JP2643699B2 - Fuzzy sensor device - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a fuzzy sensor device for detecting an ambiguous sensation such as an indoor environment.

[0002]

2. Description of the Related Art When realizing a so-called intangible sensor for detecting an ambiguous sensation of a human being, for example, a sensation about an indoor environment such as "hot" or "cold", a method of evaluating such sensation is required. It has been proposed to use a multi-input information processing technology such as fuzzy inference or a neural network.

[0003]

However, if the above-described sensor is realized using fuzzy inference, it takes time and effort to create fuzzy rules and membership functions based on a large amount of data, and fuzzy inference is required. There is a problem that a large-capacity memory is required to store the inference algorithm. On the other hand, when the above-mentioned sensor is realized by using a neural network, the evaluation algorithm is completely a black box, so that there is a problem that the characteristics cannot be adjusted according to the use environment and use conditions.

[0004] In view of these problems, the present invention can use a compact intangible sensing algorithm that does not require a large-capacity memory.
An object of the present invention is to provide a fuzzy sensor device that can adjust the algorithm according to the situation.

[0005]

In order to achieve the above object, the present invention is characterized in that a "fuzzy quantification type II" technique is employed in an intangible sensing algorithm. That is, the fuzzy sensor device of the present invention
Is a detection unit for detecting a sensory ambiguity amount;
Fuzzy data from the data collected to quantify
Calculate category weights using the method of quantification type II and obtain
Category weight and the input signal from the detector
By calculating the sample score using
A characteristic distribution representing the relationship between the output and the score is obtained.
Execute algorithm to find function approximating gender distribution
Algorithm execution means .

In an embodiment of the present invention, the algorithm
Means for executing the program includes a storage for storing the algorithm.
Unit, and stored in the storage unit according to an input from the detection unit.
By the function obtained by executing the algorithm
It consists of an arithmetic unit that converts the perceived amount of ambiguity into a numerical value and outputs it.
Is done . Further, this algorithm can be adjusted by the above function according to the use environment and use conditions.

Here, the fuzzy quantification type II will be described. This is a kind of fuzzy multivariate analysis method.
Fuzzy multivariate analysis is an extension that can handle ordinary multivariate analysis up to fuzzy numbers or fuzzy groups. The main fuzzy multivariate analysis techniques currently proposed are as follows: .

Fuzzy regression analysis A method of obtaining a possible linear regression model by treating coefficients of regression analysis as fuzzy numbers.

Fuzzy time series analysis A method of analyzing fuzzy number data given to a time series and reflecting a possibility distribution in a time series model.

Fuzzy quantification class I A method for obtaining a relation between an objective function having a real value and a qualitative explanatory variable represented by a value within the range [0, 1] in a fuzzy group of a given sample. .

Fuzzy Quantification Class II From the data of a qualitative objective variable represented by a value within the range [0, 1] and a qualitative explanatory variable represented by a value within the range [0, 1], fuzzy quantification is performed. A method for finding a linear (linear) expression that represents a group.

Fuzzy quantification III Each sample and category are quantitatively classified based on qualitative data represented by values within the range of the class [0, 1] in consideration of the membership function value of the fuzzy group. Technique.

Fuzzy quantification type IV A method of giving a numerical value such that the distance between the individuals and the familiarity have a monotonous relationship in consideration of the membership function values of the individuals belonging to the fuzzy group.

Among them, the fuzzy quantification class I is a method for explaining an external reference (objective variable) given quantitatively based on a qualitative factor (explanatory variable).

On the other hand, the fuzzy quantification type II is a method for explaining a qualitatively given external criterion (fuzzy group) based on qualitative factors (explanatory variables). FIG. 9 shows data handled in the fuzzy quantification type II. The difference from fuzzy quantification class I is that the external criterion is fuzzy group B
₁ , B ₂ , ‥‥, B _M. The purpose of fuzzy quantification class II is that each category A _i (i = 1,2,2)
形式, K) linear form with the category weights a _i as coefficients

[0016]

(Equation 1)

Thus, the structure of the external reference is best represented on the real axis, in other words, the fuzzy groups B ₁ ,..., B _M of the external reference are best separated on the real axis. ,
That is, the category weights a _i are determined. this is,
Indicates the degree of separation of the fuzzy group B _r (r = 1,..., M)
Find a _i of linear form (1) that maximizes fuzzy dispersion ratio η ²
It is to stop. Since the method of finding is known, detailed
The explanation is omitted, but in summary, "Fuzzy quantification type II
Analysis "is based on qualitative factors (explanatory variables)
Explain the external criteria (fuzzy groups) given qualitatively
Therefore, qualitative data represented by values in the range [0, 1]
From the data, the linear category characterizing the fuzzy group
Weights a _i and the category weights
Each sample (each number in FIG. 9)
value of the set of explanatory variables and external criteria corresponding to ω)
(Sample score) and calculate these sample scores
By analyzing the relationship of external criteria to
This is a method for finding features and the like.

When the category weights a _i are determined in this way, the sample weights are sampled in a linear form using them as coefficients.
A is required. Then, by plotting the membership values of the external criteria (fuzzy group) for each sample score , the data analysis results by fuzzy quantification type II
Graph showing a characteristic distribution as can be obtained. Specific examples
Will be described later.

[0019]

According to the present invention, the fuzzy quantification type II technique described above is employed in the intangible sensing algorithm, and the detection signal from the detection unit is analyzed. Therefore , to quantify the amount of sensory ambiguity
Data is collected and the fuzzy quantity
Category weights are calculated and obtained using
Category weight and the detection signal from the detection unit
Is used to calculate a sample score. Then, by plotting the data against these <br/> sample score characteristic distribution representing the relationship between the output for the sample scores
Is obtained, the function approximating the characteristic distribution is a function to quantify sensory ambiguity amount. An algorithm for quantifying the amount of sensory ambiguity is configured based on this function. This function performs an output corresponding to the input from the detector.
Used to calculate

As described above, by introducing an algorithm for quantifying the amount of sensory ambiguity obtained by the fuzzy quantification type II, the memory can store the calculation formula and the function of the sample score by the fuzzy quantification type II. The algorithm can be adjusted.

[0021]

FIG. 1 shows the structure of a fuzzy sensor device according to the present invention. This apparatus includes one or a plurality of signal detectors 11 to 1n each of which includes various detectors for generating a data signal representing an environmental state. Each signal detector 11,
, 1n are subjected to A / D conversion or the like by the corresponding input signal converters 21 to 2n, and are subjected to an algorithm.
It is taken into execution means 30. Algorithm execution means 3
0 includes a CPU31 and algorithms storage unit 32, C
The PU 31 functions as an operation unit that executes an operation described below in accordance with the algorithm stored in the algorithm storage unit 32.
Works .

The above algorithm quantifies the amount of sensory ambiguity obtained by the fuzzy quantification type II, and is configured as shown in FIG. That is, CP
In U31, when detection signals 1 to n from the signal detection units 11 to 1n are input via the signal conversion units 21 to 2n (step S1), a sample score corresponding to these inputs is calculated (step S2). ). Next, the amount of sensory ambiguity is calculated using the nonlinear function F1 for each sample score (step S3) and output as a numerical value. This output value is subjected to D / A conversion or the like by the output signal converter 41 and output from the signal output unit 42 (step S
4). This output signal is used for performing air conditioning control or the like according to the environment.

FIG. 3 shows a procedure for obtaining a nonlinear function required for the configuration of the algorithm of FIG. First, processing for normalizing input data is performed (step S11). The data is collected through sensitivity tests and questionnaires involving human beings. Next, the normalized data is analyzed by the fuzzy quantification class II (step S12), and is approximated by a higher-order nonlinear function obtained by the least square method or the like (step S1).
3).

The above algorithm can be adjusted according to the use environment and use conditions by the man-machine interface 43 connected to the fuzzy sensor device of FIG.

FIG. 4 shows "Hot" as a specific example of FIG.
1 shows a comfort sensor device for detecting a sensory ambiguity amount such as “cold”. In this device, the algorithm storage unit 32 in FIG. 1 stores a storage unit 33 storing a procedure for calculating a sample score in step S2 in FIG. 2 and a non-linear function used in step S3 in FIG. And a storage unit 34. Further, the man-machine interface 43
Is configured so that the origin of the sample score and the value of the scaling can be appropriately adjusted by an input operation of the operator in order to adjust the above algorithm according to the use environment and use conditions.

First, as a problem setting, comfort perceived by humans due to temperature and humidity is an ambiguous factor and cannot be directly measured basically. And the time and date reflecting the degree of human clothing and the time reflecting human activity are defined as measurable amounts for determining comfort, and these are the signal detection units 11 to 11 in FIG. 1n, clock 101, thermometer 102, hygrometer 10
3 to detect. Then, the degree of comfort is determined based on these measured amounts.

Next, in order to link the above-mentioned measurable amount with the degree of comfort, various feelings of the insured person are investigated by a questionnaire method. As for temperature and humidity, the height is normalized to the range [0, 1]. The date is normalized to the range of [0, 1] so that the value is closer to "1" as the summer is closer, and the time is [0] so that the value is closer to "1" as the time is closer to 2:00 pm. , 1].

FIG. 5 shows the data X thus normalized.
Examples of 1 (temperature), X2 (humidity), X3 (date), and X4 (time) are shown. In the example of FIG. 4, these data are analyzed by fuzzy quantification II. That is, the storage unit 33 of FIG.
According to the sample score calculation procedure stored in the formula (1), the category weight a
Determine _i . As a result, when the data is plotted based on the sample score S obtained by the following equation (2),
It can express the characteristics of "hot" and "cold".

S = a ₁ · X ₁ + a ₂ · X 2 + a ₃ · X 3 + a ₄ · X 4 (2) a ₁ = 0.661, a ₂ = 0.360, a ₃ = -0.306, a ₄ = -0.116 ) Sample score S of each data obtained by
Is as shown in FIG. FIG. 7 shows a plot of the data with the horizontal axis representing the sample score S and the vertical axis representing the degree of hotness Y1 (or the degree of coldness Y2).

Next, based on the data distribution of FIG. 7, a function F1 giving the degree of feeling "hot" Y1 and a function F2 giving the degree Y2 of feeling "cold" are obtained as shown in FIG. And these functions can be represented by the following equations.

F1 = (1 / π) · tan ⁻¹ ｛0.109 (S−0.3)｝ + 0.5 (3) F2 = 1−F1 = − (1 / π) · tan ⁻¹ ｛0.109 (S−0.3 )｝ + 0.5… (4) Therefore, the measurable temperature, humidity, date and time are set to [0,
1], the sample score S of each data obtained by the equation (1) is obtained, and by using the function F1 or F2, the degree of the feeling of "hot", which is the amount of sensory ambiguity , that is, the comfort level The degree can be obtained, and the closer the value of the function F1 is to 0.5, the closer to a comfortable situation.

According to the embodiment shown in FIG. 4, as described above, it is possible to realize an intangible sensor that quantifies the vague feeling of a human, and there is no limit to the number of inputs that can be measured. Further, since the sampling score S is irrelevant to scaling and the position of the origin, the man-machine interface 43 can use the scaling factor and the position of the origin as parameters for adjusting the deviation due to individual differences or the like. Furthermore, if a similar function is to be realized by fuzzy inference, it is necessary to create a membership function and fuzzy rules. Therefore, it takes time to create an algorithm, and the required memory and operation time increase. Then, only the formula (2) for calculating the sample score and the function F1 (F2 is obtained as 1-F1) are required.

Although the comfort sensor has been described as an embodiment, the present invention is not limited to this, and can be similarly applied to "sweetness" and other intangible sensing.

[0034]

As described above, according to the present invention, the fuzzy quantification type II technique is used to quantify sensory vague quantities.
Because of, it can be used compact algorithm that does not require a large-capacity memory, only a small amount of memory to store it. Also, the algorithm can be adjusted according to the situation.

[Brief description of the drawings]

FIG. 1 is a block diagram showing a configuration of a fuzzy sensor device according to the present invention.

FIG. 2 is a flowchart showing a configuration of an algorithm used in the present invention.

FIG. 3 is a flowchart showing a procedure for creating an algorithm.

FIG. 4 is a block diagram showing a comfort sensor as a specific example of the fuzzy sensor device of FIG. 1;

FIG. 5 is a diagram showing an example of normalized data in the comfort sensor of FIG. 4;

FIG. 6 is a diagram showing an example of a sample score in the comfort sensor shown in FIG. 4;

7 is a diagram showing a data distribution of a degree of feeling “hot” and “cold” in the comfort sensor of FIG. 4;

8 is a diagram showing a non-linear function that gives a degree of feeling “hot” or “cold” in the comfort sensor of FIG. 4;

FIG. 9 is a diagram showing data handled in fuzzy quantification type II.

[Explanation of symbols]

11 to 1n: signal detection unit, 21 to 2n: input signal conversion unit, 30: algorithm execution means , 31: CPU, 32
... algorithm storage, 33 ... sample score calculation procedure storage, 34 ... function storage, 41 ... output signal converter, 4
2 ... Signal output unit, 43 ... Man-machine interface, 1
01: clock, 102: thermometer, 103: hygrometer.

Claims (3)

(57) [Claims] 1. A detector for detecting a perceptual amount of ambiguity , and a category weight is calculated from data collected for quantifying the amount of ambiguity by a method of fuzzy quantification type II.
And the obtained category weight and the input from the detector
By calculating the sample score using the signal and
A characteristic distribution representing the relationship between the output and the sample score
Algorithm for obtaining a function that approximates the characteristic distribution
And characterized by a algorithm executing means for executing
Fuzzy sensor device for. Wherein said algorithm execution means, said Argo
A storage unit for storing a rhythm and a response to an input from the detection unit.
First, by executing the algorithm stored in the storage unit,
Quantify the sensory ambiguity with the given function
2. The fuzzy sensor device according to claim 1 , comprising an arithmetic unit for outputting . 3. The fuzzy sensor device according to claim 2, wherein the algorithm is adjusted by the function according to a use environment or a use condition.