JP2910727B2 - Target signal detection method and device - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a method and an apparatus for transmitting a sound wave or an electromagnetic wave and detecting a reverberation signal from a target corresponding to the transmitted signal. The present invention relates to a target signal detection method and a target signal detection device for determining whether the target signal is the target signal.

[0002]

2. Description of the Related Art In general, a target signal detection system which emits a sound wave or an electromagnetic wave, detects a reverberation signal from a target corresponding to the transmitted signal, and can determine what kind of object the reverberation signal originates from, It is used for the purpose of assisting the operator by visually recognizing the display contents on the display unit or directly hearing the echo signal with his / her ear to determine the operation based on the experience and intuition of the operator.

Conventionally, this type of target signal detection system has been disclosed in, for example, Japanese Patent Application Laid-Open No. 8-152472, in which a predetermined section of the reverberation signal is cut out and used for detection. A preprocessing unit that converts the pattern into a pattern and outputs the pattern, and the neural network that has learned the relationship between the pattern for detection and the predetermined frequency in advance determines what object the pattern for detection is reflected from. A neural network determining unit for determining is provided, whereby a process of determining whether the target is a target is performed, and the result is displayed.

[0004]

In the conventional target signal detection method, the detection rate of the reverberation signal from the target is high, but a signal such as reverberation similar to the reverberation signal from the target is erroneously detected. There was a problem in that the (erroneous determination) rate was high. This is because it is difficult to completely separate the reverberation signal from a target other than the target and the reverberation signal from the target only by performing the determination process of determining whether or not the target is the target from the reverberation signal. is there.

[0005]

According to the present invention, the results of target signal detection processing for each of a plurality of reception beams are integrated, and when the reverberation signal determined as the target satisfies a certain condition, the target signal detection processing is performed. False alarm (misjudgment) in which the determined echo signal is an echo signal from a source other than the target
Is determined comprehensively.

A target signal detection method according to the present invention is a target signal detection method for transmitting a sound wave or an electromagnetic wave and detecting a reverberation signal from the target with respect to the transmitted signal. A first step of converting and outputting a time-series signal for each reception beam; and performing frequency analysis on the time-series signal to convert the time-series signal into spectrum time-series data. A section corresponding to the length of the signal is cut out to generate a detection pattern, a determination result as to whether or not the signal is an echo signal from a target is output from the detection pattern, and a signal distance and a signal level of the cut section are output. Second output bell and Doppler
Based on the output of the second step and the output of the second step, a matrix of the azimuth (received beam), the distance, and the determination result is formed, and the determination result, the level distribution characteristic, and the Doppler distribution characteristic of all the received beams are calculated for each distance. The third step of detecting the characteristics of the signal and testing whether the determination result is erroneous.
Steps.

A target signal detecting device according to the present invention transmits a sound wave or an electromagnetic wave, receives a reverberation signal from the target signal, and detects a signal from a target. An input processing unit that converts the reception signal into a reception signal for each reception beam, and converts the reception signal into a detection pattern characterized by the spectrum of the reception signal cut out from a predetermined section, and determines whether the detection pattern is an echo signal from a target. A neural network that outputs the result of the determination, a Doppler calculation processing unit that calculates the Doppler of the detection pattern from the reception signal in the predetermined section, a level calculation processing unit that calculates a signal level and a distance, Form a matrix of received beam azimuth, distance, and discrimination results, and determine the discrimination results, level distribution characteristics, and Dopp for all received beams at the same distance. Distribution characteristic the determination result of the determination as echo signals from the target from having a target test unit to test whether determined or not erroneous.

According to the present invention, the level, Doppler and distance of the reverberation signal are calculated as the output of the target signal detection processing, in addition to the result of discrimination of the reverberation signal from what object, and each received beam is calculated. These pieces of information output every time are integrated so as to be comprehensively determined, and it is tested whether or not the signal is a reverberation signal from a source other than the target. Therefore, it is possible to separate a reverberation signal from a target other than a target similar to a target which cannot be discriminated and separated by processing using only a single reception beam and a reverberation signal from the target.

[0009]

Next, embodiments of the present invention will be described in detail with reference to the drawings. FIG. 1 is a block diagram relating to a target signal detection method and apparatus according to the present invention.

The transmitter / receiver 1 has a structure in which a plurality of transmitting / receiving elements are arranged, and transmits a sound wave or an electromagnetic wave and receives an echo signal.

The input processing unit 2 forms a reception beam by phasing the echo signal received by the transducer 1 in an arbitrary direction, and outputs a time-series signal of the amplitude value of each reception beam.

The target signal detection processing unit 3 converts the time series signal of the amplitude value of each received beam output from the input processing unit 2 into a neural network by a method disclosed in, for example, Japanese Patent Application Laid-Open No. 8-152472. It converts it into a detection pattern for input to the receiver, performs the discriminating process, and receives the result along with the Doppler representing the level of the reverberation signal, the distance, and the difference frequency from the frequency of the transmitted signal in the part where the discriminating process was performed. Output for each beam.

The target verification processing unit 4 creates a matrix of the distance, the azimuth (received beam), and the discrimination result from the output of the target signal detection processing unit 3 and, at the same distance, the total number of received beams of the discrimination result being the target. Based on the signal level of each received beam and the threshold value judgment result for the Doppler of each received beam, it is verified whether there is a false alarm (false judgment) which is a reverberation signal from other than the target, and the result is displayed. Output to the container 5.

Next, a detailed configuration of the input processing unit 2 will be described with reference to FIG. FIG. 2 is a block diagram illustrating a configuration of the input processing unit 2.

The receiving beam phasing processing unit 21
The output signals of the plurality of transmission / reception elements are phased to form a reception beam having reception directivity in an arbitrary direction. The band-pass filter 22 extracts a signal of a reception frequency band from the signal of each reception beam obtained by the reception beam phasing processing unit 21. The A / D converter 23 converts a signal of each reception beam having an analog value into a digital value, and outputs a time-series signal having an amplitude value.

Next, a detailed configuration of the target signal detection processing section 3 will be described with reference to FIG. FIG. 3 is a block diagram illustrating a configuration of the target signal detection processing unit 3.

The preprocessing unit 31 converts the time series signal of the amplitude value output from the input processing unit 2 into a detection pattern which is the input format of the neural network discrimination processing unit 33. Further, the Doppler of the detection pattern is calculated.

The level calculation processing section 32 calculates the signal level before being converted into the detection pattern by the preprocessing section 31 from the time series signal of the amplitude value output from the input processing section 2.
The neural network discrimination processing unit 33 includes the preprocessing unit 3
The detection pattern output from 1 is input, and it is determined from which object this signal is reflected.

Next, a detailed configuration of the pre-processing unit 31 will be described with reference to FIG. FIG. 4 is a block diagram illustrating a configuration of the preprocessing unit 31.

The frequency conversion processing unit 311 includes the input processing unit 2
The frequency conversion is performed on the time series signal of the amplitude value outputted from each of the short time sections, and the time series signal is converted into a spectrum time series signal. The cutout processing unit 312 cuts out the time-series signal of the spectrum output from the frequency conversion processing unit 311 in a section corresponding to a target reverberation signal length determined by the pulse width of the transmission signal, and obtains the frequency, time, and spectrum level. Output a three-dimensional pattern. The Doppler calculation processing unit 313 obtains the center of gravity from the spectrum intensity distribution of the section cut out by the cutout processing unit 312, and calculates Doppler from the deviation from the transmission signal frequency. The normalization processing unit 314 includes the cutout processing unit 3
The spectrum level of the three-dimensional pattern in the section cut out at 12 is normalized to the range of 0 to 1.

Next, with reference to FIG.
2 will be described in detail.

FIG. 5 is a block diagram showing the configuration of the level calculation processing section 32.

The detection processing section 321 square-detects the time series signal of the amplitude value output from the input processing section 2. The cutout processing unit 322 converts the time-series signal of the detected amplitude value output from the detection processing unit 321 into the cutout processing unit 31.
Cut out at a position synchronized with the section to be cut out in 2. The average calculation processing unit 323 calculates an average value of the amplitude values of the sections cut out by the cutout processing unit 322. Distance calculation processing unit 32
4 calculates the distance of the signal of the section clipped by the clipping processing unit 322.

Next, a detailed configuration of the neural network discrimination processing unit 33 will be described with reference to FIG. FIG. 6 is a block diagram illustrating a configuration of the neural network determination processing unit 33.

The neural network is hierarchical, with units between each layer being interconnected. The neural network that has learned the relationship between the input detection pattern and the output is used to determine from which object the detection pattern output from the preprocessing unit 31 is reflected.

Next, a detailed configuration of the target verification processing unit 4 will be described with reference to FIG. FIG. 7 is a block diagram illustrating a configuration of the target verification processing unit 4.

The matrix processing unit 41 creates a matrix of the azimuth (received beam), the distance, and the discrimination result (〇 for the target, x for other than the target) from the discrimination result of each reception beam output from the target signal detection processing unit. Then, it outputs the determination result, signal level, and Doppler of each reception beam at the distance including the determination result determined as the target. The target detection number determination processing unit 42, when the total number of the results determined as the target exceeds the threshold value from the determination result of each reception beam output from the matrix processing unit 41, the data determined as the target. Are all judged as false alarms (false judgments).

The level determination processing section 43 calculates an average value of the levels of the respective reception beams output from the matrix processing section 41, and calculates a logarithmic value after dividing the level of each reception beam by the average value. I do. At this time, as shown in FIG. 7, in the case of data including a reverberation signal from the target, the signal level of the reception beam including the reverberation signal from the target and the surrounding beams generally becomes higher. Therefore, the logarithmic value after dividing by the average value becomes higher in the reception beam including the reverberation signal from the target and the surrounding beams. However, in the case of reverberation signal data from other than the target, it is generally omni
Since the difference of the signal level with respect to the direction becomes smaller, the logarithmic value after dividing by the average value gathers around zero. From such characteristics, when there is no data exceeding the threshold value set for the logarithmic value, all the determination results at this distance are determined to be other than the target. Even if the data exceeds the threshold set for the logarithmic value, data that continuously spreads in the azimuth (reception beam) direction is determined to be a signal from a source other than the target.

The Doppler determination processing unit 44 calculates the absolute value of the Doppler of each received beam output from the matrix processing unit 41, and after adding 1, calculates a logarithmic value. The reason for adding 1 is to prevent the logarithmic value from becoming negative infinity.
At this time, as shown in FIG. 7, in the case of data including a reverberation signal from the target, the received beam including the reverberation signal from the target and its surrounding beams generally have a large Doppler value. Therefore, the logarithmic value obtained as described above is
The level increases in the reception beam including the reverberation signal from the target and the surrounding beams. However, in the case of reverberation signal data from a source other than the target, the logarithmic value is 0 because the difference in Doppler value generally decreases in all directions (received beam).
Gather around. From such characteristics, when there is no data exceeding the threshold value set for the logarithmic value, all the determination results at this distance are determined to be other than the target. Even if the data exceeds the threshold set for the logarithmic value, data that continuously spreads in the azimuth (reception beam) direction is determined to be a signal from a source other than the target.

Based on the results of the target detection number determination processing unit 42, the level determination processing unit 43, and the Doppler determination processing unit 44, the comprehensive determination processing unit 45 is data that satisfies the following conditions as echo signals from sources other than the target. Comprehensively determines a false alarm (false determination).

1) The result of the target detection number determination processing section 42 is
At the time of false alarm.

2) The result of the target detection number determination processing section 42 is
When there is no false alarm, but the result of both the level determination processing unit 43 and the Doppler determination processing unit 44 is a false alarm.

The display 5 displays the result of the target detection processing section 4.

Next, the signal processing or operation of the target signal detection method and apparatus of the present embodiment will be described in detail using an example applied to an active sonar using sound waves.

The basic configuration of this embodiment is the same as the block diagrams shown in FIGS. 1 to 7, and will be described using the reference numerals in FIG.
A transmitter / receiver 1 for transmitting sound waves into water and receiving the echo signal, an input processing unit 2 for performing beam phasing processing on the received echo signal, converting the received echo signal into a reception signal for each reception beam, and outputting the received signal. The received signal is cut out in a predetermined section, converted into a detection pattern characterized by the spectrum of the received signal, the center of gravity is obtained from the spectrum intensity distribution in the detection pattern, and the Doppler of the detection pattern is obtained from the deviation from the transmission frequency. Is calculated, and the relationship between the detection pattern and a desired output is input to the neural network in which the learned pattern has been learned. A target signal detection processing unit 3 that calculates and outputs a signal level and a distance from a time-series signal of an amplitude value before conversion into a pattern.
And a matrix of the azimuth (reception beam) of the reception beam direction, the distance and the discrimination result from the discrimination results output for each reception beam, and the discrimination results and level distribution characteristics of all reception beams at the same distance. A target verification processing unit 4 for verifying whether or not the determination result determined as the target from the Doppler distribution characteristics is a false alarm (false determination) from a target other than the target.
And a display 5 for displaying the result.

The input processing section 2 has the basic configuration shown in FIG. 2, and the reception beam phasing processing section 21 converts signals output from a plurality of electroacoustic transducers of the transducer 1 into predetermined signals. To form N reception beams having reception directivity. The band-pass filter 22 extracts a signal in the band of the reception frequency from each reception beam output obtained by the reception beam phasing processing unit 21. The A / D converter 23 converts each received beam output of an analog value into a digital value, and outputs a time-series signal of an amplitude value.

The target signal detection processor 3 has the same basic configuration as that shown in FIG. 3, and the preprocessor 31 converts the time series signal of the amplitude value output from the input processor 2 into a neural network discriminator. In addition to the conversion into the detection pattern which is the input format of 33, the center of gravity is obtained from the spectrum intensity distribution in the detection pattern, and the Doppler of the detection pattern is calculated from the deviation from the transmission frequency. The level calculation processing unit 32 calculates a signal level before being converted into the detection pattern by the preprocessing unit 31 and a distance from the time series signal of the amplitude value output from the input processing unit 2. The neural network discrimination processing unit 33 discriminates from what object the object is reflected from the detection pattern output from the preprocessing unit 31. The pre-processing unit 31, the level calculation processing unit 32, and the neural network discrimination processing unit 33 perform processing for each of the received beams.

The pre-processing unit 31 has the basic configuration shown in FIG. 4. The frequency conversion processing unit 311 shifts the time series signal of the amplitude value output from the input processing unit 2 by s1 samples every n samples. Is divided into short-time section data, and the short-term section data is frequency-transformed by an FFT (Fast Fourier Transform) method to calculate a spectrum level, and the time series data of the amplitude value is converted into the time series data of the spectrum. I do. Each of the time series data of this spectrum is similarly called a frame. At the time of this conversion, noise having frequency components over a wide band is converted into a frequency, whereby the energy of the noise is dispersed into each spectrum,
As a result, the noise level in the frequency domain is reduced and the signal-to-noise ratio is improved. The cutout processing unit 312 cuts out a section of a p frame corresponding to a target echo signal length determined by a time width of a transmitted pulse while shifting the section by s2 frames, and outputs a frame (time), a frequency,
Generate a three-dimensional pattern at the spectral level.

The Doppler calculation processing unit 313 detects the frequency that is the maximum spectrum level for each of the p frames from the three-dimensional pattern output from the extraction processing unit 312, and detects the detected maximum p spectrum levels. Is calculated, and the frequency value obtained from the deviation from the transmission signal frequency is defined as the Doppler of the three-dimensional pattern.

△ F = F′−F0 F ′ = ΣFMAX _f / p where FMAX indicates the frequency of the maximum spectral level, subscript f indicates the frame, p indicates the total number of frames generating the three-dimensional pattern, F ′ is the center of gravity of the spectrum of the three-dimensional pattern, F0 is the transmission signal frequency, and ΔF is the Doppler of the three-dimensional pattern.

The normalization processing section 314 includes the cut-out processing section 3
In order to match the spectrum level of the three-dimensional pattern output from 12 with the input condition of the neural network discrimination processing unit 33, level normalization is performed to a range of 0 to 1. In the normalization of the spectrum level, the maximum value SMAX of the spectrum level in the three-dimensional pattern is detected, and each spectrum level is divided by this maximum value.

A ′ _f ^b = A _f ^b / SMAX where A indicates the spectral level in the three-dimensional pattern, subscript f indicates the frame, subscript b indicates the frequency, and A ′ indicates the normalized vector level. Is shown. The normalized three-dimensional pattern is similarly called a detection pattern.

The level calculation processing section 32 has the basic configuration shown in FIG. 5, and the detection processing section 321 square-detects the time-series signal of the amplitude value output from the input processing section 2 so that the sign of the signal is changed. All must be positive. The cutout processing unit 322
From the time-series signal of the amplitude value subjected to the detection processing output from the detection processing unit 321, u samples of a section corresponding to a target reverberation signal length determined by the time width of a pulse to be transmitted are obtained.
The clipping is performed so as to be temporally synchronized with the section clipped by the clipping processing unit 312. The average calculation processing unit 323 calculates the average value of the amplitude values of the sections cut out by the cutout processing unit 322, and sets this as the level of the section cut out by the cutout processing unit 322.

S = Σz / u Here, z indicates the amplitude value after the detection processing, and S is the average of the amplitude values in the cutout section for u samples.

The distance calculation processing section 324 calculates the distance of the signal of the first sample in the section cut out by the cut-out processing section 322.

R = (C / 2) × (Pt / Fs) where C is the sound velocity in water, Fs is the sampling frequency, and Pt is one of the cut-out sections after the pulse is transmitted.
A sampling count value up to the data of the sample is shown, and R is a distance.

The neural network discrimination processing unit 33 has the basic configuration shown in FIG. 6 and has three layers (input layer,
The intermediate layer and the output layer are of a hierarchical type, and the units between the respective layers are mutually connected. However, the configuration of the neural network is not limited to this. In order to use a neural network as a discriminating circuit, it is necessary to learn a neural network using a known learning pattern in advance. The learning pattern is created by using a plurality of data for an actual target and classifying a detection pattern obtained through the processing up to the input processing unit 2 and the preprocessing unit 31 into two patterns other than the target and the target. To create.
Using this learning pattern, when a target detection pattern is input to the input layer, learning is performed so that 1 is output to a unit allocated to the target in the output layer and 0 is output to a unit allocated to a target other than the target. . When a detection pattern other than the target is input to the input layer, 0 is assigned to the unit assigned to the target in the output layer, and 1 is assigned to the unit assigned to other than the target.
Is learned so that is output. A known back propagation method is used as a learning method. The neural network discrimination processing unit 33 inputs the detection pattern output from the preprocessing unit 31 to the neural network that has been learned by the learning method, and determines whether the detection pattern is a target or a target other than the target. And outputs the result. Here, since the output of each unit in the output layer of the neural network has an analog amount of 0 to 1, the following determination condition is used.

If X> Y and XY> 0.4, the target is X <Y and YX> 0.4, other than the target If the above conditions are not satisfied, the target is not the target Here, X is the neural network output layer Let Y be the output value of the unit assigned to the target, and let Y be the output value of the unit assigned to other than the target in the neural network output layer.

Next, the target verification processing unit 4 has the same basic configuration as that of FIG. 6, and the matrix processing unit 41 determines the azimuth based on the discrimination result of each received beam output from the target signal detection processing unit 3. (Received beam) Creates a matrix of distance and discrimination result (〇 for target, x for non-target), and outputs the discrimination result, signal level, and Doppler of each reception beam at the distance including the discrimination result determined as the target. .

The target detection number determination processing section 42 calculates the total number of the results determined as the target from the determination results of the respective reception beams at the same distance output from the matrix processing section 41, as follows:
If a threshold value set in consideration of a case where a high S / N ratio echo signal is received from a target and a signal leaks into a plurality of beams is exceeded, all results determined as the target are erroneous. Judge as an alarm.

If D ≧ L1, the discrimination results are all false alarms. If D <L1, the discrimination result is valid. Here, D indicates the total number of reception beams as a result of discrimination as a target, and L1 indicates a threshold value. This is because the echo signal reflected from a uniformly distributed object such as the sea floor or the sea surface is
This is because it is widely distributed in the azimuth direction.

The level judgment processing unit 43 calculates the average value of the level of each reception beam output from the matrix processing unit 41, divides the level value of each reception beam by this average value, and calculates the logarithmic value. calculate.

En = 10 log S / μ μ = ΣS / N Here, μ indicates the average value of the level S of each reception beam, and N indicates the total number of reception beams.

The following threshold processing is performed on the value En obtained in this manner.

If En <L2, the discrimination results are all false alarms. En ≧ L2, and if nE <L3, the discrimination results are valid. Here, L2 is a threshold value set for the logarithmic value, and L3 is a signal exceeding L2. The threshold value nE set for the range of the included beams indicates the number of continuous beams exceeding the threshold value L2.

The Doppler determination processing section 44 calculates the absolute value of the Doppler of each reception beam output from the matrix processing section 41, adds 1, and then calculates the logarithmic value. (The purpose of adding 1 is to prevent the logarithmic value from becoming negative infinity.) G′n = 10log (│Gn│ + 1) Here, G indicates the Doppler of the determination result, and the suffix n Indicates a reception beam number. The following threshold processing is performed on the value G ′ obtained in this manner.

If G′n <L4, all of the determination results are false alarms. If G′n ≧ L4 and nG <L3, the determination result is valid. Here, L4 exceeds the threshold and nG exceeds threshold L4. Indicates the number of consecutive beams.

Based on the results of the target detection number determination processing unit 42, the level determination processing unit 43, and the Doppler determination processing unit 44, the overall determination processing unit 45 is data that satisfies the following conditions as echo signals from sources other than the target. Judge as false alarm.

1) The result of the target detection number determination processing section 42 is
At the time of false alarm.

2) The result of the target detection number determination processing unit 42 is
When there is no false alarm, but the result of both the level determination processing unit 43 and the Doppler determination processing unit 44 is a false alarm.

The display 5 displays the result of the target detection processing section 4.

Although the present embodiment has been described in more detail with reference to an example applied to an active sonar, it goes without saying that such a signal processing method or operation can be applied to a device using electromagnetic waves.

[0063]

According to the present invention, the characteristic of the echo signal from a source other than the target is obtained by integrating the determination result of the target signal detection processing for each of the received beams so as to make a comprehensive determination using a plurality of received beams. Can be detected, so that false alarms (erroneous determinations) in which reverberation signals from sources other than the target are determined to be from the target can be reduced.

[0064]

[Brief description of the drawings]

FIG. 1 is a block diagram showing an embodiment of the present invention.

FIG. 2 is a block diagram illustrating an example of a configuration of an input processing unit in FIG.

FIG. 3 is a block diagram illustrating an example of a configuration of a target signal detection processing unit in FIG.

FIG. 4 is a block diagram illustrating an example of a configuration of a preprocessing unit in FIG.

FIG. 5 is a block diagram illustrating an example of a configuration of a level calculation processing section in FIG.

FIG. 6 is a block diagram illustrating an example of a configuration of a neural network determination processing section in FIG.

FIG. 7 is a block diagram illustrating an example of a configuration of a target verification processing section in FIG.

[Explanation of symbols]

 DESCRIPTION OF SYMBOLS 1 Transceiver 2 Input processing part 3 Target signal detection processing part 4 Target verification processing part 5 Display 21 Receive beam phasing processing part 22 Bandpass filter 23 A / D converter 31 Preprocessing part 32 Level calculation processing part 33 Neural Network determination processing unit 41 Matrix processing unit 42 Target detection number determination processing unit 43 Level determination processing unit 44 Doppler determination processing unit 45 Total determination processing unit 311 Frequency conversion processing unit 312 Cutout processing unit 313 Doppler calculation processing unit 314 Normalization processing unit 321 Detection processing unit 322 Cutout processing unit 323 Average calculation processing unit 324 Distance calculation processing unit

────────────────────────────────────────────────── ─── Continuation of the front page (56) References JP-A-9-325182 (JP, A) JP-A-9-281223 (JP, A) JP-A-6-43238 (JP, A) 297114 (JP, A) JP-A-4-1586 (JP, A) JP-A-7-44517 (JP, A) JP-A-7-44516 (JP, A) Patent 2806814 (JP, B2) (58) Field (Int.Cl. ⁶ , DB name) G01S 7/00-7/64 G01S 13/00-15/96

Claims (5)

(57) [Claims] In a target signal detection method for transmitting a sound wave or an electromagnetic wave and detecting a reverberation signal from a target corresponding to the transmitted signal, a received signal is subjected to beam phasing processing, and a time-series signal for each received beam is output. In the first step, after converting the time-series signal into time-series data of a spectrum, a section corresponding to the length of the target reverberation signal is cut out to generate a detection pattern, and a detection pattern from the target is generated from the detection pattern. A second step of outputting a determination result as to whether or not the signal is an echo signal and outputting a signal distance, a signal level, and a Doppler of the section of the cutout, and an azimuth (receiving beam) based on the output of the second step. ) And a matrix of the distance and the discrimination result, and the discrimination result of all received beams and the level distribution characteristics and Doppler distribution characteristics of each reception distance are used to detect the characteristics of signals other than the target for each distance. And a third step of testing whether the determination result is an erroneous determination. 2. The second step is a step of dividing the time-series signal for each reception beam in the first step into time-series short-term data in a time-series manner and sequentially obtaining a spectrum level from the section data. And extracting the time-series data of the spectrum for each short time section into sections corresponding to the target reverberation signal length determined by the width of the transmission signal, and generating a three-dimensional pattern of frequency, frame (time) and spectrum level. Calculating the center of gravity of the frequency of the maximum spectral level in each frame of the three-dimensional pattern, and calculating the Doppler of the three-dimensional pattern from the deviation from the frequency of the transmitted signal; Normalizing the vector to a predetermined range to obtain a detection pattern; and A step of determining whether or not the signal is an echo signal from a target by a network; a step of calculating a detection output of the time-series signal in each of the reception beams in the first step; and the detection output is determined by a width of a transmission signal. Clipping a section corresponding to the target echo signal length, calculating the average of the signal level of the clipped section and setting the signal level of the clipped section, and calculating the signal distance of the clipped section 2. The target signal detecting method according to claim 1, further comprising the step of: 3. The third step is to input a discrimination result, a signal distance, a signal level, and a Doppler in each reception beam output from the second step, and to input an azimuth (reception beam), a distance and a discrimination result. Forming a matrix, and calculating the total number of results determined as targets from the results of determination of all received beams for each distance; and a threshold value for the signal level of all received beams for each distance from the matrix. Performing a process, performing a threshold process on the Doppler of all received beams for each distance from the matrix, performing a process on the total number of results determined as the target, and the signal level. From the result of the threshold processing and the result of the threshold processing performed on the Doppler, the threshold processing determines whether or not the determination result determined as an echo signal from the target is an erroneous determination. 3. The target signal detecting method according to claim 1, further comprising a step of performing a test. 4. A target signal detection device for transmitting a sound wave or an electromagnetic wave, receiving a reverberation signal thereof and detecting a signal from a target, performs beam phasing processing on the received reverberation signal, and receives a reception signal for each reception beam. An input processing unit that converts the received signal into a detection pattern characterized by the spectrum of the reception signal cut out from a predetermined section, and outputs a determination result as to whether or not the signal is an echo signal from a target from the detection pattern. A neural network, a Doppler calculation processing unit that calculates Doppler of the detection pattern from a reception signal of the predetermined section, a level calculation processing unit that calculates a signal level and a distance, an azimuth, a distance and a reception beam of a reception beam. A matrix of the discrimination results is formed, and the echo from the target is determined based on the discrimination results of all received beams at the same distance and the level distribution characteristics and Doppler distribution characteristics. A target test processing unit for testing whether or not the result of the discrimination determined as a signal is an erroneous discrimination. 5. A target signal detecting device for transmitting a sound wave or an electromagnetic wave, receiving a reverberation signal thereof, and detecting a signal from a target, performs beam phasing processing on the received reverberation signal, and generates a reception signal for each reception beam. An input processing unit for converting, and a neural network that converts the received signal into a detection pattern characterized by the spectrum of the received signal cut out from a predetermined section, and outputs a determination result as to whether or not the signal is an echo signal from a target from the detection pattern. A network, a Doppler calculation processing unit that calculates Doppler of the detection pattern from the reception signal of the predetermined section, a level calculation processing unit that calculates a signal level and a distance, and an azimuth, a distance, and a discrimination of a reception beam. A matrix processing unit that forms a matrix of results; and a target detection unit that calculates a total number of results determined as targets for each distance based on the matrix. Number-of-output determination processing unit, a level determination processing unit that performs threshold processing of signal levels of all received beams for each distance, and a Doppler determination processing unit that performs threshold processing of Doppler of all received beams for each distance A target signal detection device comprising: a total determination processing unit that tests whether or not the determination result determined as a reverberation signal from a target based on a result of each of the threshold value processing is an erroneous determination.