Electronic device and control method thereof

Abstract

To provide an electronic apparatus that enables independency of a plurality of signals to be easily grasped, and a control method therefor.SOLUTION: An electronic apparatus 10 includes: a control unit 12 that calculates independency numeric values representing independency of a plurality of signals from one another; and a sample acquisition unit 14 that acquires sample data on each of the plurality of signals. The control unit 12 calculates the independency numeric values based on the sample data on each of the plurality of signals.SELECTED DRAWING: Figure 4

Description

Technical Field

[0001] The present disclosure relates to an electronic device and a control method thereof.

Background Art

[0002] Conventionally, a device that accurately separates each component sound from a mixed signal in which each component sound is mixed has been known (see, for example, Patent Document 1).

Prior Art Documents

Patent Documents

[0003]

Patent Document 1

Summary of the Invention

Problems to be Solved by the Invention

[0004] In order to separate a mixed signal, it is required to grasp the independence of a plurality of signals included in the mixed signal.

[0005] The present disclosure has been made in view of the above points, and an object thereof is to provide an electronic device and a control method thereof that can easily grasp the independence of a plurality of signals.

Means for Solving the Problems

[0006] An electronic device according to some embodiments includes a control unit that calculates an independence value representing the mutual independence of a plurality of signals, and a sample acquisition unit that acquires sample data of each of the plurality of signals. The control unit calculates the independence value based on the sample data of each of the plurality of signals. By doing so, the independence of a plurality of signals can be easily represented as a numerical value based on the sample data. As a result, the independence of a plurality of signals can be easily grasped.

[0007] In an electronic device according to an embodiment, the control unit may calculate a plurality of single-signal sample expected values based on the sample data of each of the plurality of signals, calculate one product-signal sample expected value based on the product of the sample data of each of the plurality of signals, and calculate the independence value based on a difference value representing, as a positive value, the difference between the product of the single-signal sample expected values and the product-signal sample expected value. By doing so, the independence of a plurality of signals can be evaluated by a finite number of sample data. As a result, the independence of a plurality of signals can be easily grasped.

[0008] In an electronic device according to an embodiment, the control unit may calculate the single-signal sample expected value and the product-signal sample expected value at a plurality of orders, calculate the difference value at each order, and calculate the sum of the difference values at each order as the independence value. By doing so, the correlation between the independence value and the independence of a plurality of signals can be enhanced. As a result, the independence of a plurality of signals can be easily grasped.

[0009] In an electronic device according to an embodiment, the control unit may use, as the difference value, a value obtained by multiplying the difference value calculated at each order by a weighting coefficient of each order. By doing so, the correlation between the independence value and the independence of a plurality of signals can be enhanced. As a result, the independence of a plurality of signals can be easily grasped.

[0010] In an electronic device according to an embodiment, the control unit may use, as the difference value, the power of the reciprocal of each order of the difference value calculated at each order. By doing so, the correlation between the independence value and the independence of a plurality of signals can be enhanced. As a result, the independence of a plurality of signals can be easily grasped.

[0011] In an electronic device according to an embodiment, the control unit may use, as the difference value, the square or absolute value of the difference between the product of the single-signal sample expected values and the product-signal sample expected value. By doing so, the correlation between the independence value and the independence of a plurality of signals can be enhanced. As a result, the independence of a plurality of signals can be easily grasped.

[0012] In an electronic device according to an embodiment, when the independence value is equal to or less than a predetermined threshold, the control unit may determine that the plurality of signals are independent of each other. By doing so, it becomes easier to determine whether the plurality of signals are independent of each other. As a result, the independence of the plurality of signals can be easily grasped.

[0013] An electronic device according to an embodiment may further include an output unit that outputs a determination result as to whether the plurality of signals are independent of each other. By doing so, the user can quickly understand the independence of the plurality of signals. As a result, the independence of the plurality of signals can be easily grasped.

[0014] In an electronic device according to an embodiment, when the independence value is equal to or less than a predetermined threshold, the control unit may separate a mixed signal including the plurality of signals by independent component analysis, and when the independence value is greater than the predetermined threshold, the control unit may separate a mixed signal including the plurality of signals by frequency analysis. By doing so, it becomes easier to perform analysis of observation signals by independent component analysis. Also, when independent component analysis is executed with a smaller computational load than frequency analysis, since independent component analysis is likely to be used, separation of signals can be simplified.

[0015] A control method for an electronic device according to some embodiments includes obtaining sample data of each of a plurality of signals, and calculating an independence value representing the mutual independence of the plurality of signals based on the sample data of each of the plurality of signals. By doing so, the independence of the plurality of signals can be easily represented as a value based on the sample data. As a result, the independence of the plurality of signals can be easily grasped.

[0016] A control method for an electronic device according to an embodiment further includes determining that the plurality of signals are independent of each other when the independence value is equal to or less than a predetermined threshold. By doing so, it becomes easier to determine whether the plurality of signals are independent of each other. As a result, the independence of the plurality of signals can be easily grasped.

[0017] The control method of an electronic device according to an embodiment further includes separating a mixed signal including the plurality of signals by independent component analysis when the independence value is equal to or less than a predetermined threshold, and separating the mixed signal including the plurality of signals by frequency analysis when the independence value is greater than the predetermined threshold. By doing so, it becomes easier to perform analysis of the observed signals by independent component analysis. Further, when independent component analysis is executed with a smaller computational load than frequency analysis, the separation of signals can be simplified because independent component analysis is likely to be used.

Effect of the Invention

[0018] According to the present disclosure, there are provided an electronic device capable of easily grasping the independence of a plurality of signals and a control method thereof.

Brief Description of the Drawings

[0019]

Figure 1

Figure 2

Figure 3

Figure 4

Figure 5

Figure 6

Figure 7

Figure 8

Figure 9

Figure 10

Figure 11

Embodiments for Carrying Out the Invention

[0020] By observing a signal obtained by mixing a plurality of source signals under a plurality of conditions and analyzing the signals observed under each condition, the source signals before mixing can be estimated. That is, from a signal obtained by mixing a plurality of source signals, a signal separated into signals corresponding to the source signals can be estimated. A signal obtained by mixing a plurality of source signals is also referred to as a mixed signal. A signal obtained by observing a mixed signal is also referred to as an observed signal. A signal separated from a mixed signal into signals corresponding to the source signals is also referred to as a separated signal. As an analysis method, frequency analysis or independent component analysis (ICA) can be used. Frequency analysis can generate a separated signal from an observed signal using various means such as, for example, a beamformer or a frequency filter.

[0021] Independent component analysis is a statistical method that estimates source signals from observed signals and generates the estimation results as separated signals. An observed signal is a signal obtained by observing a mixed signal obtained by mixing a plurality of source signals, and is represented as a signal obtained by transforming a plurality of source signals with a mixing matrix A. Also, a separated signal is a signal obtained by estimating source signals from observed signals, and is represented as a signal obtained by transforming a plurality of observed signals with a separation matrix W. Independent component analysis can generate a separated signal from an observed signal by estimating the separation matrix W. Also, independent component analysis estimates the separation matrix W so that the error of the separated signal with respect to the source signal is reduced. Since independent component analysis does not require any information about the source signals, such as the frequency characteristics of the source signals, the spatial position information of the source signals, or the mixing conditions of the source signals, it can perform calculations with a smaller amount of information than frequency analysis, which requires information about the source signals, and can reduce the computational load.

[0022] Independent component analysis can be used in various fields such as the analysis of a mixed sound source containing multiple sound sources, for example, the analysis of life information sequences such as DNA sequences or amino acid sequences, the analysis of magnetic field signals such as brain signals, or the recognition of characteristic patterns included in images. Also, independent component analysis can be executed with a computational load less than that of frequency analysis.

[0023] In independent component analysis, assuming that the following matters (1) to (3) hold, even when the frequency or amplitude of the source signal is unknown, the separation matrix W can be estimated so that the error of the separated signal with respect to the source signal becomes small. That is, assuming that the following matters (1) to (3) hold, the estimation accuracy of the separated signal can be enhanced. (1) The number n of source signals constituting the mixed signal is known. In other words, the mixing matrix A and the separation matrix W are n×n regular matrices. (2) The source signals are statistically independent of each other. (3) Each source signal is generated from a non-Gaussian probability distribution.

[0024] However, independent component analysis can be used even when the above matters (1) to (3) do not hold for the source signal. Therefore, the separated signal generated by independent component analysis is not necessarily a highly accurate estimation result.

[0025] (Comparative Example) Here, it can be considered to directly confirm the estimation accuracy of the separated signal generated by independent component analysis. As a comparative example, a method of confirming the estimation accuracy of the separated signal by independent component analysis by setting a statistical model (probability distribution) of the source signal can be mentioned.

[0026] In this method, in order to set up a statistical model that fully represents the statistical properties of the source signal, it is necessary to collect an unrealistically large number of sample data, almost infinite. On the other hand, when setting up a statistical model based on a finite number of sample data, problems may occur due to that statistical model. For example, suppose two sine waves with different frequencies are set as a statistical model. In this statistical model, when the phase difference between the two sine waves is changed, although the two sine waves are independent of each other in the statistical model, the validity of the estimation result by independent component analysis may not be obtained. Also, the estimation result by independent component analysis may vary significantly depending on the installation location of the observation device for observing the signal. Therefore, it is difficult to confirm the estimation accuracy of the separated signal by independent component analysis by setting up an unknown statistical model of the source signal.

[0027] Also, as another comparative example, a method of introducing an index correlated with the estimation accuracy of the separated signal can be considered. As an index likely to be correlated with the estimation accuracy of the separated signal, an index representing the independence of the source signal can be mentioned. As an index representing the independence of the source signal, an index called SUC (Symmetric Uncertainty Coefficient) can be used.

[0028] SUC can be calculated by the following procedure. First, as shown in FIG. 1, signal data over a predetermined period is acquired. In this signal data, the amplitude is divided into a finite number of bins (intervals). Then, a histogram counting the frequency of the amplitude of the signal in each bin is generated. The histogram is generated for each signal data whose independence is to be evaluated. Next, an approximate distribution is calculated from the histogram of each signal data. Then, an approximate mutual information amount is calculated based on the approximate distribution of each signal data. By normalizing the approximate mutual information amount, a normalized mutual information amount is calculated. The normalized mutual information amount indicates that a plurality of signal data are independent when its value is 0, and indicates that a plurality of signal data are in a completely dependent relationship when its value is 1. The normalized mutual information amount corresponds to SUC. Details regarding the calculation of SUC can be understood by referring to the following paper. A. Kraskov, H. Stogbauer, and P. Grassberger, “Estimating mutual information,” Physical Review E, vol. 69, no. 6, 066138, 2004.

[0029] Here, assume that a first signal and a second signal known as source signals are obtained. A separated signal is generated by independent component analysis from an observed signal obtained by observing a signal obtained by mixing the known first signal and second signal. By comparing the known source signal and the separated signal, the estimation accuracy of the separated signal can be confirmed.

[0030] The difference between the source signal and the separated signal can be visually represented, for example, by the scatter diagram shown in FIG. 2. In the scatter diagram of FIG. 2, the horizontal axis and the vertical axis respectively correspond to the first signal and the second signal. The scatter diagram of FIG. 2 is generated by plotting points corresponding to the amplitude of the first signal and the amplitude of the second signal at the same time for each time. The curve drawn in the scatter diagram of FIG. 2 corresponds to the Lissajous curve of an oscilloscope. In FIG. 2, the source signal and the separated signal completely overlap. In this case, it is determined that the separated signal is estimated to match the source signal. On the other hand, when there is a difference between the source signal and the separated signal, the curve representing the source signal and the curve representing the separated signal do not overlap at least in part.

[0031] It is calculated numerically how geometrically different the curve of the scatter diagram of the source signal and the curve of the scatter diagram of the separated signal are. That numerical value is also referred to as ICAerror. ICAerror becomes 0 when the source signal and the separated signal completely match. ICAerror can also be calculated as the square value of the Frobenius norm of the difference matrix (A - W -1 ) of the mixing matrix A and the inverse matrix of the separation matrix W estimated by independent component analysis.

[0032] ICA error is calculated as the difference between the known first and second signals and the separated signals by virtually generating observed signals for the known first and second signals and applying independent component analysis to the observed signals to generate separated signals. On the other hand, for the known first and second signals, the normalized mutual information is calculated as SUC. By calculating the normalized mutual information and ICA error for the known first and second signals, a graph representing the correlation between the normalized mutual information and ICA error is created as exemplified in FIG. 3. In FIG. 3, the horizontal axis and the vertical axis represent ICA error and SUC, respectively.

[0033] The points plotted on the graph exemplified in FIG. 3 correspond to each of 60×60 = 3600 combinations obtained by changing the frequencies of the sine waves of the first signal and the sine waves of the second signal one by one from 1 Hz to 60 Hz and combining them. For these points, the correlation coefficient between ICA error and SUC was calculated to be 0.077. That is, it can be said that ICA error and SUC are hardly correlated. Therefore, it is not possible to guarantee the estimation accuracy of independent component analysis by calculating SUC.

[0034] Furthermore, in order to create a histogram used for calculating SUC, it is necessary to adjust the bin size. Therefore, it is difficult to implement the calculation of SUC in a device.

[0035] According to the method according to the comparative example described above, it is difficult to determine the validity of the separated signals generated by analyzing the observed signals by independent component analysis. In other words, it is difficult to determine whether the conditions (1) to (3) to be satisfied for the estimation accuracy of independent component analysis to reach the desired accuracy are satisfied. Therefore, the present disclosure describes an apparatus and a method capable of easily grasping the independence of a plurality of signals and determining the validity of analyzing observed signals by independent component analysis.

[0036] (Configuration example of signal analysis system 1) As shown in FIG. 4, a signal analysis system 1 according to an embodiment includes an electronic device 10, a first observer 21, a second observer 22, a first signal source 31, and a second signal source 32. The first signal source 31 transmits a first signal 41. The second signal source 32 transmits a second signal 42. The first signal 41 and the second signal 42 are also referred to as source signals. The first observer 21 observes the first signal 41 or the second signal 42, or a mixed signal in which the first signal 41 and the second signal 42 are mixed, and outputs the observed signal as a first observation signal. The second observer 22 observes the first signal 41 or the second signal 42, or a mixed signal in which the first signal 41 and the second signal 42 are mixed, and outputs the observed signal as a second observation signal.

[0037] The electronic device 10 acquires the first observation signal from the first observer 21 and the second observation signal from the second observer 22. Based on the first observation signal and the second observation signal, the electronic device 10 estimates signals corresponding to the first signal 41 and the second signal 42 respectively from the mixed signal in which the first signal 41 and the second signal 42 are mixed, and outputs the estimated signals as a first separated signal and a second separated signal. The electronic device 10 performs independent component analysis or frequency analysis to estimate the separated signals.

[0038] Further, the electronic device 10 calculates a numerical value representing the mutual independence between the observation signal obtained by observing only the first signal 41 and the observation signal obtained by observing only the second signal 42. The numerical value representing the mutual independence of a plurality of signals is also referred to as an independence score (IS). The electronic device 10 determines whether to perform independent component analysis or frequency analysis based on the independence score in order to estimate the separated signal from the mixed signal in which the first signal 41 and the second signal 42 are mixed.

[0039] The electronic device 10 includes a control unit 12, a sample acquisition unit 14, an output unit 16, and an input unit 18.

[0040] The control unit 12 acquires information from each component of the electronic device 10 and controls each component. The control unit 12 may be configured to include a processor such as a CPU (Central Processing Unit). The control unit 12 may implement various functions of the electronic device 10 by executing a predetermined program.

[0041] The control unit 12 may include a storage unit. The storage unit may store various information used for the operation of the control unit 12 or a program for realizing the functions of the control unit 12. The storage unit may function as a working memory of the control unit 12. The storage unit may be composed of, for example, a semiconductor memory or the like. The storage unit may be configured separately from the control unit 12.

[0042] The sample acquisition unit 14 is communicably connected to the first observer 21 and the second observer 22 by wire or wirelessly. The sample acquisition unit 14 acquires the first observation signal output by the first observer 21 and the second observation signal output by the second observer 22 as digital signals sampled at a predetermined sampling period. The sample acquisition unit 14 may acquire the first observation signal and the second observation signal as analog signals.

[0043] The output unit 16 outputs the information acquired from the control unit 12. The output unit 16 may notify the user of information by outputting visual information such as characters, graphics, or images directly or via an external device or the like. The output unit 16 may include a display device or may be connected to a display device by wire or wirelessly. The display device may include various displays such as a liquid crystal display. The output unit 16 may also notify the user of information by outputting auditory information such as sound directly or via an external device or the like. The output unit 16 may include an audio output device such as a speaker or may be connected to an audio output device by wire or wirelessly. The output unit 16 may include a vibration device. The output unit 16 may notify the user of information not only through visual information, auditory information, or tactile information but also by outputting information that the user can perceive with other senses directly or via an external device or the like.

[0044] The input unit 18 may include an input device that receives input from a user. The input device may include, for example, a keyboard or physical keys, or may include a touch panel or touch sensor or a pointing device such as a mouse. The input device is not limited to these examples and may include various other devices.

[0045] In this embodiment, it is assumed that the first signal 41 and the second signal 42 are audio signals. It is assumed that the first signal source 31 and the second signal source 32 are speakers or a speaking person. It is assumed that the first observer 21 and the second observer 22 are microphones. The electronic device 10 is configured to separate each source signal from a mixed signal in which a plurality of audio source signals are mixed.

[0046] As shown in FIG. 5, the electronic device 10 includes, as an output unit 16, an indicator lamp 16a, a display 16b, and a speaker 16c. The electronic device 10 includes a cross key as an input unit 18. The electronic device 10 further includes microphones as the first observer 21 and the second observer 22. That is, the first observer 21 and the second observer 22 may be included in the electronic device 10. "Audio A" and "Audio B" displayed on the display 16b correspond to the first signal 41 and the second signal 42, respectively.

[0047] (Operation example of the electronic device 10) The electronic device 10 analyzes an observation signal obtained by observing a mixed signal by independent component analysis or frequency analysis, thereby estimating a signal corresponding to a source signal included in the mixed signal and generating it as a separated signal. The electronic device 10 determines the independence of the source signal in order to determine which analysis method of independent component analysis and frequency analysis to select and execute. Hereinafter, a procedure for determining the independence of the source signal and a procedure for selecting an analysis method based on the evaluation result and generating a separated signal will be described respectively.

[0048] <Example procedure for evaluating the independence of the source signal> The control unit 12 of the electronic device 10 may execute the procedure illustrated in the flowchart of FIG. 6 as a signal determination method for determining the independence of the source signal. The procedure illustrated in the flowchart of FIG. 6 may be realized as a signal determination program that causes the processor constituting the control unit 12 to execute. The signal determination method is also referred to as a control method of the electronic device 10.

[0049] The control unit 12 acquires an observation signal and observation conditions (step S1). Specifically, the control unit 12 acquires an observation signal obtained by separately observing the source signal. When the source signal includes the first signal 41 and the second signal 42, in a state where only the first signal 41 is transmitted, the first observer 21 or the second observer 22 observes the first signal 41. Further, in a state where only the second signal 42 is transmitted, the first observer 21 or the second observer 22 observes the second signal 42. The electronic device 10 performs the observation by fixing it to either the first observer 21 or the second observer 22.

[0050] Further, the control unit 12 acquires, as observation conditions, the position or directivity of the observer that observes the source signal, and the position or directivity of the signal source that transmits the source signal.

[0051] Although the observation signal obtained by the first observer 21 observing only the first signal 41 and the observation signal obtained by the second observer 22 observing only the first signal 41 are observing the same signal, due to the difference in the position or directivity of each observer, they can be different signals from each other. Also, the observation signal obtained by the first observer 21 observing only the second signal 42 and the observation signal obtained by the second observer 22 observing only the second signal 42 can similarly be different signals from each other. As will be described later, the control unit 12 can determine independence for each observation condition. Also, the control unit 12 can estimate the separated signal based on the difference in the observation signal and the observation conditions.

[0052] The control unit 12 calculates an independence value (IS) (step S2). Specifically, the control unit 12 calculates the independence value according to the following procedure.

[0053] Assume that the probability distributions of the first signal 41 and the second signal 42 as source signals are represented by X and Y respectively. In this case, the joint distribution of the first signal 41 and the second signal 42 is represented by XY. When X and Y are statistically independent, the following relational expression (1) holds for the expected values E[X] and E[Y] of X and Y respectively, and the expected value E[XY] of XY.

Number

Number

[0054] The above formula (1) means that the joint distribution of X and Y is equal to the product of the probability density functions of X and Y respectively. However, infinite sample data is required in formula (2). Therefore, the probability distribution cannot be applied to unknown signals. Thus, instead of the above formula (2), a formula (3) in an approximated form where the expected value E[X] is approximated by the sample expected value E^[X n using finite sample data is used. Here, E^ represents that a ^ (hat) is attached above the letter E.

Number

[0055] By transforming formula (1) using formula (3), the following formula (4) is obtained.

Number

[0056] When formula (4) holds, the difference between the left side and the right side is 0. When formula (4) does not hold, a difference between the left side and the right side occurs. It can be said that the greater the absolute value of the difference between the left side and the right side of formula (4), the lower the independence between X and Y. The absolute value of the difference between the left side and the right side of formula (4) can be used as an index representing the independence between X and Y.

[0057] The control unit 12 calculates a difference value that represents, as a positive value, a value based on the difference between the left side and the right side of formula (4). The difference value is calculated as a value correlated with the absolute value of the difference between the left side and the right side of formula (4). The control unit 12 may calculate, as the difference value, the absolute value of the difference between the left side and the right side of formula (4). The control unit 12 may calculate, as the difference value, for example, a squared error e that is the square of the difference between the left side and the right side of formula (4) as represented by the following formula (5). n The control unit 12 may calculate n The square root of the squared error e, that is, the root mean square error, may be calculated as the difference value.

Number

[0058] The control unit 12 acquires a finite number of sample data from the observation signal obtained by the first observer 21 or the second observer 22 observing only the first signal 41, and sets the acquired sample data as X. Further, the control unit 12 acquires a finite number of sample data from the observation signal obtained by the first observer 21 or the second observer 22 observing only the second signal 42, and sets the acquired sample data as Y. The control unit 12 can calculate the squared error e n by applying X and Y corresponding to the acquired sample data to formula (5).

[0059] The control unit 12 may calculate an independence numerical value (IS) by the following formula (6) using the squared error e n calculated as a positive value.

Number

[0060] Equation (6) is the expected value of a single signal sample (E^[X n and E^[Y n ), and the mean square error e n calculated based on the expected value of the product signal sample (E^[(XY) n ) (n = 1 to k) represents calculating the sum as the difference value at each order. That is, the independence value (IS) is calculated by calculating the expected value of the single signal sample and the expected value of the product signal sample at multiple orders, calculating the difference value at each order, and calculating it as the sum of the difference values at each order. By calculating the independence value (IS) as the sum of the difference values at multiple orders, the correlation between the independence value and the independence of multiple signals can be enhanced. When calculating the independence value (IS) based on the sample expected value at each order, it is not limited to the sum of the difference values at each order, and various other operations may be used.

[0061] In each term on the right side of Equation (6), the value obtained by raising the mean square error e n to the power of 1 / n is added (n = 1, ···, k). 1 / n corresponds to the reciprocal of each order. That is, each term on the right side of Equation (6) corresponds to the difference value calculated at each order raised to the power of the reciprocal of each order. By raising the mean square error e n to the power of 1 / n, the expected value of some terms among the expected values of each term calculated based on X n and Y n is less likely to be dominant. The exponent of the power is not limited to this and may be a common value in each term. Also, a value that does not raise the mean square error e n to a power may be added as it is.

[0062] The coefficient α n included in each term on the right side of Equation (6) is a weighting coefficient for each term and can be arbitrarily set. That is, each term on the right side of Equation (6) corresponds to the difference value calculated at each order multiplied by the weighting coefficient of each order. The coefficient α n may be set to 1 for example, or may be set to different values in each term.

[0063] The right side of Equation (6) is in a form where terms are added with n taking values that increase by 1 from 1 to k. The combination of n values, that is, the combination of degrees, may be set arbitrarily. For example, n may be set to increase by a predetermined value such as 2, 4, 6, etc., or may be set irregularly such as 1, 4, 5, etc. The higher the degree of the sample expected value, the more easily the sample expected value is affected by noise. Therefore, the combination of degrees may be adjusted according to the SN ratio of the sample data.

[0064] In this embodiment, assume that n is set to 1, 2, 3, 4 in order to calculate the independence value. Also, assume that the coefficient α n is all set to 1. In this case, the independence value (IS) is calculated by the following Equation (7).

Equation

[0065] The control unit 12 can calculate the independence value (IS) by executing the procedure example described above. The control unit 12 determines whether the calculated independence value (IS) is less than or equal to the first threshold value (step S3). The control unit 12 sets the first threshold value in advance. The control unit 12 may obtain in advance the upper limit of the independence value (IS) required to obtain a desired estimation accuracy by independent component analysis and set it as the first threshold value. The control unit 12 may set the first threshold value based on the input from the user received by the input unit 18. The first threshold value is also referred to as a predetermined threshold value.

[0066] When the independence value (IS) is less than or equal to the first threshold (step S3: YES), the control unit 12 outputs, via the output unit 16, information indicating that there is independence between the first signal 41 and the second signal 42 (step S4). When the independence value (IS) is greater than the first threshold, that is, when the independence value (IS) is not less than or equal to the first threshold (step S3: NO), the control unit 12 outputs, via the output unit 16, information indicating that there is no independence between the first signal 41 and the second signal 42 (step S5). In the procedure of step S4 or S5, the control unit 12 may output information indicating the presence or absence of independence by changing the display color of the indicator lamp 16a. The control unit 12 may display, on the display 16b, information indicating the presence or absence of independence. The control unit 12 may output, from the speaker 16c, voice information indicating the presence or absence of independence. That is, in the procedure of step S4 or S5, the control unit 12 causes the output unit 16 to output the determination result as to whether the first signal 41 and the second signal 42 are independent of each other. After executing the procedure of step S4 or S5, the control unit 12 proceeds to the procedure of step S6. The control unit 12 does not necessarily have to execute either step S4 or S5, nor does it necessarily have to execute both steps S4 and S5.

[0067] The control unit 12 outputs the observation conditions and the calculated independence value (IS) (step S6). The control unit 12 may store the observation conditions and the calculated independence value (IS) in the storage unit. The control unit 12 may store, in the storage unit, a database associating the observation conditions with the independence value (IS). The control unit 12 may cause an external device to store a database associating the observation conditions with the independence value (IS). After executing the procedure of step S6, the control unit 12 ends the execution of the flowchart of FIG. 6.

[0068] <Example of Procedure for Evaluating Independence of Source Signals> The control unit 12 of the electronic device 10 selects a method for analyzing an observation signal obtained by observing a signal obtained by mixing source signals based on the independence numerical value (IS) calculated by executing the procedure of the flowchart in FIG. 6. The control unit 12 analyzes the observation signal by executing the selected method and estimates the separated signal. The control unit 12 may execute the procedure illustrated in the flowchart of FIG. 7 as a signal analysis method for estimating the separated signal. The procedure illustrated in the flowchart of FIG. 7 may be realized as a signal analysis program to be executed by a processor constituting the control unit 12. The signal analysis method is also referred to as a control method of the electronic device 10.

[0069] The control unit 12 acquires the observation signal and the observation conditions (step S11). Specifically, the control unit 12 acquires the observation signal from the first observer 21 and the second observer 22. The control unit 12 acquires the observation conditions by receiving an input from the user by, for example, the input unit 18.

[0070] The control unit 12 acquires the independence numerical value (IS) corresponding to the observation conditions (step S12). Specifically, the control unit 12 may acquire the independence numerical value (IS) associated with the observation conditions based on the database stored in the storage unit or the external device. When there is no independence numerical value (IS) associated with the observation conditions, the control unit 12 may proceed to the frequency analysis of the procedure in step S15 described later as error processing.

[0071] The control unit 12 determines whether the independence numerical value (IS) corresponding to the observation conditions is equal to or less than the second threshold value (step S13). The control unit 12 sets the second threshold value in advance. The control unit 12 may acquire in advance the upper limit of the independence numerical value (IS) required to obtain a desired estimation accuracy by independent component analysis and set it as the second threshold value. The control unit 12 may set the second threshold value to the same value as the first threshold value. The control unit 12 may set the second threshold value based on an input from the user received by the input unit 18. The second threshold value is also referred to as a predetermined threshold value.

[0072] When the independence value (IS) is less than or equal to the second threshold (step S13: YES), the control unit 12 performs independent component analysis to analyze the observation signal (step S14). When the independence value (IS) is greater than the second threshold, that is, when the independence value (IS) is not less than or equal to the second threshold (step S13: NO), the control unit 12 performs frequency analysis to analyze the observation signal (step S15). By executing the procedure of step S14 or S15, the control unit 12 can estimate the separated signal from the observation signal.

[0073] The control unit 12 outputs the estimated separated signal (step S16). After executing the procedure of step S16, the control unit 12 ends the execution of the flowchart in FIG. 7.

[0074] <Analysis Example> The validity of the estimation result of the separated signal obtained when the signal determination method and the signal analysis method described above are executed will be described below based on specific examples.

[0075] An example of the source signal is shown in FIG. 8. FIG. 8(A) is a graph representing the waveform of a sine wave as an example of the first signal 41. FIG. 8(B) is a graph representing the waveform of a sine wave with a frequency different from that of the first signal 41 as an example of the second signal 42. In FIG. 8, the horizontal axis represents time. The vertical axis represents the amplitude of the signal.

[0076] The control unit 12 can acquire the first observation signal and the second observation signal by observing, with the first observer 21 and the second observer 22 respectively, a signal obtained by mixing the first signal 41 and the second signal 42. An example of the observation signal is shown in FIG. 9. FIG. 9(A) is a graph showing the waveform of the first observation signal obtained by observing, with the first observer 21, the signal obtained by mixing the first signal 41 and the second signal 42. FIG. 9(B) is a graph showing the waveform of the second observation signal obtained by observing, with the second observer 22, the signal obtained by mixing the first signal 41 and the second signal 42. In FIG. 9, the horizontal axis represents time. The vertical axis represents the amplitude of the signal. The first observation signal and the second observation signal have different waveforms because the observation conditions such as the position or directivity of the first observer 21 and the observation conditions such as the position or directivity of the second observer 22 are different.

[0077] The control unit 12 estimates a separation signal based on the first observation signal and the second observation signal. An example of the separation signal is shown in FIG. 10. FIG. 10(A) is a graph showing the waveform of the first separation signal estimated as the signal corresponding to the first signal 41. FIG. 10(B) is a graph showing the waveform of the second separation signal estimated as the signal corresponding to the second signal 42.

[0078] When the first signal 41 and the second signal are known, ICAerror is calculated using the known first signal 41 and second signal 42 as source signals. On the other hand, an independence value (IS) between the known first signal 41 and the second signal 42 is calculated. In this analysis example, the correlation between ICAerror and the independence value (IS) was confirmed by the following procedure.

[0079] ICAerror is calculated by the following procedure. Based on the mixing matrix A arbitrarily set using the known first signal 41 and second signal 42 as source signals, the first observation signal and the second observation signal are virtually generated. A separation matrix W is estimated based on the observation signals. A separation signal is generated based on the estimated separation matrix W. ICAerror is calculated as the difference between the source signal and the estimated separation signal, or as a value determined based on the mixing matrix A and the separation matrix W.

[0080] On the one hand, in order to calculate the independence value (IS), an observation signal obtained by observing only a known first signal 41 and an observation signal obtained by observing only a known second signal 42 are virtually generated. The independence value (IS) is calculated based on the observation signals virtually generated, which are the observation signal obtained by observing only the known first signal 41 and the observation signal obtained by observing only the known second signal 42.

[0081] In this analysis example, 60×60 = 3600 combinations obtained by changing the frequencies of the first signal 41 and the second signal 42 one by one from 1 Hz to 60 Hz and combining them were used. The ICA error and the independence value (IS) were calculated for each combination. Then, the values calculated for each combination were plotted on the correlation graph illustrated in FIG. 11. In FIG. 11, the horizontal axis and the vertical axis represent the ICA error and the independence value (IS), respectively. For the points plotted on the graph of FIG. 11, the correlation coefficient between the ICA error and the independence value (IS) was calculated to be 0.91. That is, it can be said that the ICA error and the independence value (IS) are correlated. The correlation between the ICA error and the independence value (IS) is at least stronger than the correlation between the ICA error and the SUC, where the correlation coefficient was 0.077 in the above-described comparative example.

[0082] Since the ICA error and the independence value (IS) are correlated, the control unit 12 can select a method for analyzing the observation signal based on the independence value (IS). Specifically, when the independence value (IS) is less than or equal to a second threshold value, the control unit 12 can determine that the ICA error becomes less than or equal to a predetermined value, that is, the estimation accuracy of the separated signal by independent component analysis is high. As a result, the control unit 12 can select independent component analysis as a method for analyzing the observation signal. On the other hand, when the independence value (IS) is greater than the second threshold value, the control unit 12 can determine that the ICA error becomes greater than the predetermined value, that is, the estimation accuracy of the separated signal by independent component analysis is low. As a result, the control unit 12 can select frequency analysis instead of selecting independent component analysis as a method for analyzing the observation signal.

[0083] As described above, according to the electronic device 10, signal determination method, and signal analysis method according to the present embodiment, the independence of the source signal can be evaluated using a finite number of sample data. Specifically, it is not necessary to design a statistical model using an infinite number of sample data. Further, it is not necessary to actually confirm the estimation accuracy of the separated signal by independent component analysis by collecting mixed signals and performing independent component analysis. As a result, the work cost and work load can be reduced. Further, the independence of a plurality of signals can be simply expressed as a numerical value based on the sample data. As a result, the independence of a plurality of signals can be easily grasped.

[0084] Further, independent component analysis can be executed when it is determined that ICAerror representing the estimation accuracy by independent component analysis is equal to or less than a predetermined value. By doing so, the analysis of the observed signal by independent component analysis becomes easier to execute. When independent component analysis is executed with a smaller computational load than frequency analysis, the separation of signals can be simplified because independent component analysis is likely to be used.

[0085] Further, it is possible to numerically support the problems that may occur when the separated signal is estimated by independent component analysis.

[0086] Although the embodiments according to the present disclosure have been described based on the drawings and examples, it should be noted that those skilled in the art can make various modifications or alterations based on the present disclosure. Therefore, it should be noted that these modifications or alterations are included in the scope of the present disclosure. For example, the functions included in each component or each step can be rearranged so as not to be logically contradictory, and a plurality of components or steps can be combined into one or divided.

Description of Reference Numerals

[0087] 1 Signal analysis system 10 Electronic device (12: Control unit, 14: Sample acquisition unit, 16: Output unit (16a: Indicator lamp, 16b: Display, 16c: Speaker), 18: Input unit) 21, 22 First observer, second observer 31, 32 First signal source, second signal source 41, 42 First signal, second signal

Claims

1. A control unit that calculates an independence value representing the mutual independence of a plurality of signals, and a sample acquisition unit that acquires sample data for each of the plurality of signals are provided, wherein the control unit calculates a plurality of single-signal sample expected values based on the sample data for each of the plurality of signals at a plurality of orders, calculates one product-signal sample expected value based on the product of the sample data for each of the plurality of signals at a plurality of orders, calculates a difference value representing, at each order, the difference between the product of the single-signal sample expected values and the product-signal sample expected value as a positive value, and calculates the sum of the difference values for each order as the independence value, an electronic device.

2. The electronic device according to claim 1, wherein the control unit uses, as the difference value, the power of the reciprocal of each order of the difference between the product of the single-signal sample expected values calculated at each order and the product-signal sample expected value.

3. The electronic device according to claim 2, wherein the control unit uses, as the difference value, a value obtained by multiplying the power of the reciprocal of each order of the difference between the product of the single-signal sample expected values calculated at each order and the product-signal sample expected value by a weighting coefficient for each order.

4. The electronic device according to claim 1, wherein the control unit uses, as the difference value, the square or absolute value of the difference between the product of the single-signal sample expected values and the product-signal sample expected value.

5. The electronic device according to claim 4, wherein the control unit uses, as the difference value, a value obtained by multiplying the square or absolute value of the difference between the product of the single-signal sample expected values calculated at each order and the product-signal sample expected value by a weighting coefficient for each order.

6. The electronic device according to any one of claims 1 to 5, wherein the control unit determines that the plurality of signals are mutually independent when the independence value is less than or equal to a predetermined threshold.

7. The electronic device according to claim 6, further comprising an output unit that outputs a determination result as to whether the plurality of signals are mutually independent.

8. The control unit separates a mixed signal including the plurality of signals by independent component analysis when the independence value is less than or equal to a predetermined threshold, and separates a mixed signal including the plurality of signals by frequency analysis when the independence value is greater than a predetermined threshold. The electronic device according to any one of claims 1 to 7.

9. acquiring sample data for each of a plurality of signals, calculating a plurality of single-signal sample expected values based on the sample data for each of the plurality of signals at a plurality of orders, calculating one product-signal sample expected value based on the product of the sample data for each of the plurality of signals at a plurality of orders, Calculating a difference value, which represents, as a positive value, the difference between the product of the respective single-signal sample expected values and the product-signal sample expected value at each order; Calculating the sum of the difference values at each order as an independence numerical value representing the mutual independence of the plurality of signals A control method for an electronic device, comprising:

10. The control method for an electronic device according to claim 9, further comprising determining that the plurality of signals are independent of each other when the independence numerical value is less than or equal to a predetermined threshold value.

11. When the independence numerical value is less than or equal to a predetermined threshold value, separating a mixed signal including the plurality of signals by independent component analysis; When the independence numerical value is greater than a predetermined threshold value, separating a mixed signal including the plurality of signals by frequency analysis The control method for an electronic device according to claim 9 or 10, further comprising:

Images