CN109002810A - Model evaluation method, Radar Signal Recognition method and corresponding intrument - Google Patents
Model evaluation method, Radar Signal Recognition method and corresponding intrument Download PDFInfo
- Publication number
- CN109002810A CN109002810A CN201810861202.3A CN201810861202A CN109002810A CN 109002810 A CN109002810 A CN 109002810A CN 201810861202 A CN201810861202 A CN 201810861202A CN 109002810 A CN109002810 A CN 109002810A
- Authority
- CN
- China
- Prior art keywords
- model
- sample
- radar signal
- training
- interpretation
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2218/00—Aspects of pattern recognition specially adapted for signal processing
- G06F2218/08—Feature extraction
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01S—RADIO DIRECTION-FINDING; RADIO NAVIGATION; DETERMINING DISTANCE OR VELOCITY BY USE OF RADIO WAVES; LOCATING OR PRESENCE-DETECTING BY USE OF THE REFLECTION OR RERADIATION OF RADIO WAVES; ANALOGOUS ARRANGEMENTS USING OTHER WAVES
- G01S7/00—Details of systems according to groups G01S13/00, G01S15/00, G01S17/00
- G01S7/02—Details of systems according to groups G01S13/00, G01S15/00, G01S17/00 of systems according to group G01S13/00
- G01S7/021—Auxiliary means for detecting or identifying radar signals or the like, e.g. radar jamming signals
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
- G06F18/2411—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on the proximity to a decision surface, e.g. support vector machines
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2218/00—Aspects of pattern recognition specially adapted for signal processing
- G06F2218/12—Classification; Matching
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Artificial Intelligence (AREA)
- General Engineering & Computer Science (AREA)
- Life Sciences & Earth Sciences (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Remote Sensing (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- Radar, Positioning & Navigation (AREA)
- Computer Networks & Wireless Communication (AREA)
- Signal Processing (AREA)
- Radar Systems Or Details Thereof (AREA)
Abstract
The present invention relates to Radar Signal Recognition technical field, a kind of model evaluation method, Radar Signal Recognition method and corresponding intrument are provided.Wherein, model evaluation method includes: the training set and test set for obtaining radar signal；Feature extraction is carried out to the training sample in training set using intra-pulse feature analysis, obtains characteristic data set；Using the training process of TPOT method optimization characteristic data set, the disaggregated model for classifying to radar signal is obtained；The interpretation of the test sample in test set is calculated using LIME method；Whether can be used based on interpretation classification of assessment model.TPOT method energy Automatic Optimal trains process, can obtain the disaggregated model and manpower intervention training process of high quality, time saving and energy saving.And the interpretation of test sample is calculated by LIME method, and the availability of classification of assessment model accordingly, be conducive to the transparency for improving model, increase user to the trusting degree of model, improves the quality of the model of acquisition.
Description
Technical field
The present invention relates to Radar Signal Recognition technical fields, in particular to a kind of model evaluation method, radar signal
Recognition methods and corresponding intrument.
Background technique
With the development of modern technologies, the technology of the complexity of New Type Radar and counterreconnaissance, counter-measure is more and more mature,
Different Radar emitter pulses is identified from from intensive stream of radar pulses and is become more and more important, identification level is
Measure the important symbol of the technologically advanced degree of radar countermeasure set.
At this stage, Radar Signal Recognition, which is substantially, uses machine learning method, however most of machine learning methods are all
Be it is applicable in a specific range, need the artificial selection algorithm that goes, adjustment hyper parameter etc., if data set changes, in advance
The accuracy of survey will be affected.At the same time, although the available good prediction effect of model of machine learning training,
But in most cases, the model of acquisition is the unknowable black box in an inside, this will bring asking for the confidence level of model
Topic.
Summary of the invention
In view of this, the embodiment of the present invention provides a kind of model evaluation method, Radar Signal Recognition method and corresponding intrument,
To solve the above technical problems.
To achieve the above object, the invention provides the following technical scheme:
In a first aspect, the embodiment of the present invention provides a kind of model evaluation method, comprising:
Obtain the training set and test set of radar signal；
Feature extraction is carried out to the training sample in training set using intra-pulse feature analysis, obtains characteristic data set；
Using based on tree construction automatic flow optimization (Tree-based Pipeline Optimization Tool,
TPOT) method optimizes the training process of characteristic data set, obtains the disaggregated model for classifying to radar signal；
Utilize intelligible explanation (the Local Interpretable Model-Agnostic unrelated with model in part
Explanations, LIME) method, calculate the interpretation of the test sample in test set；
Whether can be used based on interpretation classification of assessment model.
In a kind of possible implementation of first aspect, using intra-pulse feature analysis to the training sample in training set
Carry out feature extraction, comprising:
Extract the Wavelet Ridge frequency cascade nature of training sample；
Extract the backbone section feature of the ambiguity function of training sample；
It is the feature of training sample by Wavelet Ridge frequency cascade nature and backbone section Fusion Features.
In a kind of possible implementation of first aspect, optimize the side TPOT using the automatic flow based on tree construction
Method optimizes the training process of characteristic data set, obtains the disaggregated model for classifying to radar signal, comprising:
Pretreatment is carried out to the feature that characteristic is concentrated using TPOT method choice characteristic processing operator and feature is selected
It selects, obtains treated characteristic data set；
Using the training process of the characteristic data set after TPOT method optimization processing, determine disaggregated model classifier and
Hyper parameter.
In a kind of possible implementation of first aspect, characteristic processing operator is SelectKBest, singular value decomposition
Principal component analysis and one of Variance Threshold operator.
In a kind of possible implementation of first aspect, classifier be decision tree classifier, random forest grader,
One of Gradient Boosting classifier, support vector machines, logistic regression and k nearest neighbour classification algorithm classifier.
In a kind of possible implementation of first aspect, the intelligible explanation LIME unrelated with model in part is utilized
Method calculates the interpretation of the test sample in test set, comprising:
Test sample in test set is disturbed, disturbance sample set is obtained；
Prediction result based on the similitude between the disturbance sample in disturbance sample set to disturbance sample on disaggregated model
Influence, determine interpretable model；
Prediction result corresponding feature weight of the test sample on disaggregated model is calculated using interpretable model, and will be special
Sign weight determines that test sample is directed to the interpretation of prediction result.
In a kind of possible implementation of first aspect, whether can be used, wraps based on interpretation classification of assessment model
It includes:
Judge whether same test sample is close for the feature weight of different prediction results；
If being close, and the prediction result of the test sample is correct, determine in the prediction result of disaggregated model exist be not easy
The radar signal type of differentiation, and then determine that disaggregated model is unavailable.
Second aspect, the embodiment of the present invention provide a kind of Radar Signal Recognition method, comprising:
Obtain practical radar signal sample；
Feature extraction is carried out to practical radar signal sample using intra-pulse feature analysis, obtains signal characteristic；
Based on signal characteristic, predicted using type of the trained disaggregated model to practical radar signal sample,
In, trained disaggregated model is to be commented by the model that the possible implementation of any one of first aspect or first aspect provides
Valence method is evaluated as available disaggregated model；
Calculate and export the interpretation of the prediction result of practical radar signal sample.
The third aspect, the embodiment of the present invention provide a kind of model evaluation device, comprising:
Sample acquisition module, for obtaining the training set and test set of radar signal；
Characteristic extracting module is obtained for carrying out feature extraction to the training sample in training set using intra-pulse feature analysis
Obtain characteristic data set；
Model optimization module optimizes characteristic data set for optimizing TPOT method using the automatic flow based on tree construction
Training process, obtain disaggregated model for classifying to radar signal；
Interpretation computing module, for calculating and surveying using the intelligible explanation LIME method unrelated with model in part
Try the interpretation for the test sample concentrated；
Availability judgment module, for whether can be used based on interpretation classification of assessment model.
Fourth aspect, the embodiment of the present invention provide a kind of Radar Signal Recognition device, comprising:
Sample acquisition module, for obtaining practical radar signal sample；
Characteristic extracting module is obtained for carrying out feature extraction to practical radar signal sample using intra-pulse feature analysis
Signal characteristic；
Prediction module, for being based on signal characteristic, using trained disaggregated model to the class of practical radar signal sample
Type is predicted, wherein trained disaggregated model is by the possible realization side of any one of first aspect or first aspect
The model evaluation method that formula provides is evaluated as available disaggregated model；
Interpretation computing module, the interpretation of the prediction result for calculating and exporting practical radar signal sample.
5th aspect, the embodiment of the present invention provide a kind of computer storage medium, meter are stored in computer storage medium
Calculation machine program instruction when computer program instructions are read and run by the processor of computer, executes first aspect or first party
The step of method that the possible implementation of any one of face provides.
6th aspect, the embodiment of the present invention provide a kind of electronic equipment, including processor and computer storage medium, meter
It is stored with computer program instructions in calculation machine storage medium, when computer program instructions are read out by the processor and run, executes the
The step of method that the possible implementation of any one of one side or first aspect provides.
Technical solution provided by the invention at least has the following beneficial effects: model evaluation side provided in an embodiment of the present invention
Method and device use TPOT method Automatic Optimal training process during model training, can obtain the classification mould of high quality
Type, and it is not necessarily to manpower intervention training process, it is time saving and energy saving.Meanwhile after training model, test is also calculated by LIME method
The interpretation of sample, and the availability of classification of assessment model accordingly are conducive to the transparency for improving model, increase user to mould
The trusting degree of type improves the quality of the model of acquisition.
Radar Signal Recognition method and device provided in an embodiment of the present invention is evaluated as using by above-mentioned model evaluation method
Available disaggregated model carries out the prediction of practical radar signal sample, can obtain preferable recognition result.Meanwhile it will also be practical
The interpretation of radar signal sample is exported to user, and the basis for forecasting of disaggregated model is understood convenient for user, thus to model
Quality is further assessed.
To enable above-mentioned purpose of the invention, technical scheme and beneficial effects to be clearer and more comprehensible, special embodiment below, and
Cooperate appended attached drawing, is described in detail below.
Detailed description of the invention
In order to illustrate the technical solution of the embodiments of the present invention more clearly, below will be to needed in the embodiment attached
Figure is briefly described, it should be understood that the following drawings illustrates only certain embodiments of the present invention, therefore is not construed as pair
The restriction of range for those of ordinary skill in the art without creative efforts, can also be according to this
A little attached drawings obtain other relevant attached drawings.
Fig. 1 shows a kind of structural block diagram that can be applied to the electronic equipment in the embodiment of the present invention；
Fig. 2 shows the flow charts for the model evaluation method that first embodiment of the invention provides；
Fig. 3 shows the flow chart of the step S11 of the model evaluation method of first embodiment of the invention offer；
Fig. 4 shows the flow chart of the step S12 of the model evaluation method of first embodiment of the invention offer；
Fig. 5 shows the flow chart of the step S13 of the model evaluation method of first embodiment of the invention offer；
Fig. 6 (A) to Fig. 6 (C) shows the schematic diagram of calculation result of interpretation；
Fig. 7 shows the functional block diagram of the model evaluation device of third embodiment of the invention offer；
Fig. 8 shows the functional block diagram of the Radar Signal Recognition device of fourth embodiment of the invention offer.
Specific embodiment
Below in conjunction with attached drawing in the embodiment of the present invention, technical solution in the embodiment of the present invention carries out clear, complete
Ground description, it is clear that described embodiments are only a part of the embodiments of the present invention, instead of all the embodiments.Usually exist
The component of the embodiment of the present invention described and illustrated in attached drawing can be arranged and be designed with a variety of different configurations herein.Cause
This, is not intended to limit claimed invention to the detailed description of the embodiment of the present invention provided in the accompanying drawings below
Range, but it is merely representative of selected embodiment of the invention.Based on the embodiment of the present invention, those skilled in the art are not doing
Every other embodiment obtained under the premise of creative work out, shall fall within the protection scope of the present invention.
It should also be noted that similar label and letter indicate similar terms in following attached drawing, therefore, once a certain Xiang Yi
It is defined in a attached drawing, does not then need that it is further defined and explained in subsequent attached drawing.Meanwhile of the invention
In description, term " first ", " second " etc. are only used for distinguishing description, are not understood to indicate or imply relative importance.
Fig. 1 shows the structural schematic diagram of electronic equipment 100 provided in an embodiment of the present invention.Referring to Fig.1, electronic equipment
100 include memory 102, storage control 104, one or more (one is only shown in figure) processors 106, Peripheral Interface
108, radio-frequency module 110, audio-frequency module 112, display module 114 etc..These components pass through one or more communication bus/signal
Line 116 mutually communicates.
Memory 102 can be used for storing software program and module, as in the embodiment of the present invention model evaluation method and
Corresponding program instruction/the module of device, Radar Signal Recognition method and device, processor 106 are stored in memory by operation
Software program and module in 102, it is such as provided in an embodiment of the present invention thereby executing various function application and data processing
Model evaluation method and device, Radar Signal Recognition method and device.
Memory 102 may be, but not limited to, random access memory (Random Access Memory, RAM), only
It reads memory (Read Only Memory, ROM), programmable read only memory (Programmable Read-Only
Memory, PROM), erasable read-only memory (Erasable Programmable Read-Only Memory, EPROM),
Electricallyerasable ROM (EEROM) (Electric Erasable Programmable Read-Only Memory, EEPROM)
Deng.Processor 106 and other possible components can carry out the access of memory 102 under the control of storage control 104.
Processor 106 can be a kind of IC chip, the processing capacity with signal.It specifically can be general procedure
Device, including central processing unit (Central Processing Unit, CPU), micro-control unit (Micro Controller
Unit, MCU), network processing unit (Network Processor, NP) or other conventional processors；It can also be dedicated processes
Device, including digital signal processor (Digital Signal Processor, DSP), specific integrated circuit (Application
Specific Integrated Circuits, ASIC), field programmable gate array (Field Programmable Gate
Array, FPGA) either other programmable logic device, discrete gate or transistor logic, discrete hardware components.It can be with
Realize or execute disclosed each method, step and the logic diagram in the embodiment of the present invention.
Various input/output devices are couple processor 106 and memory 102 by Peripheral Interface 108.In some implementations
In example, Peripheral Interface 108, processor 106 and storage control 104 can be realized in one single chip.In some other reality
In example, they can be realized by independent chip respectively.
Radio-frequency module 110 is used to receive and transmit electromagnetic wave, realizes the mutual conversion of electromagnetic wave and electric signal, thus with
Communication network or other equipment are communicated.
Audio-frequency module 112 provides a user audio interface, may include one or more microphones, one or more raises
Sound device and voicefrequency circuit.
Display module 114 provides a display interface between electronic equipment 100 and user.Specifically, display module 114
Video output is shown to user, and the content of these videos output may include text, figure, video and any combination thereof.
It is appreciated that structure shown in FIG. 1 is only to illustrate, electronic equipment 100 may also include it is more than shown in Fig. 1 or
Less component, or with the configuration different from shown in Fig. 1.Each component shown in Fig. 1 can using hardware, software or its
Combination is realized.In the embodiment of the present invention, electronic equipment 100 can be server, personal computer, Intelligent mobile equipment, intelligence
The equipment that energy wearable device and intelligent vehicle-carried equipment etc. have calculation processing power, can also be not limited to physical equipment, such as
It can be virtual machine, Cloud Server etc..
First embodiment
Fig. 2 shows the flow charts for the model evaluation method that first embodiment of the invention provides.Referring to Fig. 2, this method packet
It includes:
S10: the processor 106 of electronic equipment 100 obtains the training set and test set of radar signal.
The data of radar signal can be acquired by professional equipment, and the data of acquisition are divided into training set and test set, training
All include at least one radar signal sample in collection and test set, the radar signal sample in training set be known as training sample,
Radar signal sample in test set is known as test sample.
S11: the processor 106 of electronic equipment 100 carries out the training sample in training set using intra-pulse feature analysis special
Sign is extracted, and characteristic data set is obtained.
Intra-pulse feature analysis is a kind of method for extracting radar signal feature, and each training sample is after intra-pulse feature analysis
A corresponding feature vector and the feature of the training sample are calculated, the set of the feature of all training samples in training set
The corresponding characteristic data set of composing training collection.
S12: the processor 106 of electronic equipment 100 utilizes TPOT method, optimizes the training process of characteristic data set, obtains
Disaggregated model for classifying to radar signal.
TPOT is a kind of method of Automatic Optimal machine-learning process, this method be recorded in R.S.OLSON and
Paper " the TPOT:ATree-based Pipeline Optimization Tool for that J.H.MOORE was delivered in 2016
Automating Machine Learning " etc. in documents.Model training is carried out using characteristic data set, while utilizing TPOT
Method optimizes training process, can obtain the optimal disaggregated model based on current signature data set, which can be right
The classification of radar signal is predicted.
S13: the processor 106 of electronic equipment 100 utilizes LIME method, calculates the interpretable of the test sample in test set
Property.
LIME gives a kind of definition mode of interpretation of disaggregated model and calculation method, this method are recorded in
Paper " the ＂ Why Should I Trust that Ribeiro M T, Singh S and Guestrin C were delivered in 2016
You? ＂: Explaining the Predictions of Any Classifier " etc. in documents.Interpretation is to classification
A kind of mathematical description of the classification foundation of model, interpretation ξ (x) can pass through formulaDefinition, wherein G is a kind of interpretable model, and f is that signal is classified category of model and arrives
The probability of some classification, L (f, g, πx) it is a measurement πxF and the how many function of g difference in the range of definition, Ω are complicated
Property measurement.
The purpose that interpretation proposes is the course of work for facilitating people to understand disaggregated model, avoids model complete to user
It is rendered as the state of complete black box, user is generated can not to the classification results of model and really trusts.
Calculate test sample interpretation, can all test samples all calculated, also can choose therein
Several are calculated.
S14: the processor 106 of electronic equipment 100 is based on whether interpretation classification of assessment model can be used.
Interpretation based on test sample can clearly know current disaggregated model present in the assorting process
Problem, to judge whether current disaggregated model can be used, designated herein is available, refers to for actual Radar Signal Recognition
Task effectively classifies to practical radar signal sample, and the standard of availability is not construed as limiting herein.
The availability of classification of assessment model can take the mode manually evaluated, can also be using machine automatic Evaluation
Mode is not construed as limiting herein.
In short, above-mentioned model evaluation method uses TPOT method Automatic Optimal training process during model training,
The disaggregated model and manpower intervention training process of high quality can be obtained, it is time saving and energy saving.Meanwhile after training model,
The interpretation of test sample, and the availability of classification of assessment model accordingly are also calculated by LIME method, are conducive to improve mould
The transparency of type increases user to the trusting degree of model.
It should be understood that user after knowing the availability of disaggregated model, can select available model for actual
Radar Signal Recognition task can obtain preferable recognition result due to improving the quality of model.Alternatively, user can
To carry out training or update further directed to property to disaggregated model according to the evaluation result to model, not available model is turned
Available model is turned to, or is advanced optimized on the basis of available model, its quality is improved.
Fig. 3 shows the flow chart of the step S11 of the model evaluation method of first embodiment of the invention offer.Reference Fig. 3,
In a kind of embodiment of first embodiment, step S11 be may further include:
Step S110: the processor 106 of electronic equipment 100 extracts the Wavelet Ridge frequency cascade nature of training sample.
It is primarily based on formulaWavelet transformation is carried out to training sample, wherein a is
Translational movement, b are scale amount, and f (t) is training sample,For continuous wavelet basic function.
It is then based on formulaCalculate the Wavelet Ridge frequency feature ξ (b) for obtaining training sample, wherein ω0It is small
Wave centre frequency, arIt (b) is wavelet ridge.
Wavelet Ridge frequency cascade nature is finally obtained based on Wavelet Ridge frequency feature calculation.
Step S111: the processor 106 of electronic equipment 100 extracts the backbone section feature of the ambiguity function of training sample.
It is primarily based on formulaThe backbone section for obtaining ambiguity function is calculated,
In, RS (α) is detection vector, uαFor score field, CαFor uαOn relational operator, ρ uαRadial distance；
It is then based on following equation:
It can calculate and obtain backbone section featureWherein,For uαOn backbone section.It may be noted that
Backbone section feature have it is multiple,Only one of feature, remaining backbone section feature also have corresponding formula into
Row calculates, it is contemplated that the backbone section for calculating ambiguity function is characterized in a kind of existing method, therefore no longer illustrates one by one here.
Step S112: the processor 106 of electronic equipment 100 is by Wavelet Ridge frequency cascade nature and backbone section Fusion Features
For the feature of training sample.
The mode for merging two kinds of features is not construed as limiting, for example, can using the feature vector obtained after two kinds of merging features as
The feature of training sample.
Fig. 4 shows the flow chart of the step S12 of the model evaluation method of first embodiment of the invention offer.Reference Fig. 4,
In a kind of embodiment of first embodiment, step S12 be may further include:
Step S120: the processor 106 of electronic equipment 100 is using TPOT method choice characteristic processing operator to characteristic
The feature of concentration carries out pretreatment and feature selecting, obtains treated characteristic data set.
Wherein, characteristic processing operator at least can from SelectKBest, singular value decomposition principal component analysis and
A kind of operator is selected in Variance Threshold, TPOT method carries out the selection of feature operator automatically.Pretreatment includes pair
Feature vector such as is modified, converts at the processing, and feature selecting refers to select in pretreated feature vector and be suitable for training
The best features (a part of vector) that sample is classified, after the set that the best features of each training sample are constituted is handled
Characteristic data set.
Step S121: the processor 106 of electronic equipment 100 utilizes the instruction of the characteristic data set after TPOT method optimization processing
Practice process, determines the classifier and hyper parameter of disaggregated model.
Process optimization in step S121 executes automatically, can be obtained optimal disaggregated model after training, the model
With optimal classifier and optimal hyper parameter.Wherein, classifier be decision tree classifier, random forest grader,
One of Gradient Boosting classifier, support vector machines, logistic regression and k nearest neighbour classification algorithm classifier,
Optimal classifier is automatically selected by TPOT method.
The degree of automation that can be seen that TPOT from step S120 and step S121 is very high, avoids model training
In selection algorithm gone by manual type, the operation such as adjustment hyper parameter, disaggregated model classification performance that is time saving and energy saving, while obtaining
Also more preferable.
Fig. 5 shows the flow chart of the step S13 of the model evaluation method of first embodiment of the invention offer.Reference Fig. 5,
In a kind of embodiment of first embodiment, step S13 may include:
Step S130: the processor 106 of electronic equipment 100 disturbs the test sample in test set, is disturbed
Sample set.
Subtle disturbance is carried out to test sample, it is therefore intended that emulate actual radar signal.
Step S131: the processor 106 of electronic equipment 100 is based on the similitude pair between the disturbance sample in disturbance sample set
The influence of prediction result of the sample on disaggregated model is disturbed, determines interpretable model.
According to the definition of step S13 squadron interpretation, model, that is, model recited above G can be explained, it is however generally that, it can
Interpretation model is a local weighted naive model, and it is easy for calculating interpretation based on this model.Wherein, here pre-
Result, that is, disaggregated model is surveyed to the classification results of sample of signal, prediction result is to carry out feature extraction to sample of signal (to adopt
With intra-pulse feature analysis) after, it is input to output obtained in disaggregated model.
Step S132: the processor 106 of electronic equipment 100 calculates test sample on disaggregated model using interpretable model
The corresponding feature weight of prediction result, and by feature weight determine test sample be directed to prediction result interpretation.
Test sample obtains multiple prediction results after being input to disaggregated model, and each prediction result indicates a type of thunder
Up to signal, each prediction result can have different probability.Test sample, which is input to interpretable model, can calculate survey
Sample sheet is directed to the feature weight of each prediction result, this feature weight is defined as interpretation.
Fig. 6 (A) to Fig. 6 (C) shows the schematic diagram of calculation result of interpretation.Fig. 6 (A) to Fig. 6 (C) is shown pair
The case where modulation system of radar signal is identified, certainly in practice, the parameter for describing radar signal are not limited to signal
Modulation system, here only example.
Fig. 6 (A), Fig. 6 (B) and Fig. 6 (C) respectively correspond a kind of prediction result of test sample, i.e., are modulated using BPSK
Signal, using OFDM modulation signal and using COSTA modulation signal.By taking Fig. 6 (A) as an example, the number of bottom
0.41 indicates that test sample is directed to the feature weight of the signal using BPSK modulation, and this feature weight is by each feature above
(Feature1, Feature2 etc.) contribution obtains (summation).The case where Fig. 6 (B) and Fig. 6 (C), is similar with Fig. 6 (A), no longer by
One illustrates.
The interpretation that test signal is described according to features described above weight, then in step S14, disaggregated model be can be used
Property can be evaluated in the following way:
First determine whether same test sample is close for the feature weight of different prediction results.If being close,
And the prediction result of the test sample is correct, determines the radar signal class for existing in the prediction result of disaggregated model and not being easily distinguishable
Type, the prediction result that feature weight is close are exactly the radar signal type not being easily distinguishable.At this time, it is believed that disaggregated model is not
These signals can be effectively distinguished, and then determine that disaggregated model is unavailable.If not being close, it is usually expressed as some prediction knot
The feature weight of fruit is much larger than the feature weight for other prediction results, such as the feature weight in Fig. 6 (A) is noticeably greater than and schemes
In 6 (B) and Fig. 6 (C), show that the disaggregated model can effectively distinguish different prediction results, the disaggregated model is available.
It for not available disaggregated model, can be further processed, for example, the disaggregated model new based on training set, this point
The prediction result of class model only includes the radar signal type not being easily distinguishable.After training using with it is above-mentioned it is similar by the way of evaluate
Its availability shows that this few class radar signal cannot be distinguished really, user should be allowed to know this feelings if still unavailable
Condition.If new disaggregated model is available, initial disaggregated model can be updated, enable to have radar signal
Effect is distinguished.
It may be noted that the interpretable index that all can serve as classification of assessment model availability of each test sample output,
In practice, the quantity of the test sample of interpretation to be calculated can be selected according to demand.
Second embodiment
Second embodiment of the invention provides a kind of Radar Signal Recognition method, can identify the modulation methods including radar signal
Parameter including formula.This method obtains practical radar signal sample first, practical radar signal sample can by professional equipment into
Row acquisition.Then feature extraction is carried out to practical radar signal sample using intra-pulse feature analysis, obtains signal characteristic, specific side
Method can be with reference to the associated description in first embodiment.Later based on the signal characteristic extracted, trained classification mould is utilized
Type predicts the type of practical radar signal sample, wherein trained disaggregated model is provided by first embodiment
Model evaluation method is evaluated as available disaggregated model, be also possible to be evaluated as in practice certainly it is unavailable, but improve after
Available disaggregated model.The interpretation of the prediction result of practical radar signal sample is finally calculated and exported, is managed convenient for user
The basis for forecasting of disaggregated model is solved, deepens user to the confidence level of model, or user can also be helped to the quality of model
Further assessed.
It is appreciated that in some embodiments, it, can not also be defeated if user does not obtain the demand of interpretation
Interpretation out.Or user has trusted the disaggregated model by interpretation within certain time completely, it later can be with
Interpretation is calculated, no longer to reduce computational burden.
Second embodiment of the invention does not refer to place, can be not repeated to explain with reference to the corresponding contents in first embodiment
It states.
3rd embodiment
Fig. 7 shows the functional block diagram of the model evaluation device 200 of third embodiment of the invention offer.It, should referring to Fig. 7
Device include sample acquisition module 210, characteristic extracting module 220, model optimization module 230, interpretation computing module 240 with
And availability judgment module 250.
Wherein, sample acquisition module 210 is used to obtain the training set and test set of radar signal；
Characteristic extracting module 220 is used to carry out feature extraction to the training sample in training set using intra-pulse feature analysis,
Obtain characteristic data set；
Model optimization module 230 is used to optimize TPOT method using the automatic flow based on tree construction, optimizes characteristic
The training process of collection obtains the disaggregated model for classifying to radar signal；
Interpretation computing module 240 is used to calculate using the intelligible explanation LIME method unrelated with model in part
The interpretation of test sample in test set；
Availability judgment module 250 is used for whether can be used based on interpretation classification of assessment model.
The model evaluation device 200 that third embodiment of the invention provides, the technical effect of realization principle and generation is the
It has been illustrated in one embodiment, to briefly describe, 3rd embodiment part does not refer to place, can refer to corresponding in first embodiment
Content.
Fourth embodiment
Fig. 8 shows the functional block diagram of the Radar Signal Recognition device 300 of fourth embodiment of the invention offer.Referring to figure
8, which includes sample acquisition module 310, characteristic extracting module 320, prediction module 330 and interpretation computing module
340。
Wherein, sample acquisition module 310 is for obtaining practical radar signal sample；
Characteristic extracting module 320 is used to carry out feature extraction to practical radar signal sample using intra-pulse feature analysis, obtains
Obtain signal characteristic；
Prediction module 330 is used to be based on signal characteristic, using trained disaggregated model to practical radar signal sample
Type is predicted, wherein trained disaggregated model is by the possible realization of any one of first aspect or first aspect
The model evaluation method that mode provides is evaluated as available disaggregated model；
Interpretation computing module 340 is used to calculate and export the interpretation of the prediction result of practical radar signal sample
The technical effect of the Radar Signal Recognition device 300 that fourth embodiment of the invention provides, realization principle and generation
It has illustrated in a second embodiment, to briefly describe, fourth embodiment part does not refer to place, can refer in second embodiment
Corresponding contents.
5th embodiment
Fifth embodiment of the invention provides a kind of computer storage medium, and computer journey is stored in computer storage medium
Sequence instruction, when computer program instructions are read and run by the processor of computer, executes method provided in an embodiment of the present invention
The step of.The computer storage medium can be implemented as, but be not limited to memory 102 shown in fig. 1.
Sixth embodiment
Sixth embodiment of the invention provides a kind of electronic equipment, including processor and computer storage medium, computer
It is stored with computer program instructions in storage medium and executes the present invention when computer program instructions are read out by the processor and run
The step of method that embodiment provides.The electronic equipment can be implemented as, but be not limited to electronic equipment 100 shown in fig. 1.
It should be noted that all the embodiments in this specification are described in a progressive manner, each embodiment weight
Point explanation is the difference from other embodiments, and the same or similar parts between the embodiments can be referred to each other.
For device class embodiment, since it is basically similar to the method embodiment, so being described relatively simple, related place ginseng
See the part explanation of embodiment of the method.
In several embodiments provided herein, it should be understood that disclosed device and method can also pass through it
Its mode is realized.The apparatus embodiments described above are merely exemplary, for example, the flow chart and block diagram in attached drawing are aobvious
The device of multiple embodiments according to the present invention, architectural framework in the cards, the function of method and computer program product are shown
It can and operate.In this regard, each box in flowchart or block diagram can represent one of a module, section or code
Point, a part of the module, section or code includes one or more for implementing the specified logical function executable
Instruction.It should also be noted that function marked in the box can also be attached to be different from some implementations as replacement
The sequence marked in figure occurs.For example, two continuous boxes can actually be basically executed in parallel, they sometimes may be used
To execute in the opposite order, this depends on the function involved.It is also noted that each of block diagram and or flow chart
The combination of box in box and block diagram and or flow chart can be based on the defined function of execution or the dedicated of movement
The system of hardware is realized, or can be realized using a combination of dedicated hardware and computer instructions.
In addition, each functional module in each embodiment of the present invention can integrate one independent portion of formation together
Point, it is also possible to modules individualism, an independent part can also be integrated to form with two or more modules.
It, can be with if the function is realized and when sold or used as an independent product in the form of software function module
It is stored in computer-readable storage medium.Based on this understanding, technical solution of the present invention is substantially in other words to existing
Having the part for the part or the technical solution that technology contributes can be embodied in the form of software products, the computer
Software product is stored in a storage medium, including some instructions are used so that computer equipment executes each embodiment institute of the present invention
State all or part of the steps of method.Computer equipment above-mentioned includes: personal computer, server, mobile device, intelligently wears
The various equipment with execution program code ability such as equipment, the network equipment, virtual unit are worn, storage medium above-mentioned includes: U
Disk, mobile hard disk, read-only memory, random access memory, magnetic disk, tape or CD etc. are various to can store program code
Medium.
The foregoing is only a preferred embodiment of the present invention, is not intended to restrict the invention, for the skill of this field
For art personnel, the invention may be variously modified and varied.All within the spirits and principles of the present invention, made any to repair
Change, equivalent replacement, improvement etc., should all be included in the protection scope of the present invention.It should also be noted that similar label and letter exist
Similar terms are indicated in following attached drawing, therefore, once being defined in a certain Xiang Yi attached drawing, are then not required in subsequent attached drawing
It is further defined and explained.
The above description is merely a specific embodiment, but scope of protection of the present invention is not limited thereto, any
Those familiar with the art in the technical scope disclosed by the present invention, can easily think of the change or the replacement, and should all contain
Lid is within protection scope of the present invention.Therefore, the protection scope of the present invention shall be subject to the protection scope of the claims.
It should be noted that, in this document, relational terms such as first and second and the like are used merely to a reality
Body or operation are distinguished with another entity or operation, are deposited without necessarily requiring or implying between these entities or operation
In any actual relationship or order or sequence.Moreover, the terms "include", "comprise" or its any other variant are intended to
Non-exclusive inclusion, so that the process, method, article or equipment including a series of elements is not only wanted including those
Element, but also including other elements that are not explicitly listed, or further include for this process, method, article or equipment
Intrinsic element.In the absence of more restrictions, the element limited by sentence "including a ...", it is not excluded that
There is also other identical elements in process, method, article or equipment including the element.
Claims (10)
1. a kind of model evaluation method characterized by comprising
Obtain the training set and test set of radar signal；
Feature extraction is carried out to the training sample in the training set using intra-pulse feature analysis, obtains characteristic data set；
Optimize TPOT method using the automatic flow based on tree construction, optimizes the training process of the characteristic data set, used
In the disaggregated model classified to radar signal；
Using the intelligible explanation LIME method unrelated with model in part, solving for the test sample in the test set is calculated
The property released；
Evaluate whether the disaggregated model can be used based on the interpretation.
2. model evaluation method according to claim 1, which is characterized in that described to utilize intra-pulse feature analysis to the instruction
Practice the training sample concentrated and carry out feature extraction, comprising:
Extract the Wavelet Ridge frequency cascade nature of the training sample；
Extract the backbone section feature of the ambiguity function of the training sample；
It is the feature of the training sample by the Wavelet Ridge frequency cascade nature and backbone section Fusion Features.
3. model evaluation method according to claim 1, which is characterized in that described to utilize the automatic flow based on tree construction
Optimize TPOT method, optimize the training process of the characteristic data set, obtains the classification mould for classifying to radar signal
Type, comprising:
Pretreatment and spy are carried out to the feature that the characteristic is concentrated using the TPOT method choice characteristic processing operator
Sign selection, obtains treated characteristic data set；
The training process for optimizing treated the characteristic data set using the TPOT method, determines point of the disaggregated model
Class device and hyper parameter.
4. model evaluation method according to claim 3, which is characterized in that the characteristic processing operator is
One of SelectKBest, the principal component analysis of singular value decomposition and VarianceThreshold operator.
5. model evaluation method according to claim 3, which is characterized in that the classifier be decision tree classifier, with
In machine forest classified device, GradientBoosting classifier, support vector machines, logistic regression and k nearest neighbour classification algorithm
A kind of classifier.
6. model evaluation method according to any one of claims 1-5, which is characterized in that described to be managed using part
The explanation LIME method unrelated with model of solution, calculates the interpretation of the test sample in the test set, comprising:
Test sample in the test set is disturbed, disturbance sample set is obtained；
Based on it is described disturbance sample set in disturbance sample between similitude to the disturbance sample on the disaggregated model
The influence of prediction result determines interpretable model；
The corresponding feature of prediction result of the test sample on the disaggregated model is calculated using the interpretable model to weigh
Weight, and the feature weight is determined that the test sample is directed to the interpretation of the prediction result.
7. model evaluation method according to claim 6, which is characterized in that described based on described in interpretation evaluation
Whether disaggregated model can be used, comprising:
Judge whether same test sample is close for the feature weight of different prediction results；
If being close, and the prediction result of the test sample is correct, determine in the prediction result of the disaggregated model exist be not easy
The radar signal type of differentiation, and then determine that the disaggregated model is unavailable.
8. a kind of Radar Signal Recognition method characterized by comprising
Obtain practical radar signal sample；
Feature extraction is carried out to the practical radar signal sample using intra-pulse feature analysis, obtains signal characteristic；
Based on the signal characteristic, carried out using type of the trained disaggregated model to the practical radar signal sample pre-
It surveys, wherein the trained disaggregated model is to be evaluated as by model evaluation method of any of claims 1-7
Available disaggregated model；
Calculate and export the interpretation of the prediction result of the practical radar signal sample.
9. a kind of model evaluation device characterized by comprising
Sample acquisition module, for obtaining the training set and test set of radar signal；
Characteristic extracting module is obtained for carrying out feature extraction to the training sample in the training set using intra-pulse feature analysis
Obtain characteristic data set；
Model optimization module optimizes the characteristic data set for optimizing TPOT method using the automatic flow based on tree construction
Training process, obtain disaggregated model for classifying to radar signal；
Interpretation computing module, for calculating the survey using the intelligible explanation LIME method unrelated with model in part
Try the interpretation for the test sample concentrated；
Availability judgment module, for evaluating whether the disaggregated model can be used based on the interpretation.
10. a kind of Radar Signal Recognition device characterized by comprising
Sample acquisition module, for obtaining practical radar signal sample；
Characteristic extracting module is obtained for carrying out feature extraction to the practical radar signal sample using intra-pulse feature analysis
Signal characteristic；
Prediction module, for being based on the signal characteristic, using trained disaggregated model to the practical radar signal sample
Type predicted, wherein the trained disaggregated model be commented by model of any of claims 1-7
Valence method is evaluated as available disaggregated model；
Interpretation computing module, the interpretation of the prediction result for calculating and exporting the practical radar signal sample.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810861202.3A CN109002810A (en) | 2018-08-01 | 2018-08-01 | Model evaluation method, Radar Signal Recognition method and corresponding intrument |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810861202.3A CN109002810A (en) | 2018-08-01 | 2018-08-01 | Model evaluation method, Radar Signal Recognition method and corresponding intrument |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN109002810A true CN109002810A (en) | 2018-12-14 |
Family
ID=64597394
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810861202.3A Pending CN109002810A (en) | 2018-08-01 | 2018-08-01 | Model evaluation method, Radar Signal Recognition method and corresponding intrument |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109002810A (en) |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109800441A (en) * | 2019-02-01 | 2019-05-24 | 北京金山数字娱乐科技有限公司 | A kind of model output recommended method and device, model export recommender system |
| CN111563761A (en) * | 2020-01-19 | 2020-08-21 | 深圳前海微众银行股份有限公司 | Crowd expansion method, device, equipment and storage medium |
| CN111759305A (en) * | 2019-04-02 | 2020-10-13 | 纬创资通股份有限公司 | Living body detection method and living body detection system |
| WO2022012347A1 (en) * | 2020-07-14 | 2022-01-20 | International Business Machines Corporation | Predictive models having decomposable hierarchical layers configured to generate interpretable results |
| CN114064440A (en) * | 2022-01-18 | 2022-02-18 | 恒生电子股份有限公司 | Training method of credibility analysis model, credibility analysis method and related device |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106054189A (en) * | 2016-07-17 | 2016-10-26 | 西安电子科技大学 | Radar target recognition method based on dpKMMDP model |
| CN107194433A (en) * | 2017-06-14 | 2017-09-22 | 电子科技大学 | A kind of Radar range profile's target identification method based on depth autoencoder network |
-
2018
- 2018-08-01 CN CN201810861202.3A patent/CN109002810A/en active Pending
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106054189A (en) * | 2016-07-17 | 2016-10-26 | 西安电子科技大学 | Radar target recognition method based on dpKMMDP model |
| CN107194433A (en) * | 2017-06-14 | 2017-09-22 | 电子科技大学 | A kind of Radar range profile's target identification method based on depth autoencoder network |
Non-Patent Citations (1)
| Title |
|---|
| WENQIANG ZHANG等: "Radar Signal Recognition Based on TPOT and LIME", 《PROCEEDINGS OF THE 37TH CHINESE CONTROL CONFERENCE》 * |
Cited By (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109800441A (en) * | 2019-02-01 | 2019-05-24 | 北京金山数字娱乐科技有限公司 | A kind of model output recommended method and device, model export recommender system |
| CN109800441B (en) * | 2019-02-01 | 2023-03-24 | 北京金山数字娱乐科技有限公司 | Model output recommendation method and device and model output recommendation system |
| CN111759305A (en) * | 2019-04-02 | 2020-10-13 | 纬创资通股份有限公司 | Living body detection method and living body detection system |
| CN111759305B (en) * | 2019-04-02 | 2023-08-25 | 纬创资通股份有限公司 | Living body detection method and living body detection system |
| CN111563761A (en) * | 2020-01-19 | 2020-08-21 | 深圳前海微众银行股份有限公司 | Crowd expansion method, device, equipment and storage medium |
| CN111563761B (en) * | 2020-01-19 | 2024-06-07 | 深圳前海微众银行股份有限公司 | Crowd expansion method, device, equipment and storage medium |
| WO2022012347A1 (en) * | 2020-07-14 | 2022-01-20 | International Business Machines Corporation | Predictive models having decomposable hierarchical layers configured to generate interpretable results |
| US11593680B2 (en) | 2020-07-14 | 2023-02-28 | International Business Machines Corporation | Predictive models having decomposable hierarchical layers configured to generate interpretable results |
| GB2611983A (en) * | 2020-07-14 | 2023-04-19 | Ibm | Predictive models having decomposable hierarchical layers configured to generate interretable results |
| CN114064440A (en) * | 2022-01-18 | 2022-02-18 | 恒生电子股份有限公司 | Training method of credibility analysis model, credibility analysis method and related device |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109002810A (en) | Model evaluation method, Radar Signal Recognition method and corresponding intrument | |
| CN110097037B (en) | Intelligent monitoring method and device, storage medium and electronic equipment | |
| CN107846392B (en) | Intrusion detection algorithm based on improved collaborative training-ADBN | |
| CN106973057B (en) | A kind of classification method suitable for intrusion detection | |
| CN103870751B (en) | Method and system for intrusion detection | |
| WO2019108603A1 (en) | Machine learning techniques for evaluating entities | |
| CN108664893A (en) | A kind of method for detecting human face and storage medium | |
| CN109359693A (en) | A kind of Power Quality Disturbance Classification Method | |
| CN109936582A (en) | Construct the method and device based on the PU malicious traffic stream detection model learnt | |
| CN110247910A (en) | A kind of detection method of abnormal flow, system and associated component | |
| Gagne et al. | Classification of convective areas using decision trees | |
| CN110297207A (en) | Method for diagnosing faults, system and the electronic device of intelligent electric meter | |
| CN103235953B (en) | A kind of method of optical fiber distributed perturbation sensor pattern recognition | |
| CN103957116A (en) | Decision-making method and system of cloud failure data | |
| CN109034076A (en) | A kind of automatic clustering method and automatic cluster system of mechanical fault signals | |
| CN111582315B (en) | Sample data processing method and device and electronic equipment | |
| CN115617784A (en) | Data processing system and processing method for informationized power distribution | |
| CN110619216A (en) | Malicious software detection method and system for adversarial network | |
| CN113343123A (en) | Training method and detection method for generating confrontation multiple relation graph network | |
| CN107025293A (en) | A kind of second power equipment defective data method for digging and system | |
| CN110807159B (en) | Data marking method and device, storage medium and electronic equipment | |
| CN113746780A (en) | Abnormal host detection method, device, medium and equipment based on host image | |
| CN116318925A (en) | Multi-CNN fusion intrusion detection method, system, medium, equipment and terminal | |
| CN105930430A (en) | Non-cumulative attribute based real-time fraud detection method and apparatus | |
| CN106530199B (en) | Multimedia integration steganalysis method based on window type hypothesis testing |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20181214 |
|
| RJ01 | Rejection of invention patent application after publication |