WO2014156083A1

WO2014156083A1 - Vibration classification system, vibration determination device, vibration determination condition setting device, vibration classification method, and readable medium

Info

Publication number: WO2014156083A1
Application number: PCT/JP2014/001604
Authority: WO
Inventors: 宗一朗高田; 茂葛西; 三上　伸弘
Original assignee: 日本電気株式会社
Priority date: 2013-03-29
Filing date: 2014-03-20
Publication date: 2014-10-02
Also published as: JPWO2014156083A1

Abstract

A vibration classification system (1) is equipped with a determination device (10) and a determination condition setting device (50). The determination device (10) detects a first vibration and transmits first vibration waveform data representing the first vibration to the determination condition setting device (50). The determination condition setting device (50) uses the first vibration waveform data as learning data to calculate a determination condition for determining the class of the vibration, and transmits the determination condition to the determination device (10). The determination device (10) detects a second vibration, determines the class of the second vibration on the basis of the determination condition, and transmits the determination result for the second vibration to the outside.

Description

Vibration classification system, vibration determination device, vibration determination condition setting device, vibration classification method, and readable medium

The present invention relates to a vibration classification system, a vibration determination device, a vibration determination condition setting device, a vibration classification method, and a program, and in particular, a vibration classification system, a vibration determination device, a vibration determination condition setting device suitable for detection of an intruder by an intruder, The present invention relates to a vibration classification method and a program.

Patent document 1 is a glass breakage detector for crime prevention that is attached to a glass plate that partitions the outside of the room, such as a window glass and a glass door, and detects when a breaker breaks the glass plate and notifies by sound or light. Disclosure. The glass breakage detector disclosed in Patent Document 1 includes a vibration sensor unit that converts vibration of a glass plate into a voltage signal, an amplification unit that amplifies the voltage signal, and an amplitude of a predetermined frequency component extracted from the amplified voltage signal. And a vibration analysis unit that outputs a warning signal to a CPU (Central Processing Unit) when glass breakage is detected by comparing with a threshold value. When the alarm signal is input, the CPU notifies the glass breakage alarm at the output unit.

Patent Document 2 discloses a crime prevention system capable of detecting locking / unlocking of an entrance door. The security system includes a detection device and a determination device. The detection device is attached to the inner key knob portion (thumbturn) of the entrance door. The detection device incorporates an acceleration sensor that detects acceleration in the directions of three axes (X axis, Y axis, and Z axis). The acceleration sensor detects the acceleration of the inner key. When the motion of the thumb turn is detected by the acceleration sensor, the detection device notifies the determination device of the thumb turn acceleration by wireless communication.

The determination device is installed in the same room or in the same building as the entrance door to which the detection device is mounted, receives a signal transmitted from the detection device, determines whether there has been an unauthorized intrusion, and the user or security company Notice. The determination device determines the state of the thumbturn to which the detection device is attached based on the acceleration data received from the detection device. The determination device calculates the tilt angle of the thumb turn from the acceleration of the thumb turn on which the detection device is mounted, and determines whether the key is locked or unlocked based on the tilt angle. .

Patent Document 3 discloses a crime prevention device that detects unauthorized opening by an intruder based on vibration generated in a door body. The security device is provided on the indoor side surface of the door body provided at the entrance of the house. The security device includes a triaxial vibration detection sensor, a security control unit, an indicator lamp, a buzzer device, and an auxiliary lock device. The three-axis vibration detection sensor detects a vibration component in the three-dimensional direction of the vibration generated in the door body, and outputs the detection signal to the crime prevention control unit. When the detection signal from the triaxial vibration detection sensor is input, the security control unit determines whether or not the vibration generated in the door body is an unauthorized opening vibration compared with a vibration pattern registered in advance. . If it is determined that the detection signal is illegal, an abnormality detection signal is output, the indicator lamp is turned on, the buzzer device is sounded, and the thumb turn of the door body is rotated from the locked posture to the unlocked posture in the auxiliary lock device. To restrict what to do.

It should be noted that the vibration pattern generated in the door body when a normal opening / closing operation is performed on the crime prevention control unit, the vibration pattern generated in the door body when wind pressure is applied, and the vibration generated in the door body when the vehicle travels in the vicinity. The vibration generated in the door body by the installation environment, such as a pattern, is registered in advance as a normal vibration pattern. The security control unit determines whether or not the detection signal (vibration pattern) from the triaxial vibration detection sensor is a normal vibration pattern registered in advance, and if it is determined that the vibration is other than the normal vibration pattern Then, an abnormality detection signal is output as a vibration caused by unauthorized opening, that is, an abnormal vibration pattern.

The registration of the vibration pattern to the security control unit is performed in advance when the security device is shipped from the factory, but new registration or deletion can be freely performed by a contractor or user at the installation site. As a result, a normal vibration pattern suitable for the installation environment can be registered in the security control unit, and malfunction of the security device can be reduced.

Patent Document 4 discloses a window opening / closing detection system capable of detecting the opening / closing and locking state of a window using an acceleration sensor. The window opening / closing detection system includes an opening / closing detector attached to the window, and a management device connected to the opening / closing detector so as to be capable of wireless communication. The open / close detector includes: an acceleration sensor that detects movement of the window; waveform detection means that detects an output waveform of the acceleration sensor; storage means that associates and stores the output waveform of the acceleration sensor and the open / closed state of the window; and the acceleration sensor Open / close determining means for comparing the output waveform and a pre-stored waveform to determine the open / closed state of the window, and a transmission unit for transmitting the open / close information of the window to the management device.

The acceleration sensor is a two-dimensional acceleration sensor and can detect the window acceleration in the opening / closing direction of the window and the direction orthogonal thereto. The storage means stores the locking operation when locking the crescent lock on the window and the output waveform of the acceleration sensor in association with each other. The open / close detector detects the state of the crescent lock from the output waveform of the acceleration sensor.

JP-A-2005-78500 JP 2009-93470 A JP 2009-102904 A JP 2011-169041 A

However, since neither of the techniques disclosed in

Patent Documents

1 and 2 sets the determination conditions according to the installation environment, there is a problem that erroneous determination is likely to occur due to the influence of disturbance vibration factors. Patent Document 3 discloses that the malfunction of the security device can be reduced by registering the normal vibration pattern at the installation site. However, the techniques disclosed in

Patent Documents

3 and 4 detect the detected vibration pattern. The detected vibration pattern (waveform) is determined based on the (waveform) and the vibration pattern (waveform) registered in advance. In order to execute such a determination method with high accuracy, a high processing capability is required for an apparatus that performs the determination. Since a device having a high processing capability consumes a large amount of power, there is a problem that it is not preferable to use a device having a high processing capability as a determination device that requires constant power supply.

The present invention has been made in order to solve such problems, and it is possible to set the determination condition according to the installation environment, and a simple configuration with low power consumption for a device that requires constant power supply. It is an object of the present invention to provide a vibration classification system, a vibration determination device, a vibration determination condition setting device, a vibration classification method, and a program.

The vibration classification system according to the first aspect of the present invention includes a determination device and a determination condition setting device. The determination device detects a first vibration and transmits first vibration waveform data representing the first vibration to the determination condition setting device. The determination condition setting device calculates a determination condition for determining a vibration classification using the first vibration waveform data as teacher data, and transmits the determination condition to the determination device. The determination device detects a second vibration, determines a classification of the second vibration based on the determination condition, and transmits a determination result regarding the second vibration to the outside.

A vibration determination device according to a second aspect of the present invention transmits vibration detection means for detecting a first vibration and first vibration waveform data representing the first vibration to a determination condition setting device to determine a vibration classification. Communication means for receiving a judgment condition for doing so from the judgment condition setting device, and an arithmetic means. The vibration detection means detects a second vibration. The computing means determines the classification of the second vibration based on the determination condition. The communication means transmits a determination result regarding the second vibration to the outside.

A vibration determination condition setting device according to a third aspect of the present invention determines a vibration classification by using communication means for receiving first vibration waveform data from the vibration determination device, and using the first vibration waveform data as teacher data. Calculating means for calculating a determination condition for this. The communication means transmits the determination condition to the vibration determination device.

In the vibration classification method according to the fourth aspect of the present invention, the determination device detects the first vibration, transmits first vibration waveform data representing the first vibration to the determination condition setting device, and the determination condition setting device. Calculates a determination condition for determining a vibration classification using the first vibration waveform data as teacher data, transmits the determination condition to the determination device, and the determination device detects a second vibration. The classification of the second vibration is determined based on the determination condition, and the determination result for the second vibration is transmitted to the outside.

The vibration classification method according to the fifth aspect of the present invention detects a first vibration, transmits first vibration waveform data representing the first vibration to a determination condition setting device, detects a second vibration, and performs the determination. The classification of the second vibration is determined based on the determination condition received from the condition setting device, and the determination result for the second vibration is transmitted to the outside.

A program according to a sixth aspect of the present invention transmits first vibration waveform data representing a first vibration to a determination condition setting device, and classifies a second vibration based on the determination condition received from the determination condition setting device. Determine and transmit the determination result for the second vibration to the outside.

According to the present invention, the vibration classification system, the vibration determination device, and the vibration that can set the determination condition according to the installation environment and can make the device that needs constant power supply have a simple configuration with low power consumption. A determination condition setting device, a vibration classification method, and a program are provided.

1 is a schematic diagram illustrating a configuration of a vibration classification system according to a first exemplary embodiment. FIG. 3 is a time sequence diagram illustrating a vibration classification method according to the first exemplary embodiment. It is the schematic which shows the structure of the vibration classification system concerning Embodiment 2. FIG. It is a block diagram of the IIR type digital filter which the frequency band limitation part with which the vibration classification system concerning Embodiment 2 is provided is realized. FIG. 10 is a time sequence diagram illustrating a vibration classification method according to the second exemplary embodiment. It is the schematic which shows the state which installed the determination apparatus concerning Embodiment 2 in the generation | occurrence | production site of vibration, and connected the determination condition setting apparatus to the determination apparatus. 10 is a flowchart of a determination condition calculation step included in the vibration classification method according to the second exemplary embodiment. It is a conceptual diagram of unlocking vibration. It is a conceptual diagram of a locking vibration spectrum. It is a conceptual diagram of a difference spectrum. It is a conceptual diagram of linear discriminant analysis. The dividing number k is a conceptual view of a linear discriminant analysis when set to k _1. The dividing number k is a conceptual view of a linear discriminant analysis when set to k _2. It is a graph which shows the relationship between the evaluation function J and the division number k. 10 is a flowchart of a vibration determination process included in the vibration classification method according to the second exemplary embodiment. It is a difference spectrum in the case of measuring data. This is a feature space when the division number k is set to 10 for actually measured data. This is a feature space when the division number k is set to 18 for actually measured data. This is a feature space when the division number k is set to 30 for actually measured data. It is a graph which shows the relationship between the evaluation function J and the division | segmentation number k when measuring data is made into object. 10 is a table for comparing a determination result according to a comparative example and a determination result according to Embodiment 2.

Hereinafter, embodiments of the present invention will be described with reference to the drawings.

(Embodiment 1)
With reference to FIG. 1, the vibration classification system 1 according to the first exemplary embodiment includes a determination device 10 and a determination condition setting device 50. The determination device 10 includes a vibration detection unit 11, a communication unit 24, and a calculation unit 25. The vibration detection unit 11 is, for example, a vibration sensor. The communication unit 24 is a wired communication device that performs wired communication or a wireless communication device that performs wireless communication. The calculation unit 25 is a calculation device such as a CPU (Central Processing Unit), for example. The determination condition setting device 50 includes a communication unit 54 and a calculation unit 55. The communication unit 54 is a wired communication device that performs wired communication or a wireless communication device that performs wireless communication. The calculation unit 55 is a calculation device such as a CPU, for example.

判定 Determining device 10 detects vibration, determines vibration classification, and transmits the determination result to the outside. The determination condition setting device 50 calculates a determination condition for the determination device 10 to determine the vibration classification. The vibration classification system 1 is suitable for a crime prevention application that detects an intruding action by an intruder based on mechanical vibrations such as an entrance door or a window glass.

Referring to FIG. 2, the vibration classification method according to the first exemplary embodiment includes step 10 for initial setting of determination device 10 and step S20 for operating determination device 10. The initial setting step S10 includes steps S120, S130, S140, and S170. Operation step S20 includes steps S200, S210, and S220.

Each step included in the initial setting step S10 will be described. In step S120, the vibration detection unit 11 of the determination device 10 detects vibration. In step S 130, the communication unit 24 of the determination device 10 transmits vibration waveform data representing the vibration detected in step S 120 to the determination condition setting device 50. In step S 130, the communication unit 54 of the determination condition setting device 50 receives vibration waveform data from the determination device 10. In step S140, the calculation unit 55 of the determination condition setting device 50 calculates a determination condition for determining a vibration classification by using the vibration waveform data as teacher data. In step S 170, the communication unit 54 of the determination condition setting device 50 transmits the determination condition to the determination device 10. The communication unit 24 of the determination device 10 receives the determination condition from the determination condition setting device 50.

Each step included in the operation step S20 will be described. Each step included in the operation step S20 is executed by the determination apparatus 10. In the operation step S20, the determination condition setting device 50 is not necessary. In step S200, the vibration detection unit 11 detects vibration. In step S210, the calculation unit 25 determines the classification of vibration detected in step S200 based on the determination condition. In step S220, the communication unit 24 transmits the determination result regarding the vibration detected in step S200 to the outside.

According to the present embodiment, the determination condition setting device 50 calculates the determination condition based on the vibration detected by the determination device 10 in step S120, and the determination device 10 classifies the vibration detected in step S200 based on the determination condition. Determine. Therefore, the determination condition can be set according to the installation environment of the determination apparatus 10. Therefore, erroneous determination due to the influence of disturbance vibration factors is prevented.

Furthermore, according to the present embodiment, the determination device 10 and the determination condition setting device 50 are separate devices, and it is necessary to always install and always supply power at the generation site of the vibration to be determined. Only the determination device 10 is provided. Since it is not necessary to always supply power, the determination condition setting device 50 with a small power consumption constraint calculates the determination condition. Therefore, the determination condition can be calculated using a complicated algorithm. By using a complicated algorithm, the determination condition can be set so that high-precision determination can be performed without requiring high processing capability. Therefore, it is possible to make the determination device 10 that needs to constantly supply power have a simple configuration with low power consumption. As a result, it becomes easy to install the determination apparatus 10 in a general house.

In the determination device 10, the functions of the vibration detection unit 11, the communication unit 24, and the calculation unit 25 may be realized by a dedicated circuit or a dedicated device, but may be realized by a computer CPU executing a program. . In the determination condition setting device 50, the functions of the communication unit 54 and the calculation unit 55 may be realized by a dedicated circuit or a dedicated device, but may also be realized by a CPU of a computer executing a program.

(Embodiment 2)
Next, a second embodiment will be described. In the following description, descriptions of matters common to the first embodiment and matters obvious from the first embodiment may be omitted.

Referring to FIG. 3, the vibration classification system 1 according to the second exemplary embodiment includes a determination device 10 and a determination condition setting device 50. The determination condition setting device 50 includes a storage unit 53, a communication unit 54, and a calculation unit 55. The calculation unit 55 includes a frequency band limiting unit 56, a feature amount extraction unit 57, a discriminant analysis unit 58, and an optimization calculation unit 59. The frequency band limiting unit 56 performs frequency band limitation using a digital filter. The discriminant analysis unit 58 performs linear discriminant analysis. The optimization calculator 59 cooperates with the frequency band limiting unit 56, the feature amount extraction unit 57, and the discriminant analysis unit 58 to perform optimization calculation of determination conditions for the determination device 10 to determine the vibration classification. The optimization calculation unit 59 can perform matrix calculation and statistical analysis calculation. The storage unit 53 is a storage device such as a hard disk drive or a semiconductor memory.

The determination device 10 includes a vibration detection unit 11 and a main unit 20. The vibration detection unit 11 and the main unit 20 have separate casings. The main unit 20 includes an unnecessary response removal unit 21, an analog-digital (A / D) conversion unit 22, a storage unit 23, a communication unit 24, and a calculation unit 25. The calculation unit 25 includes a frequency band limiting unit 26, a feature amount extraction unit 27, and a discriminant analysis unit 28. The frequency band limiting unit 26 performs frequency band limitation using a digital filter. The discriminant analysis unit 28 performs linear discriminant analysis. The vibration detection unit 11 is, for example, a piezoelectric acceleration sensor with a built-in signal amplification circuit. The unnecessary response removing unit 21 is, for example, a bandpass filter including a resistor and a capacitor. The unnecessary response removal unit 21 is provided between the vibration detection unit 11 and the A / D conversion unit 22. The A / D converter 22 is, for example, a Σ-Δ type A / D converter. The storage unit 23 is a storage device such as a hard disk drive or a semiconductor memory.

For example, the analog-digital conversion bit number of the A / D conversion unit 22 is 12 bits, and the sampling frequency of the A / D conversion unit 22 is 5 kHz. Preferably, the number of analog-digital conversion bits of the A / D converter 22 is 12 bits or less, and the sampling frequency of the A / D converter 22 is 6 kHz or more.

The vibration classification system 1 is used, for example, for a crime prevention application that detects an intruder act by an intruder based on vibration caused by locking and unlocking the front door. The vibration classification system 1 may be referred to as a front door lock / unlock detection system. The determination device 10 may be referred to as a lock / unlock detection device. The determination condition setting device 50 may be referred to as an initial setting terminal.

An IIR (Infinite Impulse Response) type digital filter realized by the frequency band limiting unit 26 will be described. The digital filter generates output vibration waveform data y [n] by limiting the frequency band of the input vibration waveform data u [n]. The input / output relationship of discrete time data is expressed by the following equation.

FIG. 4 shows a block diagram corresponding to this input / output relationship. The delay element z ⁻¹ delays the time data by one step. The order of the digital filter represents the number of delay elements z ⁻¹ and represents the use of past discrete time data corresponding to this number. When a digital filter is expressed by z-converting the input / output relationship, it is expressed by the following equation.

Where the symbol _n a, _{n b} represents the order of the digital filter. The frequency band limiting unit 26 provides a digital filter of the number of dimensions of the feature amount used by the discrimination analysis unit 28 to determine the vibration classification. When the number of dimensions of the feature quantity is L, j takes 1,.

A monitoring frequency band can be cited as a design factor of the digital filter. The monitoring frequency band is a limited frequency band, that is, a frequency band of the output vibration waveform data y [n]. The monitoring frequency band may be referred to as a pass frequency band. Monitoring frequency bands, digital filter coefficient in the formula (2) and the digital filter order n _a, defined by n _b. The digital filter filter coefficient is shown below.

The digital filter orders n _a and _nb are preset by the manufacturer or user of the vibration classification system 1. The digital filter coefficient is calculated by the determination condition setting device 50 when the determination device 10 is initially set.

The configuration and operation of the frequency band limiting unit 56 are the same as the configuration and operation of the frequency band limiting unit 26. The frequency band limiting unit 56 and the frequency band limiting unit 26 share the same digital filter order.

Referring to FIG. 5, the vibration classification method according to the second embodiment includes step 10 for initial setting of determination device 10 and step S20 for operating determination device 10. The initial setting step S10 includes steps S100, S110, S120, S130, S140, S170, and S180. Operation step S20 includes steps S200, S210, and S220.

FIG. 6 shows a site where the vibration to be determined is generated. In step S 100, the user installs the determination device 10 at the site where the vibration to be determined is generated. Specifically, the vibration detection unit 11 is fixed to the outer frame 110 of the entrance door 100 with a double-sided tape or the like in order to detect vibration due to locking and unlocking of the lock 104 provided on the entrance door 100. The main unit 20 connected to the vibration detection unit 11 via the signal cable is fixed near the vibration detection unit 11. In step S 110, the user connects the determination condition setting device 50 to the determination device 10 (specifically, the main unit 20) using a USB (Universal Serial Bus) cable 80. The determination condition setting device 50 is preferably, for example, a notebook personal computer or a tablet terminal that has high processing capability and is suitable for carrying. Instead of performing wired communication between the determination device 10 and the determination condition setting device 50, wireless communication may be performed. When performing wireless communication, the determination condition setting device 50 is installed within a range where wireless communication with the determination device 10 is possible.

In step S120, the vibration detection unit 11 detects vibration due to locking and unlocking of the lock 104 provided in the entrance door 100. Specifically, the vibration detection unit 11 converts the vibration due to the lock of the lock 104 into an electric signal (hereinafter referred to as a lock electric signal), and the vibration due to the unlocking of the lock 104 as an electric signal (hereinafter referred to as an unlock electric signal). ). The unnecessary response removing unit 21 removes noise from the lock electric signal and the unlock electric signal by a band pass filter. The A / D converter 22 performs analog-to-digital conversion of the locked electrical signal after passing through the unnecessary response removing unit 21 (bandpass filter) into vibration waveform data (hereinafter also referred to as locked vibration waveform data). The A / D conversion unit 22 performs analog-digital conversion of the unlocked electrical signal after passing through the unnecessary response removing unit 21 (bandpass filter) into vibration waveform data (hereinafter also referred to as unlocked vibration waveform data). Locking vibration waveform data represents vibration due to locking. The unlocking vibration waveform data represents vibration due to unlocking.

In step S 130, the communication unit 24 of the determination device 10 transmits the lock vibration waveform data and the unlock vibration waveform data to the determination condition setting device 50. The receiving unit 54 of the determination condition setting device 50 receives the locking vibration waveform data and the unlocking vibration waveform data from the determination device 10.

Next, step S140 will be described. In step S140, the calculation unit 55 of the determination condition setting device 50 calculates a determination condition for determining the vibration classification using the lock vibration waveform data and the unlock vibration waveform data as teacher data.

Referring to FIG. 7, step S140 includes steps S141 to S160. The optimization calculation unit 59 executes steps S141 and S142 for i = 1 to M. Here, M represents the number of classes in the discriminant analysis. When a specific example is shown in the present embodiment, M = 2. Here, i represents a class, class i = 1 represents locking (event A), and class i = 2 represents unlocking (event B). Specifically, the optimization calculation unit 59 inputs the lock vibration waveform data (step S141), executes FFT (Fast Fourier Transform), and obtains spectrum data (hereinafter referred to as lock vibration spectrum data) from the lock vibration waveform data. Is calculated (S142). The optimization calculation unit 59 inputs the unlocking vibration waveform data (step S141), executes FFT, and calculates spectrum data (hereinafter referred to as unlocking vibration spectrum data) from the unlocking vibration waveform data (S142). In step S143, the optimization calculation unit 59 calculates difference spectrum data from the lock vibration spectrum data and the unlock vibration spectrum data.

8A, 8B, and 8C show the unlocking vibration spectrum represented by the unlocking vibration spectrum data, the locking vibration spectrum represented by the locking vibration spectrum data, and the difference spectrum represented by the difference spectrum data, respectively. In step S144, the optimization calculation unit 59 performs a peak search on the difference spectrum data, and detects peaks 71 and 72. In step S145, the optimization calculation unit 59 extracts the feature frequency candidates f ₀ , f ₁ , f ₂ ,... As the candidate of the center frequency of the feature amount extraction frequency band from which the feature amount is extracted from the difference spectrum data. The characteristic frequency candidate f ₀ is the center frequency of the frequency band 70 in which there is no difference between the unlocking vibration spectrum and the locking vibration spectrum. The characteristic frequency candidate f ₁ is the frequency at the apex of the peak 71. The feature frequency candidate f ₂ is the frequency at the apex of the peak 72. In step S146, the optimization calculation unit 59 uses the feature frequency F _j (j = 1 to L) as the center frequency of the feature quantity extraction frequency band from which the feature quantity is extracted as the feature frequency candidates f ₀ , f ₁ , f _2. Select from ... Specifically, the optimization calculation unit 59 selects L feature frequency candidates from f ₀ , f ₁ , f ₂ ,... And sets them as feature frequencies F ₁ to F _L. Here, L is the number of dimensions of the feature quantity used by the discriminant analysis unit 28 to determine the vibration classification. When a specific example is shown in the present embodiment, L = 2.

The computing unit 55 optimizes the bandwidth of the feature amount extraction frequency band by executing steps S147 to S154. The computing unit 55 executes Steps S147 to S149 for i = 1 to M, j = 1 to L, and k = 1 to N. Here, k is the number of divisions, and N is a natural number set by the manufacturer or user of the vibration classification system 1. The relationship between the detuning parameter μ _k for setting the bandwidth of the feature quantity extraction frequency band and the division number k is expressed by the following equation.

Here, B is an upper limit of the detuning parameter μ _k and is determined based on past experience. When B = 300 and N = 30, the detuning parameter μ _k has a value within the range of the following equation.

In step S147, the optimization calculation unit 59 calculates digital filter coefficients corresponding to the feature amount extraction frequency band F _j −μ _k to F _j + μ _k . The frequency band limiting unit 56 limits the frequency band of the vibration waveform data to the feature amount extraction frequency band F _j −μ _k to F _j + μ _k based on the digital filter coefficient. At this time, the monitoring frequency band of the digital filter is set to the feature amount extraction frequency band. In step S148, the feature quantity extraction unit 57 extracts the maximum amplitude of the vibration waveform data in the feature quantity extraction frequency band F _j −μ _k to F _j + μ _k as the feature quantity X _{i, j, k} . In step S _ 149, the calculation unit 55 stores the feature amounts X _{i, j, k} in the storage unit 53. Here, when the class i is 1 (locked), the locked vibration waveform data is used as the vibration waveform data. When the class i is 2 (unlocked), unlocked vibration waveform data is used as the vibration waveform data.

Here, the discriminant analysis will be described with reference to FIG. FIG. 9 is a conceptual diagram of the linear discriminant analysis method used in the present embodiment. The horizontal axis of the feature space in the upper right in FIG. 9 indicates the feature quantity X ₁ obtained from the feature quantity extraction frequency band F ₁ −μ _k to F ₁ + μ _k, and the vertical axis of the feature space represents the feature quantity extraction frequency. The feature amount X ₂ obtained from the band F ₂ −μ _k to F ₂ + μ _k is shown. Here, when the dimension number L of the feature quantity is 2, the feature space is referred to as a feature plane, and the discriminant function is a straight line. When the feature quantity is multidimensional, the discriminant function is a hyperplane. The triangle symbol corresponds to the case where the class i is 1 (locked). The circle symbol corresponds to the case where the class i is 2 (unlocked). The distribution corresponding to the case where the class i is 1 (locked) and the distribution corresponding to the case where the class i is 2 (unlocked) are roughly separated, but misidentification occurs due to a slight overlap. In the linear discriminant analysis method, variable conversion from the feature amounts X ₁ and X ₂ to the determination index z is performed so that the misclassification is minimized. The following formula is used as the conversion formula.

Here, λ ₀ , λ ₁ , and λ ₂ are referred to as linear discrimination coefficients. λ ₁ is the weight (coefficient) of X ₁ . λ ₂ is a weight (coefficient) of X ₂ . The above equation (6) may be referred to as a linear discriminant. The feature amounts X ₁ and X ₂ are aggregated into the determination index z by the conversion of the above expression, and a new distribution corresponding to the case where the class i is 1 (locked) and a new distribution corresponding to the case where the class i is 2 (unlocked) Is generated. The new distribution is shown at the lower left in FIG. Here, z = 0 is a boundary threshold value between the two new distributions. The smallest misclassification can be understood as the smallest overlap between these two distributions.

With reference to FIGS. 10A and 10B, the effect of the division number k on the linear discriminant analysis will be described. Since the detuning parameter μ _k depends on the division number k, the detuning parameter μ _k changes when the division number k changes. Changing the detuning parameter μ _k is changing the bandwidth of the feature amount extraction frequency band (monitoring frequency band). Figure 10A, in a case the division number k is k _1, shows the corresponding distribution when the distribution and class i class i corresponds to the case of 1 (locked) 2 (unlocked). Figure 10B, in a case the division number k is k _2, shows the corresponding distribution when the distribution and class i class i corresponds to the case of 1 (locked) 2 (unlocked). If the division number k is k ₁ has a large overlap of the distribution, when the division number k is k ₂ is small overlap distribution. When the division number k changes and the detuning parameter μ _k changes, the distribution separation changes. By appropriately setting the division number k and the detuning parameter μ _k , it is possible to respond robustly to cases where the peak width included in the difference spectrum is wide or the peak width is narrow. If there are many difference spectrum components included in the feature quantity extraction frequency band, high quality feature quantities can be extracted, and separation of distribution becomes large. On the other hand, if the difference spectrum component included in the feature amount extraction frequency band is small, the overlap between the distribution corresponding to the locking and the distribution corresponding to the unlocking increases, and the risk of erroneous determination increases. However, if the detuning parameter μ _k is too large, noise is extracted as a feature amount, which increases the risk of erroneous determination. Therefore, the division number k and the detuning parameter μ _k are optimized so that the risk of erroneous determination is minimized.

Referring to FIG. 7, calculation unit 55 executes steps S150 to S151 for k = 1 to N. In step S150, the optimization calculation unit 59 calculates linear discrimination coefficients λ ₀ , λ ₁ to λ _L based on the feature amounts X _{i, j, k} extracted in step S148. The linear discrimination coefficients λ ₀ and λ ₁ to λ _L calculated in step S150 may be referred to as linear discrimination coefficient candidates. In step S151, the discriminant analysis unit 58 inputs in step S141 based on the feature quantities X _{i, j, k} extracted in step S148 and the linear discriminant coefficients λ ₀ , λ ₁ to λ _L calculated in step S150. The classification of the obtained vibration waveform data is determined. In step S151, the optimization calculation unit 151 calculates a false detection rate in the determination result. Specifically, the optimization calculation unit 151 detects the detection error rate P (i = 1 | i = 2) in which the class i of the unlocking vibration waveform data is 1 (locking), and the class i of the locking vibration waveform data. 2 is calculated as an erroneous detection rate P (i = 2 | i = 1). As a result, the false detection rate P (i = 1 | i = 2) and the false detection rate P (i = 2 | i = 1) are calculated for each case where k is 1 to N.

In step S152, the optimization calculation unit 59 calculates the optimal division number k ^* that minimizes the evaluation function J represented by the following equation. The optimization calculation unit 59 estimates the optimal division number k ^* using, for example, a least square estimation method.

The evaluation function J is an average value of the false detection rate P (i = 2 | i = 1) and the false detection rate P (i = 2 | i = 1).

FIG. 11 is a graph showing the relationship between the evaluation function J and the division number k. When the division number k is the optimum division number k ^* , the evaluation function J is minimized.

In step S153, the optimization calculation unit 59 calculates the optimal detuning parameter μ _{k *} based on the optimal division number k ^* and the equation (4). That is, the optimization calculation unit 59 determines that the false detection rate (specifically, the evaluation function J) is minimum based on the determination result based on the linear discrimination coefficients λ ₀ , λ ₁ to λ _L calculated in step S150. The optimum bandwidth as the bandwidth of the feature quantity extraction frequency band is calculated.

In step S154, the optimization calculation unit 59 calculates digital filter coefficients corresponding to the feature amount extraction frequency bands F _j −μ _{k *} to F _j + μ _{k *} for each of j = 1 to L. That is, the optimization calculation unit 59 calculates the digital filter coefficient based on the feature frequencies F ₁ to F _L selected in step S146 and the optimal detuning parameter μ _{k *} (optimum bandwidth) calculated in step S153. .

The computing unit 55 executes steps S155 to S157 for i = 1 to M and j = 1 to L. In step S155, the frequency band limiting unit 56 determines the frequency band of the vibration waveform data input in step S141 based on the digital filter coefficient calculated in step S154 as the feature amount extraction frequency band F _j −μ _{k *} to F. Limit to _j + μ _{k *} . In step S156, the feature amount extraction unit 57 extracts the maximum amplitude in the feature amount extraction frequency band F _j −μ _{k *} to F _j + μ _{k *} of the vibration waveform data as the feature amount X _{i, j, k *} . In step S157, the calculation unit 55 stores the feature amounts X _{i, j, k *} in the storage unit 53. Here, when the class i is 1 (locked), the locked vibration waveform data is used as the vibration waveform data. When the class i is 2 (unlocked), unlocked vibration waveform data is used as the vibration waveform data.

In step S158, the optimization calculation unit 59 calculates linear discrimination coefficients λ ₀ , λ ₁ to λ _L based on the feature amounts X _{i, j, k *} extracted in step S156. In step S159, the discriminant analysis unit 58 uses the feature quantities X _{i, j, k *} extracted in step S156 and the linear discriminant coefficients λ ₀ , λ ₁ to λ _L calculated in step S158, in step S141. The classification of the input vibration waveform data is determined. In step S159, the optimization calculation unit 59 calculates a false detection rate in the determination result. Specifically, the optimization calculation unit 59 detects the detection error rate P (i = 1 | i = 2) in which the class i of the unlocking vibration waveform data is 1 (locking), and the class i of the locking vibration waveform data. 2 is calculated as an erroneous detection rate P (i = 2 | i = 1).

In step S160, the optimization calculation unit 59 compares the false detection rate in the determination in step S159 with the target value. For example, when both the false detection rate P (i = 1 | i = 2) and the false detection rate P (i = 2 | i = 1) are smaller than the target values (YES in step S160), step S140 is ended. Proceed to step S170. When both the false detection rate P (i = 1 | i = 2) and the false detection rate P (i = 2 | i = 1) are not smaller than the target value (NO in step S160), the process returns to step S146. Note that the evaluation function J as an average value of the false detection rate P (i = 1 | i = 2) and the false detection rate P (i = 2 | i = 1) is compared with a target value, and a step is performed based on the comparison result. You may determine whether it complete | finishes S140 and progresses to step S170 or returns to step S146.

When returning to step S146, the method of selecting the feature frequencies F ₁ to F _L from the feature frequency candidates f ₀ , f ₁ , f ₂ ,... Is changed, and steps S147 to S160 are performed again. By step S140, an optimal digital filter coefficient and an optimal linear discrimination coefficient are obtained.

Referring to FIG. 5, in step S 170, the communication unit 54 of the determination condition setting device 50 transmits the determination condition to the determination device 10. The determination condition includes the digital filter coefficient calculated in step S154 and the linear discrimination coefficient calculated in step S158. The communication unit 24 of the determination device 10 receives the determination condition from the determination condition setting device 50. In step S180, the user releases the connection between the determination device 10 and the determination condition setting device 10.

Next, each step included in the operation step S20 will be described. Each step included in the operation step S20 is executed by the determination apparatus 10. In the operation step S20, the determination condition setting device 50 is not necessary. In step S200, the vibration detection unit 11 detects vibration. Specifically, the vibration detection unit 11 converts vibration due to locking or unlocking of the lock 104 into an electric signal. The unnecessary response removing unit 21 removes noise from the electric signal by a bandpass filter. The A / D converter 22 performs analog-digital conversion of the electrical signal after passing through the unnecessary response removing unit 21 (bandpass filter) into vibration waveform data. In step S210, the calculation unit 25 determines the classification of vibration detected in step S200 based on the determination condition.

Referring to FIG. 12, step S210 includes steps S211 to S217. The computing unit 25 executes steps S211 to S213 for j = 1 to L. In step S211, the frequency band limiting unit 26 uses the frequency band of the vibration waveform data obtained in step S200 based on the digital filter coefficient transmitted by the determination condition setting device 50 as the feature amount extraction frequency band F _j -μ _{k *.} Limited to ~ F _j + μ _{k *} . In step S212, the feature amount extraction unit 27 extracts the maximum amplitude in the feature amount extraction frequency band F _j −μ _{k *} to F _j + μ _{k *} of the vibration waveform data as the feature amount X _j . As described above, the feature amount extraction processing according to the present embodiment includes step S211 in which the vibration waveform data is band-limited by the digital filter, and step S212 in which the feature amount is extracted from the vibration waveform data after the band limitation. In step S212, the absolute value of the amplitude is extracted. In step S _ 213, the calculation unit 25 stores the feature amount X _j in the storage unit 23.

In step S214, the discriminant analysis unit 28 calculates a determination index z based on the linear discrimination coefficient transmitted by the determination condition setting device 50 and the feature quantity X _j (j = 1 to L) extracted in step S212. In step S215, the discriminant analysis unit 28 compares the determination index z calculated in step S214 with the boundary threshold “0”. When the determination index z is larger than the boundary threshold (YES in step S215), the discriminant analysis unit 28 determines that the vibration classification detected in step S200 is class 1 (locked) (step S216). When the determination index z is not greater than the boundary threshold (NO in step S215), the discriminant analysis unit 28 determines that the vibration classification detected in step S200 is class 2 (unlocked) (step S217).

Referring to FIG. 5, in step S220, communication unit 24 transmits a determination result regarding the vibration detected in step S200 to the outside.

In the present embodiment, vibration classification is determined based on a plurality of feature amounts (steps S214 to S217). Therefore, the determination accuracy is improved. In the present embodiment, the vibration classification is determined based on the maximum amplitude in the feature amount extraction frequency band F _j −μ _{k *} to F _j + μ _{k *} in step S211. Therefore, the determination accuracy is maintained even if the frequency at which the amplitude is maximum changes slightly.

In the present embodiment, the bandwidth of the feature amount extraction frequency band is set so that the false detection rate is minimized (step S153). Therefore, the determination accuracy is improved. In the present embodiment, feature frequencies F ₁ to F _L are selected from the feature frequency candidates f ₀ , f ₁ , f ₂ ,... So that the false detection rate is smaller than the target value (steps S160 and S146). Therefore, a certain determination accuracy is guaranteed. In the present embodiment, noise is removed from the electrical signal representing vibration by the unnecessary response removing unit 21 (bandpass filter). Therefore, unnecessary responses (unnecessary external communication) are reduced.

Also, step S210 does not include calculations with large load and power consumption such as matrix calculation, statistical analysis calculation, fast Fourier transform, peak search calculation, and digital filter iterative calculation. Therefore, step S210 can be executed by a general-purpose microcontroller without using a CPU with high power consumption such as a DSP (Digital Signal Processor). Therefore, the power consumption of the determination apparatus 10 that needs to be constantly supplied can be suppressed.

Next, a test for automatically setting determination conditions using the vibration classification system 1 according to the present embodiment, determining vibrations based on the determination conditions, and evaluating the detection rate will be described.

FIG. 13 shows a difference spectrum between the locking vibration spectrum and the unlocking vibration spectrum obtained by actual measurement. The peak search detected two or more distinct peaks including the first peak and the second peak. 731 Hz was extracted as the frequency f ₁ of the apex of the first peak, and 1040 Hz was extracted as the frequency f ₂ of the apex of the second peak. Further, since it is also characterized by no peak, 150 Hz was extracted as the center frequency f _{0 in a} frequency band in which there is no difference between the unlocking vibration spectrum and the locking vibration spectrum.

14A to 14C show the division number k dependency of the feature space. The horizontal axis of the feature space represents a feature value X ₁ obtained from the feature amount extraction frequency band around the frequency f _1, the longitudinal axis of the feature space obtained from the feature extraction frequency band around the frequency f ₂ and indicating the feature quantity _{X 2.} FIG. 14A shows a distribution (triangle symbol) corresponding to locking and a distribution (circle symbol) corresponding to unlocking when the division number k is 10. FIG. FIG. 14B shows a distribution (triangle symbol) corresponding to locking and a distribution (circle symbol) corresponding to unlocking when the division number k is 18. FIG. 14C shows a distribution (triangle symbol) corresponding to locking and a distribution (circle symbol) corresponding to unlocking when the division number k is 30. The broken line in the figure shows a straight line when the left side of the linear discriminant (6) is 0. When the division number k is 10 and 30, the distribution corresponding to locking and the distribution corresponding to unlocking overlap each other, but when the division number k is 18, the distribution corresponding to locking and the distribution corresponding to unlocking are clear. Isolated on.

FIG. 15 is a graph showing the relationship between the evaluation function J and the division number k. The division number k was optimized using the evaluation function J. The optimal number of divisions k ^* that minimizes the evaluation function J was 15. The digital filter in this test was a secondary bandpass filter. A digital filter coefficient corresponding to the feature amount extraction frequency band 710 to 750 Hz centered on the frequency f ₁ was calculated as follows.

The digital filter coefficient corresponding to the feature amount extraction frequency band 1020 to 1060 Hz centered on the frequency f ₂ was calculated as follows.

The linear discriminant coefficient was calculated as follows.

The determination condition setting device 50 transmits the above parameters to the determination device 10 by wired communication. The determination device 10 writes the parameter in the storage unit 23 and determines vibration waveform data representing vibration based on the parameter.

FIG. 16 shows a comparison between the determination result obtained by the present embodiment and the determination result according to the comparative example. Here, in the comparative example, the amplitude of a predetermined frequency component is extracted from the vibration waveform data, and the determination is performed based on a comparison between the amplitude and a threshold value. According to the comparative example, the detection rate for unlocking vibration was 89.6%, the false detection rate was 24.8%, the false alarm rate was 10.4%, the detection rate for locking vibration was 75.2%, The detection rate was 10.4% and the false alarm rate was 24.8%. In contrast, according to the present embodiment, the detection rate for unlocking vibration is 99.8%, the false detection rate is 0.8%, the false alarm rate is 0.2%, and the detection rate for locking vibration is It was 99.2%, the false detection rate was 0.2%, and the false alarm rate was 0.8%. According to the present embodiment, high-precision detection is possible.

Moreover, according to the present embodiment, the average time required from vibration detection (step S200) to determination (step S210) was 90 ms. From detection of vibration to determination could be performed within 1 minute.

The vibration classification system 1 according to the present embodiment can be applied to detection of locking and unlocking of the entrance door 100 in the security system. Therefore, the industrial value of the vibration classification system 1 is high.

In the determination apparatus 10, the vibration detection unit 11, the unnecessary response removal unit 21, the A / D conversion unit 22, the storage unit 23, the communication unit 24, the calculation unit 25, the frequency band limiting unit 26, the feature amount extraction unit 27, and the discriminant analysis The function of the unit 28 may be realized by a dedicated circuit or a dedicated device, but may also be realized by a CPU of a computer executing a program. In the determination condition setting device 50, the functions of the storage unit 53, the communication unit 54, the calculation unit 55, the frequency band limiting unit 56, the feature amount extraction unit 57, the discriminant analysis unit 58, and the optimization calculation unit 59 are a dedicated circuit or a dedicated device. However, it may be realized by the CPU of the computer executing the program.

In the above example, the program can be stored using various types of non-transitory computer readable media and supplied to the computer. Non-transitory computer readable media include various types of tangible storage media. Examples of non-transitory computer-readable media include magnetic recording media (for example, flexible disks, magnetic tapes, hard disk drives), magneto-optical recording media (for example, magneto-optical disks), CD-ROMs (Read Only Memory), CD-Rs, CD-R / W, semiconductor memory (for example, mask ROM, PROM (Programmable ROM), EPROM (Erasable PROM), flash ROM, RAM (Random Access Memory)). In addition, the program may be supplied to the computer by various types of temporary computer readable media. Examples of transitory computer readable media include electrical signals, optical signals, and electromagnetic waves. The temporary computer-readable medium can supply the program to the computer via a wired communication path such as an electric wire and an optical fiber, or a wireless communication path.

Note that the present invention is not limited to the above-described embodiment, and can be appropriately changed without departing from the spirit of the present invention. The present invention can be applied to vibration classification other than vibration due to locking or unlocking of a lock. For example, it is possible to classify vibrations to be measured and vibrations other than noise.

Some or all of the above embodiments can be described as in the following supplementary notes, but are not limited thereto.

(Supplementary note 1) A vibration classification system including a determination device and a determination condition setting device. The determination device detects a first vibration and transmits first vibration waveform data representing the first vibration to the determination condition setting device. The determination condition setting device calculates a determination condition for determining a vibration classification using the first vibration waveform data as teacher data, and transmits the determination condition to the determination device. The determination device detects a second vibration, determines a classification of the second vibration based on the determination condition, and transmits a determination result regarding the second vibration to the outside.

(Supplementary note 2) The vibration classification system according to supplementary note 1, wherein the determination condition includes a digital filter coefficient and a linear discrimination coefficient. The determination device converts the first vibration into a first electric signal, converts the second vibration into a second electric signal, and converts the first electric signal into the first vibration waveform data. Analog-to-digital conversion means for converting the second electric signal into second vibration waveform data; and determination apparatus side communication means for transmitting the first vibration waveform data to the determination condition setting device and receiving the determination conditions; A determination device-side frequency band limiting unit that limits a frequency band of the second vibration waveform data to a plurality of feature quantity extraction frequency bands based on the digital filter coefficient; and a plurality of feature quantities from the plurality of feature quantity extraction frequency bands Determination device side feature amount extraction means, and determination device side discrimination analysis means for determining the classification of the second vibration based on the linear discrimination coefficient and the plurality of feature amounts. The determination device-side communication unit transmits the determination result for the second vibration to the outside.

(Additional remark 3) It is a vibration classification system of Additional remark 2, Comprising: The said determination apparatus side feature-value extraction means extracts the some maximum amplitude in each of these frequency bands as said some feature-value.

(Supplementary note 4) The vibration classification system according to supplementary note 1, wherein the first vibration includes a first class vibration belonging to a first class and a second class vibration belonging to a second class. The first vibration waveform data includes first class vibration waveform data representing the first class vibration and second class vibration waveform data representing the second class vibration. The determination condition setting device includes: a setting device side frequency band limiting unit that limits a frequency band of the first class vibration waveform data and the second class vibration waveform data to a plurality of frequency bands; Setting device side feature quantity extraction means for extracting a plurality of first class feature quantities from the plurality of frequency bands and extracting a plurality of second class feature quantities from the plurality of frequency bands of the second class vibration waveform data, respectively. And a setting device-side discriminating / analyzing means for determining a classification of the first class vibration waveform data and the second class vibration waveform data based on the plurality of first class feature quantities and the plurality of second class feature quantities, Optimizing the determination condition in cooperation with the setting device side frequency band limiting unit, the setting device side feature amount extraction unit, and the setting device side discriminant analysis unit Comprising the optimization calculation means for performing calculation to receive the first vibration waveform data, and a setting device side communication means for transmitting the determination condition in the determination device.

(Supplementary note 5) The vibration classification system according to supplementary note 4, wherein the determination condition includes a digital filter coefficient and a linear determination coefficient. The optimization calculation means calculates first class spectrum data from the first class vibration waveform data, calculates second class spectrum data from the second class vibration waveform data, and calculates the first class spectrum data and the first class spectrum data. Difference spectrum data is calculated from the two-class spectrum data, and a plurality of characteristic frequencies are extracted from the difference spectrum data. The setting device side frequency band limiting means limits the frequency bands of the first class vibration waveform data and the second class vibration waveform data to a plurality of first frequency bands centered on the plurality of characteristic frequencies, respectively. The setting device-side feature amount extraction unit extracts a plurality of first class first feature amounts from the plurality of first frequency bands of the first class vibration waveform data, and the plurality of second class vibration waveform data. A plurality of second class first feature values are respectively extracted from the first frequency band. The optimization calculation unit calculates a linear discriminant coefficient candidate based on the plurality of first class first feature values and the plurality of second class first feature values. The setting device side discriminating / analyzing unit determines a classification of the first class vibration waveform data and the second class vibration waveform data based on the linear discriminant coefficient candidates. The optimization calculation means calculates an optimum bandwidth as a bandwidth of the plurality of first frequency bands that minimizes the false detection rate based on a determination result based on the linear discrimination coefficient candidate, and The digital filter coefficient is calculated based on the characteristic frequency and the optimum bandwidth. The setting device side frequency band limiting means limits the frequency band of the first class vibration waveform data and the first class vibration waveform data to a plurality of second frequency bands based on the digital filter coefficient. The setting device-side feature quantity extraction unit extracts a plurality of first class second feature quantities from the plurality of second frequency bands of the first class vibration waveform data, and the plurality of second class vibration waveform data. A plurality of second class second feature quantities are respectively extracted from the second frequency band. The optimization calculation means calculates the linear discrimination coefficient based on the plurality of first class second feature values and the plurality of second class second feature values.

(Supplementary note 6) The vibration classification system according to supplementary note 5, wherein the optimization calculating unit extracts a plurality of feature frequency candidates including the plurality of feature frequencies from the difference spectrum data, and the plurality of feature frequency candidates. To select the plurality of characteristic frequencies. The setting device side discriminating / analyzing unit determines a classification of the first class vibration waveform data and the second class vibration waveform data based on the linear discrimination coefficient. The optimization calculation means changes the way of selecting the plurality of feature frequencies from the plurality of feature frequency candidates when the false detection rate in the classification determination based on the linear discrimination coefficient is not lower than a target value, and changes the digital frequency The calculation of the filter coefficient and the linear discrimination coefficient is performed again.

(Additional remark 7) It is a vibration classification system of

Additional remark

2 or 3, Comprising: The said determination apparatus is provided with a band pass filter. The analog-to-digital conversion means converts the first electric signal after passing through the band-pass filter into the first vibration waveform data, and converts the second electric signal after passing through the band-pass filter into the second vibration waveform data. Convert to

(Supplementary note 8) The vibration classification system according to

supplementary note

2 or 3, wherein the analog-digital conversion means has an analog-digital conversion bit number of 12 bits or less, and the analog-digital conversion means has a sampling frequency of 6 kHz or more.

(Additional remark 9) It is a vibration classification system in any one of additional remarks 1 thru | or 8, Comprising: The said determination apparatus determines the classification | category of the said 2nd vibration within 1 minute from the detection of the said 2nd vibration.

(Supplementary Note 10) Vibration detection means for detecting a first vibration and first vibration waveform data representing the first vibration are transmitted to a determination condition setting device, and a determination condition for determining a vibration classification is set as the determination condition. A vibration determination apparatus comprising communication means for receiving from an apparatus and calculation means. The vibration detection means detects a second vibration. The computing means determines the classification of the second vibration based on the determination condition. The communication means transmits a determination result regarding the second vibration to the outside.

(Supplementary Note 11) Communication means for receiving first vibration waveform data from the vibration determination device and calculation means for calculating a determination condition for determining a vibration classification using the first vibration waveform data as teacher data. Vibration determination condition setting device. The communication means transmits the determination condition to the vibration determination device.

(Supplementary note 12) The vibration determination condition setting device according to supplementary note 11, wherein the first vibration waveform data belongs to a first class vibration waveform data representing a first class vibration belonging to a first class and a second class. Second-class vibration waveform data representing second-class vibration. The computing means includes frequency band limiting means for limiting the frequency bands of the first class vibration waveform data and the second class vibration waveform data to a plurality of frequency bands, and the plurality of frequency bands of the first class vibration waveform data. A plurality of first class feature quantities respectively extracted from the plurality of frequency bands of the second class vibration waveform data, and a plurality of second class feature quantities respectively. Discriminant analysis means for determining a classification of the first class vibration waveform data and the second class vibration waveform data based on the feature quantity and the plurality of second class feature quantities, the frequency band limiting means, and the feature quantity extraction means And optimization calculation means for performing optimization calculation of the determination condition in cooperation with the discriminant analysis means.

(Supplementary Note 13) The determination device detects the first vibration, the determination device transmits first vibration waveform data representing the first vibration to the determination condition setting device, and the determination condition setting device transmits the first vibration waveform data. Is used as teacher data to calculate a determination condition for determining a vibration classification, the determination condition setting device transmits the determination condition to the determination device, the determination device detects a second vibration, and the determination A vibration classification method in which an apparatus determines a classification of the second vibration based on the determination condition, and the determination apparatus transmits a determination result regarding the second vibration to the outside.

(Supplementary Note 14) Based on the determination condition received from the determination condition setting device, the first vibration is detected, the first vibration waveform data representing the first vibration is transmitted to the determination condition setting device, the second vibration is detected. And determining a classification of the second vibration, and transmitting a determination result of the second vibration to the outside.

(Supplementary Note 15) A vibration classification method for calculating a determination condition for determining a vibration classification using first vibration waveform data received from a vibration determination apparatus as teacher data and transmitting the determination condition to the vibration determination apparatus .

(Supplementary Note 16) The first vibration waveform data representing the first vibration is transmitted to the determination condition setting device, the classification of the second vibration is determined based on the determination condition received from the determination condition setting device, and the second vibration is determined. A program that causes a computer to send the determination result of the above to the outside.

(Supplementary Note 17) A computer that calculates determination conditions for determining vibration classification using first vibration waveform data received from a vibration determination apparatus as teacher data, and transmits the determination conditions to the vibration determination apparatus. A program to be executed.

The present invention has been described above with reference to the embodiment, but the present invention is not limited to the above. Various changes that can be understood by those skilled in the art can be made to the configuration and details of the present invention within the scope of the invention.

This application claims priority based on Japanese Patent Application No. 2013-073709 filed on March 29, 2013, the entire disclosure of which is incorporated herein.

DESCRIPTION OF SYMBOLS 1 Vibration classification system 10 Determination apparatus 11 Vibration detection part 21 Unnecessary response removal part 22 A / D conversion part 24 Communication part 25 Calculation part 26 Frequency band restriction part 27 Feature-value extraction part 28 Discriminant analysis part 50 Determination condition setting apparatus 54 Communication part 55 Calculation Unit 56 Frequency Band Limiting Unit 57 Feature Amount Extracting Unit 58 Discriminant Analysis Unit 59 Optimization Calculation Unit

Claims

A determination device;
A judgment condition setting device,
The determination device detects a first vibration, and transmits first vibration waveform data representing the first vibration to the determination condition setting device;
The determination condition setting device calculates a determination condition for determining a vibration classification using the first vibration waveform data as teacher data, and transmits the determination condition to the determination device.
The determination device detects a second vibration, determines a classification of the second vibration based on the determination condition, and transmits a determination result of the second vibration to the outside.
The vibration classification system according to claim 1,
The determination condition includes a digital filter coefficient and a linear discrimination coefficient,
The determination device includes:
Vibration detecting means for converting the first vibration into a first electric signal and converting the second vibration into a second electric signal;
Analog-to-digital conversion means for converting the first electric signal into the first vibration waveform data and converting the second electric signal into second vibration waveform data;
A determination device side communication means for transmitting the first vibration waveform data to the determination condition setting device and receiving the determination condition;
A determination device-side frequency band limiting unit that limits a frequency band of the second vibration waveform data to a plurality of feature amount extraction frequency bands based on the digital filter coefficient;
A determination device-side feature quantity extraction means for extracting a plurality of feature quantities from the plurality of feature quantity extraction frequency bands;
A determination device-side discriminant analysis unit that determines a classification of the second vibration based on the linear discrimination coefficient and the plurality of feature amounts;
The determination apparatus-side communication unit transmits the determination result for the second vibration to the outside.
The vibration classification system according to claim 2,
The determination device-side feature quantity extraction unit extracts a plurality of maximum amplitudes in each of the plurality of frequency bands as the plurality of feature quantities.
The vibration classification system according to claim 1,
The first vibration includes a first class vibration belonging to a first class and a second class vibration belonging to a second class,
The first vibration waveform data includes first class vibration waveform data representing the first class vibration, and second class vibration waveform data representing the second class vibration,
The determination condition setting device includes:
Setting device side frequency band limiting means for limiting the frequency bands of the first class vibration waveform data and the second class vibration waveform data to a plurality of frequency bands;
A plurality of first class feature amounts are extracted from the plurality of frequency bands of the first class vibration waveform data, respectively, and a plurality of second class feature amounts are extracted from the plurality of frequency bands of the second class vibration waveform data, respectively. A setting device-side feature amount extraction means to perform,
A setting device-side discriminating / analyzing unit that determines a classification of the first class vibration waveform data and the second class vibration waveform data based on the plurality of first class feature quantities and the plurality of second class feature quantities;
In conjunction with the setting device side frequency band limiting means, the setting device side feature amount extraction means, and the setting device side discriminant analysis means, optimization calculation means for performing optimization calculation of the determination condition;
A vibration classification system comprising: a setting device-side communication unit that receives the first vibration waveform data and transmits the determination condition to the determination device.
The vibration classification system according to claim 4,
The determination condition includes a digital filter coefficient and a linear discrimination coefficient,
The optimization calculation means includes:
Calculating first class spectrum data from the first class vibration waveform data;
Calculating second class spectrum data from the second class vibration waveform data;
Calculating difference spectrum data from the first class spectrum data and the second class spectrum data;
Extracting a plurality of characteristic frequencies from the difference spectrum data;
The setting device-side frequency band limiting means limits the frequency bands of the first class vibration waveform data and the second class vibration waveform data to a plurality of first frequency bands centered on the plurality of characteristic frequencies, respectively.
The setting device side feature amount extraction means includes:
Extracting a plurality of first class first feature values from the plurality of first frequency bands of the first class vibration waveform data,
Extracting a plurality of second class first feature values from the plurality of first frequency bands of the second class vibration waveform data,
The optimization calculation means calculates linear discriminant coefficient candidates based on the plurality of first class first feature values and the plurality of second class first feature values,
The setting device side discriminant analysis means determines the classification of the first class vibration waveform data and the second class vibration waveform data based on the linear discriminant coefficient candidates,
The optimization calculation means includes:
Based on the determination result based on the linear discriminant coefficient candidate, calculate the optimum bandwidth as the bandwidth of the plurality of first frequency bands that minimize the false detection rate,
Calculating the digital filter coefficient based on the plurality of characteristic frequencies and the optimum bandwidth;
The setting device side frequency band limiting means limits the frequency band of the first class vibration waveform data and the first class vibration waveform data to a plurality of second frequency bands based on the digital filter coefficient,
The setting device side feature amount extraction means includes:
Extracting a plurality of first class second feature values from the plurality of second frequency bands of the first class vibration waveform data,
Extracting a plurality of second class second feature values from the plurality of second frequency bands of the second class vibration waveform data,
The optimization calculation means calculates the linear discrimination coefficient based on the plurality of first class second feature values and the plurality of second class second feature values.
Vibration detecting means for detecting the first vibration;
Communication means for transmitting first vibration waveform data representing the first vibration to a determination condition setting device and receiving a determination condition for determining a vibration classification from the determination condition setting device;
An arithmetic means,
The vibration detecting means detects the second vibration;
The calculation means determines the classification of the second vibration based on the determination condition,
The said communication means transmits the determination result about the said 2nd vibration to the exterior. Vibration determination apparatus.
Communication means for receiving first vibration waveform data from the vibration determination device;
Computing means for calculating a determination condition for determining a vibration classification using the first vibration waveform data as teacher data;
The communication means transmits the determination condition to the vibration determination device.
The vibration determination condition setting device according to claim 7,
The first vibration waveform data includes first class vibration waveform data representing a first class vibration belonging to a first class and second class vibration waveform data representing a second class vibration belonging to a second class,
The computing means is
Frequency band limiting means for limiting a frequency band of the first class vibration waveform data and the second class vibration waveform data to a plurality of frequency bands;
A plurality of first class feature amounts are extracted from the plurality of frequency bands of the first class vibration waveform data, respectively, and a plurality of second class feature amounts are extracted from the plurality of frequency bands of the second class vibration waveform data, respectively. Feature quantity extraction means for
Discriminant analysis means for determining a classification of the first class vibration waveform data and the second class vibration waveform data based on the plurality of first class feature quantities and the plurality of second class feature quantities;
A vibration determination condition setting device comprising: an optimization calculation unit that performs optimization calculation of the determination condition in cooperation with the frequency band limiting unit, the feature amount extraction unit, and the discriminant analysis unit.
Judgment device
Detecting the first vibration,
Transmitting first vibration waveform data representing the first vibration to a determination condition setting device;
The determination condition setting device comprises:
Calculating a determination condition for determining a vibration classification using the first vibration waveform data as teacher data;
Transmitting the determination condition to the determination device;
The determination device is
Detecting the second vibration,
Determining the classification of the second vibration based on the determination condition;
A vibration classification method of transmitting a determination result about the second vibration to the outside.
Transmitting first vibration waveform data representing the first vibration to the determination condition setting device;
Determining the classification of the second vibration based on the determination condition received from the determination condition setting device;
A non-transitory computer-readable medium storing a program for causing a computer to transmit a determination result about the second vibration to the outside.