The present invention relates to an information processing service technique. And, the present invention relates to a technique for achieving a detecting apparatus and a detecting method detecting an anomaly part of cyclic time-series data.
In a health care field, a medical field, a nursing care field and others, systems for data measurement targeting human have been increasingly used. Such a system provides a user with valuable information for the user by calculating an analysis result from resultant data and feeding back the result to the user. The data is cyclic time-series data often.
As one example of such a system, a system (finger-tapping measuring/analyzing system) simply evaluating a cognitive function and a movement function by measuring and analyzing a finger-tapping motion of the user is exemplified (see, for example, a Patent Document 1).
In this specification, the finger-tapping motion means a repeatedly-opening/closing motion of a thumb and a forefinger. By the measurement of the finger-tapping motion, the cyclic time-series data is provided. It is known that success of the finger-tapping motion depends on presence/absence of and a severity of brain dysfunction such as dementia and Parkinson's disease. It has been pointed out that the analysis result of the cyclic time-series data measured by the system has possibilities of early recognition of such a brain dysfunction of the user, evaluation for estimation of the severity and others.
Patent Document 1: Japanese Patent Application Laid-Open Publication No. 2013-109540
It has been pointed out that the analysis result of the cyclic time-series data measured by the system has possibilities of early recognition of such a brain dysfunction of the user, evaluation for estimation of the severity and others. (in the present invention, this is referred to as (A) overall data evaluation that means evaluation for entire data waveform).
However, in a case of a bad evaluation result resulted from the analysis for the cyclic time-series data of the finger-tapping motion, a reason why this evaluation result is made could not have been offered so far. In other words, for the user, it could not have been explained which part of the cyclic time-series data is anomalous to cause the bad evaluation result, and therefore, conviction has not been made.
Therefore, a technique of detecting the anomaly part of the cyclic time-series data has been needed (in the present invention, the technique is referred to as (B) partial-data anomaly evaluation that means anomaly evaluation for a part of the data waveform).
However, a result of the (B) partial-data anomaly evaluation does not often match a result of the (A) overall-data evaluation. In other words, an evaluation model of the (A) is possibly inconsistent with an evaluation model of the (B) since the evaluation model of the (A) is created based on the cyclic time-series data itself while the evaluation model of the (B) is created based on the partial data extracted from the cyclic time-series data. The inconsistency between the (A) and the (B) makes the user confused of which one is reliable, and therefore, this inconsistency should be solved.
The inconsistency between the (A) and the (B) may be caused by the following two points of view. The first point of view is inconsistency in an anomaly rate. For example, circumstances on which the anomaly is detected in the (A) while the anomaly part is not detected in the (B) or opposite circumstances on which the anomaly is not detected in the (A) while a lot of anomaly parts are detected in the (B) are conceivable. Such inconsistent circumstances should not be caused. The second point of view is inconsistency in a contribution level of a feature amount to the anomaly detection. For example, circumstances on which a feature amount contributing to the anomaly detection in the (A) is not important in the anomaly detection in the (B) or opposite circumstances on which a feature amount not contributing to the anomaly detection in the (A) is important in the anomaly detection in the (B) are conceivable. Under such circumstances, an algorithm of the anomaly detection appears to be inconsistent, and there is a risk of losing the reliability of the entire system.
Accordingly, a purpose of the present invention is to provide highly-reliable evaluations of the overall data and the partial data in order not to cause the inconsistency in the above-described two points of view.
The above and other objects and novel characteristics of the present invention will be apparent from the description of the present specification and the accompanying drawings.
As means for solving the above-described issues, techniques described in the claims are used.
As one example cited, in a detecting apparatus detecting anomaly by using the cyclic information indicating a biological body state, the detecting apparatus includes: a cyclic-information acquiring unit configured to acquire the cyclic information; a cyclic-information feature-amount calculating unit configured to calculate a feature amount of the cyclic information acquired by the cyclic-information acquiring unit; a cyclic-information anomaly detecting unit configured to detect anomaly of the cyclic information based on the feature amount calculated by the cyclic-information feature-amount calculating unit; an anomaly-rate creating unit configured to create a cyclic-information anomaly rate based on a detection result of the cyclic-information anomaly detecting unit; a partial-information creating unit configured to create partial information based on the cycle by using the cyclic information acquired by the cyclic-information acquiring unit; a partial-information feature-amount calculating unit configured to calculate a feature amount of the partial information created by the partial-information creating unit; a partial-information anomaly detecting unit configured to detect anomaly of the partial information created by the partial-information creating unit, based on the feature amount calculated by the partial-information feature-amount calculating unit and the anomaly rate created by the anomaly-rate creating unit; and an output unit configured to output information based on a detection result of the partial-information anomaly detecting unit and a detection result of the cyclic-information anomaly detecting unit.
By usage of the techniques of the present invention, highly-reliable evaluations of overall data and partial data can be provided.
In the present working examples, a technique of detecting the anomaly part of the cyclic time-series data is proposed. Hereinafter, examples of embodiments of the present invention will be explained with reference to the accompanying drawings. Note that the same components are denoted by the same reference symbols in principle throughout all the drawings for explaining the embodiments, and the repetitive explanation thereof will be omitted.
The embodiments will be explained with reference to the drawings. However, the present invention is not interpreted to be limited to the contents described in the following embodiments. It could be easily understood for those who are skilled in the art that the specific configuration of the present invention is changeable with the scope of the idea or concept of the present invention.
If there are a plurality of elements having the same or similar function, the elements are denoted with the same symbol but the different index for the explanation. However, if it is unnecessary to discriminate the plurality of elements, the index is omitted for the explanation in some cases.
Terms such as “first”, “second”, and “third” in the present specification are attached in order to discriminate the elements, and are not always limit the numbers, the orders or the contents. The numbers for the identification of the elements are used for each phrase, and the number used in one phrase does not always indicate the same structure in a different phrase. Alternatively, the element identified by one number is not prevented from also having a function of the element identified by a different number.
A position, a size, a range and others of each component illustrated in the drawings or others are not often illustrated as a practical position, size, range and others in order to easily understand the invention. Therefore, the present invention is not always limited in the position, the size, the range and others disclosed in the drawings or others.
With reference to
[Human-Data Measuring System]
The measuring system 2 is a system of measuring a hand finger motion by using a magnetic-sensor type motion sensor. A motion sensor is connected to the measuring apparatus 3. This motion sensor is worn on the hand finger of the user. The measuring apparatus 3 measures the hand finger motion by using the motion sensor, and provides measuring data including a time-series waveform signal. The terminal apparatus 4 displays various types of information including the partial-data anomaly detection result on a display screen, and receives an operational input from the user. In the first embodiment, the terminal apparatus 4 is a PC.
The cyclic time-series data anomaly-part detecting system 1 has a function of offering an anomaly-part detecting service as a service based on information processing. Functions of the cyclic time-series data anomaly-part detecting system 1 include a partial-data anomaly detecting function. The partial-data anomaly detecting function is a function of detecting an anomaly part of the cyclic time-series data measured by the measuring system 2.
To the cyclic time-series data anomaly-part detecting system 1, for example, the cyclic time-series data or others is input as input data from the measuring system 2. From the cyclic time-series data anomaly-part detecting system 1, for example, the partial-data anomaly detection result or others is output as output data to the measuring system 2. The partial-data anomaly detection result includes a partial-data anomaly level and a partial-data anomaly feature amount in addition to the partial-data anomaly detection result.
The human-data measuring system of the first embodiment is applicable to not only the facility such as the hospital and the nursing care facility, the examinee and others, but also a wide variety of general facilities and people. The measuring apparatus 3 and the terminal apparatus 4 may be configured as a combined measuring system. The measuring system 2 and the cyclic time-series data anomaly-part detecting system 1 may be configured as a combined apparatus. The measuring apparatus 3 and the cyclic time-series data anomaly-part detecting system 1 may be configured as a combined apparatus.
[Cyclic Time-Series Data Anomaly-Part Detecting System]
The control unit 101 entirely controls the cyclic time-series data anomaly-part detecting system 1, is made of a Central Processing Unit (CPU), a Read Only Memory (ROM), a Random Access Memory (RAM) and others, and achieves a data processing unit performing the partial-data anomaly detection and others, based on a software program processing. The data processing unit of the control unit 101 includes a user-information managing unit 11, a task processing unit 12, an overall-data evaluating unit 13, an overall-data/partial-data matching unit 14, a partial-data anomaly evaluating unit 15, a practice-menu determining unit 16 and a result output unit 17. The control unit 101 achieves a function of inputting the measurement data from the measuring apparatus 3, a function of processing and analyzing the measurement data, a function of outputting a control instruction to the measuring apparatus 3 or the terminal apparatus 4, a function of outputting the display data to the terminal apparatus 4, and others.
The user-information managing unit 11 performs a processing of registering and managing the user information that is input by the user to user information 41 of a DB 40, a processing of checking the user information 41 of the DB 40 when the user uses the service, and others. The user information 41 includes a user's individual attribution value, use history information, user setting information and others. The attribution value includes a sex, an age and others. The use history information is information for managing a user's use history in the service offered by the present system. The user setting information is setting information for the functions of the present service set by the user.
The task processing unit 12 is a unit of performing a processing of a task for analyzing and evaluating a motility function and others. The task is, in other words, a predetermined hand finger motion. The task processing unit 12 outputs the task to the screen of the terminal apparatus 4 in accordance with task data 42 of the DB 40. The task processing unit 12 acquires the measurement data (the cyclic information indicating the biological body state) of the task measured by the measuring apparatus 3, and stores the data as the overall data 43A into the DB 40. The overall data described here means the overall cyclic time-series data measured during predetermined time. As described above, the task processing unit 12 (cyclic-information acquiring unit) acquires the cyclic information indicating the biological body state.
The overall-data evaluating unit 13 includes an overall-data feature-amount calculating unit 13A (that is a cyclic-information feature-amount calculating unit) and an overall-data evaluating unit 13B (that is a cyclic-information anomaly detecting unit). The overall-data feature-amount calculating unit 13A calculates a feature amount representing characteristics of overall data 44A (cyclic time-series data) in accordance with the overall data 44A of the user, and stores the feature amount as an overall-data feature amount 44B into the DB 40. The overall-data evaluating unit 13B evaluates the overall data in accordance with the overall-data feature amount 44B with reference to an overall-data DB 43, and stores an evaluation result as an overall-data evaluation result 44C into the DB 40. The overall-data evaluation result 44C is made of an overall-data anomaly level 44Ca and an overall-data feature-amount contribution level 44Cb.
The overall-data/partial-data matching unit 14 is made of an anomaly-rate determining unit 14A and a feature-amount importance-level determining unit 14B. The anomaly-rate determining unit 14A creates an anomaly rate 45A (cyclic-information anomaly rate) by using the overall-data anomaly level 44Ca, and stores the anomaly-rate data into the DB 40. The feature-amount importance-level determining unit 14B creates a feature-amount importance level 45B (feature-amount importance level) by using the overall-data feature-amount contribution level 44Cb with reference to a feature-amount correspondence table 50B, and stores the importance-level data into the DB 40. Combination of the anomaly rate 45A and the feature-amount importance level 45B is overall-data/partial-data matching information 45.
The partial-data anomaly evaluating unit 15 includes a partial-data creating unit 15A (that is a partial-information creating unit), a partial-data feature-amount calculating unit 15B (that is a partial-information feature-amount calculating unit), and a partial-data anomaly detecting unit 15C (that is a partial-information anomaly detecting unit). The partial-data creating unit 15A creates partial data 46A by dividing the overall data 44A, and stores the partial data into the DB 40. The partial-data feature-amount calculating unit 15B calculates a feature amount of each of the partial data 46A, and stores the feature amount data as a partial-data feature amount 46B into the DB 40. The partial-data anomaly detecting unit 15C determines the anomaly of the partial data in accordance with the partial-data feature amount 46B with reference to the partial data acquired from the overall-data DB 43 by using the anomaly rate 45A and the feature-amount importance level 45B, and stores the anomaly data as a partial-data anomaly detection result 46C into the DB 40. The partial-data anomaly detection result 46C includes a partial-data anomaly level 46Ca, a partial-data anomaly presence/absence 46Cb, and a partial-data anomaly feature amount 46Cc.
In this manner, the partial-data anomaly detecting unit 14C creates an anomaly level of the partial information created by the partial-data creating unit 15A, information indicating whether the partial information created by the partial-data creating unit 15A is anomalous or not, and information indicating an anomaly feature amount that is a feature amount to be a cause of the detection indicating that the partial information created by the partial-data creating unit 15A is anomalous. In this case, the cyclic time-series data anomaly-part detecting system 1 creates the detailed information of the anomaly of the partial information, and therefore, can further provide the detailed information by using the information allowing the part having the anomaly to be specified.
The practice-menu determining unit 16 determines a practice menu 47 based on a practice-menu list 50D and a practice-menu correspondence table 50E by using the partial-data anomaly feature amount 46Cc, and stores the practice menu data into the DB 40. In this manner, the practice-menu determining unit 16 determines the practice menu for improving the anomaly feature amount calculated by the partial-data anomaly detecting unit 14C.
The result output unit 17 performs a processing of outputting the overall-data evaluation result 44C, the partial-data anomaly detection result 46C and the practice menu 47 onto the screen of the terminal apparatus 4. The overall-data evaluating unit 13 and the partial-data anomaly evaluating unit 15 perform a screen output processing in coordination with the practice-menu determining unit 16 and the result output unit 17. In this manner, the result output unit 17 further outputs the menu determined by the practice-menu determining unit 16. In this case, the cyclic time-series data anomaly-part detecting system 1 provides the practice menu regarding the anomaly part of the partial data, and therefore, can provide the useful information for solving this anomaly part.
The result output unit 17 outputs the overall-data evaluation result 44C, and also outputs a result as a whole, and therefore, can provide information having a plurality of view points based on the cyclic time-series data.
The data and the information stored in the DB 40 of the storage unit 102 includes the user information 41, the task data 42, the overall-data DB 43, the overall data 44A, the overall-data feature amount 44B, the overall-data evaluation result 44C, the overall-data/partial-data matching information 45, the partial data 46A, the partial-data feature amount 46B, the partial-data anomaly detection result 46C, the practice menu 47 and others. By the control unit 101, a management table 50 is stored and managed in the storage unit 102.
The administrator can set contents of the management table 50. The management table 50 includes an overall-data feature-amount list 50A for setting the feature amount of the overall data, a feature-amount correspondence table 50B for setting correspondence between the feature amount of the overall data and the feature amount of the partial data, a partial-data feature-amount list 50C for setting the feature amount of the partial data, a practice-menu list 50D for setting a candidate of the practice menu, a practice-menu correspondence table 50E for setting correspondence between the partial-data anomaly feature amount 46Cc and the practice menu and others.
[Measuring Apparatus]
Note that a mode of storing each measurement data in storage means of the measuring apparatus 3 may be applicable, or a mode of storing each measurement data in not the measuring apparatus 3 but only the cyclic time-series data anomaly-part detecting system 1 may be applicable.
[Terminal Apparatus]
[Hand Finger, Motion Sensor, Finger-Tapping Measurement]
As shown in
As the motion sensor 20, a different sensor from the magnetic sensor is applicable if the sensor can measure the distance between the two fingers. For example, the waveform of the distance between the two fingers can be provided by repeat opening and closing of the two fingers touching a tablet terminal or a touch panel type PC. Alternatively, the waveform of the distance between the two fingers can be provided by an infrared sensor detecting a shape of the hand or a position of the finger tip.
The finger-tapping motion specifically includes the following various tasks. As the motions, for example, one-hand free running motion, one-hand metronome motion, both-hand simultaneous free running motion, both-hand alternate free running motion, both-hand simultaneous metronome motion, both-hand alternate metronome motion, and others are exemplified. The one-hand free running motion means that two fingers of one hand perform the finger tapping many times as quick as possible. The one-hand metronome motion means that two fingers of one hand perform the finger tapping in synchronization with stimulation at a constant pace. The both-hand simultaneous free running motion means that two fingers of a left hand and two fingers of a right hand perform the finger tapping at the same timing. The both-hand alternate free running motion means that two fingers of a left hand and two fingers of a right hand perform the finger tapping at an alternate timing. Another motion is finger tapping following a marker.
[Motion-Sensor control Unit and Finger-Tapping Measurement]
The alternating-current generator circuit 312a creates an alternating-current voltage signal having a predetermined frequency. The current-generating amplifier circuit 312b converts the alternating-current voltage signal into an alternating current having a predetermined frequency, and outputs the alternating current to the transmitter coil unit 21. In the transmitter coil unit 21, the magnetic field is generated by the alternating current. This magnetic field generates the induced electromotive force in the receiver coil unit 22. The receiver coil unit 22 outputs an alternating current generated by the induced electromotive force. This alternating current has the same frequency as the predetermined frequency of the alternating-current voltage signal generated by the alternating-current generator circuit 312a.
The preamplifier circuit 312c amplifies the detected alternating current. The wave-detector circuit 312d performs the wave detection of the amplified signal in accordance with a reference signal 312i out of the phase adjuster circuit 312f. The phase adjuster circuit 312f adjusts a phase of the alternating-current voltage signal having the predetermined frequency or a double frequency out of the alternating-current generator circuit 312a, and outputs the signal as the reference signal 312i. The LPF circuit 312e limits a bandwidth of the wave-detected signal and outputs a resultant signal, and the amplifier circuit 312g amplifies this signal to have a predetermined voltage. Then, The output signal terminal 312h outputs an output signal corresponding to the measured waveform signal.
The waveform signal that is the output signal is a signal having a voltage value representing the distance D between the two fingers. The distance D and the voltage value are exchangeable in accordance with a predetermined calculus equation. This calculus equation can be also provided by calibration. The calibration is measured while, for example, the user holds a block having a predetermined length with two fingers of a target hand. The predetermined calculus equation is provided as an approximate curve that minimizes the error from a dataset of the voltage value and the distance value in the measurement value. Alternatively, by the calibration, a size of the user's hand may be recognized and used for normalization of the feature amount or others. In the first embodiment, the magnetic sensor is used as the motion sensor 20, and the measuring means handling the magnetic sensor is used. The present invention is not limited to this, and different detecting means and measuring means such as an acceleration sensor, a strain gauge and a high-speed camera are also applicable.
[Process Flow]
(Step S1)
First, the user operates the measuring system 2. Specifically, in the terminal apparatus 4, the initial screen is displayed on the display screen. On the initial screen, the user selects a desirable operational item. For example, an operational item for detecting/processing the anomaly data is selected. The terminal apparatus 4 transmits the instruction input information corresponding to this selection to the cyclic time-series data anomaly-part detecting system 1. Alternatively, the user can enter and register the user information such as the sex, the age or others. In this case, the termina apparatus 4 transmits the entered user information to the cyclic time-series data anomaly-part detecting system 1. The user-information managing unit 11 of the cyclic time-series data anomaly-part detecting system 1 registers the user information into the user information 41.
(Step S2)
The task processing unit 12 of the cyclic time-series data anomaly-part detecting system 1 transmits the task data for the user to the terminal apparatus 4 in accordance with the instruction input information of the step S1 and the task data 42 of the finger tapping. The task data includes one or more types of the task information regarding the hand finger motion such as the one-hand free running motion, the both-hand simultaneous free running motion and the both-hand alternate free running motion. In the terminal apparatus 4, the task information of the hand finger motion is displayed on the display screen in accordance with the received task data. The user performs the task of the hand finger motion in accordance with the task information on the display screen. The measuring apparatus 3 measures the task, and transmits the measurement data to the cyclic time-series data anomaly-part detecting system 1. The cyclic time-series data anomaly-part detecting system 1 stores the measurement data into the measurement data 42B.
(Step S3)
The overall-data feature-amount calculating unit 13A of the cyclic time-series data anomaly-part detecting system 1 calculates the overall-data feature amount 44B by using the overall data 44A in accordance with the overall-data feature-amount list 50A. Then, the overall-data evaluating unit 13B provides the overall-data evaluation result 44C by applying a statistical method such as multivariate analysis, machine learning or others to the overall-data feature amount 44B. The overall-data evaluation result 44C includes the overall-data anomaly level 44Ca and the overall-data feature-amount contribution level 44Cb.
(Step S4)
The result output unit 17 of the cyclic time-series data anomaly-part detecting system 1 transmits the overall-data evaluation result 44C to the terminal apparatus 4, and this result is displayed on the screen. In this manner, the result output unit 17 outputs the information based on the detection result made by the overall-data evaluating unit 13B. On the screen, the user can check the evaluation result of the user's cyclic time-series data.
(Step S5)
The anomaly-rate determining unit 14A of the cyclic time-series data anomaly-part detecting system 1 calculates the anomaly rate 45A in accordance with the overall-data anomaly level 44Ca.
(Step S6)
The anomaly-rate determining unit 14A of the cyclic time-series data anomaly-part detecting system 1 calculates the feature-amount importance level 45B in accordance with the overall-data feature-amount contribution level 44Cb with reference to the feature-amount correspondence table 50B.
(Step S7)
The partial-data creating unit 15A of the cyclic time-series data anomaly-part detecting system 1 creates the partial data 46A by using the overall data 44A. The partial-data creating unit 15A creates the partial information based on the cycle (such as information of one cycle) by using the overall data 44A. Then, the partial-data feature-amount calculating unit 15B calculates the partial-data feature amount 46B by using the partial data 46A in accordance with the partial-data feature-amount list 50C. Then, the partial-data anomaly detecting unit 15C provides the partial-data anomaly detection result 46C by applying the statistical method such as the multivariate analysis or the machine learning to the partial-data feature amount 46B while using the anomaly rate 45A and the feature-amount importance level 45B.
(Step S8)
The result output unit 17 of the cyclic time-series data anomaly-part detecting system 1 transmits the partial-data anomaly detection result 46C to the terminal apparatus 4, and the result data is displayed on the screen. On the screen, the user can check the anomaly part of the user's cyclic time-series data.
(Step S9)
The practice-menu determining unit 16 of the cyclic time-series data anomaly-part detecting system 1 creates the practice menu 47 in accordance with the partial-data anomaly feature amount 46Cc with reference to the practice-menu list 50D and the practice-menu correspondence table 50E.
(Step S10)
The result output unit 17 of the cyclic time-series data anomaly-part detecting system 1 transmits the practice menu 47 to the terminal apparatus 4, and the menu data is displayed on the screen. On the screen, the user can check the practice menu to be performed by the user.
[Calculation of Overall-Data Feature Amount]
The detailed example of the feature amount will be further described below. In the first embodiment, a plurality of feature amounts resulted from the waveforms of the distance, the speed and the acceleration are used. In another embodiments, note that only some feature amounts of the plurality of feature amounts may be used, or different feature amounts may be used, and the details of the definition of the feature amounts is not limited, either.
For example, the feature amount [Distance] includes a plurality of feature-amount parameters identified with identification numbers (A1) to (A11). A term in parentheses [ ] of the feature-amount parameter indicates a unit. (A1) “Maximum Amplitude of Distance” [mm] is a difference between the maximum value and the minimum value of the amplitude of the waveform of the distance (
(A3) “Average of Local Maximum Values of Distance” [mm] is an average of the local maximum values of the amplitudes of the respective cycles. (A4) “Standard Deviation of Local Maximum Values of Distance” [mm] is a standard deviation of the above-described values. (A5) “Slope (Damping Rate) of Approximate Curve of Local Maximum Points of Distance” [mm/second] is a slope of a curve of approximated local maximum points of the amplitudes. This parameter mainly represents amplitude change due to fatigue during the measurement time. (A6) “Variance Coefficient of Local Maximum Values of Distance” is a variance coefficient of the local maximum values of the amplitudes, and its unit ([−]) represents dimensionless quantity. This parameter is a value that is the normalized standard deviation by average, and therefore, personal difference in a finger length can be excluded. (A7) “Standard Deviation of Localized Local Maximal Values of Distance” [mm] is a standard deviation of local maximal values of three adjacent amplitudes.
This parameter is a parameter for evaluating a degree of amplitude variation in local short time. (A8) “Average of Local Minimum Values of Distance” [mm] is an average of the local minimum values of the amplitudes of the respective cycles. (A9) “Standard Deviation of Local Minimum Values of Distance” [mm] is a standard deviation of the above-described values. (A10) “Variance Coefficient of Local Minimum Values of Distance” is a variance coefficient of the local minimum values of the amplitudes, and its unit ([−]) represents dimensionless quantity. This parameter is a value that is the normalized standard deviation by average, and therefore, personal difference in a finger length can be excluded. (A11) “Standard Deviation of Localized Local Minimum Values of Distance” [mm] is a standard deviation of local minimum values of three adjacent amplitudes. This parameter is a parameter for evaluating a degree of localized short-time amplitude variation.
The feature amount [Speed] includes feature-amount parameters identified by the following identification numbers (A12) to (A26). (A12) “Maximum Amplitude of Speed” [m/second] is a difference between the maximum value and the minimum value of the speed in the waveform of the speed (
(A16) “Average of Local Minimum Values of Closing Speed” [m/second] is an average of local minimum values of the speed at the time of the closing motion. (A17) “Energy Balance” [−] is a ratio between a sum of squares of the speeds during the opening motion and a sum of squares of the speeds during the closing motion. (A18) “Total Energy” [m2/second2] is a sum of squares of the speed during the entire measurement time. (A19) “Variance Coefficient of Local Maximum Values of Opening Speed” [−] is a variance coefficient of the local maximum values of the speed at the time of the opening motion, and is a value that is normalized standard deviation by average. (A20) “Average of Local Minimum Values of Closing Speed” [m/second] is an average of local minimum values of the speeds at the time of the closing motion. (A21) “Number of Times of Shaking” [−] is a number provided by subtraction of the number of times of the finger tapping with the large opening/closing motion from the number of times of reciprocation changing in a sign of the waveform value of the speed between positive and negative. (A22) “Average of Distance Rate at Opening Speed Peak” [−] is an average value of rates of distances each having the maximum speed value during the opening motion provided when the amplitude of the finger tapping is set to 1.0. (A23) “Average of Distance Rate at Closing Speed Peak” [−] is an average value of the similar rates of distances each having the minimum speed value during the closing motion. (A24) “Ratio between Distance Rates at Speed Peak” [−] is a ratio between the value of (A22) and the value of (A23). (A25) “Standard Deviation of Distance Rate at Opening Speed Peak” [−] is a standard deviation of the rates of the distances each having the maximum speed value during the opening motion provided when the amplitude of the finger tapping is set to 1.0. (A26) “Standard Deviation of Distance Rate at Closing Speed Peak” [−] is a standard deviation of the similar rates of the distances each having the minimum speed value during the closing motion.
The feature amount [Acceleration] includes feature-amount parameters identified by the following identification numbers (A27) to (A36). (A27) “Maximum Amplitude of Acceleration” [m/second2] is a difference between the maximum value and the minimum value of the acceleration in the waveform of the acceleration (
(A30) “Average of Local Maximum Values of Closing Acceleration” [m/second2] is an average of the local maximum values of the acceleration during the closing motion, and is a third value of the four extremums. (A31) “Average of Local Minimum Values of Closing Acceleration” [m/second2] is an average of the local minimum values of the acceleration during the closing motion, and is a fourth value of the four extremums. (A32) “Average of Contact Time” [second] is an average of contact time in the closing state of the two fingers. (A33) “Standard Deviation of Contact Time” [second] is a standard deviation of the contact time. (A34) “Variance Coefficient of Contact Time” [second] is a variance coefficient of the contact time. (A35) “Number of Times of Zero Crossing of Acceleration” [−] is an average number of times of sign change of the acceleration value between positive and negative during one cycle of the finger tapping. This value is ideally two. (A36) “Number of Times of Cringe” [−] is a value provided by subtraction of the number of times of the finger tapping with the large opening/closing from the number of times of the reciprocation changing in the sign of the acceleration value between positive and negative during one cycle of the finger tapping.
Next,
(A43) “Skewness of Tap-Interval Distribution” [−] is a skewness of a frequent distribution of the tap interval, and represents a skewness level of the frequent distribution in comparison to normal distribution. (A44) “Standard Deviation of Local Tap Intervals” [second] is a standard deviation of three adjacent tap intervals. (A45) “Slope (Damping Rate) of Approximate Curve of Tap Intervals” [−] is a slope of a curve of an approximated tap interval. This slope mainly represents tap-interval change due to fatigue during the measurement time.
The feature amount [Phase Difference] includes a plurality of feature-amount parameters identified by the following identification numbers (A46) to (A49). (A46) “Average of Phase Difference” [degree] is an average of phase differences in the waveform of the both hands. The phase difference is an indicator for mismatch of the left-hand finger tapping with the right-hand finger tapping as shown with an angle when one cycle of the right-hand finger tapping is set to 360°. A state without the mismatch is set to 0°. (A47) “Standard Deviation of Phase Difference” [degree] is a standard deviation of the phase differences. The larger the value of (A46) or the value of (A47) is, the larger the mismatch between the both hands is, and the more the instability is. (A48) “Both-Hand Similarity” [−] is a value representing correlation provided when time mismatch is 0 in a case of application of a cross-correlation function to the right-hand and left-hand waveforms. (A49) “Time Mismatch at Maximum Both-Hand Similarity” [second] is a value representing time mismatch at the maximum correlation of (A48).
The feature amount [Marker Following] includes a plurality of feature-amount parameters identified by the following identification numbers (A50) and (A51). (A50) “Average of Delay Time from Marker” [second] is an average of delay time of the finger tapping from time at which a marker is cyclically shown. The marker is dealt with by stimuli such as visual stimulus, audio stimulus and tactual stimulus. This parameter value is based on a moment of the closing state of the two fingers. (A51) “Standard Deviation of Delay Time from Marker” [second] is a standard deviation of the delay time.
[Overall-Data Evaluation]
The overall-data evaluating unit 13B provides the overall-data evaluation result 44C representing good/bad of the overall data in accordance with the overall-data feature amount 44B calculated by the overall-data feature-amount calculating unit 13A. For example, with reference to the overall-data DB 43, an estimate equation for estimating the anomaly is provided by multiple regression analysis taking the anomaly as an objective variable and taking the plurality of feature amounts of the overall-data feature amount 44B as an explanatory variable. The anomaly is defined as an indicator that is small in the normal state while large in the anomalous state. As an example of the anomaly, a severity score of the brain dysfunction or others is exemplified, such as the Mini Mental State Examination (MMSE) representing a severity of the dementia and the Unified Parkinson's Disease Rating Scale (UPDRS) representing a severity of the Parkinson's disease. However, such a severity has characteristics in which the more the normal state is, the larger the value of the severity is, while the more the anomalous state is, the smaller the value is. For example, in the MMSE, a perfect score that is 30 points shows the highest cognitive function, and, the closer to zero the score is, the lower the cognitive function is. Therefore, the MMSE or the UPDRS is used for the anomaly after a preprocessing of inverting a positive sign and a negative sign of the MMSE or the UPDRS. Then, the overall-data feature amount 44B is assigned to the estimate equation of the multiple regression analysis, and an estimated severity score is provided as the overall-data anomaly level 44Ca.
The overall-data feature-amount contribution level 44Cb has a larger value when having larger influence on the estimation model of each feature amount, and has a smaller value when having a smaller influence of the same. For example, as the overall-data feature-amount contribution level 44Cb, an absolute value of a standardized partial regression coefficient of the estimate equation of the multiple regression analysis is designed.
In order to estimate the anomaly, not the multiple regression analysis but a similar method may be used. For example, a discriminant/regression method of simultaneously performing discrimination and regression based on a linear model may be used. Alternatively, a support vector machine regression method or a different regression method such as a neural network may be used.
The overall-data anomaly level 44Ca may be not the severity score of the brain dysfunction when being an indicator representing a mismatch level from the normal finger-tapping waveform. The overall-data feature-amount contribution level 44Cb may be not the standardized partial regression coefficient when being an indicator representing an importance level of each feature amount of the overall-data feature amount 44B in the estimate equation.
[Determination of Anomaly Rate]
The anomaly-rate determining unit 14A provides an anomaly rate 45A (R[%]) by assigning the overall-data anomaly level 44Ca (X) to a predetermined conversion function. The term “R” is designed to “0% R 100%”. The conversion function is a function that monotonically increases so that the larger the overall-data anomaly level 44Ca is, the larger the R is, and, for example, an exponent function of “R =a * exp(X−b) +c” is set. The term “a” is designed to be a large value when it is desirable to rapidly increase the R along with increase in the X, or a small value when it is desirable to monotonically increase the R along with the increase in the X. And, in this conversion function, the terms “b” and “c” are designed so as to satisfy “R=0%” when the anomaly level (after the preprocess) is the expectable minimum value (such as −30 in MMSE) or satisfy “R=Rm (0% Rm≤100%, such as Rm=50%)” when the anomaly level (after the preprocess) is the expectable maximum value (such as 0 in MMSE). In this manner, the partial-data anomaly detecting unit 15C does not detect the anomaly at all when the anomaly level is at the minimum, and detects the more anomalies when the anomaly level is larger.
Note that the overall-data evaluating unit 13B may estimate the overall-data anomaly level 44Ca out of the range of the expectable minimum to maximum values (−30 to 0 in the MMSE) of the anomaly level (after the preprocess). In this case, the overall-data anomaly level may be changed to the minimum value when being smaller than the minimum value or to the maximum value when being larger than the maximum value. Note that the conversion function may be not the exponent function if being the monotonically increase function, and may be, for example, a logarithm function, a sigmoid function, a linear function or others.
[Determination of Feature-Amount Importance Level]
With reference to the feature-amount correspondence table 50B shown in
Then, the feature-amount importance level Qk is provided by assigning the overall-data feature-amount contribution level Cj (j=1, 2, . . . and NA (the number of the overall-data feature amounts)) of the overall-data feature amount Aj to the predetermined conversion function. The conversion function is a function that monotonically increases so that the larger the overall-data feature-amount contribution level Cj is, the larger the feature-amount importance level Qk is, and is set to be, for example, an exponent function. This conversion function is designed to satisfy “Qk=1” when the Cj is the expectable minimum value or satisfy “Qk=100” that is a larger value when the Cj is the expectable maximum value. In this manner, the partial-data anomaly detecting unit 15C performs the anomaly detection without focusing the Pk when the overall-data feature-amount contribution level Cj is the minimum value, or performs the anomaly detection with more focusing the Pk when the overall-data feature-amount contribution level Cj is larger. Note that the conversion function may be not the exponent function if being the monotonically increase function, and may be, for example, a logarithm function, a sigmoid function, a linear function or others. When the above-described processes are performed to all the overall-data feature-amount contribution levels Cj, all the feature-amount importance levels Qk can be provided. Note that a plurality of Cj are corresponded to the same Qk in some cases. In this case, the Qk may be calculated after the largest Cj is selected. The present invention is not limited to this, the Qk may be calculated after the largest Cj is selected, or the Qk may be calculated from an average value of the plurality of Cj. Alternatively, when none of the Cj is corresponded to the Qk, “Qk=1” may be set as a default value.
[Creation of Partial Data]
The partial-data creating unit 15A extracts the finger-tap waveform for each cycle to provide the partial data 46A. As shown in
A later-described partial-data feature-amount calculating unit 15B also calculates the feature amounts (P19) and (P20) using the waveforms of the both hands. However, in order to calculate these feature amounts, the waveforms of the both hands in the same time zone are necessary. Therefore, one cyclic may be extracted in the right-hand waveform, and a waveform in the same time zone may be extracted from a left-hand waveform. This approach to the right hand and the left hand may be inverted.
[Calculation of Partial-Data Feature Amount]
For example, the feature amount [Distance] includes a plurality of feature-amount parameters identified with identification numbers (P1) to (P3). A term in parentheses [ ] of the feature-amount parameter indicates a unit. (P1) “Minimum Value of Distance” [mm] is the minimum value of the amplitude of the partial data. (P2) “Maximum Value of Distance” [mm] is the maximum value of the amplitude of the partial data. (P3) “Total Motion Distance” [mm] is a sum of absolute values of distance change amounts during entire measurement time of the partial data.
The feature amount [Speed] includes feature-amount parameters identified with identification numbers (P4) to (P8). (P4) “Maximum Value of Opening Speed” [m/second] is the maximum value of the speed in the opening motion of the partial data. The opening motion is a motion from the closing state of the two fingers to the maximum opening state. (P5) “Minimum Value of Closing Speed” [m/second] is the minimum value of the speed in the closing motion. The closing motion is a motion from the maximum opening state of the two fingers to the closing state. (P6) “Energy Balance” [−] is a ratio between a sum of squares of the speed in the opening motion and a sum of squares of the speed in the closing motion. (P7) “Total Energy” [m2/second2] is a sum of squares of the speed during the entire measurement time of the partial data. (P8) “Number of Times of Shaking” [−] is a count provided by subtraction of 1 that is the number of times of the finger tapping from the number of times of reciprocation changing in a sign of the waveform value of the speed between positive and negative. (P9) “Distance Rate at Opening Speed Peak” [−] is a distance at the maximum speed value in the opening motion when the amplitude of the finger tapping is set to 1.
(P10) “Distance Rate at Closing Speed Peak” [−] is a distance at the minimum speed value in the closing motion when the amplitude of the finger tapping is set to 1. (P11) “Ratio between Distance Rates at Speed Peak” [−] is a ratio between the value of (P10) and the value of (P11).
The feature amount [Acceleration] includes feature-amount parameters identified with the following identification numbers (P12) to (P17). (P12) “Maximum Value of Opening Acceleration” [m/second2] is the maximum value of the acceleration in the opening motion, and is the first value of four extremums. (P13) “Minimum Value of Opening Acceleration” [m/second2] is the minimum value of the acceleration in the opening motion, and is the second value of the four extremums. (P14) “Maximum Value of Closing Acceleration” [m/second2] is the maximum value of the acceleration in the closing motion, and is the third value of the four extremums. (P15) “Minimum Value of Closing Acceleration” [m/second2] is the minimum value of the acceleration in the closing motion, and is the fourth value of the four extremums. (P16) “Contact Time” [second] is contact time of the two fingers in the closing state. (P17) “Number of Times of Cringe” [−] is a value provided by subtraction of 1 that is the number of times of the finger tapping with the large opening/closing from the number of times of the reciprocation changing in the sign of the acceleration value between positive and negative during one cycle of the finger tapping.
The feature amount [Tap Interval] includes a feature-amount parameter identified with the following identification number (P18). (P18) “Tap Interval” [second] is time for one cyclic of the finger tapping.
The feature amount [Phase Difference] includes feature-amount parameters identified with the following identification numbers (P19) to (P20). (P19) “Phase Difference” [degree] is a phase difference of the waveforms of the both hands. When one cyclic of the right-hand finger tapping is set to 360 degrees, the phase difference is an indicator representing a difference of the left-hand finger tapping from the right-hand one as an angle. In a case without the difference, the angle is set to 0 degree. (P20) “Both-Hand Similarity” [−] is a value representing correlation in a case without time difference of 0 under application of a cross-correlation function to the waveforms of the right hand and the left hand.
The feature amount [Marker Following] includes a feature-amount parameter identified with the following identification number (P21). This feature amount is calculated with reference to a motion task following a marker. (P21) “Delay Time from Marker” [second] is delay time of the finger tapping from time at which the marker is cyclically shown. The marker is dealt with by stimuli such as visual stimulus, audio stimulus and tactual stimulus. This parameter value is based on a moment of the closing state of the two fingers.
[Partial-Data Anomaly Detection]
The partial-data anomaly detecting unit 15C detects the anomaly of the partial data 46A by using the multivariate analysis or the machine learning. As a preprocess, first, the partial-data feature amount 46B is standardized so that the average is 0 while the standard deviation is 1 as generally performed. By such standardization, weights of the feature amounts in the model resulted from the multivariate analysis or the machine learning can be prevented from being not uniform due to the difference in the range of each feature amount. Next, a feature-amount spatial distribution of the partial-data feature amount 46B is changed with reference to the feature-amount importance level (Qk) 45B that is calculated by the feature-amount importance-level determining unit 14B. As one example of a method of changing the feature-amount spatial distribution, the standardized partial-data feature amount Pk is multiplied by the feature-amount importance level Qk (k=1, 2 . . . and NP (the number of the partial-data feature amounts)). By this process, the distribution of the partial-data feature amount having the high importance level is made large in the feature-amount space, and its anomaly due to the machine learning can be easily detected. Alternatively, as another example of the method of changing the feature-amount spatial distribution, the partial-data feature amount 46B (Ak) may be assigned to an exponent function “Ak'=p*Qk*exp(Ak)” (the term “p” is a predetermined value) to change the Ak into the Ak′. In this manner, the farther from the average 0 the partial-data feature amount 46B (Ak) is, the farther the far data is. The larger the feature-amount importance level Qk is, the rapidly farther the far data is, and therefore, the data can be easily detected as the anomaly.
Then, 1-class Support Vector Machine (SVM) that is one type of the machine learning is used to perform the anomaly detection. The SVM to be a precondition of this process is a method in 2-class classification of defining a classification boundary so as to maximize a margin between the classification boundary (a hyperplane expressed by a linear function) and data of each class. However, when the classification boundary being the hyper plane cannot be separated if the classification boundary of two groups has a complicate shape, and therefore, the SVM is devised to be capable of handling the complicate-shape classification boundary by application of a kernel function. An idea of a 1-class SVM is the same as that of the 2-class classification of the SVM, but is a method of classifying the data into the anomaly data at a certain rate and other normal data in one class. A rate of outliers of the 1-class SVM is designed to be the anomaly rate 45A (R) calculated by the anomaly-rate determining unit 14A. In this manner, the higher the overall-data anomaly level 44Ca is, the larger the rate of the partial data to be detected as the anomaly is.
Note that a different method from the 1-class SVM may be applied to the detection of the anomaly of the partial data. For example, a normal distribution centering the average of the feature-amount distribution of the partial data 46A may be assumed, and data having a large distance from the center of the normal distribution may be detected as the anomaly.
[Partial-Data Anomaly Detection Result]
In the 1-class SVM, a classification score “y” is calculated, and data having a classification score y of a negative value is determined as the anomaly. A result detected by this determination is designed to be the partial-data anomaly presence/absence 46Cb. It is conceivable that the farther from 0 and the smaller the classification score y is, the larger the anomaly level is. Accordingly, this classification score y is transformed by a function of causing “z=0%” in “y=0” and “z=100%” in “y =−∞” to be asymptotic, and the term “z” is designed to be the partial-data anomaly level 46Ca. And, in order to find out which feature amount of all feature amounts contributes to the anomaly determination of the finger-tapping waveform for each cycle determined as the anomaly by the 1-class SVM, a feature amount that is far from the average by the standard deviation “SD=2.0 or more” is designed to be the partial-data anomaly feature amount 46Cc.
[Effect of Partial-Data Anomaly Evaluating Unit]
An example of the partial-data anomaly detection result 46C is shown in
Each of
As described above, by the anomaly-rate determining unit 14A and the feature-amount importance-level determining unit 14B, the anomaly detection can be performed on the partial data (the finger-tapping waveform for each cycle) while the matching with the anomaly detection result of the overall data is maintained.
[Determination of Practice Menu]
[Display Screen (1)—Menu]
As an example of a display screen of the terminal apparatus 4,
In the user information section 1501, the user information is input by the user and is registerable. If the user information is already input to an electronic health record or others, the section may be in corporation with this user information. As examples of the enterable user information, a user ID, a name, a birth date or an age, a sex, a dominant hand, disease/symptom, a note and others are exemplified. The dominant hand is selectable from and enterable as a right hand, a left hand, both hands, uncertain and others. The disease/symptom may be selectable from and enterable as, for example, an option in a list box, or may be enterable as an optional text. When this system is used in a hospital or others, not the user but a doctor or others may enter them instead of the user. The present anomaly-data processing system is also applicable to a case without the registration of the user information.
In the operational menu section 1502, operational items for functions offered as the service are displayed. The operational items include “Calibration”, “Measurement of Hand Finger Motion”, “Anomaly-Data Detection/Process”, “End” and others. When the “Calibration” is selected, the above-described calibration, in other words, the process for the adjustment of the motion sensor 20 or others to the user's hand finger is performed. A status indicating whether the adjustment is done or not is also displayed. When the “Measurement of Hand Finger Motion” is selected, the screen changes to a task measurement screen for measuring the task of the hand finger motion such as the finger tapping. When the “Anomaly-Data Detection/Process” is selected, the anomaly detection is performed to the measured data to be targeted, the anomaly-data detection result is displayed, and the screen changes to a screen for performing the process to the detected anomaly data. When the “End” is selected, the service ends.
In the setting section 1503, the user setting is enabled. For example, when there is a type of the anomaly detection item to be detected by the user, the measuring operator or the administrator, the anomaly detection item can be selected from the options and can be set. And, a process for each anomaly detection item can be selected. A threshold of the anomaly data detection or others can be also set. These setting contents are transmitted to the cyclic time-series data anomaly-part detecting system 1 through the communicating unit 105, and the cyclic time-series data anomaly-part detecting system 1 detects/processes the anomaly data with reference to the specified setting in this stage.
[Display Screen (2)—Task Measurement]
As another example,
[Display Screen (3)—Overall-Data Evaluation Result]
As still another example,
[Display Screen (4)—Anomaly-Part Detection Result]
As still another example,
The screen display method of the partial-data evaluation result shown in
[Display Screen (5)—Simultaneous Display of Overall-Data Evaluation Result and Anomaly-Part Detection Result]
It is more preferable to simultaneously display the screen of the overall-data evaluation result shown in
As the overall-data evaluation result in the simultaneous display of the content of the overall-data evaluation result and the content of the anomaly-part detection result, only the score may be displayed, only the radar chart may be displayed, both of the score and the radar chart may be displayed, or a different display method may be used. Similarly, as the anomaly-part detection result, the display is not limited to the graph display shown in
In the cyclic time-series data anomaly-part detecting system 1 of the first embodiment, the overall-data evaluating unit 13 detects the anomaly of the overall data 44A based on the overall-data feature amount 44B, and besides, creates the overall-data anomaly level 44Ca (the cyclic-information anomaly rate). The anomaly-part detecting system 1 creates the partial data 46A by dividing the overall data 44A that is the cyclic time-series data, calculates the partial-data feature amount 46B, and displays and outputs the partial-data anomaly detection result 46C that is the detection result of the anomaly of the partial data 46A, based on the partial-data feature amount 46B and the overall-data anomaly level 44Ca.
As described above, the anomaly-part detecting system 1 can avoid the results of the evaluation based on the overall data and the evaluation based on the partial data from being different from each other because of detecting the anomaly by using the overall-data anomaly level 44Ca for each piece of the partial data 46A resulted from the division of the overall data 44A.
In the cyclic time-series data anomaly-part detecting system 1 of the first embodiment, the overall-data evaluating unit 13 detects the anomaly of the overall data 44A based on the overall-data feature amount 44B, and besides, creates the feature-amount importance level 45B. The anomaly-part detecting system 1 creates the partial data 46A by dividing the overall data 44A that is the cyclic time-series data, calculates the partial-data feature amount 46B, and displays and outputs the partial-data anomaly detection result 46C that is the detection result of the anomaly of the partial data 46A, based on the partial-data feature amount 46B and the feature-amount importance level 45B.
As described above, the anomaly-part detecting system 1 can avoid the results of the evaluation based on the overall data and the evaluation based on the partial data from being different from each other because of detecting the anomaly by using the feature-amount importance level 45B for each piece of the partial data 46A resulted from the division of the overall data 44A.
The cyclic time-series data anomaly-part detecting system 1 of the first embodiment can detect the anomaly part of the overall data 44A and show it to the user by creating the partial data 46A by dividing the overall data 44A that is the cyclic time-series data, calculating the partial-data feature amount 46B, and providing the partial-data anomaly detection result 46C. In this case, the partial-data anomaly detection result 46C and the anomaly detection result of the overall data can be matched with each other by the application of the anomaly-rate determining unit 14A and the feature-amount importance-level determining unit 14B. When the overall-data evaluation result 44C is bad, the user can specifically recognize which part has the problem, from the partial-data anomaly detection result 46C. Further, since the practice menu 47 provided by the practice-menu determining unit 16 is shown, the user can recognize the practice method for improving the problem.
In the present embodiment, note that the anomaly-part detection targeting the time-series data of the finger-tapping motion has been explained. However, different data may be acceptable if the data is cyclic time-series data. For example, time-series data resulted from measurement of electrocardiographic signals, magnetocardiographic signals, pulse waves, breathing, brainwave, ambulation, eye blink, mastication and others are exemplified.
With reference to
[System]
As the facilities, various facilities such as a hospital, a health checking center, a public facility, an entertainment facility, and user's house are applicable. The system 7 is placed in the facility. As examples of the system 7 in the facility, a system 7A in a hospital H1, a system 7B in a hospital H2, and others are exemplified. Each of the system 7A and the system 7B in the respective hospitals includes the terminal apparatus 4 and the measuring apparatus 3 configuring the measuring system 2 that is the same as that of the first embodiment. A configuration of each system 7 may be the same or different. The system 7 in the facility may include an electronic-health-record managing system in the hospital or others. A measuring apparatus of the system 7 may be a dedicated device.
The server 6 is an apparatus managed by the service provider. The server 6 has a function of providing, to the facility and the user, the same partial-data anomaly detection service as that of the cyclic time-series data anomaly-part detecting system 1 of the first embodiment, as a service based on information process. The server 6 provides a service process in a client server system to the measuring system. The server 6 has a user managing function or others in addition to such a function. The user managing function is a managing function of registering and accumulating the user information of the user group, the measurement data, the analyzed evaluation data and others in the DB, the data and the information being provided from the systems in the plurality of facilities.
[Server]
The control unit 601 controls the entire server 6, is made of a CPU, a ROM, a RAM and others, and achieves the data processing unit 600 detecting the anomaly data, determining the anomaly-data process or others, based on a software program process. The data processing unit 600 includes a user-information managing unit 11, a task processing unit 12, an overall-data evaluating unit 13, an overall-data/partial-data matching unit 14, a partial-data anomaly evaluating unit 15, a practice-menu determining unit 16 and a result output unit 17.
The user-information managing unit 11 registers and manages the user information related to the user group of the systems 7 in the plurality of facilities, as the user information 41 into the DB 640. The user information 41 includes an attribution value of each user, use history information, user setting information and others. The use history information includes actual-history information in which each user has used the anomaly-part detection service in past.
[Server Management Information]
The history information item is information for managing the actual history of the user's usage of the anomaly-part detection service, and the information of date and time of each usage or others is stored in time series. As the history information item, data such as each data in the practice in this usage, that is the measurement data, the analyzed evaluation data, the anomaly-data detection result, the anomaly-data process content and others are stored. As the history information item, information of an address at which each data is stored may be stored.
[Effects and Others]
As similar to the first embodiment, the cyclic time-series data anomaly-part detecting system of the second embodiment can detect the anomaly part of the overall data 44A and show it to the user by creating the partial data 46A by dividing the overall data 44A that is the cyclic time-series data, calculating the partial-data feature amount 46B, and providing the partial-data anomaly detection result 46C. In this case, the partial-data anomaly detection result 46C and the anomaly detection result of the overall data can be matched with each other by the application of the anomaly-rate determining unit 14A and the feature-amount importance-level determining unit 14B. When the overall-data evaluation result 44C is bad, the user can specifically recognize which part has the problem, from the partial-data anomaly detection result 46C. Further, since the practice menu 47 provided by the practice-menu determining unit 16 is shown, the user can recognize the practice method for improving the problem.
In the foregoing, the present invention has been concretely described on the basis of the embodiments. However, the present invention is not limited to the foregoing embodiments, and various modifications can be made within the scope of the present invention.
The present invention is not limited to the embodiments, and includes various modification examples. For example, a part of the structure of one embodiment can be replaced with the structure of another embodiment, and besides, the structure of another embodiment can be added to the structure of one embodiment. Further, another structure can be added to/eliminated from/replaced with a part of the structure of each embodiment.
1 . . . cyclic time-series data anomaly-part detecting system, . . . measuring system, 3 . . . measuring apparatus, 4 . . . terminal apparatus
Number | Date | Country | Kind |
---|---|---|---|
2019-134881 | Jul 2019 | JP | national |
Filing Document | Filing Date | Country | Kind |
---|---|---|---|
PCT/JP2020/018855 | 5/11/2020 | WO |