Signal analysis method and system based on acquiring and identifying noise panoramic distribution model

Abstract

The present application belongs to the technical field of signal analysis, and particularly relates to a signal analysis method and system based on obtaining and identifying a noise panoramic distribution model, which comprises the following steps: S1, presetting a sample measurement duration; S2, repeatedly and continuously measuring in a rich-condition measurement environment to obtain multiple measurement results; S3, processing the measurement results of reference samples and test samples to form training data of the reference samples and the test samples, respectively; S4, taking the observability of noise as a convergence target to perform artificial intelligence model training; and S5, inputting the measurement results into the trained artificial intelligence model and outputting a result as a specific type of the sample to be identified. The present application is different from the noise processing scheme in the prior art and provides a signal analysis method based on obtaining and identifying a noise panoramic distribution model from a completely different technical perspective to solve the noise reduction problem that is difficult to handle in the prior art.

Description

Technical Field

[0001] This invention belongs to the field of signal analysis technology, and in particular relates to a signal analysis method and system based on acquiring and identifying a panoramic noise distribution model. Background Technology

[0002] To meet the diverse applications and needs across various practical fields, existing technologies offer a wide range of sample measurement methods. Theoretically, all types of sample measurements, even those that appear instantaneous, are essentially continuous measurement processes involving the integration of measured values ​​in the time domain. Depending on the type of sample being measured, different measurement durations are preset for the continuous measurement process to accumulate signal strength, ensuring that the measurement results meet the requirements of actual observation.

[0003] However, due to limitations in the measurement environment, equipment accuracy, and the inherent properties of the samples, the sample measurement results are inevitably mixed data containing both signal and noise. During continuous sample measurements, the noise, influenced by multiple factors, is almost certainly not ideal random noise, but rather nonlinear noise, often possessing a highly complex form and content. In such cases, completely resolving the noise problem requires removing the noise source or shielding the interference, but this approach is certainly impractical in real-world sample measurements. Furthermore, for complex noise, it is difficult to design denoising schemes using conventional engineering techniques that employ one or a few commonly used mathematical models.

[0004] Furthermore, signals with extremely weak strength and complex characteristics are highly likely to be submerged by noise. Conventional mathematical denoising methods struggle to handle such measurement results because it is difficult to establish a reasonable mathematical model to simulate and remove the noise mixed in the measurement results. Under stable measurement conditions, even if the signal strength can be enhanced through continuous measurement and integration, the complexity of the signal itself will still affect the extraction effect; on the other hand, continuous measurement inevitably leads to noise also being integrated, failing to improve the situation where the signal is submerged in noise, making it difficult to separate the signal from the noise. Summary of the Invention

[0005] The main objective of this invention is to provide a signal analysis method based on acquiring and identifying a panoramic noise distribution model, which aims to solve the technical problem that existing technologies are unable to analyze complex signals under ultra-low signal-to-noise ratio conditions.

[0006] To achieve the above objectives, the technical solution of this application is as follows:

[0007] A signal analysis method based on acquiring and identifying a panoramic noise distribution model includes the following steps:

[0008] S1: Preset the sample measurement duration according to the actual sample measurement requirements;

[0009] S2: Based on the sample measurement duration preset in step S1, repeated continuous measurements are performed on the reference sample and the test sample to obtain multiple measurement results; each measurement result contains a signal, as well as a noise panorama or at least a partial noise panorama formed under multiple noise observation dimensions; each continuous measurement is performed in a rich-condition measurement environment; rich conditions refer to: measurement conditions that are not aimed at maintaining the consistency of external conditions, do not involve noise suppression, are natural, and include real complex noise factors.

[0010] S3: Process the measurement results of the reference sample and the test sample to form training data for the reference sample and the test sample respectively;

[0011] S4: Based on the training data of the reference sample and the test sample, with the observability of noise as the convergence target, the artificial intelligence model is trained so that the model can identify the signal and noise from the measurement results and distinguish between the reference sample and the test sample.

[0012] S5: Input the measurement results of the sample to be identified into the trained artificial intelligence model. The output of the artificial intelligence model is the specific type of the sample to be identified.

[0013] Optionally, in step S2, during each continuous measurement of the reference sample and the test sample, a rich-condition measurement environment is created by repeatedly introducing slight perturbations. This creates multiple observation dimensions of noise during the continuous measurement of the samples, forming a noise panorama or at least a partial noise panorama in the measurement results.

[0014] Furthermore, in step S2, during each continuous measurement of the reference sample and the test sample, the method of introducing multiple slight perturbations is selected from continuous introduction, introduction at fixed intervals, or introduction at random intervals.

[0015] During each continuous measurement process, the overall measurement results of the reference sample and the test sample, as well as the signals in the measurement results, will each exhibit stable statistical characteristics; at the same time, as the noise panorama or at least a partial noise panorama is formed, the statistical distribution pattern of the noise will also tend to stabilize.

[0016] Furthermore, the form of the slight disturbance can be selected from, but is not limited to, spatial disturbance, physical disturbance, and environmental disturbance; spatial disturbance includes, but is not limited to: causing slight displacement of the measurement site, causing slight rotation of the measurement site; physical disturbance includes, but is not limited to: vibrating the measurement equipment or sample, agitating the fluid sample; environmental disturbance includes, but is not limited to: changing the ambient temperature, ambient humidity, electromagnetic field strength, and air pressure.

[0017] Further, in step S3, the steps of processing the measurement results of the reference sample and the test sample to form training data for the reference sample and the test sample respectively include:

[0018] S31. Normalize the measurement results of the reference sample and the test sample;

[0019] S32. Based on the normalization results of step S31, establish a posterior probability model framework;

[0020] The measurement results of the reference sample and the test sample, after processing in steps S31-S32, will generate training data that meets the requirements, which will be used for subsequent training of artificial intelligence models.

[0021] Furthermore, in step S4, the artificial intelligence model can be selected from, but is not limited to, artificial neural networks, perceptrons, support vector machines, Bayesian classifiers, Bayesian networks, random forest models, or clustering models.

[0022] Furthermore, in step S4, during the training process of the artificial intelligence model, the model will iteratively learn, summarize, and converge a large amount of empirical learning on the features contained in the training data that enable signal and noise recognition, as well as the features that enable the distinction between reference samples and test samples.

[0023] Among them, the features that enable signal and noise identification include the statistical distribution pattern of the measurement results of continuous measurement that conforms to the true mathematical statistical laws of the signal; the features that enable noise identification include the statistical distribution pattern of the noise panorama or at least part of the noise panorama formed during continuous measurement that approximates the true mathematical statistical laws of the noise; and the features that enable the differentiation between reference samples and test samples include the statistical distribution patterns of the continuous measurement results of the two types of samples respectively.

[0024] The present invention also provides a signal analysis system based on acquiring and identifying a panoramic noise distribution model, including a setting module, a measurement module, a processing module, a training module and an analysis module;

[0025] Based on actual sample measurement needs, the module can preset the sample measurement duration;

[0026] Based on the preset sample measurement duration of the setting module, the measurement module performs repeated continuous measurements on the reference sample and the test sample to obtain multiple measurement results. Each measurement result includes the signal, as well as a noise panorama or at least a partial noise panorama formed under multiple noise observation dimensions. Each continuous measurement is performed in a rich-condition measurement environment. Rich conditions refer to measurement conditions that are not aimed at maintaining the consistency of external conditions, do not involve noise suppression, are natural, and include real complex noise factors.

[0027] The processing module processes the measurement results of the reference sample and the test sample to form training data for the reference sample and the test sample, respectively.

[0028] Based on the training data of reference samples and test samples, the training module uses the observability of noise as the convergence target to train the artificial intelligence model, enabling the model to identify signals and noise from the measurement results and distinguish between reference samples and test samples.

[0029] The analysis module inputs the measurement results of the sample to be identified into a trained artificial intelligence model, and the output of the artificial intelligence model is the specific type of the sample to be identified.

[0030] Optionally, the measurement module includes a perturbation mechanism; during each continuous measurement of the reference and test samples by the measurement module, the perturbation mechanism creates a richly conditional measurement environment by introducing slight perturbations multiple times. This creates multiple observation dimensions of noise during the continuous measurement of the samples, forming a comprehensive noise panorama or at least a partial noise panorama in the measurement results.

[0031] Furthermore, during each continuous measurement of the reference sample and the test sample, the perturbation mechanism introduces multiple slight perturbations in a manner selected from continuous introduction, introduction at fixed intervals, or introduction at random intervals.

[0032] During each continuous measurement process of the measurement module, the overall measurement results of the reference sample and the test sample, as well as the signals in the measurement results, will exhibit stable statistical characteristics. At the same time, as the noise panorama or at least a partial noise panorama is formed, the statistical distribution pattern of the noise will also tend to stabilize.

[0033] Furthermore, the slight disturbances introduced by the perturbation mechanism can be in the form of, but are not limited to, spatial perturbations, physical perturbations, and environmental perturbations; spatial perturbations include, but are not limited to: causing slight displacement of the measurement site or slight rotation of the measurement site; physical perturbations include, but are not limited to: vibrating the measurement equipment or sample or agitating the fluid sample; environmental perturbations include, but are not limited to: changing the ambient temperature, ambient humidity, electromagnetic field strength, and air pressure.

[0034] Furthermore, the processing module includes a normalization module and a posterior probability module; wherein, the normalization module normalizes the measurement results of the reference sample and the test sample; the posterior probability module establishes a posterior probability model framework based on the normalization results, and forms training data for the reference sample and the test sample that meet the requirements, respectively, for subsequent artificial intelligence model training.

[0035] Furthermore, the artificial intelligence model can be selected from, but is not limited to, artificial neural networks, perceptrons, support vector machines, Bayesian classifiers, Bayesian networks, random forest models, or clustering models.

[0036] Furthermore, during the training process of the artificial intelligence model, the model will iteratively learn, summarize, and converge a large amount of empirical learning on the features contained in the training data that enable signal and noise recognition, as well as the features that enable the distinction between reference samples and test samples.

[0037] Among them, the features that enable signal and noise identification include the statistical distribution pattern of the measurement results of continuous measurement that conforms to the true mathematical statistical laws of the signal; the features that enable noise identification include the statistical distribution pattern of the noise panorama or at least part of the noise panorama formed during continuous measurement that approximates the true mathematical statistical laws of the noise; and the features that enable the differentiation between reference samples and test samples include the statistical distribution patterns of the continuous measurement results of the two types of samples respectively.

[0038] The beneficial effects of this application are as follows:

[0039] 1. This invention differs from existing noise processing solutions by providing a signal analysis method based on acquiring and identifying a panoramic noise distribution model from a completely different technical perspective, thereby solving the noise reduction problem that is difficult to handle with existing technologies.

[0040] The signal analysis method provided by this invention is based on mathematical statistics principles. While it does not directly separate signals from noise, it can still effectively distinguish between them and successfully identify multiple independent signals based on different measurement samples. This enables practical applications such as sample detection and material classification. Furthermore, this invention utilizes artificial intelligence technology to perform mixed modeling of noise and the signals submerged within it. Even if the noise lacks mathematical assumptions, the trained AI model can deeply uncover the hidden mathematical and statistical patterns in the measurement results, accurately obtaining the mathematical distribution model of the signal and noise.

[0041] 2. This invention integrates the signal through continuous sample measurements, improving signal observability while clearly revealing the relatively stable data distribution of the signal, which is beneficial for subsequent signal extraction and analysis. Furthermore, this invention does not impose uniform measurement conditions but actively creates a dynamic, condition-rich measurement environment by introducing slight perturbations. Noise from multiple observation dimensions will accumulate to form a comprehensive noise picture or at least a partial noise picture; that is, the accumulation of noise "samples" will almost completely cover all possibilities of the noise itself. Simultaneously, the noise distribution model will tend towards its true distribution form.

[0042] This invention can discover the mathematical and statistical laws of noise from the mixed data distribution of sample measurement results, and distinguish between noise and signal, as well as identify different types of signals, from the perspective of data distribution models. Compared with the direct noise removal and signal extraction performed in existing engineering technologies, this invention deeply explores the mathematical and statistical laws of noise and signal, avoiding the false elimination of signals by denoising operations and avoiding the influence of commonly used denoising steps on the signal itself. Therefore, this invention provides an effective solution to the problem of removing noise or extracting the signal itself from mixed sample measurement results, a problem that existing technologies cannot solve.

[0043] 3. Noise is an unavoidable influencing factor in sample measurements, and highly complex noise can even exhibit dynamic characteristics. Therefore, even with the best measurement conditions available at present, continuous sample measurements will exhibit fluctuations that are infinitely close to the real signal. However, these fluctuations can only be "statistically stable" dynamic changes near the real signal, and it is impossible to make pre-assumptions about the specific measured values ​​during dynamic changes.

[0044] However, the accumulation of measurement results through continuous sample measurements will gradually eliminate the influence of uncertainties in the aforementioned dynamic changes, leading to a stable data distribution model for the overall measurement results. This stable data distribution model represents the macroscopic set of all components interacting during the measurement process. That is, in addition to the actual signal, interference factors that may cause noise, such as environmental complexity, equipment accuracy, and the inherent influence of sampling methods, are all incorporated into the overall distribution model of the measurement results. Therefore, the overall distribution model of continuous sample measurement results can fully reflect its own characteristics. While the distribution model of the measurement results tends to stabilize, the noise panorama, or at least a partial noise panorama, formed by integrating multiple observation dimensions will also exhibit unique mathematical and statistical laws and approach the true noise distribution model. This invention distinguishes noise from signals by identifying the complete noise mathematical model, and this identification scheme will obtain more comprehensive and accurate identification results.

[0045] 4. Continuous sample measurements conducted under rich conditions enable noise from multiple observation dimensions to accumulate and form a comprehensive noise picture or at least a partial one, and the noise data distribution model tends to reflect the theoretically true distribution model of noise. In this context, this invention employs artificial intelligence technology to uncover the noise distribution model. The trained AI model can discover true features in high background noise data that meet the experimenter's analytical needs or objectives, providing more efficient mathematical calculations and outputting highly empirical and accurate analytical results in real time.

[0046] 5. This invention creates a conditionally rich measurement environment through different perturbation introduction methods to increase the noise observation dimension during continuous sample measurement, enabling the measurement results to reveal a complete noise panorama, or providing at least a partially accurate noise panorama for subsequent signal analysis. Different perturbation introduction methods have varying levels of practical difficulty and may lead to different effects in increasing the noise observation dimension. In the practical application of the technical solution involved in this invention, considering factors such as sample characteristics, sample measurement methods, and measurement accuracy requirements, experimenters can choose from the perturbation introduction methods provided by this invention according to their actual needs. The diverse perturbation introduction methods disclosed in this invention provide experimenters with a wide range of choices and, to some extent, reduce the difficulty of applying this invention, making the technical solution more valuable for widespread application. Attached Figure Description

[0047] Figure 1 A flowchart illustrating a signal analysis method based on acquiring and identifying a panoramic noise distribution model provided by the present invention;

[0048] Figure 2 Included with instruction manual Figure 1 The flowchart of step 3 in the signal analysis method shown is as follows;

[0049] Figure 3 A schematic diagram illustrating the principle of forming a noise panorama or at least a partial noise panorama during continuous measurement.

[0050] Figure 4 A system architecture diagram of a signal analysis system based on acquiring and identifying a panoramic noise distribution model provided by this invention. Detailed Implementation

[0051] The purpose of signal processing is to extract useful information from sample measurement results, such as content with research value or distinctive features that differentiate signals from others. Due to the many uncertainties in the sample measurement process, these "content with research value" and "distinctive features" often cannot be represented by independent numerical values, but are reflected by the overall statistical distribution of the signal data.

[0052] In actual sample measurements, the results inevitably contain noise in addition to the signal reflecting the true characteristics. Existing signal processing solutions either eliminate the noise from the measurement results or extract the signal. However, when the noise in the measurement results cannot be simulated using a "known" mathematical model, both noise elimination and signal extraction become extremely difficult.

[0053] The technical objective of this invention is to identify the "unknown" distribution models of signals and noise from measurement results based on statistical principles, thereby achieving effective differentiation between signals and noise. Furthermore, when measurement results originate from different measurement samples, accurate identification of sample types can be achieved by identifying differentiated data distribution models.

[0054] Regarding noise in measurement results, current techniques generally assume that the "ideal" mathematical statistical law of noise conforms to a Gaussian distribution. However, actual sample measurement processes typically cannot create an "ideal" noise environment. Furthermore, even by optimizing sample measurement conditions as much as possible through improving equipment accuracy and material purity, it is still possible to obtain measurement results with ideal analytical conditions. That is, the target signal to be analyzed may be submerged in noise due to its weak intensity, or the signal characteristics may be extremely complex and difficult to analyze. Mining data distribution models from such measurement results is often very difficult, and it may even be impossible to assume a noise distribution model at all.

[0055] In this context, the present invention achieves continuous measurement of samples under multi-dimensional perturbation environment by actively creating diverse measurement conditions. Thus, during the continuous measurement of samples, it enables comprehensive observation of noise, that is, it constructs a noise panorama or at least a partial noise panorama that can show the complete data statistical distribution model, and this data statistical distribution model will infinitely approach the true distribution of noise.

[0056] Specifically, for continuous measurements, the measurement result is equivalent to the integral of all transient measurements within the measurement duration. Due to the complexity of the noise environment, the noise level will differ for each transient measurement. From a sample and sampling perspective, the noise in each transient measurement is equivalent to a random sample from the overall noise population; random sampling cannot reflect the true characteristics. However, given that the noise exhibits specific statistical regularities and conforms to a specific data distribution model, when the integral of all transient measurements is equal to the continuous measurement result, the noise sampling range will expand to approximately the entire noise population, and the overall statistical regularities of the noise reflected in the continuous measurement results will tend to reflect the true situation. Therefore, from a statistical perspective, it is entirely theoretically feasible to reveal the statistical distribution model of noise through the construction of a panoramic view of noise in a perturbation environment.

[0057] Having acquired at least a partial overview of the noise and observing a relatively clear and stable distribution model, this invention employs artificial intelligence (AI) technology to deeply explore the statistical patterns of the noise. AI technology is an effective means applicable to various data analyses and solving empirical data processing problems. For example, deep learning models can simulate the human learning process, quickly summarizing human empirical data processing methods to achieve signal recognition and judgment. In this invention, the AI ​​model, trained on large datasets, guarantees the accuracy of its empirical analysis results, effectively identifying the mathematical distribution model of the noise and enabling subsequent specific analyses such as noise separation and signal classification.

[0058] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to embodiments. It should be understood that the specific embodiments described herein are for illustrative purposes only and are not intended to limit the invention.

[0059] Example 1

[0060] A signal analysis method based on acquiring and identifying a panoramic noise distribution model is described in the flowchart attached to the instruction manual. Figure 1 The method includes the following steps:

[0061] S1: Preset the sample measurement duration according to the actual sample measurement requirements;

[0062] In this embodiment of the invention, the sample measurement is not a transient measurement, but a measurement that lasts for a period of time, which needs to be preset before the sample measurement. For example, the preset measurement duration for spectral measurements is 3 seconds, and the preset measurement duration for electroencephalogram (EEG) signal measurements is 30 seconds. There is no explicit specification for the preset measurement duration; rather, it depends heavily on the specific sample type, measurement method, and actual measurement requirements. Specifically, for example, some signal measurements need to be sustained for a period of time to accumulate measurement values ​​and enhance the results; however, some optical measurements cannot be sustained for too long to avoid thermal damage to the sample. Therefore, when there are limitations or empirical solutions for continuous sample measurement, those skilled in the art of sample measurement can reasonably preset the measurement duration according to actual needs.

[0063] S2: Based on the sample measurement duration preset in step S1, repeated continuous measurements are performed on the reference sample and the test sample to obtain multiple measurement results; each measurement result contains a signal, as well as a noise panorama or at least a partial noise panorama formed under multiple noise observation dimensions; wherein, each continuous measurement is performed in a condition-rich measurement environment.

[0064] In this embodiment of the invention, step S1 presets the measurement duration for continuous sample measurement. For example, the measurement duration for spectral measurement is specified as 3 seconds, and the measurement duration for electroencephalogram (EEG) signal measurement is specified as 30 seconds; in step 2, continuous measurement of the corresponding samples is performed with the preset measurement duration.

[0065] In the field of signal acquisition and analysis, maintaining consistency of external conditions during sample measurement is a conventional method to reduce noise fluctuations and achieve a good signal-to-noise ratio. However, this invention does not involve setting consistent external conditions, but rather conducting repeated measurements of reference and test samples in a rich-condition measurement environment. Here, rich conditions refer to measurement conditions that are natural, do not involve noise suppression, and include real, complex noise factors, and are not aimed at maintaining consistency of external conditions.

[0066] In the embodiments of the present invention, as shown in the appendix to the specification... Figure 3 As shown, a single noise observation dimension cannot yield true and comprehensive noise observation results; that is, noise cannot exhibit complete statistical patterns that conform to its true distribution characteristics. However, this invention, by actively creating a condition-rich measurement environment, makes the noise observation dimension during continuous measurement process tend from local to comprehensive, which is conducive to forming a complete noise panorama or at least a partial noise panorama. While constructing the noise panorama, the statistical patterns of noise data will tend to their true mathematical statistical patterns. At the same time, limited by the inherent properties of the samples, the signal in the measurement results will always remain statistically unchanged.

[0067] In this embodiment of the invention, a noise panorama refers to a noise distribution model that fully reflects its theoretical true distribution model; a partial noise panorama refers to a noise distribution model that cannot fully reflect its theoretical true distribution model, but the distribution model already has the accuracy to be used for subsequent signal analysis.

[0068] Furthermore, repeatedly performing sample measurements is also considered an effective way to reduce random errors. However, in this embodiment of the invention, the main purpose of repeatedly performing continuous sample measurements is to obtain a sufficient number of measurement results for training the artificial intelligence model.

[0069] S3: Process the measurement results of the reference sample and the test sample to form training data for the reference sample and the test sample respectively;

[0070] S4: Based on the training data of the reference sample and the test sample, with the observability of noise as the convergence target, the artificial intelligence model is trained so that the model can identify the signal and noise from the measurement results and distinguish between the reference sample and the test sample.

[0071] In this embodiment of the invention, the training data of the reference sample and the test sample are randomly allocated as learning data and detection data respectively in a preset ratio. The artificial intelligence model is trained using the learning data, and the detection data is input into the trained artificial intelligence model to calculate the signal recognition result. If the signal recognition accuracy is lower than a preset threshold, training continues using the learning data; if the signal recognition accuracy is higher than the preset threshold, the artificial intelligence model is considered to have completed training.

[0072] S5: Input the measurement results of the sample to be identified into the trained artificial intelligence model. The output of the artificial intelligence model is the specific type of the sample to be identified.

[0073] In this embodiment of the invention, reference samples and test samples are used as two types of known samples. Multiple measurement results from both are processed to form training data, and the trained artificial intelligence model can effectively distinguish between the two types of known samples. When the sample to be identified is one of the two known samples, the artificial intelligence model can accurately identify the specific type of the sample.

[0074] Optionally, in step S2, during each continuous measurement of the reference sample and the test sample, a rich-condition measurement environment is created by repeatedly introducing slight perturbations. This creates multiple observation dimensions of noise during the continuous measurement of the samples, forming a noise panorama or at least a partial noise panorama in the measurement results.

[0075] In step S2, during each continuous measurement of the reference sample and the test sample, the method of introducing multiple slight perturbations is selected from continuous introduction, introduction at fixed intervals, or introduction at random intervals.

[0076] During each continuous measurement process, the overall measurement results of the reference sample and the test sample, as well as the signals in the measurement results, will each exhibit stable statistical characteristics; at the same time, as the noise panorama or at least a partial noise panorama is formed, the statistical distribution pattern of the noise will also tend to stabilize.

[0077] Furthermore, the form of the slight disturbance can be selected from, but is not limited to, spatial disturbance, physical disturbance, and environmental disturbance; spatial disturbance includes, but is not limited to: causing slight displacement of the measurement site, causing slight rotation of the measurement site; physical disturbance includes, but is not limited to: vibrating the measurement equipment or sample, agitating the fluid sample; environmental disturbance includes, but is not limited to: changing the ambient temperature, ambient humidity, electromagnetic field strength, and air pressure.

[0078] See the instruction manual appendix Figure 2 In step S3, the steps of processing the measurement results of the reference sample and the test sample to form training data for the reference sample and the test sample respectively include:

[0079] S31. Normalize the measurement results of the reference sample and the test sample;

[0080] S32. Based on the normalization results of step S31, establish a posterior probability model framework;

[0081] The measurement results of the reference sample and the test sample, after processing in steps S31-S32, will generate training data that meets the requirements, which will be used for subsequent training of artificial intelligence models.

[0082] Specifically, in this embodiment of the invention, the measurement results of the reference sample and the test sample are regarded as the measurement values ​​obtained by measuring the target composed of a complex system.

[0083] Define the measurement density function as Where S is the measurement space dimension; V is the measurement environment; then the number of systems in the measurement target is N, which is defined by equation (1):

[0084]

[0085] Define B(V) as the measurement function, then the measured value have:

[0086]

[0087] in,

[0088]

[0089] Equation (3) is the normalization condition. In this embodiment of the invention, in order to make the measurement results of the reference sample and the test sample meet the normalization condition of Equation (3), step S31 is adopted to normalize the measurement results of the reference sample and the test sample.

[0090] Since the continuous measurements of the reference sample and the test sample are repeated, the repeated process is represented in a discrete manner, and equation (2) is rewritten in ensemble form:

[0091]

[0092] Define H as the ensemble density function, then we have:

[0093] <v> ＝H <n>(5)

[0094] The statistical fluctuations of complex systems are as follows:

[0095]

[0096] In this context, for repeated measurements, δS represents the information entropy of the measurement, and δP represents the environmental change in the measurement.

[0097] Let δP be the statistical space of the noise panorama, and δS be the statistical space of the signal. Therefore, according to Bayes' theorem, we have:

[0098]

[0099] In equation (7), we define Equation (8) represents the posterior probability condition. In this embodiment of the invention, to ensure that the measurement results of the reference sample and the test sample satisfy the posterior probability condition of Equation (8), step S32 is adopted to establish a posterior probability model framework based on the normalization result obtained in step S31.

[0100] Then the estimation δn of statistical fluctuations in complex systems * for:

[0101] δn * =argmax δn P(H <n>|δn)P(δn) (9)

[0102] In this embodiment of the invention, the measurement results processed by steps S31-S32 satisfy the normalization condition of equation (3) and the posterior probability condition of equation (8). Measurement results that satisfy the above two conditions can be used to estimate the statistical fluctuations of complex systems in equation (9). Measurement results that satisfy the above two conditions will be used as training data in subsequent artificial intelligence model training steps.

[0103] In step S4, the artificial intelligence model can be selected from, but is not limited to, artificial neural networks, perceptrons, support vector machines, Bayesian classifiers, Bayesian networks, random forest models, or clustering models.

[0104] In this embodiment of the invention, the estimation of statistical fluctuations of complex systems in equation (9) above will be achieved by an artificial intelligence model.

[0105] In step S4, during the training process of the artificial intelligence model, the model will iteratively learn, summarize and converge a large amount of empirical learning on the features contained in the training data that can realize signal and noise recognition, as well as the features that can distinguish between reference samples and test samples.

[0106] Specifically, the features that enable signal recognition include the statistical distribution pattern of the measurement results of continuous measurement that conforms to the true mathematical statistical laws of the signal; the features that enable noise recognition include the statistical distribution pattern of the noise panorama or at least part of the noise panorama formed during continuous measurement that approximates the true mathematical statistical laws of the noise; and the features that enable the differentiation between reference samples and test samples include the statistical distribution patterns of the continuous measurement results of the two types of samples.

[0107] Example 2

[0108] As per the instruction manual Figure 4 As shown, a signal analysis system based on acquiring and identifying a panoramic noise distribution model includes a setting module 1, a measurement module 2, a processing module 3, a training module 4, and an analysis module 5.

[0109] Based on actual sample measurement needs, module 1 presets the sample measurement duration;

[0110] In this embodiment of the invention, the sample measurement is not a transient measurement, but a measurement that lasts for a period of time. The setting module 1 needs to preset this duration before the sample measurement. For example, the preset measurement duration for spectral measurements is 3 seconds, and the preset measurement duration for electroencephalogram (EEG) signal measurements is 30 seconds. There is no explicit specification for the preset measurement duration; it depends largely on the specific sample type, measurement method, and actual measurement requirements. Specifically, for example, some signal measurements need to continue for a period of time to accumulate measurement values ​​and enhance the results; however, some optical measurements cannot be sustained for too long to avoid thermal damage to the sample. Therefore, when there are limitations or empirical solutions for continuous sample measurement, those skilled in the art of sample measurement can reasonably preset the measurement duration according to actual needs.

[0111] Based on the preset sample measurement duration of setting module 1, measurement module 2 performs repeated continuous measurements on the reference sample and the test sample to obtain multiple measurement results. Each measurement result includes the signal and a noise panorama or at least a partial noise panorama formed under multiple noise observation dimensions. Each continuous measurement is performed in a rich-condition measurement environment. Rich conditions refer to measurement conditions that are not aimed at maintaining the consistency of external conditions, do not involve noise suppression, are natural, and include real complex noise factors.

[0112] In this embodiment of the invention, the setting module 1 presets the measurement duration for continuous sample measurement. For example, the measurement duration for spectral measurement is specified as 3 seconds, and the measurement duration for electroencephalogram (EEG) signal measurement is specified as 30 seconds; the measurement module 2 will perform continuous measurement of the corresponding samples according to the preset measurement duration.

[0113] In the field of signal acquisition and analysis, maintaining consistency of external conditions during sample measurement is a conventional method to reduce noise fluctuations and achieve a good signal-to-noise ratio. However, this invention does not involve setting consistent external conditions, but rather conducting repeated measurements of reference and test samples in a rich-condition measurement environment. Here, rich conditions refer to measurement conditions that are natural, do not involve noise suppression, and include real, complex noise factors, and are not aimed at maintaining consistency of external conditions.

[0114] In the embodiments of the present invention, as shown in the appendix to the specification... Figure 3 As shown, a single noise observation dimension cannot yield true and comprehensive noise observation results; that is, noise cannot exhibit complete statistical patterns that conform to its true distribution characteristics. However, this invention, by actively creating a condition-rich measurement environment, makes the noise observation dimension during continuous measurement process tend from local to comprehensive, which is conducive to forming a complete noise panorama or at least a partial noise panorama. While constructing the noise panorama, the statistical patterns of noise data will tend to their true mathematical statistical patterns. At the same time, limited by the inherent properties of the samples, the signal in the measurement results will always remain statistically unchanged.

[0115] In this embodiment of the invention, a noise panorama refers to a noise distribution model that fully reflects its theoretical true distribution model; a partial noise panorama refers to a noise distribution model that cannot fully reflect its theoretical true distribution model, but the distribution model already has the accuracy to be used for subsequent signal analysis.

[0116] Furthermore, repeatedly performing sample measurements is also considered an effective way to reduce random errors. However, in this embodiment of the invention, the main purpose of repeatedly performing continuous sample measurements is to obtain a sufficient number of measurement results for training the artificial intelligence model.

[0117] Processing module 3 processes the measurement results of the reference sample and the test sample to form training data for the reference sample and the test sample, respectively.

[0118] Based on the training data of reference samples and test samples, training module 4 uses the observability of noise as the convergence target to train the artificial intelligence model, so that the model can identify signals and noise from the measurement results and distinguish between reference samples and test samples.

[0119] In this embodiment of the invention, the training data of the reference sample and the test sample are randomly allocated as learning data and detection data respectively in a preset ratio. The artificial intelligence model is trained using the learning data, and the detection data is input into the trained artificial intelligence model to calculate the signal recognition result. If the signal recognition accuracy is lower than a preset threshold, training continues using the learning data; if the signal recognition accuracy is higher than the preset threshold, the artificial intelligence model is considered to have completed training.

[0120] For the measurement results of the sample to be identified, the analysis module 5 inputs them into the trained artificial intelligence model, and the output of the artificial intelligence model is the specific type of the sample to be identified.

[0121] In this embodiment of the invention, reference samples and test samples are used as two types of known samples. Multiple measurement results from both are processed to form training data, and the trained artificial intelligence model can effectively distinguish between the two types of known samples. When the sample to be identified is one of the two known samples, the artificial intelligence model can accurately identify the specific type of the sample.

[0122] Optionally, the measurement module 2 includes a perturbation mechanism 21; during each continuous measurement of the reference sample and the test sample by the measurement module 2, the perturbation mechanism 21 creates a rich-condition measurement environment by introducing slight perturbations multiple times. Thus, during the continuous measurement of the samples, multiple observation dimensions of noise are created, forming a noise panorama or at least a partial noise panorama in the measurement results.

[0123] During each continuous measurement of the reference sample and the test sample, the perturbation mechanism 21 introduces multiple slight perturbations in a manner selected from continuous introduction, introduction at fixed intervals, or introduction at random intervals.

[0124] During each continuous measurement process of measurement module 2, the overall measurement results of the reference sample and the test sample, as well as the signals in the measurement results, will exhibit stable statistical characteristics. At the same time, as the noise panorama or at least a partial noise panorama is formed, the statistical distribution pattern of the noise will also tend to stabilize.

[0125] Furthermore, the form of slight disturbance introduced by the perturbation mechanism 21 can be selected from, but is not limited to, spatial perturbation, physical perturbation, and environmental perturbation; spatial perturbation includes, but is not limited to: causing slight displacement of the measurement site, causing slight rotation of the measurement site; physical perturbation includes, but is not limited to: vibrating the measurement equipment or sample, agitating the fluid sample; environmental perturbation includes, but is not limited to: changing the ambient temperature, ambient humidity, electromagnetic field strength, and air pressure.

[0126] Processing module 3 includes a normalization module 31 and a posterior probability module 32;

[0127] The normalization module 31 normalizes the measurement results of the reference sample and the test sample; the posterior probability module 32 establishes a posterior probability model framework based on the normalization results, and generates training data for the reference sample and the test sample that meet the requirements, which are then used for subsequent artificial intelligence model training.

[0128] Specifically, in this embodiment of the invention, the measurement results of the reference sample and the test sample are regarded as the measurement values ​​obtained by measuring the target composed of a complex system.

[0129] Define the measurement density function as Where S is the measurement space dimension; V is the measurement environment; then the number of systems in the measurement target is N, which is defined by equation (1):

[0130]

[0131] Define B(V) as the measurement function, then the measured value have:

[0132]

[0133] in,

[0134]

[0135] Equation (3) is the normalization condition. In this embodiment of the invention, in order to make the measurement results of the reference sample and the test sample meet the normalization condition of Equation (3), the normalization module 31 performs normalization processing on the measurement results of the reference sample and the test sample.

[0136] Since the continuous measurements of the reference sample and the test sample are repeated, the repeated process is represented in a discrete manner, and equation (2) is rewritten in ensemble form:

[0137]

[0138] Define H as the ensemble density function, then we have:

[0139] <v> ＝H <n>(5)

[0140] The statistical fluctuations of complex systems are as follows:

[0141]

[0142] In this context, for repeated measurements, δS represents the information entropy of the measurement, and δP represents the environmental change in the measurement.

[0143] Let δP be the statistical space of the noise panorama, and δS be the statistical space of the signal. Therefore, according to Bayes' theorem, we have:

[0144]

[0145] In equation (7), we define Equation (8) represents the posterior probability condition. In this embodiment of the invention, in order to ensure that the measurement results of the reference sample and the test sample satisfy the posterior probability condition of Equation (8), the posterior probability module 32 establishes a posterior probability model framework based on the normalization result output by the normalization module 31.

[0146] Then the estimation δn of statistical fluctuations in complex systems * for:

[0147] δn * =argmax δn P(H <n>|δn)P(δn) (9)

[0148] In this embodiment of the invention, the measurement results processed by the normalization module 31 and the posterior probability module 32 can satisfy the normalization condition of equation (3) and the posterior probability condition of equation (8). The measurement results that satisfy the above two conditions can be used to estimate the statistical fluctuations of complex systems in equation (9). The measurement results that satisfy the above two conditions will be used as training data in the subsequent artificial intelligence model training steps.

[0149] Artificial intelligence models can be selected from, but are not limited to, artificial neural networks, perceptrons, support vector machines, Bayesian classifiers, Bayesian networks, random forest models, or clustering models.

[0150] In this embodiment of the invention, the estimation of statistical fluctuations of complex systems in equation (9) above will be achieved by an artificial intelligence model.

[0151] During the training process of an artificial intelligence model, the model will iteratively learn, summarize, and converge a large amount of empirical learning on the features contained in the training data that enable signal and noise recognition, as well as the features that enable the distinction between reference samples and test samples.

[0152] Among them, the features that enable signal and noise identification include the statistical distribution pattern of the measurement results of continuous measurement that conforms to the true mathematical statistical laws of the signal; the features that enable noise identification include the statistical distribution pattern of the noise panorama or at least part of the noise panorama formed during continuous measurement that approximates the true mathematical statistical laws of the noise; and the features that enable the differentiation between reference samples and test samples include the statistical distribution patterns of the continuous measurement results of the two types of samples respectively.

[0153] The above embodiments are merely illustrative of the technical solutions of the present invention and are not intended to limit it. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this invention should be included within the protection scope of this invention. Therefore, the protection scope of this invention should be determined by the scope defined in the claims.< / n> < / n> < / v> < / n> < / n> < / v>

Claims

1. A signal analysis method based on acquiring and identifying a panoramic noise distribution model, characterized in that: Includes the following steps: S1: Preset the sample measurement duration according to the actual sample measurement requirements; S2: Based on the sample measurement duration preset in step S1, repeated continuous measurements are performed on the reference sample and the test sample to obtain multiple measurement results; each measurement result contains a signal, as well as a noise panorama or at least a partial noise panorama formed under multiple noise observation dimensions; wherein, each continuous measurement is performed in a rich-condition measurement environment; rich-condition refers to: measurement conditions that are not aimed at maintaining the consistency of external conditions, do not involve noise suppression, are natural, and include real complex noise factors; S3: Process the measurement results of the reference sample and the test sample to form training data for the reference sample and the test sample respectively; S4: Based on the training data of the reference sample and the test sample, with the observability of noise as the convergence target, the artificial intelligence model is trained so that the model can identify the signal and noise from the measurement results and distinguish between the reference sample and the test sample. S5: Input the measurement results of the sample to be identified into the trained artificial intelligence model. The output of the artificial intelligence model is the specific type of the sample to be identified. In step S2, during each continuous measurement of the reference sample and the test sample, a rich-condition measurement environment is created by introducing slight disturbances multiple times, forming a noise panorama or at least a partial noise panorama in the measurement results. In step S2, during each continuous measurement of the reference sample and the test sample, the method of introducing multiple slight perturbations is selected from continuous introduction, introduction at fixed intervals, or introduction at random intervals. The form of the slight disturbance can be selected, but is not limited to, spatial disturbance, physical disturbance, and environmental disturbance; spatial disturbance includes: causing slight displacement of the measurement site, causing slight rotation of the measurement site; physical disturbance includes: vibrating the measurement device or sample, agitating the fluid sample; environmental disturbance includes, but is not limited to: changing the ambient temperature, ambient humidity, electromagnetic field strength, and air pressure. Noise panorama refers to a noise distribution model that fully reflects its theoretical true distribution model; partial noise panorama refers to a noise distribution model that cannot fully reflect its theoretical true distribution model, but the distribution model already has the accuracy to be used for subsequent signal analysis.

2. The signal analysis method based on acquiring and identifying a panoramic noise distribution model according to claim 1, characterized in that: In step S3, the steps of processing the measurement results of the reference sample and the test sample to form training data for the reference sample and the test sample respectively include: S31. Normalize the measurement results of the reference sample and the test sample; S32. Based on the normalization results of step S31, establish a posterior probability model framework; The measurement results of the reference sample and the test sample, after processing in steps S31-S32, will generate training data that meets the requirements, which will be used for subsequent training of artificial intelligence models.

3. The signal analysis method based on acquiring and identifying a panoramic noise distribution model according to claim 1, characterized in that: In step S4, the artificial intelligence model can be selected from, but is not limited to, artificial neural networks, perceptrons, support vector machines, Bayesian classifiers, Bayesian networks, random forest models, or clustering models.

4. The signal analysis method based on acquiring and identifying a panoramic noise distribution model according to claim 1, characterized in that: In step S4, during the training process of the artificial intelligence model, the model will iteratively learn, summarize and converge a large amount of empirical learning on the features contained in the training data that can realize signal and noise recognition, as well as the features that can distinguish between reference samples and test samples. Among them, the features that enable signal and noise identification include the statistical distribution pattern of the measurement results of continuous measurement that conforms to the true mathematical statistical laws of the signal; the features that enable noise identification include the statistical distribution pattern of the noise panorama or at least part of the noise panorama formed during continuous measurement that approximates the true mathematical statistical laws of the noise; and the features that enable the differentiation between reference samples and test samples include the statistical distribution patterns of the continuous measurement results of the two types of samples respectively.

5. A signal analysis system based on acquiring and identifying a panoramic noise distribution model, characterized in that: It includes a setup module, a measurement module, a processing module, a training module, and an analysis module; Based on actual sample measurement needs, the module can preset the sample measurement duration; Based on the preset sample measurement duration of the setting module, the measurement module performs repeated continuous measurements on the reference sample and the test sample to obtain multiple measurement results. Each measurement result includes the signal, as well as a noise panorama or at least a partial noise panorama formed under multiple noise observation dimensions. Each continuous measurement is performed in a rich-condition measurement environment. Rich conditions refer to measurement conditions that are not aimed at maintaining the consistency of external conditions, do not involve noise suppression, are natural, and include real complex noise factors. During each continuous measurement of the reference sample and the test sample, a rich-condition measurement environment is created by introducing slight disturbances multiple times, forming a noise panorama or at least a partial noise panorama in the measurement results. During each continuous measurement of the reference sample and the test sample, the introduction of multiple minor perturbations can be selected from continuous introduction, introduction at fixed intervals, or introduction at random intervals. The form of the slight disturbance can be selected, but is not limited to, spatial disturbance, physical disturbance, and environmental disturbance; spatial disturbance includes: causing slight displacement of the measurement site, causing slight rotation of the measurement site; physical disturbance includes: vibrating the measurement device or sample, agitating the fluid sample; environmental disturbance includes, but is not limited to: changing the ambient temperature, ambient humidity, electromagnetic field strength, and air pressure. The processing module processes the measurement results of the reference sample and the test sample to form training data for the reference sample and the test sample, respectively. Based on the training data of reference samples and test samples, the training module uses the observability of noise as the convergence target to train the artificial intelligence model, enabling the model to identify signals and noise from the measurement results and distinguish between reference samples and test samples. The analysis module inputs the measurement results of the sample to be identified into a trained artificial intelligence model, and the output of the artificial intelligence model is the specific type of the sample to be identified. Noise panorama refers to a noise distribution model that fully reflects its theoretical true distribution model; partial noise panorama refers to a noise distribution model that cannot fully reflect its theoretical true distribution model, but the distribution model already has the accuracy to be used for subsequent signal analysis.

6. The signal analysis system based on acquiring and identifying a panoramic noise distribution model according to claim 5, characterized in that: The measurement module includes a perturbation mechanism; during each continuous measurement of the reference sample and the test sample by the measurement module, the perturbation mechanism creates a rich measurement environment by introducing slight perturbations multiple times, thereby creating multiple observation dimensions of noise during the continuous measurement of the sample, forming a noise panorama or at least a partial noise panorama in the measurement results. During each continuous measurement of the reference sample and the test sample, the perturbation mechanism introduces multiple slight perturbations in a manner that can be selected from continuous introduction, introduction at fixed intervals, or introduction at random intervals. The forms of slight perturbations introduced by perturbation mechanisms include spatial perturbations, physical perturbations, and environmental perturbations; Spatial perturbations include: causing slight displacement or slight rotation of the measurement site; physical perturbations include: vibrating the measurement equipment or sample, or agitating the fluid sample; environmental perturbations include: changing the ambient temperature, ambient humidity, electromagnetic field strength, and air pressure.

7. The signal analysis system based on acquiring and identifying a panoramic noise distribution model according to claim 5, characterized in that: The processing module includes a normalization module and a posterior probability module. The normalization module normalizes the measurement results of the reference sample and the test sample. Based on the normalization results, the posterior probability module establishes a posterior probability model framework and generates training data for the reference sample and the test sample that meet the requirements, which are then used for subsequent artificial intelligence model training.

8. The signal analysis system based on acquiring and identifying a panoramic noise distribution model according to claim 5, characterized in that: Artificial intelligence models include artificial neural networks, perceptrons, support vector machines, Bayesian classifiers, Bayesian networks, random forest models, or clustering models; During the training process of an artificial intelligence model, the model will iteratively learn, summarize, and converge a large amount of empirical learning on the features contained in the training data that enable signal and noise recognition, as well as the features that enable the distinction between reference samples and test samples.

Images