Pattern extraction method and communication multiplexing method

Abstract

To solve the problem in which pattern extraction does not extract patterns accurately, is difficult to understand as a feature of data analysis, and requires a long signal length when applied to multiplexing communications.SOLUTION: In the present invention, the minimum value of division of a signal that serves as a base and a target signal is obtained. This makes it possible to calculate values of only an expected pattern by eliminating values of other patterns. It is also possible to extract hidden patterns, and it makes it possible to extract patterns from various data such as image processing and moving image processing, and to calculate feature quantities by analyzing the periodicity and continuity of data, making it easier to diagnose the data. In addition, for binary data, when the base is selected and the calculation of the present invention is used, it makes it possible to increase the number of transmission codes in communication multiplexing, perform multiplexing with a short length, and simplify processing, by performing calculations that exclude zeros in the base.SELECTED DRAWING: Figure 2

Description

Technical Field

[0001] The present invention relates to a pattern extraction method and a communication multiplexing method using a base extraction division correlation minimum method. In particular, for data such as the number of infected persons with infectious diseases, the number of customer visitors, vibration measurement, electroencephalogram, electrocardiogram, images of stained somatic tissues, the size and shape of crystals in microscopic images, images of corroded metal surfaces, and videos with markers, it is an algorithm capable of accurately extracting patterns from arbitrary data using a pattern called a base, and is a pattern extraction method for obtaining findings by accurately extracting to a specific pattern when performing data analysis.

[0002] Moreover, it is a communication multiplexing method that performs communication multiplexing using this pattern extraction for binary data.

Background Art

[0003] Autocorrelation processing performed in signal processing and statistics is often used when calculating similarity. A peak is calculated when the similarity is high.

[0004] In addition, as a means for extracting feature amounts of image patterns, there is feature amount extraction using a simultaneous normal matrix. It is possible to calculate feature amounts such as the contrast for each pixel distance.

[0005] Furthermore, when performing communication multiplexing, correlation processing using an M-sequence or a GOLD-sequence is performed for communication multiplexing. As a result, communication bandwidth is saved.

Prior Art Documents

Patent Documents

[0006]

Patent Document 1

Patent Document 2

Non-Patent Documents

[0007]

Non-Patent Document 1

Non-Patent Document 2

Summary of the Invention

Problems to be Solved by the Invention

[0008] In the conventional calculation process of similarity by convolution sum, a high correlation value is output for data with a high similarity to the basis, but the amplitude of the basis is not calculated.

[0009] When performing autocorrelation processing in statistics, knowledge was required to interpret periodicity.

[0010] In image processing, there is a texture feature quantity called a simultaneous normal matrix, and there is an index indicating the relationship with adjacent data, but it is necessary to select the positions of pixels to be compared and perform a lot of analysis, and it is difficult to say that the features of the entire image are grasped collectively.

[0011] In the conventional multiplexing of communication using the M series, there were problems that the number of pairs was small and a long signal length was required for multiplexing.

Means for Solving the Problems

[0012] In the first invention of the present application, data is sequentially cut out from the head of the data by the length of the basis, the value of each data is divided by a non-zero value of the basis, and when all the results of the division are positive values, the minimum value is calculated, and when all the results of the division are negative values, the maximum value is calculated, and arbitrary pattern extraction is performed.

[0013] The present invention obtains the minimum value of the division of the base signal and the target signal by using the correlation based on the designed base. Thereby, it is possible to exclude the numerical values of other patterns and calculate only the numerical values of the assumed pattern. Thereby, the amplitude of the base can be calculated. It is also possible to extract hidden patterns in data that are not explicitly visible.

[0014] In the second aspect of the present invention, for the data, a base with 1s consecutive is the first term, a base with one 0 between 1s is the second term, a base with two 0s between 1s is the third term, and so on, increasing the interval of 0s by one each time to create a periodically continuing base. Cut the data in order from the beginning of the data by the length of the base, divide the value of each data by the non-zero values of the base, calculate the minimum value if all the results of the division are positive values, and obtain the average for the entire data with that base, thereby extracting a periodic pattern.

[0015] In periodic analysis, autocorrelation and Fourier transform used in statistics are the main analysis methods for time series data, but special knowledge is required to understand the results. However, by periodically defining the base of the present invention, a quantitative value of a specified frequency can be calculated. The frequency in the Fourier transform is defined as a frequency. The frequency calculated from the method of the present invention is a different concept, and the present invention is a more effective method when the frequency of events is sparse compared to the Fourier transform.

[0016] In the third aspect of the present invention, for the data, a base with one consecutive 1 is the first term, a base with two consecutive 1s is the second term, increasing one by one to create a continuous base. Cut the data in order from the beginning of the data by the length of the base, divide the value of each data by the non-zero values of the base, obtain the minimum value if all the results of the division are positive values, and obtain the average value with that base, thereby extracting a continuity pattern.

[0017] It is difficult to quantitatively calculate the continuity from time-series data or frequency data. Therefore, by continuously defining the basis of the present invention, it becomes possible to quantitatively analyze whether the visitors to the store are continuously visiting. Similarly, by obtaining the frequency of physical phenomena, it becomes possible to quantitatively analyze the continuity of the phenomena.

[0018] It becomes possible to quantitatively analyze whether the visitors to the store are continuously visiting. Similarly, by obtaining the frequency of physical phenomena, it becomes possible to quantitatively analyze the continuity of the phenomena.

[0019] Also, when using the continuity analysis for the image from which the crystal boundary has been extracted, it becomes possible to obtain the average size of the crystal.

[0020] The fourth invention is used for binary data for communication multiplexing. In the calculation of the present invention, at all positions of the data obtained by repeatedly adding different bases from itself by OR sum, the cross-correlation is 0. For the data obtained by repeatedly using its own base, the head of the base for the calculation of the present invention is 1, and the correlation value of the present invention at other positions is 0. Select the base that satisfies these conditions, modulate the data by expressing it with a combination of a plurality of bases formed by combining the patterns of 1 and 0 of a plurality of binary data, obtain the OR sum of the bases superimposed by the data, and modulate by superimposing signals at the same frequency. Then, transmit and receive the data, and demodulate a plurality of binary data at the head of each original base by using the calculation of the present invention to perform communication multiplexing.

[0021] By selecting and using the basis of the present invention, it becomes possible to simultaneously transmit a plurality of signals at the same frequency band. It becomes possible to transmit synchronized signals and a plurality of signals in the same band.

Advantages of the Invention

[0022] By performing pattern extraction, it becomes possible to detect data indicating a specific pattern. It becomes possible to detect customers who visit in a specific pattern, the spread of dyed tissues, etc.

[0023] By numerically calculating periodicity from data, it becomes possible to calculate the period of the occurrence of infectious diseases from the number of infectious disease patients, and the periodicity of the stained somatic cell tissue can be calculated.

[0024] By numerically calculating continuity from data, it becomes possible to calculate the degree of continuous customer arrivals from customer data and the average size of the stained somatic cell tissue.

[0025] In the field of communication, it is possible to simultaneously transmit a plurality of data using the same bandwidth. In such a process, there are often few available patterns, but the number of available patterns increases. Also, multiplexing can be performed with a short signal length.

Brief Description of Drawings

[0026]

Figure 1

Figure 2

Figure 3

Figure 4

Figure 5

Figure 6

Figure 7

Figure 8

Figure 9

Figure 10

Figure 11

Figure 12

Mode for Carrying Out the Invention

[0027] Hereinafter, with reference to the drawings, the pattern extraction method, the periodicity analysis method, the continuity analysis method, and communication multiplexing will be sequentially described.

[0028] With reference to FIG. 1, arbitrary pattern extraction, which is the first embodiment of the present invention, will be described. In this first embodiment, it has a data reading unit 11a that reads data such as an individual's web access history, visitor number data, and images, and a base reading unit 11b that inputs a base that is a definition of a pattern. In the arithmetic processing unit 12, pattern extraction processing is performed, and recording and display are performed by a recording unit 13a that stores the processing result and a display unit 13b that displays the processing result. By using the stored and displayed data, prediction of the number of next-period visitors and detection of exactly the same pattern are performed.

[0029] FIG. 2 is an overview of the processing of the pattern extraction processing arithmetic unit 12. For arbitrary sampled data, an application of cross-correlation processing in the signal processing field is performed. In normal cross-correlation processing, the processing is performed using the sum of the multiplication of the base signal and the target signal, but in the present invention, the minimum value of the division of the base signal and the target signal is obtained. At this time, the processing of the part where the base is 0 is excluded from the processing because it is impossible to calculate. When all the division values other than 0 of the base signal are positive values, the minimum value is calculated. When all the division values are negative values, the maximum value is calculated as a negative value. When positive and negative values are mixed, the calculation is performed as 0. Thereby, extraction of an arbitrary pattern becomes possible. Negative values are values extracted with the pattern inverted.

[0030] This processing is a processing in which the multiplication processing of the correlation processing performed in the signal processing field is replaced with division, and the processing of taking the sum is replaced with the minimum value when all the division values are positive, and the maximum value is used as the calculated value when all the division values are negative.

[0031] By performing such processing, components of other patterns are excluded, data other than that involving only the specific pattern represented by the basis is excluded, the amount of the specific pattern can be quantitatively expressed, and it becomes possible to predict data after the specific pattern. For example, it is possible to predict purchasing behavior from a specific web access history pattern and the like.

[0032] Figure 3 is an example of the processing flow of one-dimensional data in the pattern extraction processing calculation unit 12. S101, S102, and S111 indicate that correlation processing of data is performed and the processing is repeated from 0 to the data length minus the basis length. S103, S104, S105, S106, S107, and S108 indicate that division of the data and the basis in the specified order is performed. Therefore, when the value of the basis specified by the basis is 0, division is not performed. Then, when all the calculation results are positive, the initial value is obtained, and when all the calculation results are negative, the maximum value is obtained.

[0033] Next, with reference to Figure 4, the periodicity analysis, which is the second embodiment of the present invention, will be described. In this second embodiment, as shown in Figure 4, in the data reading unit 21, data such as the number of new cases of infectious diseases, the number of visitors, and images are read. In the periodicity calculation processing unit 22, periodicity calculation is performed. In the display unit 23a, the processing result is displayed. In the recording unit 23b, the processing result is recorded. Since the basis is created within the processing, there is no basis reading unit, and the processing for creating the basis is performed, and the same calculation as in the first embodiment is performed. An example of the method for creating the basis is shown in Figure 5. Also, negative values are processed in advance and the calculation is performed in a state where there are no negative values.

[0034] By performing the calculation of the periodicity analysis, it is possible to accurately know the period of the number of new cases of infectious diseases.

[0035] FIG. 5 shows an example of the processing flow of two-dimensional data for periodic analysis in the periodic analysis calculation unit 22. S201 and S202 indicate the process of eliminating negative values by subtracting the minimum value of the data from all the data values when negative values exist in the data, thus making the data non-negative. S203, S204, S205, S206, S207, S211, and S212 indicate the process of creating a basis by repeatedly subtracting the basis length from 0 to the data length for each dimension of the two-dimensional data. S208 creates a basis that actually has two-dimensional periodicity using the specified index. Using the created basis, the operation of the present invention is performed in S209. Then, in S210, the average value of the calculation with that basis is set as the calculated value of the period of that basis.

[0036] Next, referring to FIG. 6, the continuity analysis, which is the third embodiment of the present invention, will be described. In this third embodiment, as shown in FIG. 6, in the data reading unit 31, data such as a binary image of a stained tissue image, visitor number data, and an image are read. In the continuity calculation processing unit 32, a periodic calculation is performed. In the display unit 33a, the processing result is displayed. In the recording unit 33b, the processing result is displayed. Since a basis is created within the process, there is no basis reading unit, and the process of creating a basis is performed, and the same operation as in the first embodiment is carried out. An example of the method of creating a basis is shown in FIG. 5.

[0037] By using continuity analysis, it becomes possible to predict how continuously visitors come based on the number of visitors, and it is also possible to know the average size of the stained tissue from the stained tissue image.

[0038] FIG. 7 shows an example of the processing flow of two-dimensional data for continuity analysis in the continuity analysis calculation unit 32. S301, S302, S303, S304, S308, and S309 are processes of creating a basis by repeatedly subtracting the basis length from 0 to the data length for each dimension. S305 creates a basis having two-dimensional continuity using the specified index. Using the created basis, the operation of the present invention is performed in S306. Then, in S307, the average value of the calculation with that basis is set as the calculated value of the period of that basis.

[0039] Next, the fourth embodiment regarding communication multiplexing will be described. In this fourth embodiment, as shown in FIG. 8, transmission data 1 to N41a, 41b, 41c, which are a plurality of binary data, are modulated by the basis extraction division correlation minimum method modulation unit 42 according to the modulation of the present invention, and are modulated into a state where the data can be transmitted by the transmission wave modulation unit (including D / A conversion) 43, and the data is transmitted. The transmitted signals are received by the reception units 45a, 45b, 45c of the respective signals, and after being demodulated by the reception wave demodulation unit (including A / D conversion), the basis extraction division correlation minimum method demodulation units 46a, 46b, 46c perform the demodulation of the present invention. The data extraction processing units 48a, 48b, 48c perform the process of extracting the leading part of the data, and the received data 1 to N49a, 49b, 49c perform the process of restoring the data transmitted by the data 41.

[0040] FIG. 9 shows an overview of the processing of the basis extraction division correlation minimum method modulation unit 42. When there are three types of bases, 1 and 0 are represented by the presence or absence of each base, and the binary data of the signal to be transmitted is created by obtaining the OR sum of each base.

[0041] FIG. 10 shows an overview of the processing of the basis extraction division correlation minimum method demodulation units 47a, 47b, 47c. The data modulated according to the present invention is received, the arithmetic processing of the present invention is performed according to the selected base, and the binary data is demodulated at the head of each base.

[0042] FIG. 11 is an example of the processing flow of the basis extraction division correlation minimum method demodulation units 47a, 47b, 47c. After obtaining the latest data in S401, the arithmetic processing of the present invention is performed in S402, S403, S404, S405, S406, S407, S408. In S402 and S403, the variables used in the repeated processing within this time are initialized. In S404 and S407, a repeated process from 0 to the base length is performed. In S405 and S406, in the process within the repeated process, the minimum value of the part other than 0 is obtained. Since it is binary data, 1 is set to True and 0 is set to False, and the data of the part where the base is other than 0 is obtained. After performing the process of the base length, in S408, when all the values in the temporary data are 1 (True), the process returns a result where the calculation result is 1 (True).

[0043] FIG. 12 shows an example of the processing in the logic operation circuits of the base extraction division correlation minimum method demodulation units 47a, 47b, and 47c. When the time t is 0 and when the time t is 1, the processing procedures for the latest data at each of the times shown in the first row are shown in the second to seventh rows. Looking at the second, third, and fourth rows, for the locations where the base is 0, an OR operation with 1 is performed, and for the locations where the base is 1, an AND operation with 1 is performed. As a result, the intermediate results of the base calculation process are shown in the fifth row. This enables the calculation by always setting the 0 part of the base to 1 (True). Finally, by obtaining the AND of all the operation results for the base length shown in the fifth and sixth rows, the operation result at that time shown in the row can be obtained.

Claims

1. A pattern extraction method for performing pattern extraction at any position, comprising the steps of: cutting off data in order from the beginning by the length of the base; dividing the value of each data by a non-zero value of the base; calculating the minimum value when all the results of the division are positive values, and calculating the maximum value when all the results of the division are negative values.

2. A pattern extraction method for extracting a periodic pattern, comprising the steps of: creating a base that periodically continues by increasing the number of 0s between 1s one by one, with the base having 1 consecutive as the first term, the base having one 0 between 1s as the second term, the base having two 0s between 1s as the third term, and so on; cutting off data in order from the beginning by the length of the base; dividing the value of each data by a non-zero value of the base; calculating the minimum value when all the results of the division are positive values; and obtaining the average of the calculated values of the entire data for that base.

3. A pattern extraction method for performing continuous pattern extraction, comprising the steps of: creating consecutive bases by increasing the number of consecutive 1s one by one, with the base having one consecutive 1 as the first term, the base having two consecutive 1s as the second term, and so on; cutting off data in order from the beginning by the length of the base; dividing the value of each data by a non-zero value of the base; calculating the minimum value when all the results of the division are positive values; and obtaining the average of the calculated values of the entire data for that base.

4. A communication multiplexing method for performing communication multiplexing, which is used for binary data for communication multiplexing, and includes the steps of: selecting a base where the minimum value at non-zero positions of the base of the data cut off by the length of the base is 0 at all positions of the data obtained by repeatedly adding the OR sum of bases different from itself, and the head of the base of the minimum value at non-zero positions of the base of the data cut off by the length of the base of the data obtained by repeatedly adding the OR sum of its own base and other bases is 1 and the values at other positions are 0; representing a plurality of binary data by a combination of a plurality of bases combined with 1 and 0 patterns; obtaining the OR sum of the bases superimposed by the data and modulating by superimposing signals at the same frequency; transmitting the data; receiving the data; dividing the value of each data by a non-zero value of the base; and demodulating a plurality of binary data by calculating the minimum value when all the results of the division are positive values.

Images