.

Here, a = 0.1 and b = 0.9. The index n is the index of the histogram, specifically n = 1 for the first histogram or the first 4096 data points. By using a low-pass, a histogram h _{gl_TP} is obtained.

  This weighting stabilizes the angle identification.

  The mixing ratio of the two source signals is determined based on the calculated angle.

Is substituted into the following equation.

  If the maximum of the histogram is at, for example, 18 °, V = 0.325 is obtained as the mixing ratio.

  Based on the mixing ratio, a subtraction signal is calculated, and in the case of two output signals, it is given as follows.

  Here, N = 1, 2,. N is the index of the source signal to be filtered out.

  Depending on whether the mixing ratio is larger or smaller than 1, a subtraction signal is calculated accordingly. At this time, more than two output signals can be used, but doing so simply results in higher computational costs. If there are not only two output signals, as in the case of a stereo signal, but more output signals, it is therefore preferable that the source signal appears most strongly to extract the source signal. One output signal is selected.

  In any of the subtraction signals calculated in this way, one source signal is shielded.

  These subtracted signals are then Fourier transformed. This is done in a block fashion, where subsequent blocks always start with a width corresponding to half the size of the data point being converted earlier. This means that the first half of one block has already been Fourier transformed as the second part of the preceding block.

  This method enables continuous signal processing and is known by the name of the superposition addition method.

  In order to minimize the leakage effect, the input block is multiplied by a window (eg, Hanning-Fister). The window function f (n) for N data points is:

Fourier transformed subtraction signal X ^{~ (hereinafter} referred to as "conversion signal".) From the remainder signal is calculated. To two converted signals X ^~, remainder signal is provided as from the first converted signal simply by subtracting the second converted signal. That is,

In the following, starting from two conversion signals X ^~ ₁ , X ^~ ₂ and one remainder signal X ^~ ₃ , this gives a total of three signals. That is why these three signals are compared with each other to extract one individual signal.

For each data point of the signal _{^{X ~ 1, X ~ 2,}} X ~ 3, usually, each signal has a data point 4096, extrema amplitude of the three signals is calculated.

Start with an array with 3x4096 data points. The numeral 3 indicates the number of converted signals and remainder signals, and 4096 indicates the number of data points that have undergone Fourier transform in one block. Focusing on the vector _X ^{~ _1,} X ^{~ _2,} the first frequency bin to the first data point of the X ^~ _3, determines the three power for comparison. At the minimum of these three values, the maximum value of the other two values is set, and the other values are set to zero.

  To illustrate this clearly, a numerical example is given.

The first value of the positions X ^to ₁ is 5, the first value of X ^to ₂ is 10, and the first value of X ^to ₃ is 15. Then, the numerical value 15 is set at 5 and the first numerical values of X ^to ₂ and X ^to ₃ are set to 0. Thus, numerical values are considered for each column. An array of 4096 columns and three rows is obtained, in which two-thirds of the values are zero. Values that are not equal to zero are randomly distributed across the individual vectors. Individual vectors for X ^~ ₁ and ^X _{~ 2} ^is referred to as ^S _{~ 1} and ^S _{~ 2,} the source signals _S 1 after the computation, _{S 2} is what is Fourier transform. The resulting vector of remainder value signals X ^- ₃ is not important any more.

The calculation of S ^~ ₁ and S ^~ ₂ is given by the following equation.

  Here, k is the number index of data points or frequency bins, and the value ranges from 1 to 2048 in the case of data points. The reason why it should be only half of the frequency bin is that, due to symmetry, the data points are doubled with the Fourier transform. In the above example, k = 1.

Thus transformed row (These were not correspond to the signal ^{_{X ~ 1, X ~ 2.}} ), The original that is, the use of individual signal ^S _~ ^{1, S} _{~ ^2,} which, The calculated source signals S1, S2 can be calculated by inverse Fourier transform. As the phase, signal ^X _~ ^{1, X} _{~ 2} phase can also be considered as it is. The S ^~ ₁ and ^S _{~ 2} therefore assigned each phase from the vector ^X _~ ^{1, X} _{~ 2.} This assignment is naturally performed based on the term index k.

Individual signal S ^{~ _1,} S ^~ ₂ or Fourier transformed individual signal S _1, S ₂ may look something a little different from the source signal for the calculation accuracy and calculation errors. That is, complete reproduction of the source signal is certainly not achieved, but the difference is subtle and usually not noticeable.

  In order to improve the separation of the individual signals, the following steps are possible.

In order to avoid jumps in recognizing the minimum, the minimum of the signals X ^to ₁ , X ^to ₂ , and X ^to ₃ may be determined according to the minimum recognizing performed previously. For example, a conditional low-pass filter may be used.

P ^t _hold (k) is the value of interest of frequency bin k, which is again the term index. The parameter b can be freely set between 0 and 1, and frequency sensing is cut off when b = 0.

Calculation of S ^~ ₁ and S ^~ ₂ is then performed by the following equation.

The parameter η (0 ≦ η ≦ 1) gives the intensity at which the signal passed through the low-pass filter enters the individual signals S ^to _m , m = 1, 2, 3,.

Further, the minimum of the converted signals X ^to _m , m = 1, 2, 4, 5,... Can be separately compared with the remainder value signals X ^to ₃ , that is, the minimum specific E _min (k), The assignment of 0 or the maximum value E _max (k) of the column is performed between the conversion signals X ^to _m and the remainder value signal, respectively. In this way, the vectors E _min1 , E _min2 , E _max1 , E _max2 are calculated. The minimum value is multiplied by a factor β (0 ≦ β ≦ 2). Depending on the choice of factors, different effects occur. For β <1, undesirable frequencies are suppressed, and for β> 1, harmonious tone formation is obtained.

  The individual signals are as follows.

In addition, the signals can be separated based on the phase state. When the amplitudes of the converted signals X ^to _m (k) are substantially the same and the remainder value signals X ^to ₃ (k) have a minimum value, the phase is considered. If these phases are the same, the maximum value is assigned to S ^~ ₃ (k), otherwise it is assigned to S ^~ ₁ (k) and S ^~ ₂ (k). This concept can of course be implemented separately for each frequency bin k.

  The following holds as appropriate.

In addition, the Fourier transformed output signals X ^to _{m, org} can be further taken into account. When the minimum value is in X ^~ _{1, org} (k), these are assigned to S ^~ ₁ (k) in the same way.

  The invention is explained in more detail on the basis of the embodiments described in the drawings.

Fig. 2 is a block diagram of a method according to the present invention for a stereo signal. It is a block diagram of the method by this invention in 2nd embodiment. It is a flowchart for specifying a mixing ratio. It is a flowchart for signal source separation. It is a figure which shows the structure for recording a stereo signal. It is a scatter diagram of data pairs. It is a scatter diagram of FT data pairs. It is a figure which shows a histogram. It is a figure for demonstrating the superposition addition method. It is a figure which shows the 3D amplitude spectrum of two conversion signals and one remainder signal. It is a figure which shows the 3D amplitude spectrum of an individual signal. It is a figure which shows a conversion signal and an individual signal in a vector format.

  FIG. 5 shows a configuration for recording a stereo signal. By way of example only, a configuration for recording a source signal by the XY-stereo method is shown, and basically the method according to the invention can be used not only in all other stereo technologies, The present invention can also be applied to a method in which more than two output signals are generated.

  Shown are four signal sources, which generate one source signal 1, 2, 3, 4 respectively. Source signal 1 is for a single singing voice, for example a single male or female singer, and source signal 2 is for a back chorus consisting of a large number of male or female singers, but with the same lyrics in the chorus. The source signal 3 is for a musical instrument, for example a piano, and the source signal 4 is for a group of instruments, although the same score is played as in the chorus. is there. An example of this case may be a violin unit playing the same melody. This example shows that one source signal can not consist of a single male singer, a single female singer, or a single instrument, but a singer or instrument. It can be composed of a large number of. Due to the wide directivity of the microphones 5, 6, the source signals 1, 2, 3, 4 are recorded at different levels, so that the mixing ratio of the source signals 1, 2, 3, 4 is always different.

  From the source signals 1, 2, 3, 4 there are thus obtained two output signals of a stereo signal. These output signals are just the starting points for the method according to the invention, so that the original source signal is not used any further.

Of course, can be used to generate more source signal also output signals X _m, in carrying out this method, the meaningful, but at least two source signals is required.

In the following, the two output signals X ₁ and X ₂ are referred to many times. However, the method according to the present invention is not limited to a stereo signal, and can basically be used for any number of output signals X _m , m = 1, 2, 3,.

FIG. 1 shows a block diagram according to a first embodiment of the method according to the invention. In the present embodiment, the two output signals X ₁ and X ₂ are considered, and the mixing ratio for the two source signals (for example, the source signals 1 and 2) is specified. One possible way to specify the mixing ratio is detailed later. For example, a mixing ratio V ₁ occurs for the source signal _{1 and a} mixing ratio V ₂ occurs for the source signal 2. Output signal _X 1, then subtracting the signal from the _{X 2 X ^ 1, X ^} 2 is calculated, at that time, the output signal _X 1, _{X 2} is subtracted based on the to mixing ratio as follows:

  Here, N represents each index of the source signal.

The subtraction signals X ^ ₁ and X ^ ₂ calculated in this way are Fourier transformed by the superposition addition method.

The output signals X ₁ and X ₂ are, of course, digital data and correspond to a simple sequence of numerical values and data points. The output signal can also be expressed as a vector having a very large number (usually tens of thousands). The same applies to the subtraction signal. The output signals X ₁ , X ₂ or the subtraction signal are processed further by the block in order to keep the demands on the calculation low and in particular to keep the time until the first individual signal is obtained short. The Fourier transform is applied to the first 4096 data points of the subtraction signals X ₁ to X ₂ , for example. Preferably, a power of 2 data points are Fourier transformed. This is because fast Fourier transform (FFT) can be applied in this example. At this time, the number 4096 is an ideal number in consideration of the calculation time and calculation resources consumed. Smaller data blocks or larger data blocks (eg, data blocks of 1024 to 2048 data points) may be employed. In this case, of course, it always means consecutive data points. By converting the subtracted signals X ^ ₁ , X ^ ₂ by Fourier transform, converted signals X ^- ₁ and X ^- ₂ are obtained. By subtracting the conversion signals X ^to ₂ from the conversion signals X ^to ₁ , the remainder signals X ^to ₃ are obtained.

The remainder signals X ^- ₃ include as many data points (for example, 4096 data points) as the converted signals X ^- ₁ and X ^- ₂ . Since these data points are in the frequency domain, they are usually also called frequency bins. The frequency bin is thus the data point of the transformed signal or residue signal. In a specific case, each converted signal / residue signal thus has 4096 frequency bins.

Next, the minimum is sought in the signal ^{_{X ~ 1, X ~ 2,}} X ~ 3, This is because with the same meaning as they have been shielded signal. Therefore, for each frequency bin, the minimum calculated from the amplitude of the signal ^{_{X ~ 1, X ~ 2,}} X ~ 3, at its minimum, maximum value among the respective frequency bins are set, these frequencies The other value for the bin is set to zero. The phase for a value different from zero is obtained, thus obtaining the individual signals S ^~ ₁ , S ^~ ₂ . These individual signals must be converted back to the time domain in order to obtain the calculated source signals S ₁ , S ₂ .

FIG. 2 shows an embodiment for calculating more source signals than the two calculated source signals S ₁ , S ₂ . For this embodiment, more than two mixing ratios are specified from the output signals X ₁ , X ₂ , X ₃ , X ₄ , thereby obtaining more subtraction signals and more conversion signals accordingly. It is done. In the converted signals X ₁ ^to ₁ , X ^to ₂ , X ^to ₄ , X ^to ₅ obtained in this way and the remainder signals X ^to ₃ calculated therefrom, the minimum is also specified, thereby individual signal ^{_{S ~ 1, S ~ 2,}} S ~ 3, S ~ 4 are calculated from these, is calculated source signals _{S 1, S 2, S 3} , the calculation of _{S 4} is performed.

  In the following, one possible method for specifying the mixing ratio is shown. Basically, any method can be used to specify the mixing ratio.

FIG. 3 shows a flow diagram for specifying the mixing ratio based on specifying the direction. Here, in step S1, the output signal _X 1, _{X 2} is the Fourier transform, which output signal is converted by the ^X _~ 1, _^org, the _{X ~ 2, org} obtained. In step S2, the amplitude value based on the following equation Fourier transformed output signals ^X _~ 1, _^org, against _{X ~ 2,} each data point or frequency bins _org is calculated.

In the next step S3, the angle θ and the magnitude Ψ l are calculated from the previously calculated values. At this time, the Fourier transformed output signals X ~ 1, org, calculation of X ~ 2, org of magnitude, by the data points or is carried out as a vector, in any event, and now the angle θ size A pair of values of Ψ l can be calculated.

In step S4, a histogram is calculated from the pair of values of θ and Ψ _l and this histogram is flattened in step S5.

  In step S6, several maximum histograms are identified by taking up a maximum number that exceeds a predetermined number or a threshold value.

  Each of these maximums is assigned an angle calculated from the maximum value according to the formula given above. Next, in step S7, the required mixing ratio is specified based on the calculated angle.

FIG. 4 shows a flow diagram for signal source separation. When converted signal ^X _~ ^{1, X} _{~ 2} and further ^X _{~ 4} and ^X _{~ 5} and the remainder value signals ^X _{~ 3} is optionally, minimum and maximum are computed for each frequency bin (step S8) . In other embodiments, further Fourier transformed output signals ^X _~ 1, _^org, by _{X ~ 2, org} and optionally ^X _{~ 3, org,} and ^X _{~ 4, org} is considered.

In step S9, as described above, the frequency bin or the maximum value of each column is set at the minimum in each frequency bin, and zero or a hold value P ^t _hold is assigned to all other values. It is done.

According to step 10, the converted signals X ^to ₁ , X ^to ₂ , X ^to ₄ , and X ^{to the} respective values different from zero of the individual signals S ^to ₁ , S ^to ₂ , S ^to ₄ , and S ^to _5. Each of the _five phases is assigned.

At this time, allocation is performed based on frequency bin numbering. When the frequency bin 28 of the individual signal S ^~ ₁ has a value different from zero, the individual signal obtains the phase of the frequency bin 28 of the transformed signal X ^~ ₁ . Since the maximum value is moved to the minimum value in each frequency bin and the remaining values are made zero, no value of the individual signal is the same as the value of the assigned conversion signal. Still the phase can be taken over.

Instead of setting to zero where it is not the minimum, the hold value ηP _hold can also be set. As a result, there is no place where the individual signal is skipped.

FIG. 6 shows a scatter diagram of the amplitudes of the two output signals X ₁ and X ₂ in a mixed state. In this example, the output signals X ₁ and X ₂ are sine wave signals that are shifted from each other. Each scatter diagram represents a pair of values (two items of values). At this time, the pair of values is formed based on the number of data points, that is, the number of data-bins. The first point of the data points of the output signal X ₁ has the first amplitude, the first data point of the output signal X ₂ have the same or different amplitudes. These two amplitudes are used to plot a single point in scatter plot 9. By using a plurality of data points for the output signal X ₁ and the output signal X ₂ , the same number of data point pairs are generated. At this time, it displays the amplitude of the output signal X ₁ on shaft 10, the amplitude of the output signal X ₂ on the shaft 11 are displayed. For example, using 4096 data points on output signal X ₁ output signal X ₂ yields 4096 data point pairs. These form a point cloud 12 in the scatter plot, which is formed from a plurality of points 13 determined from individual data point pairs.

FIG. 7 shows a correspondingly formed scatter diagram 14 except that the output signals X ₁ and X ₂ were Fourier transformed and the signal amplitudes were calculated as follows:

Accordingly, the shaft ^15, the _{X ~ 1, org,} that is, displays the magnitude of the amplitude but the output signal X1 is Fourier transform, the axis ^16, the amplitude of the _{X ~ 2, org} size Is displayed. By calculating the slope of the straight lines 17 to 18, the mixing ratio is obtained for each.

In this simple manner, the mixing ratio can be calculated only when this is not a perfect sine wave signal in actual recording. Instead of obtaining the inclination of the obtained straight line to create a scatter plot, the output signals X1, X2 after Fourier transform, the Fourier transformed output signals X ~ 1, org, from X ~ 2, org points pairs , Respectively, an angle θ and a vector ψ are calculated. Instead of a scatter diagram, a histogram is obtained from a pair consisting of an angle θ and a vector Ψ, and the angle is reduced to an integer value in the histogram, and a corresponding frequency in the angle is calculated therefrom. From each point pair, an individual frequency for the histogram results.

  The histogram is flattened using a flattening function to reduce the number of local minimum and maximum values. As the flattening function, for example, the following function described above can be applied.

  A histogram flattened in this way is shown in FIG. The angle is shown on the axis 19 in degrees, ie from 0 ° to 90 °, while the frequency is shown on the axis 20 respectively. The histogram 21 has an absolute true maximum 22 and a plurality of local maximums 23, 24 and 25.

  When ordered according to the number of frequencies, a maximum of 22 has the largest value, followed by a maximum of 25, 24 and 23. Of these maxima, a pre-determined number is selected from the head for the most frequent maxima, for example, two maxima, maxima 22 and 25 are selected, or the frequency or threshold is exceeded. You can continue to select the maximum as long as possible. If this threshold value is set to 10, for example, the maximum 22, 25, and 24 are applicable in the subsequent calculation, but the maximum 23 below the threshold is not applicable.

  Each maximum has one angle. The maximum 22 is, for example, 18 °, and the maximum 25 is 72 °.

  At this time, the angle is given by the following equation.

  From the angles thus identified, one mixing ratio for each angle is obtained by the following equation.

  N is the index of the source signal to be extracted. Thereby, the specification of the mixing ratio based on the direction information is completed.

The calculation of the subtraction signals X ^ ₁ and X ^ ₂ is performed by the above-described equation. In particular, the subtraction signals X ^ ₁ and X ^ ₂ can be calculated in blocks. Thus can be performed for each set of predetermined for the data points, for example, each of 4096 data points of the output signals X _1, X ₂ is calculated in the subtraction signal _X ^ _1, X ^ ₂ by using the mixing ratio The The subtraction signals X ^ ₁ and X ^ ₂ are Fourier-transformed by a superposition addition method as shown in FIG.

Shown are data blocks 26, 27, 28, 29 consisting of 2048 data points of subtraction signal X ₁ or subtraction signal X ₂ , respectively. Here, the number 2048 comes from where it is half of the data points used for the Fourier transform. The number of data points in the data blocks 26-29 thus comes from the number of data points in the data blocks 30-33.

  The data block 30 is composed of data blocks 26 and 27 and has a number corresponding to twice the number of data points of the data block 26 or 27. This data block 30 is a data block to be Fourier transformed. The data block 31 includes data blocks 27 and 28, and the data block 32 includes a data block 28 and a data block subsequent thereto. The data block 27 is included in the data blocks 30 and 31, and the data block 28 is included in the data blocks 31 and 32.

  Prior to the Fourier transform, the data blocks 30, 31, 32 and 33 are further multiplied by a Hanning window. Thereby, the leakage effect in the Fourier transform can be minimized.

FIG. 10 shows a 3D-amplitude spectrum of the converted signals X ^- ₁ and X ^- ₂ residue value signals X ^- ₃ . On the axis 35, each frequency bin number, that is, a point assigned a serial number in the signal vector, is shown, while the axis 36 shows the size. Each size is displayed on the axis 37. Position 38 has signals | X ^to ₂ |, position 39 has signals | X ^to ₃ |, and position 40 has signals | X ^to ₁ |.

The signal shown in FIG. 10 was derived from two pure sine wave signals. In an actual signal, more or less all frequency bins are full, but a sine wave signal is particularly suitable for easy viewing. As can be seen immediately from what is shown in FIG. 10, the signals | X ^to ₁ |, | X ^to ₂ |, | X ^to ₃ | each have only three different height peaks, while the others Then it has only a zero state.

Figure 11 shows a 3D- amplitude spectrum of the individual signal ^S _~ ^{1, S} _{~ 2} residue signal mixing ^S _{~ 3.} The axis 35 again displays frequency bins and the axis 36 displays size. On the axis 42, the individual signals contained therein are given. The position 43 has an individual signal | S ₁ ^to ₁ |, the position 44 has a remainder signal | S ^to ₃ |, and the position 45 has an individual signal | S ^to ₂ |. Individual signal ^| _S ~ 1 ^{|, |} order to be able to reverse Fourier transform, discrete signal _{^{| | S ~ 2 S ~ 1}} |, | S ~ 2 | to each different position from the zero, converted signal ^{X ~} Further, a phase can be applied from ₁ and X ^to ₂ , respectively.

FIG. 12 schematically shows that the individual signals are acquired from the converted signals X ^- ₁ and X ^- ₂ and the remainder value signals X ^- ₃ . Vector 46, in this example, converted signal X ^~ is partially shown here ones _1, vector 47 in response to this, the numerical value of the converted signal X ^~ _2, vector 48, the remainder signal X a ^value between ₃ and vector 49, shows the corresponding frequency bins.

From these, the individual signals S ₁ ^to ₁ shown in the vector 50, the individual signals S ₂ ^to ₂ shown in the vector 51, and the remainder value signals S ₃ ^to ₃ shown in the vector 52 come out as follows.

  For frequency bin 1, the number of 5 and 7 is greater than 1, so the minimum is in vector 48. In frequency bin 1, the corresponding vectors 50, 51, 52 are thus set to the following values: That is, at the position of the minimum value 1, the maximum value 7 is set. This value must therefore be set to frequency bin 1 of vector 52. The remainder of frequency bin 1, ie the values in vectors 50 and 51, are set to zero. Looking at the minimum is thus done column by column, data point or frequency bin. These expressions are the same.

  Correspondingly, the column values for frequency bins 2, 3, 4, 5, etc. are calculated, resulting in the values shown in FIG.

  Vector 50 collects values or minimums for vector 46, vector 51 collects for vector 47, and vector 52 collects for vector 48.

The phases of the converted signals X ^- ₁ and X ^- ₂ are still assigned to the vectors 50 and 51 of the individual signals S ^- ₁ and S ^- ₂ . The frequency bin 4 of the vector 50 will be described by taking this point as an example only. For the frequency bin 4 of the vector 50, the phase of the frequency bin 4 is obtained from the vector 46. This is because vector 50 forms the minimum of vector 46, and therefore there is a numerical value there. Correspondingly, the same frequency bin phase is passed.

  In all the embodiments of the figure, more than two signals can be used instead of the two output signals, the two converted signals or the individual signals explicitly mentioned above, and further as described at the beginning of the specification. All embodiments can also be used in view of the description with respect to the figures, and thus this point is not explicitly excluded.

1 source signal 2 source signal 3 source signal 4 source signal 5 microphone 6 microphone 7 mixing ratio specification 8 signal separation 9 scatter diagram 10 axis 11 axis 12 point cloud 13 point 14 scatter diagram 15 axis 16 axis 17 straight line 18 straight line 19 axis 20 axis 21 Histogram 22 Maximum 23 Maximum 24 Maximum 25 Maximum 26 Data block 27 Data block 28 Data block 29 Data block 30 Data block 31 Data block 32 Data block 33 Data block 34 3D-Amplitude spectrum 35 Axis 36 Axis 37 Axis 38 Position 39 Position 40 Position 41 3D-Amplitude Spectrum 42 Axis 43 Position 44 Position 45 Position 46 Vector 47 Vector 48 Vector 49 Frequency Bin 50 Vector 51 Vector 52 Vector

Claims (20)

At least at least two output signals are formed from two source signal _{(1,2,3,4) (X 1, X} 2, X 3, X 4) at least two separate signals from ^{_(S ~} 1, _S ~ ₂ , S ^to ₃ , S ^to ₄ )
a) Mixing ratios (V ₁ , V ₂ , V ₃ , V ₄ ) of at least two source signals ( ₁ , ₂ , ₃ , ₄ ) on the output signals (X ₁ , X ₂ , X ₃ , X ₄ ) ) Is identified,
b) The output signal (X ₁ , X ₂ , X ₃ , X ₄ ) is multiplied by the factor formed from the mixing ratio (V ₁ , V ₂ , V ₃ , V ₄ ), and the subtraction signal (X The output signals (X ₁ , X ₂ , X ₃ ) so that one source signal (1, ₂ , ₃ , 4) is eliminated in each of {circumflex over ()} ₁ , X ^ ₂ , X ^ ₃ , X ^ ₄ , X ₄ ), at least two subtraction signals (X ^ ₁ , X ^ ₂ , X ^ ₃ , X ^ ₄ ) are calculated,
c) the subtraction signal _{(X ^ 1, X ^ 2} , X ^ 3, X ^ 4) is converted signals by being Fourier transform ^{_{(X ~ 1, X ~ 2}} , X ~ 4, X ~ 5) occurs ,
d) A remainder signal (X ^- ₃ ) is determined from the converted signals (X ^- ₁ , X ^- ₂ , X ^- ₄ , X ^- ₅ ),
e) the converted signal ^{_{(X ~ 1, X ~ 2}} , X ~ 4, X ~ 5) and the at least two separate signals on the basis of a remainder signal ^{_{(X ~ 3) (S ~}} 1, S ~ 2, S ~ ₃ , S ^~ ₄ ) are calculated,
f) said individual signals ^{_{(S ~ 1, S ~ 2}} , S ~ 3, S ~ 4) is converted by an inverse Fourier transform, to calculate the processed source signal _{(S 1, S 2, S} 3, S 4) A method characterized by being made. The method of claim 1, wherein
The method is characterized in that the mixing ratio (V ₁ , V ₂ , V ₃ , V ₄ ) is specified by specifying a direction. The method according to claim 1 or 2,
The specification of the mixing ratio (V ₁ , V ₂ , V ₃ , V ₄ )
Fourier transforming the output signals (X ₁ , X ₂ , X ₃ , X ₄ );
Calculating an amplitude value of the Fourier transformed output signals ^{_{(X ~ 1, org, X}} ~ 2, org, X ~ 3, org, X ~ 4, org),
Obtaining a histogram (21) from the amplitude value, or specifying a slope of a straight line (17, 18) of the amplitude value in the scatter diagram (14);
A method characterized by comprising: The method of claim 3, wherein
The histogram (21) is multiplied by a flattening function. The method of claim 4, wherein
As a flattening function, in particular

A hanning window is used. A method according to any one of claims 3 to 5,
Method according to claim 1, characterized in that the angular interval in the histogram (21) is 1 °. The method according to any one of claims 3 to 6, wherein
Method according to claim 1, characterized in that a local extreme value of the histogram (21), in particular a local maximum (22, 23, 24, 25) is specified. The method of claim 7, wherein
A method characterized in that one angle is assigned to each maximum (22, 23, 24, 25) starting from an absolute true maximum (22). The method of claim 8, wherein
The method according to claim 1, wherein the angle assignment is terminated when a predetermined number is reached or when the angle is less than a threshold value. The method according to any one of claims 3 to 9,
Method characterized in that the histogram (21) is weighted by a low pass. The method of claim 10, wherein
The low pass is identified from a histogram previously determined in time. 12. The method according to any one of claims 1 to 11, wherein
Calculated two output signals _(X 1, _{X 2)} the subtraction signal when the _{(X ^ 1, X ^ 2} , X ^ 3, X ^ 4) is, when the mixing ratio is smaller than 1, the mixing ratio The second output signal (X ₂ ) multiplied by (V _N ) is subtracted from the _first output signal (X ₁ ), and otherwise, the first output signal (X ₁ ) A method characterized in that the reciprocal of the mixing ratio (1 / V _N ) is multiplied and subtracted from the second output signal (X ₂ ). The method according to any one of claims 1 to 12,
Method for the Fourier transform, characterized in that a power of 2 is used for the data points, in particular 1024, 2048 or 4096 data points. 14. A method according to any one of claims 1 to 13,
The method, characterized in that the data points for the Fourier transform are taken into account according to a quasi-continuous and superposition method. 15. A method according to any one of claims 1 to 14,
A remainder value signal (X ^- ₃ ) is defined as a subtraction signal of the conversion signal. The method according to any one of claims 1 to 15,
At least one transformed signal ^{_{(X ~ 1, X ~ 2}} , X ~ 4, X ~ 5) and / or the individual signals by the specific extreme remainder value signal ^{_{(X ~ 3) (S ~}} 1, S ~ 2, S ^~ ₃ , S ^~ ₄ ) are defined. 17. A method according to any one of claims 1 to 16,
The frequency of the individual signals (S ^~ ₁ , S ^~ ₂ , S ^~ ₃ , S ^~ ₄ ) is identified by specifying a minimum. The method according to claim 16 or 17, wherein
Within the converted signals (X ^to ₁ , X ^to ₂ , X ^to ₄ , X ^to ₅ ) and the remainder signal (X ^to ₃ ), the individual signals (S ^to ₁ , S ^to ₂ , S ^to ₃ , S) ^~ ₄ ) to generate, for each frequency bin (49), the minimum is found, the maximum value of each frequency bin (49) is set at that position, and the remaining value is set to zero A method characterized by being made. The method of claim 18, wherein
Phases of the converted signals (X ^- ₁ , X ^- ₂ , X ^- ₄ , X ^- ₅ ) with respect to the amplitude values of the individual signals (S ^- ₁ , S ^- ₂ , S ^- ₃ , S ^- ₄ ) A method characterized in that a value is determined. 20. A method according to any one of claims 1 to 19,
The number of source signals (S ₁ , S ₂ , S ₃ , S ₄ ) to be calculated and / or one or more source signals ( ₁ , ₂ , ₃ , ₄ ) themselves are given in advance. A method characterized by that.

Images