JP2017001522A - Vibration source detection device, detection system, and method - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a device for estimating the position of an object even in the case that a signal which can estimate the position of the object can be observed only intermittently in time.SOLUTION: A detection system for a ship is a detection system including a detection part for detecting vibrations from an object existing underwater to acquire vibration information, a determination part for determining a state of a ship, and an estimation part for acquiring information of the object by using vibration information at timing when the determination part determines that a state of the ship can estimate information of the object on the basis of the vibration information.SELECTED DRAWING: Figure 2

Description

  The present invention relates to a vibration source detection device, system, and method used for, for example, ships.

  There are a wide variety of systems for observing vibration signals (source signals) emitted from underwater objects such as whales and fish, such as underwater passive sonar, and estimating the position of the object. As such an example, for example, there is JP-A-10-203486 (Patent Document 1).

Japanese Patent Laid-Open No. 10-203486

  Patent Document 1 describes a method of suppressing the noise of a ship when using an automatic marine vessel maneuvering device and an underwater acoustic device to less than the allowable noise of the underwater acoustic equipment.

  However, when the ship must be driven at high speed, a signal that can estimate the direction and position of the object may be observed only intermittently. For example, because the ship is moving at a very high speed, detectors (hydrophones, microphones, and other elements used for sonar) rise from the surface of the water and receive signals that can estimate the direction and position of objects. It is a case where it cannot be done. Or it is the case where the noise accompanying the drive of the ship is too large and the signal of the detector is saturated. As an example of a large noise, for example, the noise when the hull collides with the water surface due to the vertical movement of the hull can be considered.

  In the conventional technique such as Patent Document 1, a technique that can specify the position of a target object even in such a case is not disclosed. Therefore, in the conventional system, for example, when a signal that can estimate the position of the object can be observed only intermittently due to a situation where the ship must be driven at high speed, the position of the object cannot be estimated. .

  Therefore, the present invention provides an apparatus for estimating the position of an object even when a signal that can estimate the position of the object can be observed only intermittently in time.

  In order to solve the above-described problems, one aspect of the present invention is a detection system for a ship, which detects vibration from an object in water and obtains vibration information, and determines the state of the ship. And a estimator for obtaining information on the object using vibration information at a timing when the state of the ship is determined to be able to estimate the object information based on the vibration information. System.

  As a typical example, the timing at which the state of the ship is determined to be able to estimate the information of the object based on the vibration information is the timing at which the detection unit is in water and the vibration information can be detected. Including at least. In other words, the information on the object cannot be estimated at the timing when the detection unit gets out of the water and vibration information cannot be obtained, or when the detection unit collides with the water surface and the signal is saturated. Judge that it is possible.

  Another aspect of the present invention is a vibration source detection system device that is provided in a movable body that can move in water or on water, and detects the direction of the object by detecting vibration from the object in water, A detection unit that detects vibration from the object and obtains vibration information, and a determination unit that determines whether or not the state of the moving body is a direction inferable state capable of estimating the direction of the object. And an estimator that estimates the direction of the vibration source object viewed from the moving body using the vibration information detected when the direction can be estimated among the vibration information detected by the detector. Device.

  Another aspect of the present invention is a detection method for estimating information on an object based on vibration information using a detection unit that detects vibration from an object in water and obtains vibration information in a ship moving on the water. In this detection method, the information on the object is estimated by eliminating the vibration information at the timing when the detection unit floats from the water surface and the timing when the detection unit comes into contact with the water surface.

  In a typical embodiment of the present invention, a detector that obtains vibrations from an underwater object and estimates information about the object is configured as a passive sonar.

  In a preferred specific example of the present invention, in order to determine the state of the moving body or the ship, the vibration spectrum detected by the detector or the information of an acceleration or pressure sensor provided separately is used.

  In a preferred embodiment of the present invention, a state transition model of the moving body or ship is used to determine the state of the moving body or ship.

  In a preferred embodiment of the present invention, the state transition model of the moving body or ship is prepared according to the type of driving (operation) command information of the moving body or ship.

  In a preferred embodiment of the present invention, the moving body or ship is a ship that moves at a high speed, for example, a small-sized high-power ship or a water-jet propelled ship.

  Still another aspect of the present invention is a detection unit that is provided in a moving body that is propelled by a water jet and detects underwater vibration, a determination unit that determines whether the moving body is in a state in which direction estimation is possible, and a detection The vibration source detection apparatus includes: an estimation unit that estimates the direction of the vibration source as viewed from the moving body using vibration information detected when the direction of the vibration information is detected in the vibration detection information.

  According to the present invention, it is possible to estimate the position of an object even when a signal that can estimate the position of the object can be observed only intermittently in time. Problems, configurations, and effects other than those described above will be clarified by the following description of embodiments.

It is the block diagram which showed the hardware constitutions of the estimation system of an Example. It is a functional block diagram of the estimation system of an Example. It is a flowchart at the time of estimation of the estimation system of an Example. It is a top view which shows the example of a result output process. It is a table | surface figure which shows the data structure of the own ship state transition model. It is a flowchart of the direction estimation process of an Example. It is a flowchart of the locus | trajectory estimation process of an Example. It is a table | surface figure which shows the other example of the data structure of the own ship state transition model. It is a flowchart at the time of the own ship state transition model learning of the estimation system of an Example. It is the block diagram which showed the other example of the hardware constitutions of the estimation system of an Example. It is the functional block diagram which showed the other example of the estimation system of an Example. It is a table | surface figure which shows the other example of the data structure of the own ship state transition model.

  Hereinafter, examples will be described with reference to the drawings. In the structures of the invention described below, the same portions or portions having similar functions are denoted by the same reference numerals in different drawings, and redundant description may be omitted.

  In the following embodiments, for example, when the direction and position of an underwater object cannot be specified by a signal of a passive sonar, for example, when the detection unit gets out of water and vibration information cannot be obtained, It is determined that the information on the object cannot be estimated from the signal at the timing when the signal is saturated due to collision with the water surface. This is because the information on the underwater object cannot be accurately estimated at such timing. Therefore, the passive sonar signal is not used at such timing. For this reason, in the following embodiments, the signal of the passive sonar is obtained discretely in time, and the object information is estimated. At the timing when the signal of the passive sonar cannot be obtained, the information of the object obtained discretely in time is externally or internally swept to complement the information. Various known techniques can be used for the external sweep or internal sweep. In the following examples, more specific examples will be described.

  In this embodiment, when driving a ship at high speed, etc., a vibration source detection device capable of estimating the position of an object even when a signal capable of estimating the position of the object can only be observed intermittently An example will be described. This embodiment is, for example, an underwater passive sonar system mounted on a ship.

  FIG. 1 is a block diagram illustrating a hardware configuration of the estimation system 100 according to the present embodiment.

  An estimation system 100 according to the present embodiment includes elements 101_1 to 101_N such as a hydrophone and a microphone, a multi-channel AD converter 102, a central processing unit 103, a user interface unit 104, a storage medium 105, and a volatile memory 106. . The user interface unit 104 includes an output device such as a display and an input device such as a keyboard.

  The elements 101_1 to 101_N convert sound waves in water or air into analog voltage values and send them to the multi-channel AD converter 102. The multi-channel AD converter 102 converts a multi-channel analog voltage value into a multi-channel digital signal and sends it to the central processing unit 103. The central processing unit 103 can be operated by the user interface unit 104, and collected data and control information can be stored in the storage medium 105 and the volatile memory 106.

  In general, the elements 101_1 to 101_N are arranged at positions exposed in water or in the air in order to acquire sound wave signals. In the case of a sonar of a ship, for example, it is one or a plurality of places on the bottom or the side of the ship.

  FIG. 2 is a functional block diagram of the estimation system 100 of the present embodiment. As a specific configuration example, the signal processing unit 2000 includes a signal acquisition unit 201, a direction estimation unit 202, a trajectory estimation unit 203, a result output unit 205, and an own ship state estimation unit 206. These functions are realized using the central processing unit 103, the user interface unit 104, the storage medium 105, and the volatile memory 106. Or when the said function is implement | achieved only by hardware, the exclusive circuit structure which implement | achieves the said process is shown.

  The signal acquisition unit 201 receives the multi-channel input signal z (n, t) from the multi-channel AD converter 102 and buffers z (n, t) at every sampling time interval Δsec. However, z (n, t) means an input signal of the element n at time t seconds.

  The signal acquisition unit 201 receives the buffered input signal sequence over the past M times [z (n, t- (M-1) δ), z (n, t- (M-2) δ), ..., z (n, t-δ), z (n, t)] is multiplied by a window function and subjected to a short-time Fourier transform to obtain a multi-channel frequency domain signal x (n, τ) = [x (n, τ, 1), ..., x (n, τ, F)] are calculated and output. Where δ is the sampling time interval, F is the number of frequency numbers of the short-time Fourier transform, τ is the range of the target time of the multiplication of the window function and the short-time Fourier transform (from t- (M-1) δ to t This represents the frame number when (time zone) is regarded as one time frame.

  The own ship state estimation unit 206 receives the multi-channel frequency domain signal x (n, τ) as input, executes own ship state estimation processing, and outputs the own ship state estimation result c (τ) at time τ.

  The direction estimation unit 202 receives the multi-channel frequency domain signal x (n, τ) and the own ship state estimation result as input, and executes the direction estimation process, and the direction estimation result y (τ) = [y (τ, 1), ..., y (τ, D)] are output. Where D is the number of discretized direction numbers. As is well known, the direction estimation process estimates the direction of a sound source based on the time difference between signals from a plurality of elements.

  The trajectory estimation unit 203 inputs the direction estimation results [y (τ- (B-1)), y (τ- (B-2)), ..., y (τ)] at multiple times and the own ship state estimation results The trajectory estimation process is executed, and the trajectory estimation result Traj (τ) = [θ (τ- (B-1)), θ (τ- (B-2)),…, θ ( τ)] is output.

  The result output unit 205 presents Traj (τ) to the user. The presentation method includes, for example, image information presentation via the user interface unit 104, presentation by a braille display, presentation by voice, printing of image information via a printer, and the like.

  FIG. 3 is a flowchart at the time of estimation of the estimation system 100 of the present embodiment.

  First, in S301, the estimation system 100 according to the present embodiment determines whether or not a stop operation has been performed. If the measurement stop operation has not been performed, the estimation system 100 executes the processing from S302. If the measurement stop operation is performed, the estimation system 100 stops.

  Next, in the signal acquisition process S302, the signal acquisition unit 201 buffers the multi-channel input signal z (n, t), and buffers the input signal sequence [z (n, t- (M-1) δ). , z (n, t- (M-2) δ), ..., z (n, t-δ), z (n, t)], perform window function multiplication and short-time Fourier transform to obtain multiple channels Output frequency domain signal x (n, τ) = [x (n, τ, 1),..., X (n, τ, F)].

  Next, in own ship state estimation processing S303, the own ship state estimation unit 206 estimates the own ship state based on the multi-channel frequency domain signal x (n, τ). In order to estimate the ship's own state, an algorithm such as a Viterbi algorithm, which is a state estimation method of a hidden Markov model (HMM), is used (described with reference to FIG. 5). The own ship state is obtained, for example, as a symbol of “landing”, “floating”, or “descent”.

  Next, in direction estimation processing S304, the direction estimation unit 202 performs sound source direction estimation based on the multi-channel frequency domain signal x (n, τ) and the own ship state estimation result. The direction estimation result is obtained as a two-dimensional signal y (τ, θ), which is called a direction histogram, and represents the volume of the sound source in each direction at each time.

  Next, in the trajectory estimation process S305, the trajectory estimation unit 203 performs the direction estimation results [y (τ- (B-1)), y (τ- (B-2)), ..., y (τ) at multiple times. ] Is input, and the trajectory estimation result Traj (τ) = [θ (τ- (B-1)), θ (τ- (B-2)),..., Θ (τ)] is output.

  Next, in result output processing S305, the result output unit 205 outputs the trajectory estimation result Traj (τ), and the process returns to S301.

  FIG. 4 is an example in which the result output unit 205 of the present embodiment presents the trajectory estimation result Traj (τ) using the user interface unit 104.

  In addition to the direction estimation result 410 and the trajectory estimation result 420, the corresponding own ship state is also presented at the same time, so that the cause may be due to a specific own ship state in the time when there is no accurate direction estimation result. It becomes clear, and there is an effect that the user can pay attention only to the time of the own ship state corresponding to the accurate direction estimation result.

  FIG. 5 is an example of the data structure of the own ship state transition model used in the own ship state estimation process S303 of the present embodiment. This data structure is stored in, for example, the storage medium 105 and used for processing as appropriate.

  In this embodiment, the own ship state transition model is a hidden Markov model, and has S own ship states s (1),..., S (S). Each ship state corresponds to symbols such as “landing”, “floating”, and “descent”. Each ship's state is the probability distribution dist (i) of the vibration spectrum generated in the i-th state s (i) and from the i-th state s (i) to the j-th state s (j) in the next time frame. Has transition probability q (i, j) to transition. The probability distribution dist (i) of the frequency spectrum is, for example, a mixed Gaussian model (GMM), and dist (i) is C (i) clusters [c (i, 1),…, c (i, C (i) )]have. The c_l-th cluster c (i, c_l) is defined by the spectrum mean μ (i, c_l) (Equation 1) and the spectrum covariance matrix Σ (i, c_l) (Equation 2).

  For example, as a simple example, the ship's state is “landing” in which the element 101 (or ship) is present in the water, “floating” in which the element is out of the water surface, and the element descends on the water surface and collides with the water surface. Suppose that it has a status of “descent”. The multi-channel frequency domain signal x (n, τ) has a unique frequency spectrum in each state. For example, in the “landing” state, there are spectra of own ship noise (engine sound entering through water, friction sound with water, etc.) and underwater sound (reflection sound from an object, etc.). Further, in the “floating” state, which is a state of floating from the surface of the water, the spectrum is only the noise (engine sound) of the ship, and underwater sound cannot be obtained. In the “descent” state, a large noise is instantaneously generated due to the collision between the hull and water. The current state can be estimated to some extent using such a unique spectrum, but in this example, the ship's state transition model is further used to reflect the transition probability from a certain state to a certain state, and the state with high accuracy Make an estimate.

  That is, the state of the ship can be considered as a transition model in which a predetermined state is changed with a certain rule or cycle. In the own ship state transition model, for example, the probability of transition to the “descent” state is high after the “floating” state. After the “descent” state, a model having a high probability of transition to the “landing” state may be created by experiments or simulations of the physical movement of the ship.

  As described above, the own ship state estimation unit 206 estimates the own ship state based on the multi-channel frequency domain signal x (n, τ) and the own ship state transition model.

  FIG. 6 is a flowchart of the direction estimation process S304 of the present embodiment.

  First, the direction estimation unit 202 of the present embodiment determines whether or not the own ship state is “landing” in S601. If it is “landing”, the element 101 can observe a vibration signal (source signal) emitted from an object in water and estimate the position of the object. Therefore, if the state is “landing”, S602 and subsequent steps are executed, and if not, the process ends. That is, as described with reference to FIG. 5, in the “floating” state, the element 101 cannot obtain a signal from an underwater object. In the “descent” state, the element 101 cannot accurately acquire a signal due to a large noise. Therefore, the direction estimation process is not performed in the “floating” and “descent” states.

  Next, the direction estimation unit 202 calculates an inter-element covariance matrix R (f, τ) at each frequency f using Equation 3 in S602. However, γ is a constant from 0 to 1.

  Next, the direction estimation unit 202 calculates a direction histogram by Equation 4 in S603, and ends.

  Where a (f, θ) = [a (f, θ, 1), ..., a (f, θ, N)] represents the steering vector in direction θ at frequency f, and each of a (f, θ) The element a (f, θ, n) is calculated by Equation 5 by simulation based on the element arrangement. However, η (n) is the difference in arrival time of the source signal between the element 1 and the element n, and FreqMax is a frequency corresponding to the frequency number F.

  In this example, the minimum variance beamformer method is used for sound source direction estimation. However, for the sound source direction estimation, the delay sum array method, the minimum dispersion beamformer method, GCC-PHAT method, SRP-PHAT method, MUSIC Method, ESPRIT method, SPIRE method, etc., and any of them may be used.

  As described above, this embodiment is configured to perform the direction estimation process when the state is “landing”. As a result, as shown in FIG. 4, the direction estimation results 410 are obtained discretely.

  FIG. 7 is a flowchart of the trajectory estimation process S305 of the present embodiment.

  Specifically, the trajectory estimation calculates the true sound source direction θ˜ (τ) by executing a Kalman filter using the sound source direction θ (τ) constituting the trajectory hypothesis as an observation signal. For example, the system model and the observation model of the Kalman filter are expressed by Equation 6 and Equation 7, respectively. Here, Δτ represents a time frame interval, R_ξ represents a certain covariance matrix, and σ_ψ represents a certain variance value.

  Here, instead of the Kalman filter, another tracking algorithm such as a particle filter may be used. First, the trajectory estimation unit 203 according to the present exemplary embodiment predicts a trajectory according to Equation 8 in S701.

  Next, the trajectory estimation unit 203 of the present embodiment determines whether or not the own ship state is “landing” in S702. If “landing”, S703 is executed, otherwise S704 is executed. In step S703, the trajectory estimation unit 203 according to the present exemplary embodiment updates the trajectory estimation value according to Equation 9, and ends.

  In step S704, the trajectory estimation unit 203 of the present example stores the predicted value in the estimated trajectory value according to Equation 10, and the process ends.

  As described above, the present embodiment is configured to perform a locus update process when the state is “landing”. On the other hand, when the state is not “landing”, since the direction estimation result y (τ) at time τ cannot be obtained, the predicted value is stored as an estimated trajectory.

  In the present embodiment, the position of the object can be estimated by the above configuration even when the ship must be driven at high speed.

  In the block diagram of FIG. 2, the own ship state estimation result c (τ) is input to both the direction estimation unit 202 and the trajectory estimation unit 203. In the process of FIG. 7, the trajectory estimation unit 203 detects the state based on the own ship state estimation result c (τ) in process S702, but does not use the own ship state estimation result c (τ) in S702. Since the direction estimation result y (τ) from the direction estimation unit 202 is not obtained, it can be known that the state is not “landing”. In this case, it is not necessary to input the own ship state estimation result c (τ) to the trajectory estimation unit 203.

  As another modification, the direction estimation unit 202 always performs direction estimation processing, always inputs the direction estimation result y (τ) to the trajectory estimation unit 203, and the trajectory estimation unit 203 uses the own ship state estimation result c ( Filtering may be performed using τ). In this case, it is not necessary to input the own ship state estimation result c (τ) to the direction estimation unit 202.

  In the above simple example, three states of “landing”, “floating”, and “descent” are illustrated, but it may be further subdivided based on the classification of the probability distribution of the vibration spectrum. It is.

  In the present embodiment, an example of a vibration source detection apparatus that estimates the position of an object even when the ship's ship state and the frequency of state transition are different using a ship drive command will be described. In this embodiment, a state transition model is selected based on a control signal used for ship operation. Vessel drive command changes the state of the vessel, such as output (engine) control command, steering command (rudder drive signal for boats with rudder, nozzle injection direction control signal for water jet propulsion vessels, etc.) Includes general instructions for

  FIG. 8 is an example of the data structure of the own ship state transition model of the present embodiment. A different own ship state transition model u (k) is assigned to each own ship driving command k and stored in the database. Each own ship state transition model u (k) is a hidden Markov model, and has S (k) states s (k, 1),..., S (k, S (k)). The own ship state transition model u (k) includes the probability distribution dist (k, i) of the frequency spectrum of vibration generated by the own ship drive command k in the i-th state s (k, i), and the own ship drive command. k has a transition probability q (k, i, j) for transition from the i-th state s (k, i) to the j-th state s (k, j) in the next time frame. The probability distribution dist (k, i) of the frequency spectrum is, for example, a mixed Gaussian model (GMM), and dist (k, i) is C (k, i) clusters [c (k, i, 1),…, c (k, i, C (k, i))]. The c_l-th cluster c (k, i, c_l) is defined by the spectrum mean μ (k, i, c_l) (Equation 11) and the spectrum covariance matrix Σ (k, i, cl) (Equation 12) .

  Since this embodiment has the own ship state transition model u (k) for each own ship drive command k in this way, the own ship state suitable for the own ship drive instruction k based on the information of the own ship drive instruction k. A transition model u (k) can be selected and used. That is, in this embodiment, the vibration spectrum and transition probability representing the own ship state are determined for each own ship drive command k.

  FIG. 9 is a flowchart at the time of learning the own ship state transition model of the estimation system 100 of the present embodiment. For example, the state transition of the ship is learned by actually maneuvering the ship. The estimation system 100 repeats the processing from S801 onward for all own ship drive commands.

  First, in the own ship drive method presentation process S801, the estimation system 100 uses the user interface unit 104 to present the own ship drive command corresponding to the own ship state transition model to be learned to the user. In response to this, the user is expected to start and continue the proposed ship drive command.

  Next, in the signal acquisition process S802, the signal acquisition unit 201 buffers the multi-channel input signal z (n, t), and buffers the input signal sequence [z (n, t− (M−1) δ). , z (n, t- (M-2) δ), ..., z (n, t-δ), z (n, t)], perform window function multiplication and short-time Fourier transform to obtain multiple channels Output frequency domain signal x (n, τ) = [x (n, τ, 1),..., X (n, τ, F)].

  Next, in the own ship state transition model learning process S803, the own ship state estimating unit 206 learns the own ship state transition model based on the multi-channel frequency domain signal x (n, τ). For learning the own ship state transition model, an algorithm such as a Baum-Welch algorithm, which is a hidden Markov model (HMM) learning method, is used.

Next, the estimation system 100 determines whether learning is completed in S804. Specifically, the probability distribution dist (k, i) and transition probability q (k, i, j) of the frequency spectrum of own ship state transition model u (k) converge, and the amount of change compared to the previous iteration When it becomes small, it is judged that learning is completed. If it is determined that the process has been completed, the processes subsequent to S801 are repeatedly executed for the next own ship drive command. Otherwise, the process returns to the signal acquisition process S802.
In the present embodiment, the position of the object can be estimated even when the own ship state and the frequency of the state transition are different depending on the ship drive command.

  In the present embodiment, an example of a vibration source detection apparatus that estimates the position of an object even when the correlation between the own ship state and the vibration signal is small will be described. For example, detection information of an acceleration sensor and an atmospheric pressure sensor is used.

  FIG. 10 is a diagram illustrating a hardware configuration of the estimation system 100 according to the present embodiment.

  The estimation system 100 of this embodiment includes elements 101_1 to 101_N such as a hydrophone and a microphone, a multi-channel AD converter 102, a central processing unit 103, a user interface unit 104, a storage medium 105, a volatile memory 106, an acceleration sensor 1001, It is composed of an atmospheric pressure sensor 1002.

  The elements 101_1 to 101_N convert sound waves in water or air into analog voltage values and send them to the multi-channel AD converter 102. The multi-channel AD converter 102 converts a multi-channel analog voltage value into a multi-channel digital signal and sends it to the central processing unit 103.

  FIG. 11 is a functional block diagram of the estimation system 100 of the present embodiment.

  As a specific configuration example, the signal processing unit 11000 includes a signal acquisition unit 201, a direction estimation unit 202, a trajectory estimation unit 203, a result output unit 205, and an own ship state estimation unit 206. These functions are realized using the central processing unit 103, the user interface unit 104, the storage medium 105, and the volatile memory 106. Or when the said function is implement | achieved only by hardware, the exclusive circuit structure which implement | achieves the said process is shown.

  The signal acquisition unit 201 receives the multi-channel input signal z (n, t) from the multi-channel AD converter 102 and buffers z (n, t) at every sampling time interval Δsec. However, z (n, t) means an input signal of the element n at time t seconds. Buffered input signal sequence over the past M times [z (n, t- (M-1) δ), z (n, t- (M-2) δ),…, z (n, t-δ ), z (n, t)], a window function multiplication and a short-time Fourier transform, and a multi-channel frequency domain signal x (n, τ) = [x (n, τ, 1), ..., x (n, τ, F)] is calculated and output. Where δ is the sampling time interval, F is the number of frequency numbers of the short-time Fourier transform, τ is the range of the target time of the multiplication of the window function and the short-time Fourier transform (from t- (M-1) δ to t This represents the frame number when (time zone) is regarded as one time frame.

  The own ship state estimation unit 1101 receives the sensor values of the acceleration sensor 1001 and the atmospheric pressure sensor 1002 as input, executes own ship state estimation processing, and outputs the own ship state estimation result c (τ) at time τ. In this case, for example, the acceleration sensor 1001 detects the vertical movement of the hull, and detects a state such as “landing”, “floating”, and “descent”. Further, the pressure sensor 1002 detects a state such as whether or not it is underwater. As in the first and second embodiments, the own ship state estimation unit 1101 estimates an own ship state using an algorithm such as a Viterbi algorithm, which is a hidden Markov model (HMM) state estimation method. The own ship state is obtained, for example, as a symbol of “landing”, “floating”, or “descent”.

  The direction estimation unit 202 receives the multi-channel frequency domain signal x (n, τ) and the own ship state estimation result as input, and executes the direction estimation process, and the direction estimation result y (τ) = [y (τ, 1), ..., y (τ, D)] are output. Where D is the number of discretized direction numbers.

  The trajectory estimation unit 203 inputs the direction estimation results [y (τ- (B-1)), y (τ- (B-2)), ..., y (τ)] at multiple times and the own ship state estimation results The trajectory estimation process is executed, and the trajectory estimation result Traj (τ) = [θ (τ- (B-1)), θ (τ- (B-2)),…, θ ( τ)] is output.

  The result output unit 205 presents Traj (τ) to the user. The presentation method includes, for example, image information presentation via the user interface unit 104, presentation by a braille display, presentation by voice, printing of image information via a printer, and the like.

  In the present embodiment, the own ship state is estimated based on information from the sensors 1001 and 1002, but a vibration spectrum may be used in combination as in the first and second embodiments. If the sensor's information volume is large and the information accuracy is high, it may be possible to estimate the ship's state using only the sensor without using the state transition model, but considering the cost of maintaining the sensor, the state transition There is an advantage of using a model.

  FIG. 12 is an example of the data structure of the own ship state transition model of the present embodiment.

  A different own ship state transition model u (k) is assigned to each own ship driving command k and stored in the database. Each own ship state transition model u (k) is a hidden Markov model, and has S (k) states s (k, 1),..., S (k, S (k)). The own ship state transition model u (k) has an acceleration sensor value in the first dimension and an atmospheric pressure sensor value in the second dimension, which is generated by the own ship drive command k in the i-th state s (k, i). Probability distribution dist (k, i) of vector (sensor vector) and own ship drive command k change from i-th state s (k, i) to j-th state s (k, j) in the next time frame Has transition probability q (k, i, j) to transition. The probability distribution dist (k, i) of the sensor vector is, for example, a mixed Gaussian model (GMM), and dist (k, i) is C (k, i) clusters [c (k, i, 1),…, c (k, i, C (k, i))]. The c_l-th cluster c (k, i, c_l) is defined by the sensor vector mean μ (k, i, c_l) (Equation 13) and the sensor vector covariance matrix Σ (k, i, cl) (Equation 14) Is done.

  Since the present embodiment performs the transition of the own ship state transition using the sensor input such as the acceleration and the atmospheric pressure different from the vibration input as described above, even if the correlation between the own ship state and the vibration signal is small, The position of the object can be estimated. In the example shown in FIG. 12, the sensor information indicating the own ship state and the transition probability are determined for each own ship driving command k.

  In addition, this invention is not limited to an above-described Example, Various modifications are included. For example, the above-described embodiments have been described in detail for easy understanding of the present invention, and are not necessarily limited to those having all the configurations described. Further, a part of the configuration of one embodiment can be replaced with the configuration of another embodiment, and the configuration of another embodiment can be added to the configuration of one embodiment. Further, it is possible to add, delete, and replace other configurations for a part of the configuration of each embodiment.

  Each of the above-described configurations, functions, processing units, processing means, and the like may be realized by hardware by designing a part or all of them with, for example, an integrated circuit. Each of the above-described configurations, functions, and the like may be realized by software by interpreting and executing a program that realizes each function by the processor. Information such as programs, tables, and files for realizing each function can be stored in a recording device such as a memory, a hard disk, or an SSD (Solid State Drive), or a recording medium such as an IC card, an SD card, or a DVD.

  Further, the control lines and information lines indicate what is considered necessary for the explanation, and not all the control lines and information lines on the product are necessarily shown. Actually, it may be considered that almost all the components are connected to each other.

100: estimation system, 101: element, 102: multi-channel AD converter, 103: central processing unit, 104: user interface unit, 105: storage medium, 106: volatile memory

Claims (15)

A detection system for a ship,
A detection unit that detects vibration from an object in water and obtains vibration information;
A determination unit for determining the state of the ship;
An estimation unit that obtains information on the object using the vibration information at a timing at which the state of the ship is determined by the determination unit to be able to estimate information on the object based on the vibration information;
Having a detection system. The timing at which the state of the ship is determined to be able to estimate the information on the object based on the vibration information includes at least the timing at which the detection unit is in water and the vibration information can be detected.
The detection system according to claim 1. The determination unit
Using a model representing the signal from the detector and the state transition of the ship,
Determining the state of the ship;
The detection system according to claim 1. The determination unit
Using information obtained by sensors provided in the ship and a model representing the state transition of the ship,
Determining the state of the ship;
The detection system according to claim 1. The determination unit
For each pattern of the ship drive command, select a model representing the state transition of the ship, and using the selected model,
Determining the state of the ship;
The detection system according to claim 1. A vibration source detection device that is provided in a movable body that can move in water or on water, and detects the direction of the object by detecting vibration from the object in water,
A detection unit for detecting vibration from the object and obtaining vibration information;
A determination unit that determines whether or not the state of the moving body is a direction estimation possible state capable of estimating the direction of the object;
Of the vibration information obtained by the detection unit, an estimation unit that estimates the direction of the object using vibration information detected in the state in which the direction can be estimated;
A vibration source detection device having: The vibration source detection device according to claim 6,
Whether the determination unit is in a state in which the direction can be estimated using a signal from at least one of the detection unit and another sensor provided in the moving body and a model representing a time transition of the state of the moving body. A vibration source detection device characterized by determining whether or not. The vibration source detection device according to claim 7,
The determination unit selects the model in accordance with a driving command for the moving body. The vibration source detection device according to claim 8,
A drive command presentation unit for presenting a ship drive command to the user;
The vibration source detection apparatus, wherein the determination unit learns the model corresponding to the driving command to be presented. The vibration source detection device according to claim 6,
The vibration source detection apparatus which has a result output part which displays the estimated direction of the said target object and the said direction estimation possible state simultaneously using the same time axis. The vibration source detection device according to claim 6,
The determination unit determines whether or not the direction can be estimated using at least one of acceleration and atmospheric pressure information. In a ship moving on the water, using a detection unit that detects vibration from an object in water and obtains vibration information, a detection method for estimating information on the object based on the vibration information,
A detection method characterized in that the information on the object is estimated by eliminating the vibration information at a timing when the detection unit ascends from the water surface and a timing when the detection unit contacts the water surface.   13. The detection method according to claim 12, wherein a signal from the detection unit is used to determine the timing, or a state transition model of the ship is used in addition thereto.   13. The detection method according to claim 12, wherein a signal from at least one of an acceleration sensor and a pressure sensor is used to determine the timing, or a state transition model of the ship is used in addition thereto.   13. The detection method according to claim 12, wherein a state transition model of the ship prepared for each of the ship drive command and the ship drive command pattern is used to determine the timing.

Images