JP4269149B2 - Passive sonar simulation signal generation method and apparatus - Google Patents

Description

[0001]
BACKGROUND OF THE INVENTION
The present invention relates to a sonar simulation signal generator for generating simulation data used for performance evaluation, operation training, and the like of a passive sonar.
[0002]
[Prior art]
Conventionally, as a technique in this field, a method of simulating a signal by preparing a filter having a required spectral level characteristic and passing white noise therethrough is known. Generally, examples of sonar simulation targets include narrow-band narrowband signals, broadband broadband signals, and underwater background noise, but this method can be applied to any simulation (see Patent Documents 1 and 2).
[0003]
[Patent Document 1]
Japanese Patent Laid-Open No. 02-278177 [Patent Document 2]
Japanese Patent Laid-Open No. 01-305386
For example, in Patent Document 1, as a broadband signal simulation, a spectrum distribution of a broadband signal to be simulated is prepared, and a transmission characteristic equivalent to this is realized using a variable characteristic filter, and a simulated signal is transmitted by passing white noise through this. Is generated.
[0005]
This method can be applied not only to a broadband signal but also to a simulation of a narrowband signal and a background noise. If a filter having a spectral level characteristic of each signal or noise is realized, it is necessary to pass a white noise through the filter. Simulated data can be obtained.
[0006]
In addition, when individually generating simulated data of signals and noise, for example, as in Patent Document 2, it is possible to obtain simulated data on which a plurality of signals are superimposed by linearly adding the simulated data.
[0007]
FIG. 8 is a diagram showing a configuration of a conventional passive sonar simulation signal generator. Spectral level characteristic unit 1a that outputs spectral level characteristic information, white noise generation unit 2a, spectral shape filter unit 3a, passive sonar simulation data unit 4a, passive sonar signal processing unit 5a, lofagram 6a, The passive sonar signal processor 5a includes an FFT (Fast Fourier Transform) unit 51a, a normalization processing unit 52a, and a gradation conversion unit 53a.
[0008]
Let S (f) be the spectral level characteristic unique to the ship to be simulated. Here, the spectral level characteristic S (f) represents the narrow band signal, the broadband signal, and the background noise as one spectral level characteristic. A spectrum shape filter F (f) having the same pass characteristic as S (f) is constituted by the output S (f) of the spectrum level characteristic unit 1a, and white noise n (t) generated in the white noise unit 2a is added thereto. The output d (t) is stored as simulated data r (t) in the simulated data section 4a.
[0009]
The simulated data r (t) is subjected to signal processing by the passive sonar signal processing unit 5a, whereby time frequency analysis data (lofagram) is obtained. The passive sonar signal processing unit 5a performs high-speed Fourier transform on the simulated data r (t) by the FFT unit 51a, normalizes the converted data by the normalization processing unit 52a, and converts the normalized signal by the gradation conversion unit 53a. Tone conversion is performed to generate a lofagram 6a corresponding to the spectral level characteristic S (f).
[0010]
[Problems to be solved by the invention]
The conventional passive sonar simulation signal generator is configured to generate a simulation signal from one normally distributed random number and filter as white noise, and a single random number is used for signal and noise variations in the frequency domain. As a result, there is a problem in that the frequency distribution and line width variation cannot be set for each of a plurality of signals included in the same data, and the simulation accuracy is insufficient. is there.
[0011]
In addition, when using simulated signal generation based on normally distributed random numbers, variations in signal and noise in the frequency domain are expressed by Fourier transform of normally distributed random numbers, so the degree of variation is statistically constant and depends on the signal. It cannot be set individually, and cannot cope with various level dispersions and line width variations observed in actual signals, and there is a problem that the simulation accuracy is insufficient.
[0012]
As described above, in the conventional passive sonar simulation signal generator, normally distributed random numbers that are considered to be close to actual background noise are used as white noise, but in actual data, the signal and noise have various variations, and normal In many cases, it differs from lofagrams generated from distributed random numbers. That is, variations in the level and line width of the narrowband signal often differ from one signal to another, and the influence of the appearance on the observer is large, so a more faithful simulation is required.
[0013]
(the purpose)
An object of the present invention is to provide a passive sonar simulation signal generation method and apparatus capable of improving the performance of simulation accuracy and generating a passive sonar simulation signal closer to actual data.
Another object of the present invention is to improve the simulation accuracy of a passive sonar simulation signal and to provide a simulated sonar image closer to actual data.
[0014]
[Means for Solving the Problems]
The sonar simulation signal generation method of the present invention is a sonar simulation signal generation method for generating simulation data of a sonar, and for each signal and noise to be simulated, simulation data is generated using time series random numbers having different probability density distributions. It is characterized in that simulation data is obtained by generating and adding these, or level dispersion and line width variation in the frequency domain are set and simulated by a probability density distribution in the time domain.
[0015]
The sonar simulation signal generator of the present invention is a sonar simulation signal generator for generating sonar simulation data, and generates time series random numbers having different probability density distributions for each signal and noise to be simulated. A simulation data generation unit (for example, 101 to 401, 102 to 402, 103 to 403, and 104 to 404 in FIG. 1) that generates simulation data for each unit and the output of the simulation data generation unit are added to output simulation data. A signal synthesizer (for example, 50 in FIG. 1), or generate time series random numbers with different probability density distribution for each signal or noise to be simulated, and generate simulated data for each signal or noise Simulation data generation unit (for example, 101-401, 102-402, 103-403, 104-404 in FIG. 1), and the level of the simulation data in the frequency domain The variation of dispersion and line width, set by the probability density distribution in the time domain, characterized by simulated.
[0016]
More specifically, a sonar simulation signal generator that generates simulated data of a sonar, and a plurality of random number generators that generate time series random numbers of different probability density distributions for each signal and noise to be simulated (For example, 21 to 41 in FIG. 7), and a plurality of filters (for example, 26 to 46 and 27 to 47 in FIG. 7) having the output of the plurality of random number generation units as input and having pass characteristics of respective average spectral shapes, A signal synthesizer (for example, 5 in FIG. 7) for synthesizing simulated data by synthesizing the outputs of the plurality of filters, and further, level dispersion of the time series random number, signal center frequency, It has a parameter table (for example, 1 in FIG. 7) that holds setting information of the line width and the SN ratio, and the random number generator generates time series disturbances by the level dispersion set in the parameter table. Set the level dispersion, and sets the pass characteristics of the filter by the central frequency set and the line width and the SN ratio in the parameter table.
[0017]
Also, a sonar simulation signal generator for generating sonar simulation data, which includes a plurality of random number generators (for example, FIG. 2) that generate time series random numbers having different probability density distributions for each signal and noise to be simulated. 21 to 41), outputs of the plurality of random number generation units, and a plurality of filters (for example, 22 to 42 and 24 to 44 in FIG. 2) having pass characteristics of respective average spectral shapes, The output of the filter is input, and the combined gain multiplier (for example, 23 to 43 and 25 to 45 in FIG. 2) that multiplies and outputs the combined gain, and the output of the combined gain multiplier is combined to generate simulated data. A signal synthesizing section (for example, 5 in FIG. 2) for synthesizing, and further, setting information on the level dispersion of the time series random number, the center frequency and line width of the signal, and the SN ratio are maintained. A parameter table (for example, 1 in FIG. 2) for setting the time series random number level dispersion of the random number generator by the level dispersion set in the parameter table, and the center frequency and line set in the parameter table A pass characteristic of the filter is set according to a width, and the combined gain is set according to an S / N ratio set in the parameter table.
[0018]
FIG. 1 is a diagram showing the principle of a specific example of the present invention. Simulated target data of this example is composed of a total of n-1 pieces of narrowband and broadband signals S ₁ to S _n-1 and the background noise simulated data S _n.
A filter (spectral shape filter) 102 having a spectrum shape of the signal S ₁ is prepared, and a simulated signal 1 data is generated by passing a time series random number 1 for simulating the signal S ₁ generated by the random number generation unit 101. To do. Filter 202 also having spectral shape of the signal S _2, by passing the time series random number 2 for the simulation of the signal S ₂ to the random number generator 201 generates, and generates a signal 2 simulated data. Similarly, for the total n−1 signals, a total of n−1 simulated data is generated using the filter 302 and the random number generation unit 301. Also like the signal it is also the background noise S _n, to generate a background noise simulated data using the filter 402, the random number generation unit 401.
[0019]
The signal _{S 1 ~S n-1} simulated data of the generated simulated data, the synthetic gain _{G 1 ~G n-1} in Synthesis gain calculation unit 103, 203, 303 corresponding to the average SN ratio of the respective signals After calculating and multiplying by the multiplication units 104 to 304, the signal synthesis unit 50 performs linear addition. With respect to the background noise simulated data S _n, the synthesized gain G _n of the base of the SN ratio of each signal is calculated by combining the gain calculation unit 403 is multiplied by the multiplication section 404 similarly adds the signal combining unit 50. Random number generators 101 to 401 for simulating signals select and generate time-series random numbers having a dispersion characteristic that can most accurately reproduce the level dispersion and line width dispersion of the signal in the frequency domain.
[0020]
(Function)
In general, the appearance of signals on the lofagram, such as level dispersion and line width variation, is determined by the dispersion of the result of Fourier transform of a time series random number. When the normal distribution is used for the time series random number, the level dispersion on the lofagram follows the Rayleigh distribution. Therefore, when the average power P of the signal is given, the level dispersion Σ ² in the frequency domain is
Σ ² = {(4 / π) −1} P
And is uniquely determined. The variation in the line width, so is represented as level variance of the noise through a spectrum shape filter, likewise proportional to the sigma ^2. Therefore, when normally distributed random numbers are used in the simulation of a plurality of signals, the simulation results for signals with the same SN ratio and line width are statistically the same level dispersion and line width variation.
On the other hand, in the present invention, a simulated signal is generated using a different time series random number for each signal. As a result, simulated data is generated using random numbers having individual probability density distributions as time-series random numbers corresponding to the signal Si having the level variance Σ _i ² , so that the characteristics of the spectral shape filter itself are constant. However, the level dispersion in the frequency domain and the variation in line width are different for each signal.
[0021]
DETAILED DESCRIPTION OF THE INVENTION
(Description of configuration)
Next, embodiments of the method and apparatus for generating a passive sonar simulation signal of the present invention will be described in detail with reference to the drawings.
[0022]
FIG. 2 is a block diagram showing this embodiment. Referring to FIG. 2, the embodiment of the present invention generates simulation parameters 2, 3, and 4 in parameter table 1 for setting and storing parameters to be simulated, and time-series random numbers that are sources of signals and noise. Random number generators 21, 31, 41, filter generators 22, 32, 42 that generate (configure) spectral shape filters, synthetic gain calculators 23, 33, 43 that calculate a composite gain, and multipliers 24, 34 , 44 and 25, 35, 45, a signal synthesizer 5 for synthesizing simulated data, and a simulated data storage 6 for storing simulated data.
[0023]
The signal of this embodiment is composed of _n-1 signals S _{1 to} S _n−1 , and a total of n simulation parameters are obtained by adding a background noise parameter thereto. In the parameter table 1, as signal simulation parameters, the S / N ratio corresponding to the signal strength, the center frequency corresponding to the center of the line and the line width corresponding to the thickness, and the level dispersion information corresponding to the variation between them are respectively shown. Is set. As background noise simulation parameters, level frequency characteristics over the entire band to be simulated and level dispersion information corresponding to level variations are set.
[0024]
The random number generation units 21, 31, 41, the filter generation units 22, 32, 42, and the combined gain generation units 23, 33, 43 are the level dispersion and center of each of the simulation parameters 2, 3, 4 of the setting information of the parameter table 1 It is set by the line width variation and SN ratio corresponding to the frequency and thickness. Further, the outputs of the random number generation units 21, 31, 41 are multiplied by the spectral shape filters (characteristics) of the filter generation units 22, 32, 42 in the multiplication units 24, 34, 44. The outputs to which the filters are applied by the multipliers 24, 34, and 44 are multiplied by the combined gains by the multipliers 25, 35, and 45. The outputs of the multipliers 25, 35, and 45 are combined with a combined gain that achieves a predetermined S / N ratio in the signal combining unit 5 and accumulated in the simulated data unit 6 as final simulated data.
[0025]
The functions of the random number generators 21, 31, 41 and the filter generators 22, 32, 42 relating to the generation of simulated data of each signal and background noise will be described below.
The simulation data generation of the signal S ₁ is performed by the simulation parameter 2, the random number 1 generation unit 21, the filter 1 generation unit 22, the G ₁ calculation unit 23, and the multiplication units 24 and 25. The random number 1 generation unit 21 generates a random number of probability density distribution according to the setting of the level dispersion of the simulation parameter 2. As an embodiment of the level dispersion setting, the random number generation unit generates the following discrete time series r ₁ (t _i , d) as N stages of setting values d = 1, 2,.
[0026]
r ₁ (t _i , d) = sign (p (t _i )) * | p (t _i ) | ^ad
p (t _i ) = N (0, σ ² ), a> 0
p (t _i ): normal distribution with an average of 0 and variance σ ² , sign: sign, a: constant filter 1 generator 22 has a spectrum of signal S ₁ according to the setting of the center frequency and line width of simulation parameter 2 The shape SPL _S1 (f _i ) is calculated, and an FIR filter having pass characteristics equivalent to this is generated (configured), and applied (multiplied) to r ₁ (t _i , d). As an embodiment for calculating the spectrum shape, the filter shape is approximated to the normal distribution shape as shown in FIG. 3 and given as follows.
SPL _S1 (f _i ) = exp [− (f _i −f _c1 ) ² / 2h ₁ ² ]
The same applies to f _c1 : center frequency of the signal S ₁ and h ₁ : generation of simulated data of the line width signals S _{2 to} S _n−1 of the signal S ₁ .
[0027]
Next, simulation data generation of background noise is performed by the simulation parameter 4, the random number n generation unit 41, the background noise filter generation unit 42, the G _n generation unit 43, and the multiplication units 44 and 45.
The random number n generation unit 41 generates a random number of probability density distribution according to the setting of the level dispersion of the simulation parameter 4. Setting and a random number generator level dispersion, the same configuration as the simulated data generated signals S _1, discrete time series r _{n (t} i, _d) for the background noise to generate.
[0028]
Frequency characteristics of the simulated parameter 4 is discretely represent the spectral levels characteristic of the background noise to be set to a typical value of _SPL AN point M _{(f i) (i = 1,2} ... M). The background noise filter generation unit generates (configures) an FIR filter having a pass characteristic equivalent to that of SPL _AN (f _i ), and applies (multiplies) it to r _n (t _i , d).
The simulation data S _{1 to} S _n−1 and S _n generated as described above are given synthesis gains by the synthesis gain generation units 23, 33, and 43, and are aggregated and added by the signal synthesis unit 50.
[0029]
(Description of operation)
The operation of the passive sonar simulation signal generation method and apparatus according to the present embodiment will be described below.
Simulated data generation signals S ₁ is performed as follows. Based on the setting of the level dispersion of the simulation parameter 2, the random number 1 generation unit 21 generates a discrete time series random number having a predetermined probability density distribution.
Next, based on the setting of the center frequency and the line width, the filter 1 generation unit 22 obtains a spectrum shape that realizes the set center frequency and line width, and tap coefficients (FIR filters having band pass characteristics equivalent to this) (TAP coefficient) is calculated, and an FIR filter is constructed and applied to the discrete time series random number, thereby obtaining a simulated signal of S ₁ having an average spectral shape as shown in FIG.
[0030]
Next, based on the setting of the S / N ratio, the G ₁ calculation unit 23 calculates the combined gain G ₁ and multiplies the filter output. Since the S / N ratio is the level ratio between the background noise and the signal, if the background noise addition gain G _n is determined, G ₁ is calculated based on the given G _n . If G _n is not determined, G ₁ and G _n are calculated together.
Simulated data generation is performed in the same manner for the signals S _{2 to} S _n−1 .
[0031]
Simulated data generated background noise S _n is carried out as follows. Based on the setting of the level dispersion of the simulation parameter 4, the random number n generation unit 41 generates a discrete time series random number having a predetermined probability density distribution.
Next, based on the setting of the frequency characteristic, the filter 1 generation unit 22 calculates the TAP coefficient of the FIR filter having the band pass characteristic equivalent to the set frequency characteristic, configures the FIR filter, and applies it to the discrete time series random number To do. For example, when the frequency characteristic as shown in FIG. 4 is given as a simulation parameter, broadband noise having a substantially equivalent average spectral shape is generated.
The composite gain is multiplied by the filter output using _Gn obtained by calculating the composite gain of the signal simulation data.
[0032]
The simulated data S _n simulated data S ₁ ~S _n-1 and the background noise for each signal generated by the above, all of the signal combining unit 50 by linear addition, the average spectrum as a final 5 The simulated data 60 having the data is obtained.
(Other embodiments)
Next, another embodiment of the present invention will be described. In the embodiment shown in FIG. 2, the S / N ratio, the center frequency, one line width, and level dispersion are used as the simulation parameters of the signal, but it is conceivable to use other parameters.
For example, as the line width of the parameters representing the spectrum shape of the signal, setting information for a plurality of line widths is prepared, and an approximation method other than the normal distribution shape can be used for the spectrum shape.
[0033]
FIG. 6 is a diagram showing an embodiment for generating a spectrum shape in this case. In the present embodiment, as shown in the figure, as k line widths LW (hi) for each of k spectrum levels h _i (i = 1, 2,... K),
LW (h ₁ )> LW (h ₂ )>...> LW (h _k )
A plurality of set values of line widths having the above relationship are prepared, and these are used to smoothly interpolate using spline interpolation or the like, thereby obtaining a continuous spectrum shape.
In this case, a set value of a plurality of line widths is stored in the parameter table, and the filter generation unit 22 is configured to calculate the TAP coefficient of the FIR filter according to the set value and the spectrum shape approximation method.
[0034]
As another embodiment, a configuration in which the signal-to-noise ratio of the signal is preliminarily taken into the spectrum shape and the combined gain calculation unit is unnecessary is conceivable.
FIG. 7 is a block diagram showing a configuration example of the present embodiment. In the present embodiment, the synthesis gain calculation units 23 to 43 and each multiplication unit shown in FIG. 2 are omitted, and the SN ratio of the simulation parameter is taken into the filter generation unit 22 so that the synthesis gains G _{1 to} G _n are obtained. A spectral shape of each multiplied filter is generated. In this case, for example, the spectral shape SPL ₁ (f _i ) of the signal S ₁ is
SPL (f _i ) = G ₁ exp [− (f _i −f _c1 ) ² / 2h ₁ ² ]
f _c1 : the center frequency of the signal, h ₁ : the line width of the signal S ₁ , and a filter is generated (configured). The spectral shapes of the filters of the other signals S _{2 to} S _n−1 and the background noise S _n are generated in the same manner.
[0035]
As another embodiment, time series random numbers are not generated each time according to the level dispersion setting, but a plurality of types of random value data are prepared in advance and prepared according to the level dispersion setting value. A method of selectively using the random value that has been stored is conceivable. In this case, a parameter table in which a plurality of random number data having a predetermined probability density distribution are registered is prepared, and the random number generation unit 21 acquires random number data corresponding to the set level distribution and outputs the random number data to the filter generation unit. It is realized with.
[0036]
【The invention's effect】
According to the present invention, since a simulated signal is generated using time-series random numbers having different probability density distributions for each signal, a random number that can best simulate the level dispersion of the signal is independently selected and generated. Therefore, it is possible to independently simulate the level variation for each signal of the simulation data including a plurality of signals.
[0037]
Similarly, by using time-series random numbers with different probability density distributions for each signal, random numbers that can best simulate the variation in the line width of the signal can be independently selected and generated. It is possible to independently simulate variations in line width for each signal of the included simulation data.
[0038]
Furthermore, because it is a method of generating simulated background noise using time series random numbers with a probability density distribution different from that for signal simulation, random numbers that can best simulate background noise level dispersion are selected and generated independently of the signal Therefore, it is possible to simulate the level dispersion of the signal and the background noise independently.
[Brief description of the drawings]
FIG. 1 is a diagram illustrating the principle of a specific example of the present invention.
FIG. 2 is a block diagram showing an embodiment of the present invention.
FIG. 3 is a diagram showing a filter 1 pass characteristic and an average spectrum shape of a signal S ₁ .
4 is a diagram showing the average spectral shape of the filter n pass characteristics and background noise S _n.
FIG. 5 is a diagram showing an average spectrum shape of simulated data.
FIG. 6 is a diagram showing another embodiment.
FIG. 7 is a diagram showing still another embodiment.
FIG. 8 is a diagram showing a conventional technique.
[Explanation of symbols]
1 parameter table 2, 3, 4 simulation table 5, 50 signal synthesis unit 6, 60 simulation data unit 22-42, 26-46 filter generation unit 21-41, 101-401 random number generation unit 23-43, 103-403 synthesis Gain calculation unit 102 to 402 filter

Claims (8)

Passive sonar simulation signal generation method for generating simulated data of passive sonar, generating simulated data using time series random numbers with different probability density distribution for each signal and noise to be simulated, and adding them in passive sonar simulation signal generating method characterized by obtaining simulated data. A passive sonar simulation signal generating method for generating simulated data for passive sonar, the variation in the level distribution and line width in the frequency domain, signal and each noise to be simulated target in the time domain, when the different probability density distribution A passive sonar simulation signal generation method characterized by setting and simulating using a sequence random number . A passive sonar simulation signal generator for generating simulated data for passive sonar, generates a time-series random number different probability density distribution signal and each noise to be simulated target, and generates the simulated data for each signal and noise simulation A passive sonar simulation signal generator comprising: a data generation unit; and a signal synthesis unit that adds the output of the simulation data generation unit and outputs simulation data. A passive sonar simulation signal generator for generating simulated data for passive sonar, generates a time-series random number different probability density distributions for each signal or noise becomes simulated target, and generates the simulated data for each signal or noise simulation A passive sonar simulation signal generator comprising a data generation unit, wherein level dispersion and line width variation in the frequency domain of simulated data are set and simulated by a probability density distribution in the time domain. A passive sonar simulation signal generator for generating simulated data of a passive sonar, a plurality of random number generators for generating time series random numbers of different probability density distributions for each signal and noise to be simulated, and the plurality And a plurality of filters having pass characteristics of respective average spectral shapes, and a signal synthesis unit for synthesizing simulated data by synthesizing the outputs of the plurality of filters. A passive sonar simulation signal generator. It has a parameter table that holds setting information of the level dispersion of the time series random number, the center frequency and line width of the signal, and the SN ratio, and the level of the time series random number of the random number generation unit by the level dispersion set in the parameter table 6. The passive sonar simulated signal generator according to claim 5, wherein dispersion is set, and the pass characteristic of the filter is set based on a center frequency, a line width, and an SN ratio set in the parameter table. A passive sonar simulation signal generator for generating simulated data for passive sonar, for each signal and noise to be simulated target, a plurality of random number generation unit for generating a time-series random numbers each different probability density distributions, said plurality A plurality of filters having pass characteristics of respective average spectral shapes, and a combined gain multiplying unit that receives the outputs of the plurality of filters and multiplies each combined gain and outputs the filters. A passive sonar simulation signal generator, comprising: a signal synthesis unit for synthesizing simulated data by synthesizing outputs of the synthesis gain multiplication unit. It has a parameter table that holds setting information of the level dispersion of the time series random number, the center frequency and line width of the signal, and the SN ratio, and the level of the time series random number of the random number generation unit by the level dispersion set in the parameter table The dispersion is set, a pass characteristic of the filter is set according to a center frequency and a line width set in the parameter table, and the combined gain is set according to an S / N ratio set in the parameter table. 8. A passive sonar simulation signal generator according to item 7.

Images