KR100883185B1 - System and Method for Detecting Abnormality of Biosignal using Neural Network with Weighted Fuzzy Membership Funtions - Google Patents

Abstract

The present invention relates to an apparatus and method for extracting abnormal data of a biological signal for detecting a biological signal and extracting abnormal data of the detected biological signal. More particularly, the present invention relates to a weighted fuzzy membership function-based neural network algorithm enhanced in a detected biological signal. It is to extract the fuzzy characteristics of the bio-signals accurately by applying the and to extract the abnormal data of the bio-signals from the detected bio-signals by comparing the fuzzy characteristics of the extracted bio-signals with the stored fuzzy rules.

An apparatus for extracting abnormal data of a biosignal using a weighted fuzzy membership function based neural network algorithm includes a biosignal detection unit detecting a biosignal of a biosignal detection target and a wavelet coefficient by performing wavelet transform on the detected biosignal A fuzzy characteristic extractor extracting a fuzzy characteristic of a biosignal by applying a wavelet coefficient extracting unit extracting a signal, a weighted fuzzy membership function-based neural network algorithm, and a fuzzy characteristic of the extracted biosignal And an abnormal data extracting unit for comparing the stored fuzzy rules to extract abnormal data of the biosignal and accumulating and storing the abnormal data of the extracted biosignal.

Abnormality of bio signal, fuzzy neural network, bio signal

Description

System and Method for Detecting Abnormality of Biosignal using Neural Network with Weighted Fuzzy Membership Funtions}

1 is a block diagram of an apparatus for extracting abnormal data of a biosignal using a weighted fuzzy membership function based neural network algorithm according to an embodiment of the present invention;

2 is a hierarchical diagram showing a four-level wavelet transform applied to the present invention;

3 is a detailed configuration diagram of a fuzzy characteristic extraction unit for extracting a fuzzy characteristic of a biosignal shown in FIG. 1;

4 is an exemplary diagram illustrating an i th weighted fuzzy membership function set of a hyperbox node according to the present invention;

5 is an exemplary diagram illustrating an example in which a membership function and a weight of a hyperbox node according to the present invention are adjusted;

Figure 6 is an exemplary view showing an embodiment incorporating the fuzzy characteristics of the weighted fuzzy belonging function completed learning in accordance with the present invention,

7 is a neural network structure diagram for performing a weighted fuzzy membership function based neural network algorithm applied to the present invention;

8 is a flowchart illustrating a method of extracting abnormal data of a biosignal using a weighted fuzzy membership function based neural network algorithm according to an exemplary embodiment of the present invention.

<Description of the symbols for the main parts of the drawings>

110: biosignal detection unit 120: wavelet coefficient extraction unit

130: fuzzy characteristic extraction unit 140: fuzzy rule storage unit

150: abnormal data extraction 160: abnormal data storage unit

170: abnormal data display unit 180: abnormal data transmission unit

The present invention relates to an apparatus and method for detecting a biosignal and extracting abnormal data of the detected biosignal, and more particularly, applying a weighted fuzzy membership function-based neural network algorithm to a biosignal and detecting the biosignal. By accurately extracting the fuzzy characteristics of the bio-signals and comparing the fuzzy characteristics of the extracted bio-signals with pre-stored fuzzy rules, only the abnormal data of the bio-signals are extracted.

In general, a device for detecting a biosignal or analyzing a detected biosignal is a device for measuring a physiological signal generated in a living body such as an electroencephalogram, an electrocardiogram, safety, blood pressure, or an induced electromyography, and measures a biosignal of a patient. Stored and used for medical applications.

In order to apply a biosignal detected by a biosignal detection apparatus to a patient's care, an algorithm for accurately extracting a biosignal (abnormal data) from the detected biosignal is required. Only the abnormal data is extracted from the detected biosignal. A device is needed to do this.

The present invention has been made in accordance with the above-described requirements, and an object of the present invention is to accurately extract the fuzzy characteristics of a biosignal by detecting a biosignal and applying a weighted fuzzy membership function based neural network algorithm to the detected biosignal. An apparatus and method for extracting abnormal data of a biological signal using a weighted fuzzy membership function-based neural network algorithm for extracting and storing only abnormal data of a biological signal by comparing the fuzzy characteristics of the extracted biological signal with previously stored fuzzy rules are provided.

An apparatus for extracting abnormal data of a biosignal using a weighted fuzzy membership function based neural network algorithm according to an embodiment of the present invention for achieving the above object includes a biosignal detection unit for detecting a biosignal of a biosignal detection subject, and the detected signal. A wavelet coefficient extracting unit extracting wavelet coefficients by performing wavelet transform on a biosignal, a fuzzy feature extracting unit extracting fuzzy characteristics of a biosignal by applying a weighted fuzzy membership function-based neural network algorithm to the extracted wavelet coefficients; And an abnormal data extracting unit for extracting abnormal data of the biological signal by comparing the fuzzy characteristics of the extracted biosignal with a pre-stored fuzzy rule and an abnormal data storing unit for accumulating and storing abnormal data of the extracted biosignal. Characterized in that made.

In addition, a method for determining an abnormality of a biosignal using a weighted fuzzy membership function based neural network algorithm according to an embodiment of the present invention includes detecting a biosignal of a biosignal detection subject, and performing wavelet transform on the detected biosignal. Performing a wavelet coefficient extraction step of extracting wavelet coefficients, a fuzzy characteristic extraction step of extracting fuzzy characteristics of a biosignal by applying a weighted fuzzy membership function-based neural network algorithm to the extracted wavelet coefficients, and extracting the extracted biosignal And extracting abnormal data of the biosignal by comparing the fuzzy characteristics of the biosignal and the previously stored fuzzy rules, and storing the abnormal data of the biosignal by accumulating and storing the abnormal data of the detected biosignal. It features.

The above-described contents of the present invention will become more apparent through the following detailed description with reference to the accompanying drawings, and thus, those skilled in the art to which the present invention pertains may easily implement the technical idea of the present invention. will be. In addition, in describing the present invention, when it is determined that the detailed description of the known technology related to the present invention may unnecessarily obscure the gist of the present invention, the detailed description thereof will be omitted.

Hereinafter, exemplary embodiments of the present invention will be described in detail with reference to the accompanying drawings.

1 is a block diagram of an apparatus for extracting abnormal data of a biosignal using a weighted fuzzy membership function based neural network algorithm according to an exemplary embodiment of the present invention.

Referring to FIG. 1, the apparatus for extracting abnormal data of a biosignal may include a biosignal detector 110 that detects a biosignal of a biosignal detection target, and a wavelet coefficient extracting wavelet coefficients by performing wavelet transform on the detected biosignal. An extractor 120 and a fuzzy feature extractor 130 for extracting a fuzzy characteristic of a biosignal by applying an enhanced weighted fuzzy membership function-based neural network algorithm described below to the extracted wavelet coefficients; The abnormal data extraction unit 150 extracts the abnormal data of the biological signal by comparing the fuzzy characteristics and the fuzzy rules previously stored in the fuzzy rule storage unit 140 and the abnormal data storing to accumulate and store the abnormal data of the extracted biological signal. The unit 160 is made.

The biosignal detection unit 110 may be configured as a halter for measuring the electrocardiogram of the target of the biosignal detection, and detects EEG, blood pressure, safety, EMG, and the like. The biosignal detection unit 110 is configured with a known general biosignal detection device.

As shown in FIG. 2, the wavelet coefficient extractor 120 uses the wavelet coefficients in the frequency bands for levels 3 and 4 by using the wavelet transform divided into four levels. Wavelet coefficients in the frequency bands for levels 1 and 2 are less characteristic than levels 3 and 4 and contain no noise.

The wavelet transform is a signal processing method capable of simultaneous analysis in the time-frequency domain, and the wavelet coefficient extracted by the wavelet transform indicates a similarity with respect to the mother wavelet. In addition, the wavelet transform multiplies a scale function and a mother function by an input signal and divides the low-pass filter and the high-pass filter.

As shown in FIG. 3, the fuzzy characteristic extracting unit 130 generates a hyperbox node having n fuzzy sets for the extracted wavelet coefficients, and generates the hyperbox node by the enhanced hyperbox node output equation. A membership function calculation unit 131 for obtaining a belonging function and a vertex and weight adjustment unit 131 for calculating a belonging function having a new vertex and weight by adjusting the membership function of the hyperbox node and its weight by a learning algorithm described later. And a fuzzy characteristic integrating unit 135 incorporating the fuzzy characteristics of the membership function having the new vertices and weights. The fuzzy characteristic extracting unit 130 has a weighted purge belonging function based neural network structure to be described later, and the belonging function calculating unit 131, the vertex and weight adjusting unit 133, and the fuzzy characteristic integrating unit 135 are substantially A hyperbox layer of a weighted fuzzy membership function based neural network structure to be described later is configured.

The membership function calculation unit 131 calculates a membership function belonging to a hyperbox node, and generates a hyperbox node having n fuzzy sets by a Random (B ₁ ) operation. The Random (B ₁ ) operation is for generating a hyperbox B ₁ , and has a random weight in the range of 0.45 to 0.55 and has n (number of inputs) membership functions. Arbitrary center point v _i must be within the following range.

On the other hand, as shown in Figure 4, v ₀ and v ₄ is a fixed value, the connection weight w _li is set to zero.

The w _li is hyperbox node B _l and class node C _i represents the connection weight between, in the state between the hyperbox node B _l and class node C _i that are not connected, but set to an initial value 0, and the hyper box nodes B _l When w and class node C _i are connected, w _li is set to 1. Here, C _i may have more than one connection from the hyperbox node while B _l is restricted to having only one connection to the class node.

는 도 4에 도시된 바와 같이, vmin ~ vmax 의 범위 내로 가정한다.The v _i represents the center point of the membership function, and v ₁ ㆍ v ₂ · v ₃ represents the center points of the large, medium, and small membership functions, respectively, as shown in FIG. 4. These center points are adjusted during the learning process and center points v ₀ and v ₄ are fixed values. Input data

Is assumed to be in the range of vmin to vmax, as shown in FIG.

The average value of the membership functions belonging to the hyperbox node is calculated by an Output (B ₁ ) operation. With the nth input Ah = (a ₁ , a ₂ , ...., a _n ) with _n features in hyperbox B _l , the output of B _l (value for class node classification, mean value of membership functions) is In general, the following equation is defined as: < 1 >

Equation <1>

의 출력 값을 하기 수학식 <2>(강화된 하이퍼박스 노드 출력식)과 같이 정의한다.The reasoning mechanism performed to obtain the Output (B _l ) value is described in Takagi & Sugeno model [T. Takagi and M. Sugeno, "Fuzzy Identification of Systems and Its Application to Modeling and Control," IEEE Trans., Syst., Man. , Cyben., Vol. 15, pp. 116-132, 1985. In the embodiment according to the present invention in order to enhance the inference performance

The output value of is defined as in Equation <2> (Enhanced Hyperbox Node Output Expression).

Equation <2>

의 i번째 집합의 소속함수이며, j=1, 2, 3 은 각각 대, 중, 소를 나타낸다. 각 소속함수

의 모양은 3개의 정점(vj-1, vj, vj+1)을 가지는 삼각형으로 되어 있는데 학습에 의한 소속함수의 강도(strength)를 나타내는 소속함수 가중치 W_j(0 ≤ W_j ≤ 1 , 초기값 0.45 ≤ W_j ≤ 0.55)를 갖는다. Μ _j is the hyperbox node

The membership function of the i th set of, where j = 1, 2, and 3 represent large, medium, and small, respectively. Each member function

The shape of is a triangle with three vertices (vj-1, vj, vj + 1). The weight of the membership function representing the strength of the membership function by learning W _j (0 ≤ W _j ≤ 1, initial value) 0.45 ≦ W _j ≦ 0.55).

In FIG. 4, the triangle of the shaded portion, which is a weighted fuzzy function, is shown by v _j-1 , W _j , and v _{j + 1} .

In addition, Len (μ _j ) represents a length for measuring μ _j and is defined as in Equation <3>.

Equation <3>

The enhanced hyper box node equation is based on the principle that less Len (μ _j ) has more aggregate information.

와 입력

간의 차이를 나타낸다. 만일 E_j가 인접 E_{jㅁ 1}보다 크면 적은 것이 선택된다.)The vertex and weight adjusting unit 132 uses an Adjust (B _j ) operation. The i th membership function with the input a _i and the membership functions of v _i and W _j are adjusted by the value of μ _j (a _i ) (j = 1,2,3). As a result of the Adjust (B _j ) operation, a new vertex new (v _j ) and a new weight new (W _j ) are set as in Equations <4> and <5>, where α and β are between 0 and 1. Is the learning rate range of

And input

Indicates a difference. If E _j is greater than neighboring E _{j ㅁ 1} , less is chosen.)

Equation (4)

Equation <5>

More detailed Adjust (B _j ) calculation process is performed according to a learning algorithm described later.

5 shows examples before and after Adjust (B _j ) calculation for the weighted fuzzy-proportional function of the i-th set of inputs a _i and B _l .

,

) 프로시저(procedure)를 이용하여 정점 및 가중치의 위치를 조정하고 하이퍼박스 노드를 클래스 노드에 연결한다. 강화된 가중퍼지 소속함수 기반 신경망 알고리즘 및 학습 알고리즘의 일예는 <표 1> 과 같다. The learning algorithm is Learning () which is a learning procedure function.

,

Use procedures to adjust vertex and weight positions and connect hyperbox nodes to class nodes. An example of the enhanced weighted fuzzy membership function based neural network algorithm and learning algorithm is shown in Table 1.

TABLE 1

Fuzzy characteristic integration section 133 for integrating the fuzzy nature of the membership function incorporates a fuzzy membership function having the characteristics of the new vertex and the weight by μ _b ⁱ (x) operation.

As shown in FIG. 6, the fuzzy characteristic integration of the membership function by the μ _b ⁱ (x) operation is a bounded sum (BSWFM) of the weighted fuzzy membership function, denoted by a bold line, 6>.

Equation (6)

The fuzzy characteristics of the three weighted fuzzy membership functions that have been learned by the above process are integrated by the above equation. The fuzzy characteristic of the membership function integrated by the above equation is the fuzzy characteristic of the biosignal finally input to the abnormal data extraction unit 150.

The fuzzy rule storage unit 140 stores fuzzy rules extracted by the weighted fuzzy membership function algorithm of the present invention for the normal and abnormal signal samples of the biosignal.

The abnormal data extraction unit 150 is a fuzzy rule for the abnormal data of the bio-signal among the fuzzy rules previously stored in the fuzzy rule storage unit 140 by the fuzzy characteristic of the bio-signal extracted by the fuzzy characteristic extraction unit 130 In comparison, the fuzzy characteristic of the biosignal extracted by the fuzzy characteristic extractor 130 and the biosignal waveform corresponding thereto are extracted and stored in the abnormal data storage 160.

The abnormal data storage unit 160 stores abnormal data of the biological signal and its surrounding data separately for each subject and the biological signal. When the abnormal data extracting unit 140 extracts the abnormal data of the biosignal 20 seconds after the start of extraction, the peripheral data is input to the abnormal data extracting unit 140 between 17 seconds and 23 seconds, for example. Means a fuzzy characteristic and a biosignal waveform corresponding thereto.

In addition, as shown in Figure 1, the abnormal data extraction apparatus of the biological signal according to the present invention is abnormal data for separating and displaying the abnormal data of the biological signal stored in the abnormal data storage unit 160 for each subject and the biological signal The display unit 170 may further include an abnormal data transmitter 180 for transmitting abnormal data of the biosignal to a remote location for a wireless or wired communication network.

7 is a neural network structure diagram for performing a weighted fuzzy membership function based neural network algorithm applied to the present invention.

Referring to FIG. 7, the neural network structure based on the weighted fuzzy membership function may include the following equations (1) to fuzzy characteristics of wavelet coefficients input to the input layer 310 for inputting wavelet coefficients and the input layer 320. The hyperbox layer 320 for extraction by the equation (6) and the learning algorithm, and the class layer 330 for classifying the fuzzy characteristics extracted by the hyperbox layer 320.

The input layer 310 is composed of n input nodes and inputs n feature patterns to each node.

은 단 하나의 클래스 노드에 연결되며 n개의 입력 노드를 위한 n개의 퍼지 집합을 갖는다.

의 i번째 퍼지집합은

로 표시되며, 도 4에 도시된 바와 같이, 3개의 가중 소속함수를 갖는다. The hyperbox layer 320 is composed of m hyperbox nodes to which n input nodes of the input layer 310 are connected. Where each hyperbox node

Is connected to only one class node and has n fuzzy sets for n input nodes.

I fuzzy set of

As shown in FIG. 4, it has three weighted membership functions.

The output layer 330 is composed of p class nodes for classifying the fuzzy characteristics extracted by the hyperbox layer 320. Each class node is connected to one or more hyperbox nodes. In the neural network structure based on the weighted fuzzy membership function illustrated in FIG. 8, the h th input pattern is I _h = {A _h = (a ₁ , a ₂ , ...., a _n ), class}, where class is a diagnosis And A _h means pattern data for different n features. Also, undefined features are marked with null.

8 is a flowchart illustrating a method of extracting abnormal data of a biosignal using a weighted fuzzy membership function based neural network algorithm according to an exemplary embodiment of the present invention.

As shown in FIG. 8, the method for determining an abnormality of a biosignal using a weighted fuzzy membership function based neural network algorithm according to an embodiment of the present invention may include detecting a biosignal of a biosignal detection subject (S910). And a wavelet coefficient extraction step (S920) of performing wavelet transform on the detected biosignal to extract the wavelet coefficients, and applying the enhanced weighted fuzzy membership function based neural network algorithm to the extracted wavelet coefficients to purge the biosignals. A fuzzy characteristic extraction step (S930) of extracting a characteristic, an abnormal data extraction step (S940) of extracting abnormal data of a biosignal by comparing a fuzzy characteristic of the extracted biosignal and a pre-stored fuzzy rule, and the extracted abnormal data An abnormal data storage step (S950) of accumulating and storing the data is performed.

The biosignal detection step (S910) is a step of detecting a biosignal such as an electrocardiogram, an EEG, a blood pressure, a safety degree, an electromyogram, and the like by the biosignal detection unit 110.

The wavelet coefficient extraction step S920 is a step of extracting wavelet coefficients for inputting the biosignals detected in the biosignal detection step S910 into the input layer of the neural network structure by performing wavelet transform. As described above, the present invention uses features consisting of wavelet coefficients in the frequency bands for levels 3 and 4 using wavelet transforms divided into four levels.

The fuzzy characteristic extraction step (S930) generates a hyperbox node having n fuzzy sets for the wavelet coefficients extracted in the wavelet coefficient extraction step (S920) by the Random (B ₁ ) operation and the equation <2. A function of calculating a belonging function belonging to the hyperbox node by &lt; RTI ID = 0.0 &gt; and &lt; / RTI &gt; a new vertex by adjusting the belonging function of the hyperbox node and its weight by the equations &lt; 4 &gt; And a vertex and weight adjustment step of obtaining a membership function having a weight, and a fuzzy characteristic integration step of integrating the fuzzy characteristics of the membership function having the new vertex and weight by the above equation.

The fuzzy rule extracted by performing the biosignal detection step (S910) through the fuzzy characteristic extraction step (S930) uses only two fuzzy rules generated from two wavelet coefficients, thereby simplifying implementation in a mobile environment and an embedded system. . In addition, the enhanced weighted fuzzy membership function-based neural network algorithm has a high recognition rate because it amplifies the membership function where the features of the input pattern are concentrated and weakens the membership function where the features are distributed.

Accordingly, the apparatus and method for extracting abnormality data of a biosignal according to the present invention have a feature that is simple to implement and has an excellent ability to determine whether an abnormality of a biosignal is present.

The abnormal data extraction step (S940) of the biosignal compares a fuzzy characteristic of the biosignal extracted in the fuzzy characteristic extraction step (S930) and a fuzzy rule for previously stored abnormal data, and when the biosignals are similar, The fuzzy characteristic is judged as abnormal data. The pre-stored fuzzy rule refers to a fuzzy rule extracted by performing the biosignal detection step S910 to the fuzzy characteristic extraction step S930 on normal and abnormal signal samples. Normal and abnormal samples can be obtained from the MIT / BIH arrhythmia database in the case of ECG signals.

The abnormal data storage step (S950) stores the abnormal data of the biological signal extracted from the abnormal data extraction step (S940) of the biological signal and its surrounding data for each subject and the biological signal.

The apparatus and method for extracting abnormal data of a biosignal according to the present invention can extract the exact fuzzy characteristics of a biosignal by using the enhanced weighted purge belonging function based neural network algorithm, and is detected from the biosignal detection apparatus. Since only the fuzzy characteristics of the abnormal biological signals are extracted from the biological signals and stored, they are more efficient in terms of data management and data storage capacity than in the case of storing all biological signals.

As described above, those skilled in the art will understand that the present invention can be implemented in other specific forms without changing the technical spirit or essential features. Therefore, it should be understood that the embodiments described above are all illustrative and not restrictive. The scope of the present invention should be construed as being included in the scope of the present invention all changes or modifications derived from the meaning and scope of the claims to be described later rather than the detailed description and equivalent concepts thereof.

An apparatus and method for extracting abnormal data of a biosignal using a weighted fuzzy membership function based neural network algorithm according to the present invention accurately extract abnormal data of a biosignal by an enhanced weighted fuzzy membership function based neural network algorithm and perform abnormality among detected biosignals. Only data can be automatically separated and saved.

Claims (16)

A biosignal detection unit detecting a biosignal of a biosignal detection subject; A wavelet coefficient extracting unit configured to extract wavelet coefficients by performing wavelet transform on the detected biosignal; A fuzzy characteristic extracting unit extracting fuzzy characteristics of a biosignal by applying an enhanced weighted fuzzy membership function based neural network algorithm to the extracted wavelet coefficients; An abnormal data extracting unit configured to extract abnormal data of the biological signal by comparing the fuzzy characteristics of the extracted biosignal with a pre-stored fuzzy rule; And An abnormal data extraction apparatus of a biological signal using a weighted fuzzy membership function based neural network algorithm, wherein the abnormal data storage unit accumulates and stores the abnormal data of the extracted biological signal. The method of claim 1, wherein the bio-signal detection unit, An apparatus for extracting abnormal data of a biological signal using a weighted fuzzy membership function based neural network algorithm, which detects any one of electrocardiograms such as electrocardiogram, brain wave, blood pressure, safety, and electromyogram. The method of claim 1, wherein the wavelet coefficient extracting unit extracts wavelet coefficients in the frequency bands for levels 3 and 4 by wavelet transform of four levels of the biosignal using a weighted fuzzy membership function based neural network algorithm. Abnormal data extraction device. The method of claim 1, wherein the fuzzy characteristic extraction unit, A belonging function calculator for generating a hyperbox node having n fuzzy sets for the extracted wavelet coefficients and obtaining a belonging function belonging to the hyperbox node by an enhanced hyperbox node output expression; A vertex and weight adjustment unit that adjusts the membership function of the hyperbox node and its weight by a learning algorithm to obtain a membership function having a new vertex and weight; And An apparatus for extracting abnormal data of a biological signal using a weighted fuzzy membership function based neural network algorithm, comprising: a fuzzy characteristic integration unit integrating the fuzzy characteristics of a membership function having the new vertex and weight. The method of claim 4, wherein the enhanced hyperbox node output formula,

(here,

Is a hyperbox node,

Is the i th fuzzy set of the hyperbox nodes,

Is the membership function of the ith fuzzy set,

Is the weight of the membership function,

Is the length to measure the size of the membership function.) An apparatus for extracting abnormal data of a biological signal using a weighted fuzzy membership function based neural network algorithm, which is defined by an equation such as: The method of claim 4, wherein the vertex and weight adjustment unit,

An apparatus for extracting abnormal data of a biosignal using a weighted fuzzy membership function based neural network algorithm, wherein the membership function and its weight of the hyperbox node are adjusted by a learning algorithm. The method of claim 4, wherein the fuzzy integration unit,

An apparatus for extracting abnormal data of a biosignal using a weighted fuzzy membership function based neural network algorithm, wherein the fuzzy characteristics of the membership function having the new vertices and weights are integrated by the following equation. The method of claim 1, wherein the abnormal data storage unit comprises: An apparatus for extracting abnormal data of a biological signal using a weighted fuzzy membership function based neural network algorithm, wherein the abnormal data of the extracted biological signal and its surrounding data are classified and stored for each subject and the biological signal. A biosignal detection step of detecting a biosignal of a biosignal detection subject; A wavelet coefficient extraction step of extracting wavelet coefficients by performing wavelet transform on the detected biosignal; A fuzzy feature extraction step of extracting a fuzzy characteristic of a biosignal by applying an enhanced weighted fuzzy membership function based neural network algorithm to the extracted wavelet coefficients; An abnormal data extraction step of extracting abnormal data of the biosignal by comparing the fuzzy characteristics of the extracted biosignal and a previously stored fuzzy rule; And And an abnormal data storing step of accumulating and storing abnormal data of the extracted biological signal. 16. The abnormal data extraction method of a biological signal using a weighted fuzzy membership function based neural network algorithm. The method of claim 9, wherein the detecting of the biosignal comprises: A method for extracting abnormal data from a biosignal using a weighted fuzzy membership function based neural network algorithm, which detects any one of ECG, EEG, blood pressure, safety, and EMG. 10. The method of claim 9, wherein the wavelet coefficient extraction step extracts wavelet coefficients in the frequency bands for levels 3 and 4 by wavelet transform of four levels of the biosignal using a weighted fuzzy membership function based neural network algorithm. Abnormal data extraction method. 10. The method of claim 9, wherein the purge characteristic extraction step, Generating a hyperbox node having n fuzzy sets for the extracted wavelet coefficients and calculating a membership function belonging to the hyperbox node by an enhanced hyperbox node output formula; A vertex and weight adjustment step of adjusting a membership function of the hyperbox node and its weight by a learning algorithm to obtain a membership function having a new vertex and a weight; And And a fuzzy characteristic integration step of integrating the fuzzy characteristics of the membership function having the new vertices and weights. The method of claim 12, wherein the enhanced hyperbox node output formula,

(here,

Is a hyperbox node,

Is the i th fuzzy set of the hyperbox nodes,

Is the membership function of the ith fuzzy set,

Is the weight of the membership function,

Is the length to measure the size of the membership function.) Abnormal data extraction method of a biosignal using a weighted fuzzy membership function based neural network algorithm, characterized in that defined by the equation. The method of claim 12, wherein adjusting the membership function and weights,

The abnormal data extraction method of the biosignal using a weighted fuzzy membership function based neural network algorithm, characterized in that for adjusting the membership function and its weight of the hyperbox node by a learning algorithm. The method of claim 12, wherein the fuzzy integration step,

An abnormal data extraction method of a biosignal using a weighted fuzzy membership function based neural network algorithm, characterized by integrating the fuzzy characteristics of the membership function having the new vertices and weights by the following equation. The method according to any one of claims 9 to 15, The abnormal data storage step, The abnormal data extraction method of the biological signal using the weighted fuzzy membership function-based neural network algorithm, characterized in that for storing the extracted abnormal data and the surrounding data of each of the subject and the biological signal.

Images