KR100665352B1 - Method of decoding encoded information of taste information or feeling of taste transferred from video image - Google Patents

Abstract

The present invention provides a method of first reducing the dimension by PCA and encoding the input encoding information based on the representative taste by FCMA after encoding the expression factor for the taste of taste information or video image. That is, the present invention provides a method of digitally decoding a taste of a taste information or a video image, the method comprising: performing encoding according to a representative expression of a representative taste and obtaining and storing a center value of the representative taste by clustering; Receiving the taste information that is encoded and transmitted and mapping the pattern of data to dimensionally reduced data that can be visually confirmed by a human by principal component analysis; And obtaining the center value of the representative taste by clustering the coded taste information having reduced dimensions, and comparing the stored center value with the center value of the stored representative taste to obtain the degree of belonging to the center value of the representative taste of the input taste information. To judge.

Description

Method of decoding encoded information of taste information or feeling of taste transferred from video image}

BRIEF DESCRIPTION OF THE DRAWINGS It is a figure for demonstrating the taste tetrahedron structure of hening.

FIG. 2 is a diagram for describing representation and encoding of taste information using adjective expression factors.

3 is a diagram for explaining a method of encoding taste information according to the present invention.

4 is a diagram for explaining the application of metadata in MPEG during transmission in the method of encoding taste information according to the present invention.

5 is a graph for explaining clustering of basic representative tastes.

6 is a view for explaining a process of obtaining a center value of a representative taste by reducing the dimension and clustering of various representative flavors through the PCA applied to the present invention.

7 is a block diagram illustrating a transmission and decoding process of encoded data according to the present invention.

<Description of main parts of drawing>

60 ... Representation taste 62, 72 ... Factor expression

64, 74 ... Code 65, 75 ... PCA

66, 76 ... clustering 68 ... Representative center of taste

70 ... Taste Information 78 ... Center Values for Stored Representative Flavors

80 ... small speed determination

The present invention relates to a method of first reducing the dimension by PCA and then restoring the input encoding information based on the representative taste by FCMA, after encoding the expression factor for the taste of taste information or video image. By encoding the taste emotions of images or images with a predetermined expression factor, and encoding dozens of representative tastes used by humans by storing only the central value of the representative tastes using fuzzy clustering method. Even if the taste data comes in, after encoding the expression factors for the taste feeling of taste information or image image, which can be restored by obtaining the degree of belonging of the taste to be expressed by the central value stored in advance, To reduce the dimensions and restore the input encoding information based on the representative taste by FCMA. .

In addition to the conventional video and audio, Japan Post has decided to embark on the research and development of communication technology that delivers smell, taste, and texture at high speed. Although communication means that appeal to sight and hearing, such as telephone and television, are common, the development of information transmission technology by human five senses of taste, smell, and touch is still unexplored. When the five senses information communication is realized, it is expected to be applied to welfare and education fields such as telemedicine. It also enables more natural and realistic communication.

 Five senses Information communication requires the development of technology that interprets smell, taste, and texture as information.

 Taste is the taste of the action of the chemical receptors on the tongue, oral mucosa and larynx. These chemical receptors contain taste buds, taste organs that taste. The taste buds are most common in the papilla of the tongue, with an average of 2 to 5,000 taste buds in adults, with one taste bud consisting of about 50 to 130 cells, many of which taste. It acts as a receptor. The surface of the taste buds has very fine villi, and there are tiny pores called micropores that pass through the ion channels in the cell membrane to make them taste. Large molecules are signaled in cells that bind to specific receptors on the cell surface.

 As there are three primary colors of red, green, and blue in light, there is a raw rice that is basic to taste. In order for a taste to be raw, there must be a taste receptor site different from that of other raw materials, the quality of the taste must be different from the other raw materials, and it can satisfy the three conditions that it cannot produce the taste even if several raw materials are mixed. You have to.

 In the Orient, five flavors (5) corresponding to Yin-Yang Five Elements were referred to as sweet, bitter, salty, sour, and spicy. In the West, four kinds of sweet, bitter, salty, and sour, corresponding to the four-way theorem, were considered as basic tastes. . In recent years, sweet, bitter, salty, sour, and umami (such as meat broth, tastes like MSG) are recognized as 5 raw rice that meets the original conditions. However, in addition to the raw rice mentioned above, people feel various 'flavors' such as spicy and astringent tastes. These tastes are not perceived by taste receptors like raw rice, but are complex senses in free nerves in the mouth or mucous membrane of the tongue. For example, spicy taste is a kind of painful sensation in the free nerves in the oral cavity. Spicy foods are known to contain different kinds of spicy ingredients. For example, pepper is capsaicin, pepper is piperine, mustard isothiocyanate, ginger is ginger, and garlic is allicin. Astringent taste is a sense of contraction of mucous membranes when a component containing polyphenols has a strong hydrogen bond with mucous membrane protein of tongue. Tannins in black tea, beer, and wine are typical astringents. When you eat peppermint, you feel cold. This is the result of the paralysis of the tongue's mucous membranes due to the cooling effect that occurs when the peppermint component absorbs the heat of dissolution when dissolved in saliva. Therefore, humans do not determine taste with only five taste receptors, but depend on the temperature of food, the strength of food, and the taste of food eaters. In addition, the aromas in the food excite the sense of smell, and determine the taste, including all the feelings that the eyes feel and the touch. In other words, taste is not only a sense of taste, but also a synthesis of emotions including all senses such as touch, sight, and smell.

 In order to transmit this information through communication, development of an encoding technique and a compression technique for transmitting taste information and a technique of accurately analyzing the information on the receiving side and reproducing the human taste and aroma are indispensable. In other words, there is a need for a method of expressing taste in order to transmit it over a communication.

  Again, according to the Japan Postal Communications Bureau, the realization of the five senses ICT is expected to be "five to ten years later." Future challenges are expected to secure communication quality over the Internet and transport technologies in wideband networks such as optical fibers. Human senses such as smell, taste, and touch should also be expressed as digital images and transmitted as in the case of visual and audio video and audio.

In general, when the taste information is defined in various languages, since there are so many kinds, it is almost impossible to database all the individual tastes that humans feel. Accordingly, the present applicant proposes a method of mapping an emotion of taste or a taste of an image to a predetermined expression factor, encoding the same, and restoring the input encoding information to a representative taste using FCMA. However, the problem at hand is that if the expression factor of taste is about 40 ~ 50, it can be expressed in numerous tastes by combining them, but at least 40 expression factors determined by the way of expressing human emotion, The decision process requires a calculation in at least 40 dimensions, and the center value of the representative taste is also present in more than 40 dimensions, which requires a large memory and increases the calculation time and calculation amount.

The technical problem to be achieved by the present invention is to encode the expression factor for the taste feeling of taste information or video image, as in the case of video and audio data by visual and auditory, and then first, PCA (Principal Component Analysis) An object of the present invention is to provide a method of reconstructing a large-dimensional encoded data into 2-3 dimensions by an algorithm and then using a fuzzy c-means algorithm (FCMA).

The taste information used here is defined as not only the five elements but also all the information felt in the mouth. If the expression factor of the taste is about 40-50, it is possible to express many tastes by combining them. Even if the way of expressing taste is different for each human being, it is clustered by the common denominator of the way of expressing taste. Even if the two flavors are mixed, a combination of flavors can be expressed because the degree of belonging gives a greater degree of belonging to the two representative flavors. According to this method, even if the expression factor is 40 or more dimensions, it can be reduced to 2 or 3 dimensions using the PCA algorithm, thus requiring only a small amount of memory for the representative taste. It can be considered that there is almost no loss of data because it can be expressed more than 98% of the original data if it is reduced to about three dimensions.

The method proposed in the present invention is expected to be applicable to MPEG-7 and MPEG-21 using meta data for the texture representation of multimedia data, and is a basic technology for standardizing taste information processing technology. have.

In order to achieve the above technical problem, the present invention provides a method of digitally decoding a taste of a taste information or a video image, wherein the encoding is performed according to a representative expression of a representative taste, and the center value of the representative taste is clustered by clustering. Obtaining and storing; Receiving the taste information that is encoded and transmitted and mapping the pattern of data to dimensionally reduced data that can be visually confirmed by a human by principal component analysis; And obtaining the center value of the representative taste by clustering the coded taste information having reduced dimensions, and comparing the stored center value with the center value of the stored representative taste to obtain the degree of belonging to the center value of the representative taste of the input taste information. And determining, by encoding the taste information, classifying various forms of language expression as taste expression factors by which humans feel and speak; Mapping a code to the classified taste factor; It provides a digital decoding method comprising the step of expressing the code and the concentration or degree of belonging of one or more taste factors representing a taste in an integer or a real type.

Preferably the clustering is performed using a FCMA (Fuzzy c-means algorithm), by the following equation

...수학식 12

... Equation 12

...수학식 13 a) 수학식 12에 의해 초기 소속함수 Z(0) 결정 및 클러스터 j와 패턴 i의 소속도 W_ij 을 계산하고

Equation 13 a) Determination of the initial membership function Z (0) and the membership degree W _ij of the cluster j and the pattern i by

b) calculate the cluster center Z (k + 1) by

c) ∥ Z (k) -Z (k + 1) ∥

If ∥Z (k) -Z (k + 1) ∥> ε, k = k + 1 and go back to the beginning. Otherwise, the algorithm is terminated to obtain cluster membership.

Preferably, when the actual encoded taste information is transmitted, the encoded data of the taste factor is arranged as metadata after the image data compressed by MPEG-7 or MPEG-21.

Hereinafter, with reference to the accompanying drawings, the configuration of a preferred embodiment of a method for reducing the dimension after the encoding of the expression factor for the taste feeling of the taste information or image image of the present invention and to restore the input encoded taste information based on the representative taste And the operation will be described in detail.

First, prior to describing the expression factor for digitally expressing the taste information by the emotional expression of the present invention, the existing taste expression technology will be described.

There has been much research to classify and express taste.

The German psychologist H. Henning, as shown in Fig. 1, thought that four basic flavors were placed at the vertices of a triangular pyramid, and that various kinds of tastes could be expressed on a surface or as a point inside. This is called the tetrahedron of Henning's taste.

There are many sweet substances such as sugars, alcohols and glycerol having OH groups, and some of them do not have OH groups like saccharin. The sweet substance has a hydrogen donor and a hydrogen acceptor in common, and is 2.5 to 4Å apart from each other. Sour taste is the taste of hydrogen ions, anion also affects the sour taste. Even at the same pH, the acid strength is in the order of acetic acid <formic acid <lactic acid <nitric acid <hydrochloric acid. Bitter substances are abundant in inorganic compounds such as magnesium chloride, and alkaloids such as brucin and caffeine are typical of organic compounds. Representative salts with salty taste are sodium chloride. Sodium chloride, potassium chloride, sodium iodide, etc., each have a different taste because salty taste depends on both the cation and anion of the salt. Taste occurs mainly on the surface of the tongue.

There are differences in susceptibility to the four basic taste substances on the surface of the human tongue and on each part of the soft palate. VB calling psychologists in the United States seuneun 1974 saltiness and sweetness intensity of the stimulation of the lowest in the front part of the tongue, sour lowest at the edges of the tongue, bitter is the lowest in the palate. Increasing the stimulus concentrations, however, resulted in a measurement of four basic tastes at any site. In addition, the intensity of the stimulus to the taste changes with the temperature of the liquid to taste, the temperature showing the lowest intensity of the stimulus is usually 22 to 32 ℃, depending on the nature of the taste. The fifth fifth taste receptor is called taste receptor for mono sodium glutamate (MSG) called Umami, which is also known as a chemical widely used in Asia to enhance the taste of food. . So far, there are four things that have been identified as the main tastes of food. Sweet, salty, sour, and bitter tastes have been steadily raised by many scholars that umami should be classified into unique tastes, just like the other four, and equally important as other tastes. According to Nirupa Chaudhari, a research team at the University of Miami Medical School in the United States, our bodies are very sensitive to glutamate. It is also an ingredient commonly found in sex foods. Glutamate is also known to act as a neurotransmitter (Neurotransmitter) in our brain. Neurons in our bodies (neurons) are present in a variety of receptors that detect glutamate. These receptors act as a kind of chemical doorway and can theoretically be used by cells on the tongue to detect glutamate in food. Umami was first tasted in 1908 by Kikunae Ikeda of the Tokyo Imperial Univ.

The tastes of human taste are only sweet, sour, bitter, salty and umami. However, the taste of human beings is diverse, and the taste is not only taste in taste buds, but also a sense of comprehensiveness accompanied by feelings and smells. In the present invention, it is intended to express using a variety of languages (adjectives) that humans feel and speak. However, we used language that has the taste that each receptor feels and the emotional meaning that humans feel. Although there are many kinds of these languages, it is possible to express and distinguish tastes, including the unique feelings of human beings, and classify basic adjective receptors of taste, that is, expression information expression factors, as shown in the table below.

[table]

number Taste factor number Taste factor number Taste factor One only 20 Alcoholic 39 strong 2 written 21 hot 40 weak 3 God 22 burnt 41 Chewy 4 braided 23 fresh 42 strong 5 Fill 24 unpleasant 43 Tasteless 6 Young 25 fat 44 Stubborn 7 scale 26 watery 8 Stinging 27 Subtle 9 spicy 28 venomous 10 Metallic 29 Crunchy 11 solid 30 Sue 12 light 31 neat 13 cool 32 fresh 14 Warm 33 Bitter 15 cold 34 Light 16 Steamed 35 thick 17 Crunchy 36 oily 18 Melting 37 aged 19 Murky 38 aromatic

Using the 44 adjective expression factors used in the table above, most of the flavors can be expressed.

2 is a configuration diagram for describing the expression and encoding of taste information using an adjective expression factor. When some taste information comes in, it shows an example of accepting and encoding only the expression factor of the taste information among numerous adjective expression factors. As shown in FIG. 2, the taste information expression factor is digitally encoded and transmitted as transmission data. Digital encoding by taste information expression factor refers to a process of changing taste into emotional expression factor and converting it into integer type or real number type. Even though there are about 40 expression factors in total, the length of the entire code is not so large because only a few expression factors have a taste. The encoding converts the number and density of the corresponding expression factor into integer type or real number type, and the number of expression factors corresponding to the front part shows the length of the entire code. Also at the end is a sign indicating the intensity or intensity of the image. For example, if the number of arguments is 7, the total length of the code has a length of 18 bits by adding 2x8 = 16 bits, 1 bit indicating the number of factors and 1 bit indicating the total density.

Here, when the number of arguments is n, the length m of the entire code is

m = 2 × n + 2. The length of the sign does not require a separate compression algorithm, since the number of arguments may not exceed 20 at most, even if it is a complex taste or taste. In addition, since only the expression factor is encoded, it may be said that the compression effect is included.

As shown in Fig. 3, the coding method of taste information expresses the corresponding number of factors, the number of factors, the small storage speed, and the total concentration in integer form.

If it is desired to transmit taste information by wire or wireless, it is possible to use MPEG-7 or MPEG-21 as shown in FIG. In FIG. 4, it is proposed to apply metadata in MPEG by an encoding method. For example, if two young lovers in a restaurant are tasting the wine in an image of an image, the adjectives are "god", "sweet", "cool", and "fresh". The encoding then becomes 3 1 13 32. Also, if you want to express the miso stew that your mother cuts off, you can express it with adjectives such as "light", "savory", "dark", "salt", and "cool". However, the taste can be distinguished by adjusting the concentration and flavor characteristics corresponding to each factor. If you code this taste of Doenjang-jjigae, it would be 34 5 35 4 13. It is understood that the actual encoding length is not so large because only the corresponding expression factors are encoded, not all dozens of factors.

FIG. 5 is a diagram representing certain clusters by mapping certain representative flavors onto two dimensions. In FIG. 5, each cluster is formed by emotions commonly felt by humans. Even though one person expresses emotions according to individual tastes without using expression factors of general feelings about the taste of C group, the other clusters are expressed in other clusters. The membership is closer to C, so it may be judged to belong to the C group. In particular, the present invention overcomes the ambiguity of human emotion by using a fuzzy clustering method (Fuzzy c-means algorithm) that exhibits robust clustering ability against ambiguity of human emotion.

In the present invention, the taste information is classified and represented using Principal Component Analysis (PCA) and FCMA (Fuzzy c-means algorithm). Determination of clustering and affiliation of taste information using PCA and FCMA is as follows. In order to express taste information, it is necessary to first define representative tastes that exist around us. However, there is a problem that the definition of taste that humans feel may vary according to ethnicity, region, and individual taste. However, if the definition of taste that humans feel in common is defined without being excluded, the definition of expression that feels differently for each individual may not be a big problem.

However, since there are so many kinds of flavors, a lot of representative flavors may be determined. In particular, in the case of taste, numerous representative flavors may exist because the sense of touch and feeling in the mouth other than the five elements must be expressed. In this case, it takes up a lot of memory because it has to be in the place of decoding the center value for this representative taste. In particular, when FCMA is used for decoding, if there are 40 expression factors, it has to take up a lot of memory because it has to have a center value for the representative taste in the 40-dimensional space. This problem may exist as a constraint in practical applications.

In order to overcome this problem, first, a method of compressing data using the PCA method, which is a method of mapping encoded data into a two-dimensional or three-dimensional space, is applied.

That is, in the present invention, the PCA method (Principal component analysis) and the FCM (Fuzzy c-means) algorithm has a correlation between the visual analysis and the quantitative analysis with the purpose of making the multi-dimensional data 2 We want to compress the data by dimming it into ~ 3 dimensions. To this end, we first obtain a three-dimensional eigenvector through the PCA method, and then implement the FCM algorithm in three dimensions using data whose dimensions are reduced by this eigenvector. Through this, even if new data comes in, it can be visually expressed, and through the center of the cluster that correlates with the visual data, the degree of belonging between each cluster of newly input data can be obtained.

6 is a block diagram illustrating a clustering process for obtaining center values of various representative flavors after dimension reduction through PCA. 6 is encoded 64 through the factor representation 62 of the various representative flavors 60, the encoded data is mapped in 2-3 dimensions through the PCA algorithm 65, and then FCMA (Fuzzy c-means algorithm) The process of clustering the taste information 66 and obtaining the center value 68 of the representative taste is shown by using. That is, the center value 68 of clusters can be obtained using FCMA. Here, by using the PCA algorithm, the center value for each taste can be reduced to 2-3 dimensions.

7 is a configuration diagram of a process of transmitting and decoding encoded data. In FIG. 7, when new taste information data 70 comes in, it is encoded 74 through the factor expression 72 of the taste and transmitted, clustered 76 of the taste information to obtain the center value of the representative taste, and the stored representative. Compared to the center value 78, the degree of belonging 80 between the clusters of taste can be obtained, so that it is possible to find out which cluster the taste information belongs to. In addition, even when a plurality of taste information data are mixed, it is possible to determine which cluster is mixed through the membership analysis. This process is called the decoding process of taste, and this also reduces the encoded data transmitted using the PCA algorithm 75 to 2-3 dimensions and compares it with the center value of 2-3 dimensions of the representative taste already stored to taste. Express the degree of belonging of information.

Hereinafter, the PCA (Principal Component Analysis) algorithm and FCMA (Fuzzy c-means algorithm) applied to the present invention will be described in more detail.

The PCA method is an algorithm that maps patterns of multidimensional data into two-dimensional or three-dimensional data that can be visually identified by humans, as shown in Equation (1).

Where X is the input data received by the sensor. Where N is the number of patterns and L is the number of sensors. This N × L data is now changed to the N × D data form as shown in Equation (2).

Where Y is mapping data and D is the number of dimensions desired. If k = 1, 2, ..., n, Equation 3 is used to calculate the average for each sensor.

를 빼어감에 따라 계산되어 진다.Where n is the number of sensors and m is the number of patterns. Using the average data, Equation 4 is used to obtain a covariance matrix for the input. The covariance matrix C is the mean vector in the input pattern X so that each element, C _ij, is _centered over all classes.

Calculated by subtracting

Next, the eigenvalues λ _k and eigenvectors u _k for k = 1, 2, ..., n are obtained, and then defined as in Equation 5.

Select the largest eigenvalues λ ₁ , λ _2, and λ ₃ from the calculated λ _k (if 3-D mapping), and use the eigenvectors u ₁ , u _2, and u ₃ as the principal components for multidimensional data. Mapping in three dimensions.

In general, mapping in three dimensions allows for mapping without significant loss of data in most cases with some characteristics.

This PCA method allows for clustering analysis visually. However, when new data is acquired, a separate algorithm is required to quantitatively determine which cluster is close to or belongs to which cluster.

To this end, in the present invention, three-dimensional data obtained based on the three-dimensional eigenvectors obtained using the PCA algorithm are obtained from the center of the cluster using the FCM algorithm, and the data inputted using the obtained center of the cluster. Apply the method of obtaining membership of.

Basically, FCM algorithm is unsupervise learning, but if standard data about taste is defined in advance, it can be decided as same cluster for same data. It is possible.

Quantifying the expression using a separate k-means algorithm or FCM algorithm has a different correlation between the visually analyzed data and the data obtained by the FCM algorithm because the dimensions of the PCA method are different from those of the FCM algorithm. Have. The use of the FCM using the results of the PCA enables to obtain the degree of belonging between each cluster of newly input data through the center of the cluster that correlates with the visual data represented by the PCA.

The FCM algorithm for quantitative analysis uses data mapped by the PCA process as an input vector and displays y, a component of Y _i , as a vector.

Where n is the number of input patterns and represents a matrix reduced in three dimensions by PCA. If it is displayed again as an input pattern, it is as shown in Equation 7-1.

Equation 7-1

The center of the cluster for the input pattern set is represented by Equation 7-2 below.

[Equation 7-2]

Where c is the number of cluster centers, _i of i is the number of input patterns, and _j of j is the number of clusters. Therefore, the membership degree W _ij of the center of the i-th pattern and the j-th cluster is expressed as follows.

Each of the above patterns may have different membership and clusters. And, as shown in Equations 9 and 10, the sum of the degree of belonging to one pattern is normalized to 1, and each pattern belonging degree finds the pattern of the cluster having the minimum objective function J _m .

D _ij in the objective function J _m is the Euclidean distance between the input pattern and the center of the cluster. This measures the similarity for each pattern. Where m is a membership weighting exponent and m 〓 2 is usually chosen because there is no theoretical and justified rule for choosing this value.

 The solution of the objective function of the FCM algorithm proposed by Bezdek is as follows.

FCMA is an iterative algorithm that approximates the minimum value of the objective function of Equation 13 without being solved analytically by Equations 12 and 13. Therefore, the learning values of equations 12 and 13 converge to optimal values.

Since this algorithm is an unsupervised pattern recognition technique, a corresponding learning procedure is required. This learning procedure brings the degree of belonging to the cluster center of the input pattern and indicates the distance of each input pattern to the cluster center.

The process of FCM algorithm for learning is summarized as follows.

1. Determination of initial membership function Z (0) and calculation of membership degree W _ij of cluster j and pattern i through Equation 12

2. Calculation of cluster center Z (k + 1)

3. Calculation of Z (k) -Z (k + 1) ∥

If ∥Z (k) -Z (k + 1) ∥ε = k = k + 1, go back to the beginning, otherwise terminate the algorithm.

Through the PCA process, the dimension reduction results were clustered again using the FCM algorithm to obtain a sense of belonging, visual effects, and quantitative results of new data.

Through this process, the center of the cluster with respect to the standard patterns obtained in advance and the cluster membership of each standard pattern can be obtained. The cluster centers obtained here are stored in the computer in advance and used to calculate the degree of belonging to an unknown sample when an unknown sample is input.

Where S is an unknown sample vector and j = 1, 2,... , c and the degree of belonging between the center of the standard pattern and the sample pattern can be obtained by the following equation (15).

In Equation 15, V _j represents a degree of belonging of the sample vector to the j th cluster.

As described above, after encoding the expression factor for the taste feeling of taste information or video image of the present invention, first, the dimension is reduced by PCA and the input encoding information is restored based on the representative taste by FCMA. It is possible to reduce the amount of memory to store the representative taste.

As described above, as in the case of video and audio data by visual and auditory sound according to the present invention, after encoding the expression factor for the taste feeling of taste information or video image, the dimension is first reduced by PCA and The method of restoring the input encoding information based on the representative taste can restore the taste information without using much memory. However, the method of expressing the expression factors presented in the present invention is not limited thereto, and it will be apparent to those skilled in the art that various modifications may be made, and thus the description of each modification will be omitted. In the present invention, the part of sensing the taste information is not described in detail, but when the taste information processing technology of the present invention is connected to a sensor that senses the taste information, the taste information can be processed digitally, and the taste information is remotely obtained. There is an effect that can determine.

Claims (4)

In the method of digitally decoding the taste of taste information or video image, Performing encoding according to the factor expression of the representative flavor, and obtaining and storing the center value of the representative flavor by clustering; Receiving the taste information that is encoded and transmitted and mapping the pattern of data to dimensionally reduced data that can be visually confirmed by a human by principal component analysis; And The center value of the representative taste is obtained by clustering the coded taste information of which the dimension is reduced, and compared to the center value of the stored representative taste to obtain the degree of belonging to the center value of the representative taste of the input taste information. And determining the encoding of the taste information. Classifying various forms of language expression as taste expression factors that humans taste and speak; Mapping a code to the classified taste factor; A digital decoding method comprising the step of expressing the code and the concentration or the degree of belonging of one or more taste factors representing one taste in an integer or real form. delete The method of claim 1, wherein the clustering is performed using a fuzzy c-means algorithm (FCMA),

... Equation 12

... Equation 13 a) Determining the initial membership function Z (0) and calculating the membership degree W _ij of the cluster j and the pattern i by using Equation 12 b) calculate the cluster center Z (k + 1) by c) ∥ Z (k) -Z (k + 1) ∥ If ∥Z (k) -Z (k + 1) ∥> ε, then go back to the beginning and return to the beginning, otherwise terminate the algorithm to obtain cluster membership. [4] The method of claim 1 or 3, wherein when the actual encoded taste information is transmitted, encoded data of a taste expression factor is arranged as metadata after the image data compressed by MPEG-7 or MPEG-21. Digital information decoding method of taste information.

Images