Disclosure of Invention
The embodiment of the application provides a method, a system, electronic equipment and a storage medium for denoising a gas absorption spectrum signal, which are used for at least solving the problem that the gas absorption spectrum signal obtained by a traditional denoising method in the related art is low in precision.
In a first aspect, an embodiment of the present application provides a method for constructing a gas absorption spectrum signal filter, where the method includes:
Acquiring an absorption spectrum signal of a target gas;
generating a training data set and a verification data set according to the absorption spectrum signals based on a data enhancement technology;
constructing an absorption spectrum signal filter based on a one-dimensional convolutional neural network and a frequency channel attention mechanism;
and training the absorption spectrum signal filter through the training data set and the verification data set to obtain the absorption spectrum signal filter of the target gas.
In some of these embodiments, the generating training data sets and verification data sets from the absorbance spectrum signals based on data enhancement techniques includes:
normalizing the absorption spectrum signals and taking an average value to obtain base signals;
simulating to generate a noise-free signal based on the acquisition condition of the absorption spectrum signal;
the training data set and the validation data set are generated from the base signal and the noise-free signal.
In some of these embodiments, said generating said training data set and said validation data set from said base signal and said noise-free signal comprises:
Preprocessing the base signal and the noiseless signal to obtain a preprocessed base signal and a preprocessed noiseless signal;
based on a cross validation method, the preprocessing base signal and the preprocessing noise-free signal are randomly divided into a training data set and a validation data set according to a preset proportion.
In some embodiments, the preprocessing the base signal and the noiseless signal to obtain a preprocessed base signal and a preprocessed noiseless signal includes:
multiplying the base signal by a preset amplification coefficient and a preset translation coefficient respectively, and superposing Gaussian noise to obtain the preprocessed base signal;
And respectively multiplying the noiseless signal by a preset amplification factor and a preset translation factor to obtain the preprocessing noiseless signal.
In some of these embodiments, the absorption spectrum signal filter includes a feature extraction block, an attention block, and a feature construction learning block; the construction of the absorption spectrum signal filter based on the one-dimensional convolutional neural network and the frequency channel attention mechanism comprises the following steps:
the feature extraction block is constructed and comprises a plurality of one-dimensional convolution blocks with different input and output channel data and is used for generating an initial feature map of the input data;
Constructing the attention block based on the frequency channel attention mechanism, wherein the attention block is used for generating a weight matrix according to a preset weight rule, and obtaining a weighted feature map based on the weight matrix and the initial feature map;
And constructing the feature learning block, wherein the feature learning block is used for outputting denoising data corresponding to the input data based on the weighted feature map.
In some of these embodiments, the acquiring the absorption spectrum signal of the target gas includes:
Based on the TDLAS technology, obtaining direct absorption spectrum signals and wavelength modulation spectrum signals corresponding to target gases with different concentrations;
and obtaining the absorption spectrum signal according to the direct absorption spectrum signal and the wavelength modulation spectrum signal, wherein the absorption spectrum signal is a one-dimensional spectrum signal.
In a second aspect, an embodiment of the present application provides a method for evaluating performance of a gas absorption spectrum signal filter, where the method includes:
Acquiring direct absorption spectrum signals and wavelength modulation spectrum signals of target gases with different concentrations;
Constructing a test dataset based on the direct absorption spectrum signal and the wavelength-modulated spectrum signal;
Inputting the test data set into a gas absorption spectrum signal filter constructed based on the method of the first aspect to obtain denoising data;
And according to the denoising data, assessing the denoising effect of the gas absorption spectrum signal filter.
In a third aspect, an embodiment of the present application provides a system for constructing a gas absorption spectrum signal filter, where the system includes: a signal acquisition module, a data set generation module, a filter construction module and a filter training module, wherein,
The signal acquisition module is used for acquiring an absorption spectrum signal of the target gas, wherein the absorption spectrum signal comprises a direct absorption spectrum signal and a wavelength modulation spectrum signal;
The data set generating module is used for generating a training data set and a verification data set according to the absorption spectrum signals based on a data enhancement technology;
The filter construction module is used for constructing an absorption spectrum signal filter based on a one-dimensional convolutional neural network and a frequency channel attention mechanism;
the filter training module is used for training the absorption spectrum signal filter through the training data set and the verification data set to obtain the absorption spectrum signal filter of the target gas.
In a fourth aspect, an embodiment of the present application provides a computer device, including a memory, a processor, and a computer program stored in the memory and capable of running on the processor, where the processor implements the method for constructing a gas absorption spectrum signal filter according to the first aspect when executing the computer program.
In a fifth aspect, an embodiment of the present application provides a computer readable storage medium, on which a computer program is stored, which when executed by a processor, implements a method for constructing a gas absorption spectrum signal filter according to the first aspect.
Compared with the related art, the method for constructing the gas absorption spectrum signal filter provided by the embodiment of the application has the advantages that the absorption spectrum signal of the target gas is obtained, the training data set and the verification data set are generated according to the absorption spectrum signal based on the data enhancement technology, the absorption spectrum signal filter is constructed based on the one-dimensional convolutional neural network and the frequency channel attention mechanism, the absorption spectrum signal filter is trained through the training data set and the verification data set, the absorption spectrum signal filter of the target gas is obtained, and the problem that the gas absorption spectrum signal accuracy obtained by the traditional denoising method is low is solved. The gas absorption spectrum signal filter constructed based on the method can effectively remove low-frequency optical noise mixed in a signal base line while retaining high-frequency details of the signal, and obtain a signal with higher signal-to-noise ratio, thereby improving the measurement accuracy of the absorption spectrum signal. Meanwhile, a data enhancement method is utilized to construct a simulation data set to complete training, and the problem of scarcity of gas absorption experimental data is solved.
Detailed Description
The present application will be described and illustrated with reference to the accompanying drawings and examples in order to make the objects, technical solutions and advantages of the present application more apparent. It should be understood that the specific embodiments described herein are for purposes of illustration only and are not intended to limit the scope of the application. All other embodiments, which can be made by a person of ordinary skill in the art based on the embodiments provided by the present application without making any inventive effort, are intended to fall within the scope of the present application.
It is apparent that the drawings in the following description are only some examples or embodiments of the present application, and it is possible for those of ordinary skill in the art to apply the present application to other similar situations according to these drawings without inventive effort. Moreover, it should be appreciated that while such a development effort might be complex and lengthy, it would nevertheless be a routine undertaking of design, fabrication, or manufacture for those of ordinary skill having the benefit of this disclosure, and thus should not be construed as having the benefit of this disclosure.
Reference in the specification to "an embodiment" means that a particular feature, structure, or characteristic described in connection with the embodiment may be included in at least one embodiment of the application. The appearances of such phrases in various places in the specification are not necessarily all referring to the same embodiment, nor are separate or alternative embodiments mutually exclusive of other embodiments. It is to be expressly and implicitly understood by those of ordinary skill in the art that the described embodiments of the application can be combined with other embodiments without conflict.
Unless defined otherwise, technical or scientific terms used herein should be given the ordinary meaning as understood by one of ordinary skill in the art to which this application belongs. The terms "a," "an," "the," and similar referents in the context of the application are not to be construed as limiting the quantity, but rather as singular or plural. The terms "comprising," "including," "having," and any variations thereof, are intended to cover a non-exclusive inclusion; for example, a process, method, system, article, or apparatus that comprises a list of steps or modules (elements) is not limited to only those steps or elements but may include other steps or elements not expressly listed or inherent to such process, method, article, or apparatus. The terms "connected," "coupled," and the like in connection with the present application are not limited to physical or mechanical connections, but may include electrical connections, whether direct or indirect. The term "plurality" as used herein means two or more. "and/or" describes an association relationship of an association object, meaning that there may be three relationships, e.g., "a and/or B" may mean: a exists alone, A and B exist together, and B exists alone. The character "/" generally indicates that the context-dependent object is an "or" relationship. The terms "first," "second," "third," and the like, as used herein, are merely distinguishing between similar objects and not representing a particular ordering of objects.
The embodiment provides a method for constructing a gas absorption spectrum signal filter. Fig. 1 is a flowchart of a method for constructing a gas absorption spectrum signal filter according to an embodiment of the present application, as shown in fig. 1, the flowchart includes the steps of:
Step 101, acquiring an absorption spectrum signal of a target gas.
In some of these embodiments, step 101 comprises:
step 1012, based on the TDLAS technique, obtaining direct absorption spectrum signals and wavelength modulation spectrum signals corresponding to target gases with different concentrations.
Step 1013, obtaining an absorption spectrum signal according to the direct absorption spectrum signal and the wavelength modulation spectrum signal, wherein the absorption spectrum signal is a one-dimensional spectrum signal.
With a gas sensor based on TDLAS technology, direct Absorption Spectrum (DAS) signals and Wavelength Modulation Spectrum (WMS) signals of the gas are measured. The obtained absorption spectrum signals are one-dimensional spectrum signals, the wavelength range covers absorption peaks to be detected of the gas, and each group of signals consists of a preset number (for example, 2000) of sampling points.
Step S102, based on the data enhancement technology, a training data set and a verification data set are generated according to the absorption spectrum signals.
In some of these embodiments, step S102 specifically includes:
step S1021, normalizing and averaging the absorption spectrum signals to obtain a base signal.
And respectively measuring absorption spectrum signals corresponding to a plurality of groups (for example, 1000 groups) of gases with different concentrations based on two measurement modes of WMS and DAS. And normalizing and averaging the multiple groups of gas absorption data to remove time random electronic noise, and extracting signals of two measurement modes only containing low-frequency optical noise as a base signal.
Step S1022, based on the acquisition condition of the absorption spectrum signal, a noise-free signal is generated by simulation.
A noise-free signal under the same acquisition conditions as the base signal is generated using an analog program.
Step S1023, generating a training data set and a verification data set according to the base signal and the noise-free signal.
The data set generated by the base signal is used as a training set, and the data set generated by the noise-free signal is used as a development set. The same proportion of data is acquired from the training set and the development set for model training and verification, respectively. The training set and the development set are the same size, such as data sets of 50000×2000.
In some of these embodiments, step S1023 specifically includes:
step S201, preprocessing the base signal and the noise-free signal to obtain a preprocessed base signal and a preprocessed noise-free signal.
Noisy and noiseless data sets required for training are generated by data enhancement techniques.
In some embodiments, step S201 specifically includes:
in step S2011, the base signal is multiplied by a preset amplification factor and a preset translation factor, and gaussian noise is superimposed, so as to obtain a preprocessed base signal.
In step S2012, the noiseless signal is multiplied by the preset amplification factor and the preset translation factor, respectively, to obtain a preprocessed noiseless signal.
The base signal and the noise-free signal are multiplied by an amplification factor and a shifting factor, respectively. Adjusting the signal amplitude through an amplification factor to simulate the absorption intensity signals of gases with different concentrations; the horizontal position of the signal is adjusted through the translation coefficient so as to simulate wavelength drift generated by long-term operation of the laser. Gaussian noise is superimposed on the data set constructed from the base signal to simulate random electronic noise.
Fig. 2 is a schematic diagram of a signal amplifying process according to an embodiment of the present application, and as shown in fig. 2, the amplitude of the signal is adjusted by an amplifying factor, so as to obtain a plurality of signals with the same horizontal position and different amplitudes, so as to simulate the absorption intensity signals of gases with different concentrations.
Fig. 3 is a schematic diagram of signal translation processing according to an embodiment of the present application, as shown in fig. 3, in which the signal level is adjusted by a translation coefficient to obtain a plurality of signals with the same amplitude and different level positions, so as to simulate wavelength drift generated by long-term operation of a laser.
Step S202, based on a cross-validation method, the preprocessing base signal and the preprocessing noise-free signal are randomly divided into a training data set and a validation data set according to a preset proportion.
The data sets are randomly divided into training data sets and verification data sets according to a set proportion (4:1) by adopting a cross verification method (such as Hold-Out cross verification method).
The training is completed by constructing the data set by using the data enhancement method, so that the problem of scarcity of experimental data of gas absorption is solved.
With continued reference to fig. 1, after the data set of the target gas absorption spectrum signal is acquired, step S103 is continued.
Step S103, an absorption spectrum signal filter is constructed based on the one-dimensional convolutional neural network and the frequency channel attention mechanism.
A one-dimensional convolutional neural network and a frequency channel attention mechanism are utilized to construct a gas absorption spectrum filter model, and the model is suitable for data measured by the WMS method and the DAS method.
The filter comprises three parts, namely a feature extraction block, an attention block and a feature learning block. The input data are measured WMS and DAS absorption signals, and the input data are one-dimensional data. Each row in the dataset is entered into the model, then the input size may be 1 x 2000.
In some of these embodiments, the absorption spectrum signal filter includes a feature extraction block, an attention block, and a feature construction learning block, and step S103 includes:
In step S1031, a feature extraction block is constructed, where the feature extraction block includes a plurality of one-dimensional convolution blocks with different input/output channel data, and is used to generate an initial feature map of the input data.
The feature extraction block comprises 3 one-dimensional convolution blocks, and each convolution block sequentially comprises 4 parts of a convolution layer, CELU activation functions, layer normalization and one-dimensional maximum pooling. Wherein the convolutional layer parameter is set to Kernel size=2, stride=2; the max pooling layer parameter is set to Kernel size=2, stride=2.
The number of input and output channels of the three convolution blocks are set to (1, 25), (25, 100) and (100, 200), respectively. The 1×2000 data is input to the feature extraction block, and the output data is a feature map of 200×31.
The convolutional neural network is used as a supervised deep learning method, can extract the feature map layer by layer, effectively extracts and utilizes the effective information in the training data, and has higher performance and lower calculation complexity.
Step S1032, constructing an attention block based on the frequency channel attention mechanism, wherein the attention block is used for generating a weight matrix according to a preset weight rule, and obtaining a weighted feature map based on the weight matrix and the initial feature map.
The attention block is built based on a frequency channel attention mechanism (FCANet) including DCT weight calculation, full connection layer, reLU activation function, full connection layer, and Sigmoid activation function. The K-order DCT expansion formula F (K) for the timing signal F (n) is:
where N is the length of the timing signal f (N). And taking the 2-order DCT expansion coefficient of each channel of the feature map as compressed channel information to generate a weight matrix through super-parameter optimization. The number of input and output units of the two full connection layers may be set to (200, 12) and (12, 200), respectively. The weight matrix is multiplied by the feature map output by the feature extraction to obtain a new feature map after the weight is overlapped, namely a weighted feature map, and the size of the feature map is kept unchanged.
The frequency channel attention mechanism is introduced into the convolutional neural network, so that the feature map extracted by the convolutional neural network can be effectively screened, and secondary information is ignored, thereby enhancing the generalization capability of the model. The frequency channel attention mechanism expands global average pooling in the classical channel attention mechanism into DCT expansion, and enhances the performance of an attention module.
Step S1033, a feature learning block is constructed, and the feature learning block is configured to output denoising data corresponding to the input data based on the weighted feature map.
The feature learning block includes, in order, a full connection layer, CELU activation functions, a Dropout layer, a full connection layer, and CELU activation functions. The feature learning block is used for learning the extracted image features of the feature map, capturing the connection between channels and further carrying out denoising processing. The parameter settings for the two fully connected layers may be (6200, 2000) and (2000 ), respectively. Dropout rate=0.2. The output of the feature extraction block is a denoised signal of size 1 x 2000.
FIG. 4 is a schematic diagram of a denoising process of a gas absorption spectrum signal filter model according to an embodiment of the present application, wherein as shown in FIG. 4, an initial feature map is obtained by convolution blocks of three different channel numbers for an input signal, and a weight matrix is produced by an attention block; multiplying the initial feature map by the weight matrix to obtain a weighted feature map, namely a dimension reduction data; the feature learning block learns the extracted image features of the feature map, performs denoising processing, and outputs a denoising signal.
Step S104, training the absorption spectrum signal filter through the training data set and the verification data set to obtain the absorption spectrum signal filter of the target gas.
Training data sets and validation data sets input the filter model to begin training. The training parameters may be set to: batch size=500; LEARNING RATE = 1e-4; epochs = 800; optimizer=adam.
The loss function adopted by the absorption spectrum signal filter model training is a Mean Square Error (MSE) function, and the loss function is as follows:
Wherein n is the total number of sampling points, yi is the i standard value, namely the data value corresponding to the noise-free signal, and f (wxi +b) is the i predicted value, namely the data value corresponding to the denoising signal obtained based on the absorption spectrum signal filter.
And judging the difference between the predicted value obtained based on the absorption spectrum signal filter and the standard value corresponding to the noiseless signal based on the loss function, wherein the smaller the loss function is, the better the robustness of the constructed filter model is.
FIG. 5 is a graph comparing the denoising effect of a direct absorption spectrum signal according to an embodiment of the present application, as shown in FIG. 5, the average signal-to-noise ratio of the direct absorption spectrum signal before filtering is 2.29dB, the standard deviation is 4.29ppm, and after denoising by the target gas absorption spectrum signal filter constructed as described above, the average signal-to-noise ratio is 12.89dB, and the standard deviation is 0.421ppm.
Fig. 6 is a graph comparing denoising effects of a wavelength modulation spectrum signal according to an embodiment of the present application, as shown in fig. 6, the average signal-to-noise ratio of the wavelength modulation spectrum signal before filtering is 7.06dB, the standard deviation is 78.6ppb, and after denoising by the above-constructed target gas absorption spectrum signal filter, the average signal-to-noise ratio is 21.83dB, and the standard deviation is 30.40ppb.
In summary, the gas absorption spectrum signal filter constructed based on the steps can overcome the defects that the traditional filtering algorithm is easy to lose high-frequency details in signals and has poor filtering capability on low-frequency optical noise. The gas absorption spectrum signal filter constructed based on the steps can effectively remove low-frequency optical noise mixed in a signal base line while retaining high-frequency details of signals, so that higher signal-to-noise ratio and measurement accuracy are obtained. And after the model is trained, parameters are not required to be manually adjusted, and the denoising efficiency is high.
Through the steps, the absorption spectrum signal of the target gas is obtained, the training data set and the verification data set are generated according to the absorption spectrum signal based on the data enhancement technology, the absorption spectrum signal filter is constructed based on the one-dimensional convolutional neural network and the frequency channel attention mechanism, and the absorption spectrum signal filter is trained through the training data set and the verification data set to obtain the absorption spectrum signal filter of the target gas, so that the problem that the accuracy of the gas absorption spectrum signal obtained by the traditional denoising method is low is solved. The gas absorption spectrum signal filter constructed based on the method can effectively remove low-frequency optical noise mixed in a signal base line while retaining high-frequency details of the signal, and obtain a signal with higher signal-to-noise ratio, thereby improving the measurement accuracy of the absorption spectrum signal. Meanwhile, a data enhancement method is utilized to construct a simulation data set to complete training, and the problem of scarcity of gas absorption experimental data is solved.
The embodiment also provides a performance evaluation method of the gas absorption spectrum signal filter, fig. 7 is a flowchart of the performance evaluation method of the gas absorption spectrum signal filter according to the embodiment of the application, as shown in fig. 7, and the method includes:
In step S701, direct absorption spectrum signals and wavelength modulation spectrum signals of target gases with different concentrations are obtained.
Based on WMS and DAS two measurement modes, multiple groups of gas absorption signals with different concentrations are respectively measured
Step S702, constructing a test dataset based on the direct absorption spectrum signal and the wavelength-modulated spectrum signal.
A test dataset is constructed based on the data obtained by the two measurement modes. The dimensions of each row of the test dataset are the same as those of the training dataset described above.
In step S703, the test data set is input into the gas absorption spectrum signal filter constructed based on the method of the first aspect, so as to obtain denoising data.
And inputting the test data set into a trained gas absorption spectrum signal filter for testing.
Step S704, according to the denoising data, the denoising effect of the gas absorption spectrum signal filter is evaluated.
The denoising effect of the gas absorption spectrum signal filter is checked and evaluated through the signal-to-noise ratio of the denoising data.
And storing the trained filter model, and integrating the filter model into a gas sensor system to directly filter signals measured by the sensor. After the model is trained, parameters do not need to be manually adjusted, and denoising efficiency is high.
It should be noted that the steps illustrated in the above-described flow or flow diagrams of the figures may be performed in a computer system, such as a set of computer-executable instructions, and that, although a logical order is illustrated in the flow diagrams, in some cases, the steps illustrated or described may be performed in an order other than that illustrated herein.
The embodiment also provides a system for constructing a gas absorption spectrum signal filter, which is used for implementing the above embodiment and the preferred implementation manner, and the description is omitted. As used below, the terms "module," "unit," "sub-unit," and the like may be a combination of software and/or hardware that implements a predetermined function. While the means described in the following embodiments are preferably implemented in software, implementation in hardware, or a combination of software and hardware, is also possible and contemplated.
Fig. 8 is a block diagram of a construction system of a gas absorption spectrum signal filter according to an embodiment of the present application, as shown in fig. 8, the system includes: a signal acquisition module 81, a data set generation module 82, a filter construction module 83, and a filter training module 84, wherein,
The signal acquisition module 81 is configured to acquire an absorption spectrum signal of the target gas, where the absorption spectrum signal includes a direct absorption spectrum signal and a wavelength modulation spectrum signal.
The data set generating module 82 is configured to generate a training data set and a verification data set according to the absorption spectrum signal based on the data enhancement technique.
The filter construction module 83 is configured to construct an absorption spectrum signal filter based on the one-dimensional convolutional neural network and the frequency channel attention mechanism.
The filter training module 84 is configured to train the absorption spectrum signal filter by using the training data set and the verification data set to obtain the absorption spectrum signal filter of the target gas.
In some of these embodiments, the data set generation module 82 includes: a base signal production model, a noise-free signal production model, and a dataset acquisition model.
The base signal production model is used for normalizing the absorption spectrum signals and taking the average value to obtain base signals;
And the noiseless signal production model is used for generating the noiseless signal in a simulation mode based on the acquisition conditions of the absorption spectrum signals.
A data set acquisition model for generating a training data set and a validation data set from the base signal and the noise-free signal.
In some of these embodiments, the dataset acquisition model comprises: the preprocessing model and the dataset build model.
And the preprocessing model is used for preprocessing the base signal and the noiseless signal to obtain a preprocessed base signal and a preprocessed noiseless signal.
The data set construction model is used for randomly dividing the preprocessing base signal and the preprocessing noiseless signal into a training data set and a verification data set according to a preset proportion based on a cross verification method.
In some of these embodiments, the preprocessing model includes: a base signal preprocessing model and a noise-free signal preprocessing model.
The base signal preprocessing model is used for multiplying the base signal by a preset amplification coefficient and a preset translation coefficient respectively and superposing Gaussian noise to obtain a preprocessed base signal;
the noiseless signal preprocessing model is used for respectively multiplying the noiseless signal by a preset amplification coefficient and a preset translation coefficient to obtain a preprocessed noiseless signal.
In some of these embodiments, the filter construction module 83 includes: the device comprises a feature extraction block building module, an attention block building module, a feature learning block building module and a filter model building module.
The feature extraction block construction module is used for constructing a feature extraction block, and the feature extraction block comprises a plurality of one-dimensional convolution blocks with different input and output channel data and is used for generating an initial feature map of the input data;
The feature learning block construction module is used for constructing an attention block based on a frequency channel attention mechanism, wherein the attention block is used for generating a weight matrix according to a preset weight rule and obtaining a weighted feature map based on the weight matrix and an initial feature map.
The filter model building module is used for building a feature learning block, and the feature learning block is used for outputting denoising data corresponding to input data based on the weighted feature map.
And the filter model construction module is used for obtaining the absorption spectrum signal filter according to the feature extraction block, the attention block and the feature building learning block.
In some of these embodiments, the signal acquisition module 81 includes: the system comprises an acquisition module and a construction module.
And the acquisition module is used for acquiring direct absorption spectrum signals and wavelength modulation spectrum signals corresponding to the target gases with different concentrations based on the TDLAS technology.
The construction module is used for obtaining an absorption spectrum signal according to the direct absorption spectrum signal and the wavelength modulation spectrum signal, wherein the absorption spectrum signal is a one-dimensional spectrum signal.
Through the system, the signal acquisition module 81 acquires the absorption spectrum signal of the target gas, wherein the absorption spectrum signal comprises a direct absorption spectrum signal and a wavelength modulation spectrum signal, the data set generation module 82 generates a training data set and a verification data set according to the absorption spectrum signal based on a data enhancement technology, the filter construction module 83 constructs an absorption spectrum signal filter based on a one-dimensional convolutional neural network and a frequency channel attention mechanism, and the filter training module 84 trains the absorption spectrum signal filter through the training data set and the verification data set to obtain the absorption spectrum signal filter of the target gas, so that the problem that the gas absorption spectrum signal obtained by a traditional denoising method is low in accuracy is solved. The gas absorption spectrum signal filter constructed based on the method can effectively remove low-frequency optical noise mixed in a signal base line while retaining high-frequency details of the signal, and obtain a signal with higher signal-to-noise ratio, thereby improving the measurement accuracy of the absorption spectrum signal. Meanwhile, a data enhancement method is utilized to construct a simulation data set to complete training, and the problem of scarcity of gas absorption experimental data is solved.
The above-described respective modules may be functional modules or program modules, and may be implemented by software or hardware. For modules implemented in hardware, the various modules described above may be located in the same processor; or the above modules may be located in different processors in any combination.
The present embodiment also provides an electronic device comprising a memory having stored therein a computer program and a processor arranged to run the computer program to perform the steps of any of the method embodiments described above.
Optionally, the electronic device may further include a transmission device and an input/output device, where the transmission device is connected to the processor, and the input/output device is connected to the processor.
Alternatively, in the present embodiment, the above-described processor may be configured to execute the following steps by a computer program:
s1, acquiring an absorption spectrum signal of the target gas.
S2, based on a data enhancement technology, generating a training data set and a verification data set according to the absorption spectrum signals.
S3, constructing an absorption spectrum signal filter based on the one-dimensional convolutional neural network and the frequency channel attention mechanism.
And S4, training the absorption spectrum signal filter through the training data set and the verification data set to obtain the absorption spectrum signal filter of the target gas.
It should be noted that, specific examples in this embodiment may refer to examples described in the foregoing embodiments and alternative implementations, and this embodiment is not repeated herein.
In one embodiment, fig. 9 is a schematic diagram of an internal structure of an electronic device according to an embodiment of the present application, and as shown in fig. 9, there is provided an electronic device, which may be a server, and an internal structure diagram of which may be shown in fig. 9. The electronic device includes a processor, a memory, a network interface, and a database connected by a system bus. Wherein the processor of the electronic device is configured to provide computing and control capabilities. The memory of the electronic device includes a nonvolatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, computer programs, and a database. The internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage media. The database of the electronic device is for storing data. The network interface of the electronic device is used for communicating with an external terminal through a network connection. The computer program is executed by a processor to implement a method of constructing a gas absorption spectrum signal filter.
It will be appreciated by those skilled in the art that the structure shown in fig. 9 is merely a block diagram of a portion of the structure associated with the present application and is not limiting of the electronic device to which the present application is applied, and that a particular electronic device may include more or fewer components than shown, or may combine certain components, or have a different arrangement of components.
Those skilled in the art will appreciate that implementing all or part of the above described methods may be accomplished by way of a computer program stored on a non-transitory computer readable storage medium, which when executed, may comprise the steps of the embodiments of the methods described above. Any reference to memory, storage, database, or other medium used in embodiments provided herein may include non-volatile and/or volatile memory. The nonvolatile memory can include Read Only Memory (ROM), programmable ROM (PROM), electrically Programmable ROM (EPROM), electrically Erasable Programmable ROM (EEPROM), or flash memory. Volatile memory can include Random Access Memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in a variety of forms such as Static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double Data Rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link (SYNCHLINK) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), among others.
It should be understood by those skilled in the art that the technical features of the above-described embodiments may be combined in any manner, and for brevity, all of the possible combinations of the technical features of the above-described embodiments are not described, however, they should be considered as being within the scope of the description provided herein, as long as there is no contradiction between the combinations of the technical features.
The above examples illustrate only a few embodiments of the application, which are described in detail and are not to be construed as limiting the scope of the application. It should be noted that it will be apparent to those skilled in the art that several variations and modifications can be made without departing from the spirit of the application, which are all within the scope of the application. Accordingly, the scope of protection of the present application is to be determined by the appended claims.