JP2001283134A - Method for deciding and predicting state - Google Patents

Abstract

PROBLEM TO BE SOLVED: To objectively and exactly analyze a state based on chaos attracter, and to judge and predict the state even from data having any irregular cycle. SOLUTION: A chaos attracter is prepared from the time-sequential data of a state, and the orbit cycle of the chaos attracter is measured, and the present state and a future state are predicted from the fluctuation of the orbit cycle.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a method for determining and predicting a state, and more particularly to a method for determining a current state from nonlinear time-series data such as biological information and predicting a future state. It is.

[0002]

2. Description of the Related Art Measuring and analyzing biological information such as acceleration, vibration, pulse, blood pressure, body temperature, blood glucose level, respiration, myoelectricity, electrocardiogram, blood flow, and brain waves to judge and predict a person's health condition. Has been attempted in the past. Since the biological information is nonlinear data, chaos analysis, which has been widely used as an analysis method for nonlinear data, has been used for analyzing biological information. Chaos analysis includes analysis methods such as fractal dimension analysis, Lyapunov exponent analysis, and chaos attractor analysis. Among them, chaos attractor analysis is generally used as a means for analyzing biological information.

For example, in Japanese Patent Application Laid-Open No. Hei 6-217951, an attractor is generated from a measured pulse wave, a plurality of attractor patterns stored in a memory are compared with the attractor, and a similarity between the shape and structure of the attractor is determined. An invention for judging the physical condition of a person has been proposed. However, it is difficult to objectively determine the similarity of the shape of the attractors, and the means of the present invention cannot accurately determine the physical condition of a person.

[0004] In Japanese Patent Publication No. 6-9546, a chaos attractor is created from time-series data of pulse waves and / or heart radio waves, and the chaos attractor is further processed to obtain a Lyapunov exponent to diagnose mental and physical abnormalities. Have been proposed. Certainly, the proposed invention is effective when it is based on data having a fairly regular cycle such as a pulse wave or a heart wave, but a satisfactory diagnostic result is obtained when it is based on data having an irregular cycle. There is a problem that it cannot be obtained.

[0005]

SUMMARY OF THE INVENTION The present invention has been made in view of such a conventional problem, and it is possible to objectively and accurately analyze a state from a chaos attractor and to convert data having an irregular period. It is an object of the present invention to provide a method capable of judging and predicting a state based on the information.

[0006]

According to the present invention, a chaotic attractor is created from time-series data of states, the orbital period of the chaotic attractor is measured, and the current state is determined from the fluctuation of the orbital period. In addition, a state determination / prediction method characterized by predicting a future state is provided.

Here, when the chaos attractor is created in a number space of three or more dimensions, a specific plane is set in the number space, and the time from passing through this specific plane until passing next is defined as the above-mentioned time. The orbital period of the chaos attractor may be used, or when the chaotic attractor is created in a two-dimensional number space, the angular period obtained by performing polar coordinate transformation with a desired point in the number space as the origin is the orbit of the chaos attractor. It may be a cycle. As the origin of the polar coordinate conversion, it is preferable to use the average coordinates of the components of the chaos attractor from the viewpoint of simplicity.

[0008] Further, since a large fluctuation can be objectively and accurately analyzed, the fluctuation of the orbital period of the chaos attractor is analyzed by a detrend fluctuation analysis (DFA: Defa).
It is desirable to carry out by trended Fluctuation Analysis).

If the time series data relating to biological information is used, it is possible to judge and predict the health condition of a person. In terms of easiness of measurement, such biological information is preferably acceleration data or vibration data during walking, and among acceleration data, data relating to the traveling direction is more preferable.

[0010]

BEST MODE FOR CARRYING OUT THE INVENTION The present inventors have developed a method that can objectively and accurately analyze a state from a chaos attractor, and can determine and predict a state based on data having an irregular period. As a result of diligent studies, the inventors focused on the orbital period of the chaos attractor, and found that objective and accurate analysis of the state can be performed from the fluctuation of the orbital period, leading to the present invention.

That is, a major feature of the determination / prediction method of the present invention lies in that the current state is determined from fluctuations in the attractor orbital period and the future state is predicted. By using this method, it is possible to objectively and accurately extract differences between chaotic attractors that cannot be extracted by the conventional chaotic attractor analysis using the pattern matching method. Hereinafter, the determination / prediction method of the present invention will be described in detail based on specific examples.

FIG. 1 is a block diagram showing an embodiment of the judgment / prediction method of the present invention. The time-series data detected by the detecting means 1 is sent to the data processing / analyzing means 2.
The data processing / analyzing means 2 creates a chaos attractor from the obtained time-series data, measures the orbital period of the chaotic attractor, and detects its fluctuation. When storing the detected fluctuation as base data for associating a specific fluctuation with a specific state, the detected fluctuation data is sent to the database unit 3. On the other hand, when the state is determined / predicted from the detected fluctuation, the detected fluctuation data is sent from the data processing / analysis means 2 to the state determination / prediction unit 4, where various data stored in the database unit 3 are stored. A comparison is made with the fluctuation base data, the state is determined from the matched or approximated fluctuation base data, and the determination / prediction result is output to the outside by the output means 5.

For example, if the data to be collected is acceleration data or vibration data when a person walks, an acceleration sensor or a vibration sensor having a transmission unit is attached to the person and the time series data of the state is collected. What is necessary is just to collect by a signal etc.
Further, the detection means, the analysis means, and the output means may be integrally carried around the body.

For the measurement of the orbital period of the chaotic attractor, for example, when the chaotic attractor is drawn in a three-dimensional or more number space, a plane is set in the number space, and after the attractor orbit passes through the plane, the next time. What is necessary is just to measure the time until it passes. FIG. 2 shows an embodiment in which a chaos attractor is drawn in a three-dimensional number space. One attractor trajectory coming out of the surface of the plane A travels in the front direction of the paper surface, makes one rotation halfway, travels in the reverse direction of the paper surface, and passes through the plane A from the back surface to the front surface of the plane A. The time from exiting plane A to returning to plane A is defined as the orbital period. Of course, the orbital period may be a part of the attractor trajectory. In this case, by adjusting the set position of the plane A, it is possible to measure the trajectory cycle of a desired portion of the attractor trajectory.

When a chaotic attractor is drawn in a two-dimensional number space, for example, an angular period may be obtained by polar coordinate conversion using a desired point in the number space as an origin, and this angular period may be used as an orbital period. According to such a method, the orbital period can be measured by relatively simple processing. FIG. 3 shows an embodiment in which a chaos attractor is drawn in a two-dimensional number space.
Draw a horizontal line S rightward from point O of the chaos attractor,
The time required for the attractor trajectory to rotate 360 ° in the opposite direction to the clock and return to the reference line with the horizontal line S as the reference line is defined as the trajectory cycle. At this time, it is recommended to use an average coordinate point of the components of the chaos attractor as the origin of the polar coordinate conversion because it can be easily calculated. In addition,
Depending on the shape of the attractor trajectory, the calculated average coordinate point may exist in the attractor trajectory, but in such a case, a point off the attractor trajectory near the average coordinate point may be set as the origin. .

FIG. 4 shows the actually measured acceleration data of walking by the subject. FIG. 4 is a diagram showing a temporal change in acceleration due to walking with the horizontal axis representing time and the vertical axis representing acceleration. FIG. 4 (a) shows acceleration data in a healthy state, and FIG. This is acceleration data. As can be understood from these figures, the difference between the acceleration data waveforms is not clearly apparent.

Next, a chaotic attractor drawn in a two-dimensional number space based on this data is shown in FIG. FIG. 5A shows a chaotic attractor in a healthy state, and FIG. 5B shows a chaotic attractor in a poor physical condition. When the states are classified using the pattern matching method from the shapes of these chaos attractors, it is highly likely that both are determined to be in the same state, that is, there is no difference. Have difficulty.

Therefore, the average value of the components of the chaos attractor shown in FIG. 5 was obtained, and the orbital period of the chaos attractor was measured by polar-transforming the chaos attractor with the average value coordinate point as the origin. Specifically, a horizontal line S is drawn rightward from the point O of the chaos attractor, and the time required for the attractor trajectory to rotate 360 ° in the opposite direction to the clock and return to the reference line using the horizontal line S as a reference line is defined as a trajectory cycle. And Here, in the analysis of walking information by the conventional DFA, since one walking cycle from the left or right foot landing to the next same foot landing is exclusively measured and used, only one data is obtained per one walking. However, the attractor trajectory makes one round of a large circle in one walk from the landing of the right foot to the next landing of the right foot, and then makes three rounds of a small circle. Thus, four orbit periods are obtained per one walk. As a result, if the number of measurement is the same,
An additional effect is obtained that the measurement time can be reduced to 1/4 of the conventional case.

FIG. 6 shows the measurement results of the attractor orbital period. FIG. 6A shows the attractor trajectory cycle in a healthy state, and FIG. 6B shows the attractor trajectory cycle in a poor physical condition. FIG.
As can be seen by comparing (a) and (b), differences in walking data due to differences in physical condition, which could not be determined by the pattern matching method of the chaos attractor, can be found from fluctuations in the period of the attractor trajectory. Therefore, if the fluctuation of the attractor trajectory period in various states is stored as a database, the fluctuation characteristics of the database and the measured fluctuation characteristics are sequentially compared to determine the state at that time by detecting the coincidence. You can do it.

Further, a sign of poor physical condition can be found from the fluctuation of the attractor trajectory cycle. FIG. 7 shows the orbital period of the chaos attractor on the day before the subject became ill. The fluctuation of the attractor trajectory cycle in FIG. 7 is large as compared to a healthy state, but smaller than a state of poor physical condition. Therefore, if the fluctuation of the attractor trajectory cycle, which has maintained stability, starts to increase, it can be predicted that even if the individual does not have any subjective symptoms, his physical condition is currently deteriorating, and that he will eventually become ill.

As a method of analyzing the fluctuation of the attractor trajectory, conventionally known analysis methods such as frequency conversion such as DFA and Fourier transform, wavelet analysis, and multifractal analysis can be used. DFA is desirable because it can analyze accurately.

An outline of the DFA analysis method is as follows. First, a fluctuation system is divided by a predetermined window size, and a waveform is linearly approximated for each window. Then, the absolute value of the deviation from the linear approximation is integrated, and this integrated value is used as the magnitude of the fluctuation. The magnitude of the fluctuation is plotted on the vertical axis, the window size is plotted on the vertical axis, and the magnitude of the fluctuation with respect to each window size is plotted, and the slope and y-intercept are used as indices for state determination / prediction.

FIG. 8 shows the result of analyzing the fluctuation of the attractor orbital period shown in FIGS. 6A and 6B by DFA.
In FIG. 8, the horizontal axis represents the window size and the vertical axis represents the magnitude of the fluctuation, and the healthy state and the poor physical condition are represented by a solid line and a broken line, respectively. The slope of the broken line when the physical condition is poor is smaller than that of the solid line when healthy. Therefore, in this case, it is possible to determine the health and physical condition mainly from the change in the inclination of the line.

Using the vibration data during walking as the time series data of the state, the fluctuation of the attractor orbit cycle is calculated by DF.
The result analyzed in A is shown in FIG. The vibration during walking was measured using a vibration switch (manufactured by Koshin). Since the output of the vibration switch is a digital signal of "1" and "0", the signal is added and averaged, smoothing is performed, and a chaos attractor is drawn. Like FIG. 8, FIG. 9 shows the healthy state and the poor physical condition by a solid line and a broken line, respectively, with the horizontal axis representing the window size and the vertical axis representing the fluctuation magnitude.
Since the y-intercept and the slope are different between the solid line at the time of health and the broken line at the time of poor physical condition, it is possible to determine the health and poor physical condition from these differences.

In the above-described embodiment, the method of judging and predicting the health condition is described by using the acceleration and vibration during walking as the time series data of the condition. However, the method of judging and predicting the condition of the present invention is not limited to this. Instead, it is possible to determine / predict the states of these machines / appliances from time-series data such as vibration, sound, heat generation, etc. of motors and engines. Furthermore, the weather condition can be determined and predicted from time-series data such as temperature, humidity, and wind direction.

[0026]

According to the state determination / prediction method of the present invention, a chaos attractor is created from time series data of a state, the orbital period of the chaotic attractor is measured, and the fluctuation of the orbital period is analyzed. Since it is possible to judge and predict the difference, it is possible to objectively extract differences that could not be found by the conventional pattern matching method, and to objectively and accurately analyze the state, and furthermore, based on data having irregular periods. Can also determine and predict the state.

[Brief description of the drawings]

FIG. 1 is a block diagram illustrating an embodiment of a state determination / prediction method according to the present invention.

FIG. 2 is a diagram showing an embodiment when a chaos attractor is drawn in a three-dimensional number space.

FIG. 3 is a diagram showing an embodiment when a chaotic attractor is drawn in a two-dimensional number space.

FIG. 4 is a diagram showing a change in acceleration during walking.

FIG. 5 is a chaotic attractor diagram drawn in a two-dimensional number space from an acceleration change during walking.

FIG. 6 is a diagram showing a change in an attractor orbit cycle.

FIG. 7 is a diagram showing a change in an attractor orbit cycle.

FIG. 8 is a diagram showing the result of analyzing the fluctuation of the attractor orbital period by DFA.

FIG. 9 is a diagram showing the result of analyzing the fluctuation of the attractor orbital period by DFA.

[Explanation of symbols]

 DESCRIPTION OF SYMBOLS 1 Detection means 2 Data processing / analysis means 3 Database part 4 State judgment and prediction part 5 Output part

Continued on the front page (72) Inventor Takeshi Ogawa 22-22 Nagaikecho, Abeno-ku, Osaka City, Osaka F-term (reference) 4C038 VA12 VA20 VB01 VC20 5B049 BB41 DD00 EE00 FF01 GG09

Claims (8)

[Claims] 1. Creating a chaotic attractor from time-series data of a state, measuring the orbital period of the chaotic attractor, determining the current state from the fluctuation of the orbital period, and predicting a future state. Judgment / prediction method of characteristic state. 2. The chaos attractor is created in a number space of three dimensions or more, a specific plane is set in the number space, and the time from passing through the specific plane to passing the chaos attractor is defined by the chaos attractor. The method for judging and predicting a state according to claim 1, wherein the orbital period is set. 3. The state according to claim 1, wherein the chaotic attractor is created in a two-dimensional number space, and an angular period obtained by performing a polar coordinate transformation with a desired point in the number space as an origin is defined as an orbital period of the chaos attractor. Judgment and prediction method. 4. The state judging / predicting method according to claim 3, wherein an average coordinate point of a component of the chaos attractor is used as an origin of the polar coordinate conversion. 5. The state determination / prediction method according to claim 1, wherein the fluctuation of the orbital period of the chaos attractor is analyzed by detrend fluctuation analysis. 6. The state determination / state according to claim 1, wherein the time-series data relates to biological information.
Forecasting method. 7. The method according to claim 6, wherein the biological information is at least one of acceleration data and vibration data during walking. 8. The method according to claim 7, wherein the acceleration data relates to a traveling direction.