JPH0516753B2 - - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION [Field of Industrial Application] The present invention relates to a target signal detection device for detecting a target signal from a noise signal whose amplitude characteristics exhibit an arbitrary distribution.

[Prior art] In radar observation, in order to detect a target object to be captured (hereinafter referred to as a target), it is necessary to suppress unnecessary reflected waves from the vicinity of the object (hereinafter referred to as clutter) and to suppress signals from the target object. The problem is how to detect the target signal (hereinafter referred to as target signal). There are two types of Kuratsuta: Grand Kuratsuta, which is caused by reflections from the ground surface, and Sea Tsuta, which is caused by reflections from the sea surface.
There are weather clutter caused by reflection from clutter and rain clouds.

Conventional target signal detection devices define an appropriate noise distribution and detect a target signal based on this noise distribution.

[Problems to be Solved by the Invention] By the way, the conventional target signal detection method has a problem in that the target signal cannot be detected sufficiently when the noise distribution is other than the specified one.

The present invention has been made to solve the above-mentioned problems, and an object of the present invention is to provide a target signal detection device that detects a target signal from a noise signal exhibiting an arbitrary noise distribution.

[Means for Solving the Problems] Therefore, the present invention provides a signal amplification means for linearly amplifying or logarithmically amplifying the input signal x, and a dispersion value for calculating the dispersion value σ ² of the output signal Q amplified by the signal amplification means. A calculation means, a nonlinear calculation means that outputs a nonlinear signal obtained by dispersion value σ ² to the k power (k is an arbitrary real number), compares the nonlinear signal with a preset threshold, and selects a nonlinear signal larger than the threshold as a target. A target signal detection device is constituted by a target signal detection means that outputs a signal.

[Operation] In the target signal detection device having the above configuration, the signal amplification means linearly amplifies or logarithmically amplifies the input signal x, that is, the target signal including a noise signal exhibiting an arbitrary noise distribution, and the variance value calculation means amplifies the input signal x. The nonlinear calculation means calculates the variance value σ ² of the output signal Q, and the nonlinear calculation means calculates the variance value σ ² by k
The target signal detecting means compares the nonlinear signal with a preset threshold value, and outputs a nonlinear signal larger than the threshold value as a target signal.

[Example] Hereinafter, an example of the present invention will be described in detail with reference to the accompanying drawings.

First, the signal processing theory of the target signal detection device according to the present invention will be explained.

The target signal detection device according to the present invention detects a target signal from a noise signal whose amplitude characteristics exhibit an arbitrary distribution. Here, the amplitude intensity x has a Weibull distribution and a lognormal distribution ( A case of exhibiting a Lognormal distribution will be explained.

The amplitude intensity x follows the Weibull distribution, and its probability density function P _C (X) is P _C (X) = (c/b) (x/b) ^Ci EXP [- (X/b) _C
]...(1) (Here, b is a scale parameter and c is a shape parameter.) The average value <X> of the amplified signal X obtained by linearly amplifying the received signal is: <X>=∫ ^∞ ₀ X・P _C (X) =bΓ(1/C+1)...(2). Here, Γ(z) is a gamma function, and Γ(Z)=∫ ^∞ ₀ t ^Z-1 EXP(-t)dt...(3). Further ^, the root mean square value <X ² > of the amplified signal X _is <X ² >= _∫ ^{∞ 0} Therefore, the variance value σ ² is σ ² =〈X ² 〉−〈X ² 〉 = b ² (Γ(2/C+1)−Γ ² (1/C+1))
...(5) becomes. This variance value σ ² is raised to the k power (K is any real number)
and detect the target signal.

Next, FIG. 1 is a block diagram of a target signal detection apparatus according to the present invention. In FIG. 1, a linear amplification circuit 10 (signal amplification means) amplifies the signal shown in FIG. 3, that is, an amplified signal X obtained by amplifying a received signal having an amplitude of Each is input to the calculation means 30. The mean square value calculation means 20 includes a square calculation circuit 21, a delay circuit 22 configured by, for example, a shift register, etc.;
It is composed of an integration circuit 23 and a division circuit 24, and the root mean square value of the signal X shown in the second equation <X ² >
Calculate. That is, the squaring circuit 21 squares the amplitude x of the signal X to obtain an amplitude x ² , and the delay circuit ²² samples and quantizes the signal The sample values of X ² ₁ , X ² ₂ , ...,
X ² _o is taken out, and the integrating circuit 23 extracts n sample values X ² ₁ ,
The integrated value _o 〓 _i ⁼¹ X _{2 i} ...( ⁶ ) of X ^{2 2 , ...} , 〉 〈X ² 〉=1/n 〓 ⁱ⁼¹ X ² _i ...(7) is calculated.

Further, the average value square calculation means 30 is a delay circuit 3
1. It is composed of an integration circuit 32, a division circuit 33, and a square calculation circuit 34, and calculates the square value <X> ² of the average value of the signal X shown in equation (2). That is, the delay circuit 31 samples and quantizes the signal
n sample values X ₁ , X ₂ , ..., corresponding to the amplitude x of
X _o is taken out, and the integrating circuit 32 extracts n sample values X ₁ ,
The integrated value _o 〓 ⁱ⁼¹ X _i ...(8) of X ₂ , ..., X _o is calculated, and the division circuit 33 divides the integrated value by n to obtain the average value of the integrated value 1/n _o 〓 ^{i= 1} X _i ...(9) is calculated, and the square calculator 34 squares the average value to obtain the square value of the average value of the signal X 〈X〉 ² 〈X〉 ² = (1/n _o 〓 ⁱ⁼¹ X _i ) ² ...(10) is calculated.

Next, the subtraction circuit 40 calculates the variance value σ ² =〈X ² 〉−〈X〉 ² (11) based on the root mean square value〈X〉 ² and the square value of the mean value〈X〉 ² . .

FIG. 2 is a diagram showing the variance value σ ² obtained as described above. In Figure 2, linear amplifier circuit 1
Even if the amplitude of the target signal OS and the amplitude of the noise signal are almost the same (second
(see Figure a), the variance value σ ² output from the subtraction circuit 40 has a large value in the portion corresponding to the target signal OS,
Also, the value of the part corresponding to the noise signal NS is almost constant and the fluctuation range is small (see Figure 2b), the target signal
The variance value σ ² corresponding to the OS can be easily detected.

FIG. 4 is a diagram for explaining the signal in FIG. 2b, and based on this FIG.
The signal in Figure b will be explained.

FIG. 4a, corresponding to FIG. 2a, is a diagram showing changes in the received amplified signal with respect to _distance after passing the received signal through a linear _or logarithmic amplifier. The amplitude of OS is approximately 1/(r _b ~
r _a ) was assumed to be obtained with equal amplitude.

First, the target signal OS in Figure 2 a and the OS in Figure 4 a
I will explain the difference between Now, if we consider that the radar transmission pulse width is short and the distance resolution is high, and the target object is a ship, the reflected signals will be transmitted from multiple locations such as the ship's belly, bridge, chimney, and mast. The information will be obtained sequentially depending on the distance from the radar to the reflection location. Figure 2 a
In this way, the OS can
This figure shows an example of a plurality of reflected signals (signals similar to a sine wave whose amplitudes differ depending on the reflection intensity) that are received according to the distance to the reflection point (that is, according to the time).

The OS in Figure 4a is an example of a simplified target.
A signal example is shown in which one reflecting plate (radio wave reflector) is provided at a distance r _a from the radar, and a received signal obtained from this reflecting plate is linearly or logarithmically amplified.

As a specific example, let's take the radar transmission pulse width as
If it is 80n sec, the distance between sections r _a to r _b in FIG. 4a will be approximately 12 m (meters), which corresponds to the distance resolution of the radar.

NS in Fig. 2a and Fig. 4a are the same noise signals, for example, noise signals following arbitrary distributions such as Weibull distribution or lognormal distribution, but the former is a received signal, and the latter is a received signal according to the linear or logarithmic distribution. Shows a signal passed through an amplifier.

When the target signal is as shown in Figure 2a or Figure 4a,
Considering the pulse height distribution probabilities around the average value <X> of each of the target signal and the noise signal, the pulse height distribution probability for the noise signal is generally as shown in FIG. 4b. This indicates that the probability that the peak value of the noise signal takes a value close to the average value is relatively high. That is, the value of the standard deviation σ _NS is relatively small.

Further, the wave height distribution probability for the target signal is as shown in FIG. 4c. This indicates that the peak value of the target signal has a high probability of taking a value far from the average value. That is, the value of standard deviation σ _OS is relatively large. Therefore, when comparing the dispersion value of the noise signal and the dispersion value of the target signal, there is a difference between the two, and the latter value is larger than the former value.

This will be explained below using specific numerical examples.

First, the amplitude strength of the received signal follows the Weibull distribution, and the variance value σ ² when this received signal is linearly amplified is
is expressed by the above equation (5).

Here, as an example of Weibull distributed noise, there is sea clutter from the sea surface, and as an example of the actually measured values of this sea clutter, there is an example of b=4.8 and C=1.8 in equation (5). When these measured values are substituted into Equation (5), a numerical example of the noise variance value σ ² _WN is expressed by Equation (43).

σ ² _WN = 4.9 ...(43) Also, as an example of linear amplification of the target signal, in the range r _a to r _b shown as OS in Figure 4 a,
Considering an equal amplitude signal of 1/(r _{b −ra} ), the above r _{b −ra}
= 12 m and radar pulse width = 80 n sec, the signal dispersion value σ ² _OS is expressed by equation 44).

σ ² _OS = (r _b − r _a ) ² /12=12 (44) Therefore, the value of equation (44) is larger than the value of equation (43).

What is important here is that there is a magnitude relationship in which the dispersion value of the target signal after receiving and amplification is larger than the dispersion value of the noise signal.

Since the actual target is different from the reflector in this numerical example, it is normal that such a large difference cannot be obtained. However, if there is even a slight difference, as will be described later, by raising each value to the K power, this difference can be increased, and it becomes possible to distinguish between the two and detect the target signal.

In Figure 2b, the calculation is calculated from the received amplified signal of each reflected signal of the target (reflected signal equivalent to when multiple reflectors are installed at regular intervals) obtained from one target in Figure 2a. Dispersion value signals are combined and shown as a single target signal,
This relationship holds true even if the noise distribution characteristics change or the amplifier becomes a logarithmic amplifier. Note that in this case, the sampling number n is set to an optimal value.

In this embodiment, in order to more easily detect the dispersion value σ ² corresponding to the target signal OS, the dispersion value σ ² is converted to the nonlinear amplification circuit 50 (nonlinear amplification means).
Add to.

This nonlinear amplifier circuit 50 outputs a nonlinear signal Z Z = (σ ^2k ) (12) obtained by dispersion value σ ² to the kth power (k is a real number). Note that the nonlinear amplifier circuit 5
As a specific example of 0, a square amplification circuit where k=2 is known. The nonlinear signal Z output from the nonlinear amplifier circuit 50 is a noise signal compared to the variance value σ ² output from the subtraction circuit 40, as shown in FIG. 2C.
The fluctuation of the part corresponding to NS is suppressed, and the target signal
The portion corresponding to OS is greatly amplified, making it easier to detect the variance value σ ² corresponding to the target signal OS, that is, to set the threshold value.

Further, the target signal detection circuit 60 (target signal detection means) compares the nonlinear signal Z output from the nonlinear amplifier circuit 50 with a preset threshold manually or automatically, and detects a nonlinear signal Z larger than the threshold. Output as a target signal corresponding to OS.

The above-mentioned threshold value Z _TH is set manually or automatically, but in the case of traction, it is set while observing the nonlinear signal Z with an A scope, a B scope, a PPI scope, or the like. In addition, when automatic, based on the average value <Z> of the nonlinear signal Z and the variance value Var around the average value <Z>, the noise signal NS is converted into the target signal OS.
False alarm probability or target signal OS
The detection probability is set to a predetermined value.

The average value <Z> and the root mean square value <Z ² > of the nonlinear signal Z are as follows, assuming that the probability density function of the nonlinear signal Z is P(Z), <Z>=∫ ^∞ ₀ Z・P(Z)dZ ……( 13)〈Z ² 〉＝∫ ^∞ ₀ Z ²・P(Z)dZ ...(14) Then, the fractional value Var around the average value 〈Z〉 of the nonlinear signal Z is Var＝〈Z ² 〉−〈 Z〉 ² ...(15) becomes. Also, the false alarm probability Pfa is Pfa=∫ ^∞ _0TH P(Z)dZ ...(16), which is the detection probability of the target signal OS when the amplitude t of the target signal OS is superimposed on the nonlinear signal Z. Pa
is Pa=∫ ^∞ _ZTH P(Z+t)d(Z+t)...(17). Therefore, false alarm probability Pfa or detection probability Pa
The threshold value Z TH is set so that Z _TH becomes a predetermined value. Next, a specific value of the threshold value Z _TH calculated as described above will be explained. That is, the subtraction circuit 40
M is ^the m ^- th power of the variance value σ ² which is the output ^of , the constants are K and A, then one of the following values is obtained.

K・〈M〉 …(18) K・√ ^〈2〉 …(19) K・√ ^〈2〉 −〈〉 ² …(20)〈M〉＋K・√ ^〈2〉 −〈〉 ² … … ^{( 21 )} K. √ ^〈2〉 −〈〉 ² +A……(25)〈M〉+K・√ ^〈2〉 −〈〉 ² +A……(26) K・(〈M〉+√ ^〈2〉 −〈〉 ² )+A ...(27) Next, the amplitude intensity X has a lognormal distribution (Log-
A case in which the distribution has a normal distribution will be explained.

Log normal for the received signal
The distribution is becomes. Here, m is the median value of the received signal X, and ρ is the average-to-median ratio, ρ=<X>/m (29). _The ^average value <X> of ^the received signal X ² 〉=∫ ^∞ ₀ X ² P(X; m, ρ)dx=m ² ρ ⁴
...(31) becomes. Therefore, the variance value σ ² is σ ² =〈X ² 〉−〈X〉 ² =m ² ρ ² (ρ ² −1) (45). In the actual circuit, the root mean square value <X ² > of the amplified signal X obtained by linearly amplifying the received signal and the square value of the average value <X> ² are the same as in Equations 7 and 10, so their explanation is omitted. do.

The variance value σ ² when the amplitude strength of the received signal follows a lognormal distribution and the received signal is linearly amplified is shown by the above equation (45).

Here, an example of log-normal distributed noise is sea clutter from waves on the sea surface, and the actual measured values in this case are m = 1.3 and ρ = 1.7. Substituting these measured values into equation (45), the noise variance σ ² _LN
A numerical example of is shown in equation (46).

σ ² _LN ≒9.2 ...(46) Also, as an example of linearly amplifying the target signal, in the case of the OS shown in Figure 4 a, the variance value σ ² _OS calculated using equation (44) as a numerical example is =12 is used.

Even in the case of this lognormal distribution, the variance value of the noise signal in equation (44) is larger than the variance value of the noise signal in equation (46).

Next, in order to widen the dynamic range when the received amplitude strength of the clutter is considerably larger than that of the target and a linear amplifier would be saturated,
A logarithmic amplifier is also used in the present invention.

If Kuratsuta's probability distribution follows the Rayleigh distribution, its standard deviation is proportional to the mean, so
There is a property that the dispersion value of a signal obtained by logarithmically amplifying this clutter is constant.

However, subsequent research has shown that when the radar transmission pulse width is reduced (i.e., when the distance resolution is improved), the probability distribution of clutter does not follow the Rayleigh distribution, but changes to another distribution such as the Weibull distribution or the lognormal distribution. It became clear that this would happen.

Since the present invention is directed to such arbitrarily distributed clutter, the dispersion value of a signal obtained by logarithmically amplifying clutter is usually not constant.

However, when calculating the dispersion value of a signal obtained by logarithmically amplifying the received signal, there is a difference in the value between the target object and the noise, that is, the former value is larger than the latter value, so this difference is emphasized. By doing so, target signals are detected.

Although the linear amplifier circuit 10 is used as the signal amplifying means in this embodiment, it is also possible to use a logarithmic amplifier circuit or the like.
An amplifier circuit having arbitrary characteristics may be used. When the received signal has a Weibull distribution, the logarithmic amplifier circuit 10
The output Y is as follows: Y=a ₀ ln(b ₀ X)...(32) Therefore, the average value <Y> of the logarithmically amplified signal Y
is <Y>=∫ ^∞ ₀ a ₀ ln (b ₀ x) P _C (X) dX = a ₀ ln (b ₀ b) − a ₀ /Cγ (33). Here γ is Euler's function and γ=0.5772...
It is. Also, the root mean square value of the logarithmically amplified signal Y <Y ² >
_{is 〈} Y ² _〉 ^{= ∫ ∞} _{0 a} ² ₀ ln ² ₍ ^b ₀ ² / ₀ /C ² (π ² /6 + γ ² ) ...(34). Therefore, the variance value σ ² is σ ² =〈Y ² 〉−〈Y〉 ² =a ² / ₀ /c ² ·π ² /6 (35). In the actual circuit, the root mean square value <Y ² > and the square value of the average value <Y> ² are respectively, <Y ² >=1/n _o 〓 ⁱ⁼¹ Y ² _i ...(36) <Y> ² = (1/n _o 〓 ⁱ⁼¹ Y _i ) ² ...(37), and the variance value σ ² is σ ² =〈Y ² 〉−〈Y〉 ² =1/n _o 〓 ⁱ⁼¹ Y ² _i − (1/n _o 〓 ⁱ⁼¹ Y _i ) ² ...(38).

Next, when the received signal has a lognormal distribution, the output Y of the logarithmic amplifier circuit 10 is as follows: Y=a ₀ ln (b ₀ X) (39). Therefore, the average value <Y> of the logarithmically amplified signal Y
is 〈Y〉=∫ ^∞ _{0 a 0} ln ₍ b _{0 The} value _〈 Y ² 〉 is 〈Y ² 〉=∫ ^∞ ₀ a ² ₀ ln ² ₍ ^{b 0} ) becomes. Therefore, the variance value σ ² is σ ² =〈Y ² 〉−〈Y〉 ² =2a ² ₀ lnρ (42). In an actual circuit, the root mean square value <Y ² > of the logarithmically amplified signal Y obtained by logarithmically amplifying the received signal and the square value of the average value <Y> ² are the same as equations (36) and (37). The explanation will be omitted.

The amplitude intensity of the received signal follows the Weibull distribution, and the variance value σ ² when this received signal is logarithmically amplified is shown by the above equation (35).

Consider the sea clutter as an example of noise in this case, and the actual measured values in this case are C = 1.8, a ₀ = 1
There is an example. When these measured values are substituted into Equation (35), a numerical example of the noise variance value σ ² _WL is expressed by Equation (47).

σ ² _WL ≒0.5 ...(47) Also, as an example of logarithmically amplifying the target signal, assuming a saturated signal with uniform amplitude, the OS in Figure 4a is
A numerical example can be used. Since the fractional value in this case is σ ² _OS =12 as calculated by equation (44), the variance value of the target signal is larger than the variance value of the noise signal in equation (47).

Next, the amplitude strength of the received signal follows a lognormal distribution, and the variance value σ ² when this received signal is logarithmically amplified
is expressed by the above equation (42).

As an example of the noise in this case, consider the sea clutter from the waves and sea surface, and the actual measured value in this case is ρ.
There is an example where =1.7 and a ₀ =1. When these measured values are substituted into Equation (42), a numerical example of the noise variance value σ ² _LL is expressed by Equation (48).

σ ² _LL = 1.1 ...(48) Similarly, as an example of logarithmically amplifying the target signal, assuming a saturated signal with a uniform amplitude, When compared with the value in equation (44), the variance value of the target signal is larger than the variance value of the noise signal in equation (48).

In this embodiment, a case has been described in which the target signal S is detected from a noise signal exhibiting a Weibull distribution and a lognormal distribution, but a target signal can be detected even if the target signal S is a noise signal exhibiting an arbitrary distribution.

[Effects of the Invention] As explained above, according to the present invention, by outputting the dispersion value of a noise signal whose amplitude characteristics exhibit an arbitrary distribution, and further nonlinearly amplifying the dispersion value,
Since the magnitude of the dispersion value corresponding to the target signal or the non-linearly amplified dispersion value is larger than the magnitude of the dispersion value corresponding to the noise signal, by setting an appropriate threshold value, the amplitude of the target signal can be adjusted to be less than the noise signal. Even if the amplitude of the target signal is smaller than that of the signal, the target signal can be easily detected.

[Brief explanation of the drawing]

FIG. 1 is a block diagram of a target signal detection device according to the present invention, and FIG. 2 shows changes in received signals over time.
FIG. 3 is an explanatory diagram of a dispersion value and a new dispersion value obtained by raising the dispersion value to the k power. FIG. 3 is an explanatory diagram showing changes in the received signal over time. FIG. 4 is a diagram for explaining the signal of FIG. 2b. DESCRIPTION OF SYMBOLS 10... Linear amplifier circuit, 20... Root mean square value calculation means, 21... Square calculation circuit, 22, 31... Delay circuit, 23, 32... Integration circuit, 24, 33... Division circuit, 30... Square value calculation of average value means, 34... square calculation circuit, 40... subtraction circuit, 50... nonlinear amplification circuit, 60... target signal detection circuit.

Claims (1)

[Claims] 1. A target signal detection device for detecting a target signal from an input signal x mixed with noise exhibiting an arbitrary distribution, comprising: signal amplification means for linearly or logarithmically amplifying the input signal x; , an output signal Q amplified by the signal amplification means
variance value calculation means for calculating a variance value σ 2 of the variance value σ ² ; nonlinear calculation means for outputting a nonlinear signal that is the variance value σ ² raised to the k power (k is any real number); and the nonlinear signal and a preset threshold value. A target signal detection device comprising target signal detection means for comparing the nonlinear signals and outputting a nonlinear signal larger than the threshold as a target signal. 2. The threshold value in the target signal detection means is determined based on the average value of the value M obtained by raising the variance value σ ² to the m power obtained by the variance value calculation means and the variance value around the average value of M. The target signal detection device according to claim 1, wherein the false alarm probability or detection probability of the target signal is set to a predetermined value. 3. The threshold value is the value K·<M> obtained by multiplying the average value <M> of the value M, which is the m-th power of the variance value σ ² obtained by the variance value calculating means, by a constant K. The target signal detection device described in . 4 The threshold value is the root mean square value of the value M obtained by raising the variance value σ ² obtained by the variance value calculation means to the m power <M ² >
The target signal detection device according to claim 1, wherein the square root of √ ^<2> is multiplied by a constant K: K·√ ^<2> . 5. The threshold value is calculated by calculating the variance value (<M ² >−<M> ² ) around the average value <M> of the value M, which is the m-th power of the variance value σ ² obtained by the variance value calculation means, The target signal detection device according to claim 1, wherein the square root of the dispersion value is √ ^〈2〉 −〈〉 ² multiplied by a constant K: K·√ ^〈2〉 −〈〉 ² . 6. The threshold value is calculated by calculating the variance value (<M ² >−<M> ² ) around the average value <M> of the value M, which is the m-th power of the variance value σ ² obtained by the variance value calculation means, Furthermore, the square root of the variance value √〈 ² 〉−〈〉 ² is multiplied by a constant K,
The target signal detection device according to claim 1, wherein the average value <M> is further added to a value <M>+K·√ ^<2> −<> ² . 7. The threshold value is calculated by calculating the variance value (<M ² >−<M> ² ) around the average value <M> of the value M, which is the m-th power of the variance value σ ² obtained by the variance value calculation means, A patent claim that is the square root of the variance √ ^〈2〉 −〈〉 ² , the average value〈M〉, and then multiplied by a constant K: K・(〈M〉+√ ^〈2〉 −〈〉 ² ) The target signal detection device according to item 1. 8 The threshold value is the value obtained by multiplying the average value M of the value M obtained by multiplying the variance value σ ² obtained by the variance value calculation means by the constant K, and further adding the constant A, K・〈M〉+A. A target signal detection device according to claim 1. 9 The threshold value is the root mean square value of the value M obtained by raising the variance value σ ² obtained by the variance value calculating means to the m power <M ² >
Multiply the square root √〈 ² 〉 by the constant K, and then add the constant A
The target signal detection device according to claim 1, wherein the value obtained by adding K·√< ² >+A. 10 The threshold value is calculated by calculating the variance value (〈M ² >−〈M〉 ² ) around the average value <M> of the value M, which is the mth power of the variance value σ ² obtained by the variance value calculation means, The target according to claim 1, which is a value obtained by multiplying the square root of the dispersion value √ ^〈2〉 − ^〈 〉 ² by a constant K and further adding a constant A: K・√〈2〉−〈〉 ² +A Signal detection device. 11 The threshold value is calculated by calculating the variance value (〈M ² >−〈M〉 ² ) around the average value <M> of the value M, which is the mth power of the variance value σ ² obtained by the variance value calculation means, The square root of the variance √ ^〈2〉 −〈〉 ² is multiplied by the constant K, the average value〈M〉 is added, and the constant A is further added to the value 〈M〉+K・√ ^〈2〉 −〈〉 ² +A. A target signal detection device according to claim 1. 12 The threshold value is calculated by calculating the variance value (〈M ² >−〈M〉 ² ) around the average value <M> of the value M, which is the mth power of the variance value σ ² obtained by the variance value calculation means, The square root of the variance √ ^〈2〉 −〈〉 ² , the average value〈M〉 is added, the sum of the square root of the variance value and the average value is multiplied by a constant K, and the value obtained by adding the constant A is K・( The target signal detection device according to claim 1, wherein <M>+√ ^<2> −<> ² )+A.