JP2006224836A - Extraction method for alternating current underwater electric field signal or alternating current magnetic signal accompanying with vessel and device - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a signal processing method and a device extracting an alternating current underwater electric field signal or an alternating current magnetic signal, i.e., a detection signal buried in noise regarding detection of a vessel. <P>SOLUTION: The alternating current underwater electric field signal is taken-in from an electric field sensor or the alternating current magnetic signal is taken-in from a magnetic sensor at a predetermined sampling frequency and frequency band area division is performed by discrete wavelet conversion DSP 45 to separate the detection signal and the colored noise. Thereafter, the detection signal is extracted by removing the white noise by DSP47 for noise canceling part. Further, since a noise canceler body 47b is enlarged to an unsteady signal such as the detection signal, it includes Kalman filter making filter coefficient of an FIR filter as a state parameter, successively adjusts the filter coefficient to the optimum value in every data point and extracts the signal buried in the white noise. <P>COPYRIGHT: (C)2006,JPO&NCIPI

Description

  The present invention relates to a signal processing technique for extracting an AC underwater electric field signal or an AC magnetic signal associated with a ship, which is a detection signal buried in noise.

  In a ship that is navigating, current flows in water due to corrosion and corrosion prevention of the hull, and an electromagnetic field is generated. This electromagnetic field has the property of changing periodically (very low frequency) by the rotation of the propeller of the ship.

  Conventionally, in the present applicant, as a detection method based on detection of an AC underwater electric field signal or an AC magnetic signal caused by corrosion or corrosion prevention of the navigating ship, discrete measurement data obtained from an electric field or magnetic field sensor system is used. Thus, a method of removing noise by performing autocorrelation processing and performing frequency analysis using Fourier analysis has been adopted.

  Some noise cancellers for removing white noise use an adaptive filter. Under the assumption that the adaptive filter is stationary for both signal and noise, the filter coefficients are self-adjusted little by little so as to minimize the mean square error with the signal (desired signal) to be extracted from the observed signal as input. It is a filter that outputs an optimum estimated value. An algorithm for obtaining the optimum filter coefficient is called an adaptive algorithm, and typical examples include an LMS algorithm and an SDM algorithm. As an example of a noise canceller using the adaptive filter, there is a so-called adaptive line enhancer that uses a sine wave disturbed by white noise as an observation signal and extracts an original signal, in this case, a sine wave from the observation signal. FIG. 5 shows a configuration diagram of the adaptive line enhancer.

The adaptive line enhancer of FIG. 5 includes an FIR (finite impulse response) filter 30 and an adaptive algorithm (LMS algorithm or the like) 31. The FIR filter 30 includes N (N: integer) delay operators 30a and N filters. It has a coefficient part 30b. y _k−1 , y _k−2 ,..., y _k−N are the output signals of the first to N-th delay operators 30a at discrete time k, h _{1, k} , h _{2, k} ,. h _{N, k} is a filter coefficient of the filter coefficient unit 30b in the first to _Nth stages at the discrete time k. Filter coefficients _{h 1, k, h 2,} k, ..., h N, k is varied in a predetermined adaptive algorithm according to the error _{e k} of the input signal _{y k} and an output z ^ _k of the FIR filter 30.

In this case, for example, when the sine wave signal z _k disturbed by the white noise v _k is used as the input signal y _k , the original signal (sine wave) in which the white noise is suppressed is extracted from the input signal as the output z ^ _k. can do.

  However, depending on the above prior art, colored noise having autocorrelation cannot be removed by autocorrelation processing, and detection signals (such as AC underwater electric field signals or AC magnetic signals caused by ship corrosion and corrosion prevention) In the time-integrated Fourier analysis, if the time interval in which the detection signal exists is short with respect to the integration time interval, the frequency spectrum of the detection signal cannot always be extracted from the unsteady signal that exists locally in time. In addition, in the noise canceller using the conventional adaptive line enhancer as shown in FIG. 5, the signal and the noise are assumed to be stationary, and the AC underwater electric field signal or AC magnetic signal that exists locally in time like the detection signal is used. The power spectrum of the observation signal changes with time, and it cannot be said that the optimum filter coefficient is always obtained by the adaptive algorithm.

  Then, this invention aims at providing the alternating current underwater electric field signal or alternating current magnetic signal extraction method accompanying the ship which can extract the alternating current underwater electric field signal or alternating current magnetic signal which is a non-stationary signal in broadband noise effectively. To do.

  Other objects and novel features of the present invention will be clarified in embodiments described later.

In order to achieve the above object, an AC underwater electric field signal or an AC magnetic signal extraction method associated with a ship according to the present invention performs detection of a ship. When extracting AC magnetic signal as detection signal,
Color signal under measurement environment and frequency band where the detection signal exists by taking AC underwater electric field signal from electric field sensor at predetermined sampling frequency or AC magnetic signal from magnetic sensor and dividing frequency band by discrete wavelet transform The frequency band of noise is separated, and further, the detection signal is extracted by applying a noise canceller, which is a digital filter for removing white noise, to the discrete measurement data after the frequency band division. .

  In the AC underwater electric field signal or AC magnetic signal extraction method, the noise canceller extracts unsteady signals that are locally present in the time, such as the detection signal, in order to extract unsteady signals locally. It is preferable that the optimum filter coefficient is sequentially obtained for each point, and a desired detection signal buried in white noise is output by the FIR filter constituted by the optimum filter coefficient.

  The noise canceller may preliminarily apply white noise removal target data to an adaptive filter, obtain a filter coefficient once by a predetermined adaptive algorithm, and normalize the filter coefficient as an initial value of the filter coefficient that is the state variable. .

An AC underwater electric field signal or an AC magnetic signal extraction device according to the present invention is configured to extract an AC underwater electric field signal or an AC magnetic signal resulting from corrosion or corrosion prevention of a ship underwater as a detection signal in order to detect a ship. ,
An electric field sensor for detecting an AC underwater electric field signal or a magnetic sensor for detecting an AC magnetic signal;
AC signal underwater from the electric field sensor at a predetermined sampling frequency, or AC magnetic signal from the magnetic sensor, frequency band division by discrete wavelet transform and color in the measurement environment where the detection signal exists Discrete wavelet transform means for separating the frequency band of noise;
A noise canceller, which is a digital filter for extracting the detection signal by removing white noise from the discrete measurement data after the frequency band division, is provided.

  In the AC underwater electric field signal or AC magnetic signal extraction device, the noise canceller includes an FIR filter to which the discrete measurement data after the frequency band division is input, and an optimum filter for each data point using a filter coefficient of the FIR filter as a state variable. And a Kalman filter that sequentially obtains the coefficients, and may be configured to output a desired detection signal buried in white noise.

  In the AC underwater electric field signal or AC magnetic signal extraction device, the noise canceller previously applies white noise removal target data to an adaptive filter, obtains a filter coefficient once by a predetermined adaptive algorithm, and normalizes the filter coefficient, The initial value of the filter coefficient that is the state variable may be used.

  ADVANTAGE OF THE INVENTION According to this invention, the alternating current underwater electric field signal or alternating current magnetic signal accompanying the ship which is a nonstationary signal can be extracted for ship detection from broadband noise.

  DESCRIPTION OF THE PREFERRED EMBODIMENTS Embodiments of an AC underwater electric field signal or AC magnetic signal extraction method and apparatus associated with a ship will be described below with reference to the drawings as the best mode for carrying out the present invention.

  1 is a schematic configuration diagram of a signal extraction apparatus according to an embodiment of the present invention, FIG. 2 is an explanatory diagram of the signal extraction procedure, FIG. 3 is a configuration diagram of a noise canceller main body 47b of FIG. 1, and FIG. It is a flowchart including the flow of internal data (filter coefficient etc.).

  As shown in FIG. 1, the signal extraction apparatus includes an electric field or magnetic sensor 41, an amplifier 42 that amplifies a detection output signal of the sensor 41, an AD converter 43, and a memory (data buffer) as a memory (data buffer). RAM) 44, a discrete wavelet transform DSP (Digital Signal Processor) 45 as a discrete wavelet transform means, a memory (RAM) 46 as a data buffer after band division by the discrete wavelet transform DSP 45, and a noise canceller. A DSP (Digital Signal Processor) 47 for noise cancellation unit as a (Noise Canceller) and a display 48 for processing waveform display are provided.

  The noise canceling unit DSP 47 includes an adaptive filter 47a as an adaptive line enhancer and a noise canceller body 47b including an FIR filter and a Kalman filter as digital filters for removing noise. The noise canceller main body 47b incorporates the Kalman filter having the filter coefficient of the FIR filter as a state variable in order to extend the application range for a non-stationary signal such as a detection signal. In addition, the Kalman filter requires an initial value of the state variable. The adaptive filter 47a is provided for setting an initial value of each filter coefficient (Kalman filter state variable) of the FIR filter in the noise canceller main body 47b. Details of the noise canceller main body 47b will be described later with reference to FIG.

  The electric field or magnetic sensor 41 detects an AC underwater electric field signal or an AC magnetic signal as a detection signal due to corrosion or corrosion prevention of the ship in water in order to detect a navigating ship. In the case of a sensor, an AC underwater electric field signal is detected, and in the case of a magnetic sensor, an AC magnetic signal is detected. This detection output signal is amplified by the amplifier 42. The AD converter 43 takes in the sensor output signal (analog signal) amplified by the amplifier 42 at a predetermined sampling frequency and outputs it as a digital signal. The measured data of the digital signal is stored in the memory 44. This is the discrete measurement data 10 of the AC underwater electric field signal or AC magnetic signal of FIG. 2 captured by the sensors 41 to AD converter 43 at a predetermined sampling frequency, and the frequency band by the discrete wavelet transform 11 by the discrete wavelet transform DSP 45. Is divided into octave divisions, and becomes discrete data 12 after band division.

  This discrete wavelet transform 11 can separate the detection signal and the colored noise. For example, when the colored noise has a different frequency band from the target detection signal, the frequency band can be removed. If the sampling frequency is 20 Hz, Level-1 is a band from 5 Hz to 10 Hz, Level-2 is a band from 2.5 Hz to 5 Hz, and Level-3 is from 1.25 Hz to 2.5 Hz. It is divided like a band.

  Next, in FIG. 2, the discrete data 12 after each band division is input to the noise canceling unit 13 (the noise canceling unit DSP 47 in FIG. 1), and white noise is removed by the noise canceling unit 13 in each band.

  As shown in FIG. 1, the noise canceling unit DSP 47 includes a noise canceller main body 47b and an adaptive filter 47a for setting an initial value thereof. First, the adaptive filter 47a is applied to the band-divided discrete data 12 which is white noise removal target data, and a filter coefficient calculated by an adaptive algorithm (LMS algorithm or the like) in the data is obtained. Then, an initial value of the state variable of the Kalman filter used in the noise canceller main body 47b (that is, the initial value of the filter coefficient of the FIR filter 20 in FIG. 3 described later) is obtained by scaling (normalizing) the filter coefficient by the L1 norm. Then, the discrete data 12 after band division is applied to the noise canceller main body 47b again to obtain discrete data (AC signal extraction) 14 after white noise removal in each band. Moreover, the adaptive filter 47a used here uses an adaptive line enhancer as a prior art shown in FIG. FIG. 4 shows a flowchart including the flow of internal data (filter coefficients and the like) of the signal extraction procedure described above.

  FIG. 3 is a configuration diagram of the noise canceller main body 47b in FIG. In the noise canceller main body 47b, the FIR filter 20 is a digital filter that removes noise, and includes N (N: integer) delay operators 20a and N filter coefficient units 20b. The filter coefficient values of these filter coefficient units 20b are sequentially adjusted to optimum values for each observation data by the Kalman filter 21. In order to apply the Kalman filter 21 in this way, it is necessary to construct a state model composed of two equations called a state equation and an observation equation.

First, the observed value (input signal) y _{k at} the discrete time k is expressed as the following equation (1).

y _k = z _k + v _k (1)
Here, z _k is an original signal (in other words, detection signal: AC underwater electric field signal or AC magnetic signal), and v _k is broadband noise (white noise).

Furthermore, the observed values _{y k-1, y k-} 2, ..., from _{y k-N,} becomes a material obtained by extracting the _{z k} as When z ^ _k equation (2) by the FIR filter 20, (1 ) Is expressed by the following equation (3) using equation (2).

但し、ｈ_１,ｋ，ｈ_２,ｋ，…，ｈ_Ｎ,ｋは離散時刻ｋでの１段目〜Ｎ段目のフィルタ係数部３０ｂのフィルタ係数である。

Here, h _{1, k} , h _{2, k} ,..., H _{N, k} are filter coefficients of the first to _Nth stage filter coefficient units 30b at the discrete time k.

Next, Equation (3) is expressed by a state space model composed of a state equation and an observation equation. The purpose of the application of the Kalman filter of the present invention is to estimate the optimum filter coefficient for extracting the original signal by removing the white noise, so that h _{i, k} is the state variable X _k (vector quantity) and the state equation is A state space model such as the following equation (5) is constructed as the following equation (4) and the observation equation.

The Kalman filter 21 is applied to this state space model, and the original signal data at the time k is extracted by the equation (2) using the estimated state variable _Xk . The extracted original signal data (processed waveform) is displayed on the display 48.

  In this manner, the noise canceling unit DSP 47 uses the Kalman filter 21 with the filter coefficient of the FIR filter 20 as a state variable in order to extract a non-stationary signal that exists locally in the time buried in the noise such as the detection signal. An optimum filter coefficient is sequentially obtained for each data point (every discrete time k), and a desired detection signal buried in white noise can be output by the FIR filter 20 configured by the optimum filter coefficient.

  According to this embodiment, the following effects can be obtained.

(1) By taking an AC underwater electric field signal from an electric field sensor at a predetermined sampling frequency or an AC magnetic signal from a magnetic sensor and dividing the frequency band by a discrete wavelet transform 11, the frequency band where the detection signal exists and the measurement environment It is possible to remove the colored noise by separating the frequency band of the colored noise.

(2) Extracting the detection signal buried in the white noise by applying a noise canceling unit DSP 47, which is a digital filter for white noise removal, to the discrete measurement data after the frequency band division. Can do.

(3) The noise canceling unit DSP 47 sequentially obtains the optimum filter coefficient for each data point by the Kalman filter 21 using the filter coefficient of the FIR filter 20 as a state variable, and the FIR filter 20 configured by the optimum filter coefficient It is possible to effectively suppress noise and extract a nonstationary signal that exists locally in time, such as the detection signal, buried in the noise.

  In the noise canceling unit DSP 47 of FIG. 3, the adaptive filter 47a is applied to obtain filter coefficients calculated by the adaptive algorithm in the data, and these filter coefficients are L1 norm (sum of absolute values of filter coefficient values). The initial value of the state variable of the Kalman filter used in the noise canceller main body 47b is scaled (normalized) by the above, but may be normalized by a predetermined value other than the L1 norm.

  Although the embodiments of the present invention have been described above, it will be obvious to those skilled in the art that the present invention is not limited to these embodiments, and various modifications and changes can be made within the scope of the claims.

It is a schematic block diagram which shows embodiment of the alternating current underwater electric field signal or alternating current magnetic signal extraction method and apparatus accompanying the ship which concerns on this invention. It is explanatory drawing of the signal extraction procedure shown to embodiment of this invention. It is a block diagram of the noise canceller main body in the noise cancellation part of FIG. It is a flowchart including the flow of the internal data of the signal extraction procedure shown in the embodiment. It is a block diagram of the adaptive line enhancer of a prior art.

Explanation of symbols

DESCRIPTION OF SYMBOLS 10 Discrete measurement data 11 Discrete wavelet transform 12 Discrete data after band division 13 Noise cancellation part 14 Discrete data after white noise removal 20, 30 FIR filter 20a, 30a Delay operator 20b, 30b Filter coefficient part 21 Kalman filter 41 Electric field sensor or magnetic sensor 42 Amplifier 44 Memory (Measurement data buffer)
45 DSP for discrete wavelet transform
46 memory (buffer for data after band division)
47 DSP for noise canceling part
47a Adaptive filter (Adaptive line enhancer)
47b Noise canceller body 48 Display

Claims (6)

In order to detect a ship, an AC underwater electric field signal or an AC magnetic signal extraction method associated with a ship that extracts an AC underwater electric field signal or an AC magnetic signal caused by corrosion or corrosion prevention of the ship underwater as a detection signal,
Color signal under measurement environment and frequency band where the detection signal exists by taking AC underwater electric field signal from electric field sensor at predetermined sampling frequency or AC magnetic signal from magnetic sensor and dividing frequency band by discrete wavelet transform The frequency band of noise is separated, and the detection signal is extracted by applying a noise canceller, which is a digital filter for removing white noise, to the discrete measurement data after the frequency band division. AC underwater electric field signal or AC magnetic signal extraction method for ships.   The noise canceller sequentially extracts the optimum filter coefficient for each data point by a Kalman filter using the filter coefficient as a state variable in order to extract an unsteady signal that exists locally in the time, such as the detection signal. 2. The method of extracting an AC underwater electric field signal or AC magnetic signal associated with a ship according to claim 1, wherein a desired detection signal buried in white noise is output by an FIR filter constituted by coefficients.   The noise canceller preliminarily applies the white noise removal target data to an adaptive filter, obtains a filter coefficient by a predetermined adaptive algorithm, and normalizes the filter coefficient as an initial value of the filter coefficient that is the state variable. An AC underwater electric field signal or AC magnetic signal extraction method for a ship according to claim 2. In order to detect a ship, an AC underwater electric field signal or an AC magnetic signal extraction device associated with a ship that extracts an AC underwater electric field signal or an AC magnetic signal caused by corrosion or corrosion prevention of the ship underwater as a detection signal,
An electric field sensor for detecting an AC underwater electric field signal or a magnetic sensor for detecting an AC magnetic signal;
AC signal underwater from the electric field sensor at a predetermined sampling frequency, or AC magnetic signal from the magnetic sensor, frequency band division by discrete wavelet transform and color in the measurement environment where the detection signal exists Discrete wavelet transform means for separating the frequency band of noise;
An AC underwater electric field signal or AC magnetic signal extraction apparatus for a ship, comprising: a noise canceller, which is a digital filter that removes white noise from the discrete measurement data after frequency band division and extracts the detection signal.   The noise canceller includes an FIR filter to which discrete measurement data after the frequency band division is input, and a Kalman filter that sequentially obtains an optimum filter coefficient for each data point using the filter coefficient of the FIR filter as a state variable. 5. A device for extracting an AC underwater electric field signal or an AC magnetic signal accompanying a ship according to claim 4, wherein a desired detection signal buried in noise is output.   The noise canceller preliminarily applies the white noise removal target data to an adaptive filter, obtains a filter coefficient once by a predetermined adaptive algorithm, and normalizes the filter coefficient as an initial value of the filter coefficient that is the state variable. 6. An AC underwater electric field signal or AC magnetic signal extraction device for a ship according to claim 5.

Images