High-precision signal conditioning system and conditioning method

Abstract

The application relates to the technical field of high-precision signals and signal processing and optimization, and particularly relates to a high-precision signal conditioning system and a conditioning method, the system comprising a signal conversion module, a filtering processing module, an isolation amplification module and a VAE-LSTM combined algorithm module; the signal conversion module comprises a current-frequency conversion circuit, a voltage-frequency conversion circuit and a capacitance-voltage conversion circuit; the filtering processing module comprises a low-pass filter and a band-pass filter; noise and interference in the signal are removed, the quality and stability of the signal are improved; the isolation amplification module isolates and amplifies the signal; the VAE-LSTM combined algorithm module combines the advantages of a variational autoencoder VAE and a long short-term memory network LSTM, performs deep feature extraction, noise filtering, time series modeling and prediction on the signal. The application can effectively capture deep patterns of time series, provides a stable and reliable basis for anomaly detection and signal prediction, enhances the robustness of dynamic signal processing, realizes real-time anomaly detection, provides reliable support for equipment state monitoring and fault early warning, and improves the intelligent level of signal processing.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of signal processing and optimization, in particular to a high-precision signal conditioning system and a conditioning method. BACKGROUND

[0002] With the development of modern signal processing technology, the accurate conditioning and optimization of dynamic signals play an important role in industrial control, communication systems, medical monitoring and other fields. However, the traditional signal conditioning method mainly relies on fixed parameter filters and simple mathematical models, which is difficult to deal with complex nonlinear signals and time series characteristics. Especially in the following aspects, there are significant deficiencies: 1. Noise interference problem: signals in complex environments are often disturbed by various noises, such as random noise and structured noise in sensor collected signals, and existing methods often cause signal feature loss while effectively denoising. 2. Traditional signal conditioning system lacks real-time response capability to dynamic changes in signals, especially in signal processing with significant time series characteristics (such as industrial vibration signals and medical ECG signals), which shows insufficient prediction ability. 3. Limited abnormality detection capability: in industrial and medical key application scenarios, signal anomaly detection is crucial. However, the existing technology relies on fixed thresholds for abnormal signal detection, lacks adaptive detection mechanism based on deep features, and has high false positive rate and false negative rate.

[0003] In recent years, machine learning-based signal processing technology has gradually emerged, among which, the variational autoencoder (VAE) has significant advantages in signal feature extraction and noise reduction; the long short-term memory network (LSTM) performs well in time series modeling and prediction. However, the joint application of VAE and LSTM in dynamic signal conditioning is still in its infancy, and how to realize the soft and hardware collaborative optimization under the support of hardware conditioning module and improve the real-time performance and robustness of the system is still an important problem to be solved in the existing technology. SUMMARY

[0004] 1. Technical problems to be solved

[0005] The purpose of the present application is to solve the problems of noise interference, insufficient dynamic response capability and poor abnormality detection effect in the existing dynamic signal conditioning technology, and to provide a high-precision signal conditioning system and a conditioning method.

[0006] 2. Technical solutions

[0007] The present application provides a high-precision signal conditioning system, comprising:

[0008] Signal conversion module: including current frequency conversion circuit, voltage frequency conversion circuit and capacitance voltage conversion circuit; converting different forms of signals into standard forms for subsequent processing.

[0009] The filtering processing module includes a low-pass filter and a band-pass filter, which is used to remove noise and interference in the signal and improve the quality and stability of the signal.

[0010] The isolation amplification module is used to isolate and amplify the signal, thereby improving the safety, stability and anti-interference ability of the system.

[0011] The VAE-LSTM joint algorithm module includes a VAE module and an LSTM module. The VAE module is used to extract the latent features of the signal and filter out noise, and the LSTM module is used to model and predict the time series of the latent features extracted by the VAE module. The VAE-LSTM joint algorithm module combines the advantages of variational autoencoder (VAE) and long short-term memory network (LSTM), and can extract deep features of the signal, filter out noise, and model and predict the time series.

[0012] As an optional solution of the present application, the current frequency conversion circuit converts the input current signal into a frequency signal. It has the characteristics of high precision, high stability and strong anti-interference ability, and can accurately convert the tiny current change into a recognizable frequency change.

[0013] The voltage frequency conversion circuit converts the input voltage signal into a frequency signal. It not only improves the anti-interference ability of the signal, but also facilitates subsequent digital signal processing and transmission.

[0014] The capacitance voltage conversion circuit converts the capacitance change output by the capacitance sensor into a voltage signal. It converts the tiny change of the capacitance into a measurable voltage signal.

[0015] As an optional solution of the present application, the low-pass filter allows low-frequency signals to pass through while suppressing high-frequency signals.

[0016] The band-pass filter only allows signals within a certain frequency range to pass through while suppressing signals of other frequencies.

[0017] Further, the VAE module (variational autoencoder) extracts the latent features of the signal and filters out noise. The VAE module consists of an encoder and a decoder. The encoder maps the input signal to the mean and variance in the latent space, and generates a latent vector from the latent space by random sampling. The decoder decodes the latent vector into a reconstructed signal. By minimizing the error between the reconstructed signal and the original signal and the difference between the distribution of the latent vector and the prior distribution, the VAE can learn the effective feature representation of the signal and filter out noise to a certain extent.

[0018] The LSTM (Long Short-Term Memory) module performs time series modeling and prediction on the latent features extracted by the VAE module. The LSTM module receives the latent feature sequence output by the VAE module, and learns and models these features through its internal memory cells and gating mechanisms. The LSTM predicts future signal trends based on historical feature information, providing strong support for further processing and analysis of the signal.

[0019] As an optional solution of the present application, the VAE module generates the latent distribution parameters of the signal through an encoder, and the mean and log variance are respectively:

[0020] μ = f μ (x; θ e ), logσ 2 = f σ2 (x; θ e ); where μ and logσ 2 are the mean and log variance of the latent distribution, respectively, and θ e is the trainable parameter of the encoder. x represents the signal input to the encoder of the VAE module; f u (◆) is a function determined by the encoder, which takes the input signal x and the trainable parameter θ e of the encoder as input, and outputs the mean u of the latent distribution. f σ2 (◆) is a function determined by the encoder, which takes the input signal x and the trainable parameter θ e of the encoder as input, and outputs the log variance of the latent distribution.

[0021] As an optional solution of the present application, the VAE module calculates the latent feature z through sampling in the latent space:

[0022]

[0023] where ∈ is a standard normal distribution noise randomly sampled; σ = exp(0.5·logσ 2 ). z represents the latent feature calculated by sampling in the latent space. N(0, I) represents a multivariate standard normal distribution. Where 0 represents the mean vector, indicating that the mean of the distribution is 0 in each dimension; I is the identity matrix, indicating the covariance matrix, meaning that each dimension is independent and the variance is 1.

[0024] As an optional solution of the present application, the LSTM module updates the time series state of the signal through the following steps:

[0025] Forget gate: f t = σ(W f ·[h t-1 , z t ]+b f );

[0026] In the formula, f t This is the output of the forget gate, with a value in the range [0,1], representing the cell state C of the previous time step. t-1 The proportion of those forgotten; W f This is the weight matrix of the forget gate; b f It is the bias vector of the forget gate; h t-1 It is the hidden state of the previous time step; z t It is the input feature at the current time step; It is the sigmoid activation function.

[0027] Input gate: i t =σ(W i ·[h t-1 ,z t ]+b i ),

[0028] In the formula, i t It is the output of the input gate, with a value range of [0,1], representing the degree of influence of the current input on the cell state; It is a candidate value for the cell state at the current time step; W i W c These are the weight matrices for the input gate and the candidate cell states, respectively; b i ,b c These are the bias vectors for the input gate and the candidate cell state values, respectively; It is the hyperbolic tangent activation function, with a value range of [-1, 1].

[0029] Cell state: In the formula, C t It represents the cell state at the current time step; C t-1 It represents the cell state at the previous time step.

[0030] Output gate: o t =σ(W o ·[h t-1 ,z t ]+b o ),h t =o t ·tanh(C t );

[0031] In the formula, o t This is the output of the output gate, with a value range of [0,1], representing the cell state C. t For hidden state h t The output ratio; h t It is the hidden state at the current time step; W fW i ,W o ,W c , and b f ,b i ,b c ,b o are trainable parameters of the LSTM.

[0032] As an optional solution of the present application, the LSTM module utilizes the hidden state h t to predict the output value of the signal

[0033]

[0034] where W y is the weight matrix of the prediction layer, b y is the bias parameter. t represents the time step. h t is the hidden state of the LSTM module at time step t. W y is the weight matrix of the prediction layer. b y is the bias vector of the prediction layer.

[0035] As an optional solution of the present application, the latent space dimension z of the VAE module is compressed to 10%-30% relative to the dimension of the input signal, and the calculation of the latent space dimension z is performed by the following formula:

[0036]

[0037] As an optional solution of the present application, the LSTM module uses a sliding window technique to dynamically model the time series signal, and the window length is T w , and each update is performed according to the following formula:

[0038]

[0039] where Tw is the length of the sliding window; t is the time step index; y t represents the actual value at time step

[0040] ^

[0041] t, i.e., the true observation data of the time series signal at that time. y t represents the predicted output at time step t; L LSTM is the loss function value based on the LSTM model. ||·|| represents a certain norm operation.

[0042] As an optional solution of the present application, the LSTM module uses a threshold to judge abnormal signals based on the deviation of the prediction error and the actual signal, and the deviation calculation formula is:

[0043] If δ t > δ 阈值 , an abnormal signal is determined. In the formula, δ t represents the deviation value between the prediction error and the actual signal at time step t. δ 阈值 represents a pre-set fixed value as a standard limit for determining whether the signal is abnormal.

[0044] As an optional solution of the present application, the VAE-LSTM joint algorithm module dynamically adjusts the parameters of the filtering module through the following feedback adjustment mechanism:

[0045] Based on the signal prediction result of the LSTM, the frequency range of the filter is optimized;

[0046] Based on the signal reconstructed by the VAE, the zero offset compensation of the signal conversion module is adjusted.

[0047] As an optional solution of the present application, it also includes an oscillator module for generating sine wave, sawtooth wave and square wave signals to provide reference signals;

[0048] As an optional solution of the present application, it includes a constant current source module for providing stable bidirectional current to ensure the current stability of signal processing;

[0049] The present application provides a high-precision signal conditioning system and a conditioning method thereof, comprising the following steps:

[0050] S1, signal acquisition and conversion: the input signal is converted by the signal conversion module, including current frequency I / F conversion, voltage frequency V / F conversion or capacitance voltage C / V conversion;

[0051] S2, signal preprocessing: the converted signal is input into the filtering processing module to remove high-frequency noise or interference signals;

[0052] S3, signal amplification and isolation: the preprocessed signal is electrically isolated and amplified by the isolation amplification module to enhance the signal strength and prevent interference;

[0053] S4, feature extraction and noise reduction: the VAE module is used to extract the latent features of the signal, the latent distribution parameters of the signal are generated by the encoder, the feature representation is generated by sampling the latent space, and the signal is reconstructed by the decoder to filter out noise;

[0054] S5, time series modeling and prediction: the feature representation generated by the VAE module is input into the LSTM module, and the future value of the signal is predicted based on the dynamic modeling formula of the time series using the hidden state;

[0055] S6, abnormality detection and feedback optimization: according to the deviation between the predicted value and the actual value, the abnormal signal is determined; according to the feedback result, the frequency range of the filter or the zero offset compensation of the signal conversion module is adjusted;

[0056] S7, signal output: output the optimized signal for external devices to use.

[0057] 3. Beneficial effects:

[0058] The present application combines variational autoencoder (VAE) and long short-term memory network (LSTM) to build a high-precision signal conditioning system, which can effectively solve the problems of long-term feature capture, noise suppression and anomaly detection in complex dynamic signal processing. The system has the following significant technical effects:

[0059] 1. Long-term feature capture ability is significantly improved: In the LSTM module of the present application, the cell state is the core storage unit of long-term memory, which can dynamically record important information in the time series signal. It ensures the long-term retention of key features in the signal, effectively captures the deep patterns of time series, and provides a stable and reliable foundation for anomaly detection and signal prediction.

[0060] 2. Robustness of dynamic signal processing is enhanced: The present application combines the feature extraction capability of VAE module. VAE maps the input signal to the latent feature space through the encoder and decoder and reconstructs it. The reconstruction process of VAE retains the key features of the signal while removing complex noise. The LSTM module uses the latent features extracted by VAE to model and predict time series combined with historical state information. It can maintain high-precision output in complex and multi-noise environments, and is particularly suitable for scenes with obvious dynamic characteristics such as industrial vibration signals, communication signals and medical signals.

[0061] 3. Real-time anomaly detection can be realized: The system can quickly identify signal anomalies, significantly reducing false positive and false negative rates, and providing reliable support for device state monitoring and fault warning.

[0062] 4. The intelligent level of signal processing is improved: The present application realizes dynamic optimization of signal conditioning through feedback mechanism: adjusting the frequency range of the filter according to the prediction results of LSTM; adjusting the zero offset compensation of the signal conversion module based on the signal features reconstructed by VAE. The filter and signal conversion modules of the system can adapt to dynamic environments to ensure signal output quality. The system can automatically adjust parameters according to real-time changes in signal environment, improving the intelligence and stability of signal processing.

[0063] 5. Improve the overall technical effect of the system: The system can be applied to industrial monitoring, communication systems, medical signal processing and navigation systems, etc. to provide efficient solutions for complex dynamic signal processing. BRIEF DESCRIPTION OF DRAWINGS

[0064] Figure 1 is the structural block diagram of the present application;

[0065] Figure 2 is a high-precision signal conditioning method flow chart. DETAILED DESCRIPTION

[0066] Example 1: Reference Figure 1 The present application provides a high-precision signal conditioning system, comprising:

[0067] Signal conversion module: including current frequency conversion circuit, voltage frequency conversion circuit and capacitance voltage conversion circuit; responsible for converting different forms of signals into standard forms for subsequent processing.

[0068] Filter processing module: including low-pass filter and band-pass filter; used to remove noise and interference in the signal, improve the quality and stability of the signal.

[0069] Isolation and amplification module: used for signal isolation and amplification; to improve the safety, stability and anti-interference ability of the system. In many practical applications, the signal source and the subsequent processing circuit may be at different potentials, or there may be large electrical interference. The isolation and amplification module uses isolation technology such as photoelectric isolation, magnetic isolation, etc. to isolate the input signal from the output signal electrically, effectively preventing mutual interference between signals and the spread of electrical faults. At the same time, the module amplifies the signal to meet the requirements of the subsequent processing circuit for signal amplitude. The isolation and amplification module is composed of input circuit, isolation unit and output circuit. The input circuit is responsible for preprocessing the input signal, such as impedance matching, filtering, etc.; the isolation unit uses optical coupler, transformer and other isolating devices to convert the input signal into optical signal or magnetic signal for transmission, and then convert it back to electrical signal; the output circuit amplifies and adjusts the isolated signal to output the required signal.

[0070] VAE-LSTM joint algorithm module: including VAE module and LSTM module; VAE module is used to extract the latent features of the signal and filter out noise; LSTM module is used to model and predict the time series of the latent features extracted by VAE module. VAE-LSTM joint algorithm module combines the advantages of variational autoencoder (VAE) and long short-term memory network (LSTM), which can extract deep features, filter noise, and model and predict time series of signals.

[0071] In this technical solution, the high-precision signal conditioning system can realize high-precision processing, analysis and prediction of various complex signals through the cooperative work of signal conversion module, filter processing module, isolation and amplification module and VAE-LSTM joint algorithm module, providing strong technical support for many fields of application.

[0072] Further, the core function of the current-to-frequency conversion circuit is to convert the input current signal into a frequency signal. In practical applications, many sensors output current signals, such as photoelectric sensors, pressure sensors, etc. In order to facilitate the processing and transmission of digital systems, it is necessary to convert these current signals into frequency signals. Its working principle is based on the combination of current-to-voltage conversion and voltage-to-frequency conversion. First, the input current is converted into a voltage signal through a precise resistor, and then the voltage signal is converted into a frequency signal using a special voltage-to-frequency conversion chip. This conversion method has the characteristics of high precision, high stability and strong anti-interference ability, and can accurately convert small current changes into recognizable frequency changes.

[0073] The voltage-to-frequency conversion circuit is mainly used to convert the input voltage signal into a frequency signal. In some analog signal acquisition systems, the voltage signal output by the sensor may have a wide range and high precision requirements. This circuit accurately converts the input voltage signal into a frequency signal through a specific circuit structure and algorithm. Its conversion process is based on the principle of integrator and comparator, which integrates the input voltage and generates a pulse signal when the integrated value reaches a certain threshold, thereby realizing the conversion from voltage to frequency. This conversion method not only improves the anti-interference ability of the signal, but also facilitates subsequent digital signal processing and transmission.

[0074] The capacitance-to-voltage conversion circuit is used to convert the capacitance change output by the capacitance sensor into a voltage signal. In some physical quantity measurements, such as displacement, pressure, humidity, etc., capacitance sensors are often used. These sensors reflect the changes of the measured quantities through the changes of capacitance. The capacitance-to-voltage conversion circuit converts the small changes of capacitance into measurable voltage signals through clever circuit design. Common implementation methods include bridge circuit and charge amplifier circuit. The bridge circuit uses the change of capacitance to change the balance state of the bridge, thereby outputting a voltage signal related to the change of capacitance; the charge amplifier circuit amplifies the charge on the capacitor to convert the capacitance change into a voltage change.

[0075] Further, the function of the low-pass filter is to allow low-frequency signals to pass through while suppressing high-frequency signals. In the actual signal acquisition process, due to environmental noise, power interference and other factors, high-frequency noise often mixes into the signal. The low-pass filter filters out these high-frequency noise through its specific frequency response characteristics, only retaining the low-frequency useful signal. Common types of low-pass filters include Butterworth filter, Chebyshev filter and elliptic filter, etc. The Butterworth filter has flat passband and gradually declining stopband characteristics, suitable for occasions where the amplitude of signals in the passband is required to be relatively flat; the Chebyshev filter has equal ripple characteristics in the passband or stopband, which can improve the selectivity of the filter to a certain extent; the elliptic filter has the narrowest transition band and the smallest stopband attenuation, suitable for occasions where the performance of the filter is required to be high.

[0076] The function of a band-pass filter is to allow signals within a specific frequency range to pass through while suppressing signals of other frequencies. In some applications, it is necessary to extract signals within a specific frequency range, such as carrier signals in communication systems, specific frequency components in biomedical signals, etc. By designing appropriate center frequency and bandwidth, a band-pass filter can accurately extract the required signal and effectively suppress interference of other frequencies. A band-pass filter can be composed of a low-pass filter and a high-pass filter, or a special band-pass filter design method such as a double-T network, a Wheatstone bridge circuit, etc.

[0077] Further, the main function of the VAE module (Variational Autoencoder) is to extract the latent features of the signal and filter out noise. The variational autoencoder is a deep learning architecture based on a probabilistic model, which learns the latent feature representation of the signal by mapping the input signal to a low-dimensional latent space. In this process, the VAE module can automatically model and filter out noise in the signal, resulting in a more pure feature representation. The VAE module consists of an encoder and a decoder. The encoder maps the input signal to the mean and variance in the latent space, and generates a latent vector from the latent space by random sampling; the decoder decodes the latent vector into a reconstructed signal. By minimizing the error between the reconstructed signal and the original signal and the difference between the distribution of the latent vector and the prior distribution, the VAE can learn the effective feature representation of the signal and filter out noise to some extent.

[0078] The LSTM (Long Short-Term Memory) module is used to model and predict the time series of the latent features extracted by the VAE module. Long Short-Term Memory is a recurrent neural network specifically designed for processing time series data, which can effectively solve the problems of gradient disappearance and gradient explosion in traditional recurrent neural networks, thus better capturing long-term dependencies in time series data. In a high-precision signal conditioning system, the LSTM module receives the latent feature sequence output by the VAE module and learns and models these features through its internal memory cells and gating mechanisms. LSTM predicts future signal trends based on historical feature information, providing strong support for further processing and analysis of signals. For example, in industrial process monitoring, LSTM can predict the operating state of equipment based on historical signal data, and detect potential hidden faults in advance.

[0079] Example 2:

[0080] The difference between this embodiment and Example 1 is that the VAE module generates the latent distribution parameters of the signal through the encoder, and the mean and log variance are respectively:

[0081] μ=f μ (x; θe ) ; logσ 2 = f σ2 (x; θ e ) ; where μ and logσ 2 are the mean and log-variance of the latent distribution, respectively, and θ e are the trainable parameters of the encoder. x represents the signal input to the VAE module encoder, which can be various types of data such as time series signals, image data, etc., and is the starting input of the entire processing flow, and the encoder will generate the corresponding latent distribution parameters based on this input signal. f u (◆) is a function determined by the encoder, which takes the input signal x and the trainable parameters θ e of the encoder as input, and outputs the mean u of the latent distribution. The specific form of this function depends on the network structure and design of the encoder, and its role is to map the input signal x to the mean position of the latent space through a specific calculation method. σ2 (◆) is a function determined by the encoder, which takes the input signal x and the trainable parameters θ e of the encoder as input, and outputs the log-variance of the latent distribution. Its specific implementation form depends on the architecture of the encoder, which is used to calculate the log-variance of the distribution corresponding to the input signal x in the latent space.

[0082] Preferably, the VAE module calculates the latent feature z through sampling in the latent space:

[0083]

[0084] where ∈ is a standard normal distribution noise randomly sampled. z represents the latent feature calculated by sampling in the latent space. It is obtained by adding the product of the standard deviation σ and the random noise ∈ to the mean u. As a key intermediate variable connecting the encoder and the decoder in the VAE model, z contains the feature information of the input data, and the decoder will reconstruct data similar to the original input based on z. N(0, I) represents a multivariate standard normal distribution. Among them, 0 represents the mean vector, indicating that the mean of the distribution is 0 in each dimension; I is the identity matrix, indicating the covariance matrix, meaning that each dimension is independent of each other, and the variance is 1. In practical applications, if the latent space is multi-dimensional, then ε is sampled from this multivariate standard normal distribution, and z will also take values in the multi-dimensional latent space accordingly.

[0085] Preferably, the LSTM module updates the time series state of the signal through the following steps:

[0086] Forget gate: f t = σ(W f · [h t-1 , z t ] + bf );

[0087] Input gate: i t = σ(W i · [h t-1 , z t ] + b i ),

[0088] Cell state:

[0089] Output gate: o t = σ(W o · [h t-1 , z t ] + b o ), h t = o t · tanh(C t ); where t denotes the time step, used to identify different time instances in the time series. When processing time series data, data is input into the LSTM module sequentially, and each time step has corresponding input, state update, etc. h t-1 represents the hidden state of the LSTM module at time step t-1. The hidden state carries information from previous time steps and participates in various calculations at the current time step, helping the model remember past information. c t-1 is the cell state of the LSTM module at time step t-1. The cell state is an important carrier of long-term memory information in LSTM, which transmits information between different time steps and retains key information during the update process. w f is the weight matrix of input x t to the forget gate, which determines the degree of influence of input x t on the output of the forget gate. b f is the bias vector of the forget gate, which provides a learnable constant term for the calculation of the forget gate, helping the model better fit the data. W i is the weight matrix of input x t to the input gate, controlling the effect of input x t on the output of the input gate. b i is the bias vector of the input gate, adding a learnable constant offset to the calculation of the input gate. W c is the weight matrix of input x t to the cell state update part, affecting the contribution of input to the cell state update. b c is the bias vector of the cell state update part, providing a constant term for the calculation of the cell state update. W o is the weight matrix of input x t to the output gate, controlling the effect of input x tThe influence on the output of the output gate. o is the bias vector of the output gate, providing an adjustable constant term for the calculation of the output gate. t is the output of the forget gate at time step t. The forget gate decides which information in the previous cell state c t-1 needs to be retained and which information needs to be forgotten, with the value range in [0-1], calculated by the Sigmoid function. The closer to 1 indicates the more information retained, and the closer to 0 indicates the more information forgotten. t is the output of the input gate at time step t. The input gate controls which information in the current input x t needs to be added to the cell state. It is calculated by the Sigmoid function, with the value range in [0-1], used to determine the retention degree of input information. ～ t is the candidate cell state generated at time step t. It is calculated based on the current input x t and the hidden state h t-1 at the previous time step, transformed by a tanh function, with the value range in [-1, 1], containing new information of the current input. t is the updated cell state at time step t. It is calculated by the forget gate output, the previous cell state, the input gate output, and the candidate cell state. It inherits part of the information of the previous cell state and adds new information of the current input, and is the main carrier of long-term memory in the LSTM module. t is the output of the output gate at time step t. The output gate decides which information in the current cell state c t needs to be output as the hidden state of the current time step. It is calculated by the Sigmoid function, with the value range in [0-1], used to control the proportion of output information. t is the hidden state of the LSTM module at time step t. It is calculated by the output gate o t and the cell state c t transformed by the tanh function, containing useful information of the current time step, which will be used as one of the inputs for the next time step, and also can be used as the output of the model for subsequent tasks.

[0090] Preferably, the LSTM module uses the hidden state h t to predict the output value of the signal where t represents the time step.h. t is the hidden state of the LSTM module at time step t. W y is the weight matrix of the prediction layer, belonging to the trainable parameters of the prediction layer. Its function is to combine the hidden state ht Perform a linear transformation using matrix multiplication W. y h t This is used to adjust the weights of each feature in the hidden state, so that the model can better capture the relationship between the hidden state and the final output value, thereby making reasonable predictions about the output value. y This is the bias vector of the prediction layer, and these are the trainable parameters of the prediction layer. The bias vector b y This provides a learnable constant offset for the prediction result. After passing through the weight matrix W... y For hidden state h t After performing a linear transformation, add the bias vector b. y This allows the model to adjust its predictions more flexibly under different data distributions and features, which helps improve the model's prediction accuracy and generalization ability.

[0091] Preferably, the latent space dimension z of the VAE module is compressed to 10%-30% relative to the dimension of the input signal, and the latent space dimension z is calculated using the following formula:

[0092]

[0093] Preferably, the LSTM module uses a sliding window technique to dynamically model the time series signal, with a window length of Tw, and each update is performed according to the following formula:

[0094] Where Tw is the length of the sliding window, which determines the amount of time-series data involved in each LSTM module modeling step. t is the time step index, used to iterate through each time point within the sliding window. y t This represents the actual value at time step t.

[0095] ^

[0096] That is, the actual observed data of the time series signal at that specific moment. t This represents the predicted output at time step t, which is the prediction made by the LSTM module for the current time step t based on historical data within the sliding window. LSTM This is the loss function value based on the LSTM model. ||·|| represents a certain norm.

[0097] ^

[0098] The operation, used here to measure the actual value y t and predicted value y t The distance or difference between them.

[0099] Preferably, the LSTM module uses a threshold to determine abnormal signals based on the deviation between the prediction error and the actual signal. The deviation calculation formula is as follows:

[0100] If δ t > δ 阈值 , then determine an abnormal signal. In the formula, δ t represents the deviation value between the prediction error and the actual signal at time step t. δ 阈值 represents a pre-set fixed value as the standard limit for determining whether the signal is abnormal.

[0101] Preferably, the VAE-LSTM joint algorithm module dynamically adjusts the parameters of the filtering module through the following feedback regulation mechanism:

[0102] LSTM-based signal prediction results Optimizing filter frequency range: In the process of signal processing, the frequency range of the filter directly affects the filtering effect of the signal. Different frequency components of the signal may contain different information, while there may also be noise interference. The prediction results of LSTM not only contain estimates of future signal values, but also imply the frequency characteristics of the signal. The process of optimizing the frequency range of the filter is a dynamic and adaptive process. The system first analyzes the possible frequency components and interference in the signal based on the prediction results of LSTM. Then, through specific algorithms and control logic, the parameters of the filter are adjusted to optimize the frequency range. This process involves adjusting the type of filter (such as low-pass filter, high-pass filter, band-pass filter, etc.), order, and cutoff frequency, etc. For example, if the prediction results show that there is interference in a specific frequency range in the signal, the system may select a band-pass filter with an appropriate order and accurately adjust its cutoff frequency to effectively filter out the interference in that frequency range while preserving the useful components of the signal.

[0103] Based on the VAE reconstructed signal, adjust the zero offset compensation of the signal conversion module: During the signal conversion process, due to the influence of non-ideal characteristics of circuit elements, environmental noise and other factors, the signal may have a zero offset error, that is, there is a fixed deviation between the output signal and the input signal. This zero offset error will affect the accuracy and reliability of the signal, especially in some applications that require high accuracy of the signal, such as sensor signal acquisition, precision measurement, etc. The VAE reconstructed signal can be used as a reference to detect and correct this zero offset error. By comparing the original signal and the reconstructed signal, the system can calculate the size and direction of the zero offset error. If the reconstructed signal shows that there is a zero offset error, the system will adjust the zero offset compensation parameters of the signal conversion module according to this error information. The method of adjusting the zero offset compensation involves adjusting the hardware parameters or software algorithms in the signal conversion module. In terms of hardware, it may be necessary to adjust the gain, bias voltage and other parameters of the amplifier to eliminate the zero offset error. In terms of software, it may be necessary to use digital signal processing algorithms such as adaptive filtering, calibration algorithms, etc. to correct the signal. For example, in a sensor signal acquisition system, the analog signal output by the sensor may have a zero offset error after analog-to-digital conversion (ADC). The system can reconstruct the converted digital signal by VAE, and then adjust the bias voltage or digital correction coefficient of the ADC according to the difference between the reconstructed signal and the original signal, so as to compensate for the zero offset error and improve the accuracy of signal acquisition.

[0104] Preferably, it also includes an oscillator module for generating sine wave, sawtooth wave and square wave signals to provide a variety of reference signals for the system.

[0105] Preferably, it also includes a constant current source module for providing stable bidirectional current to ensure the current stability of signal processing. It meets the needs of different signal processing circuits.

[0106] Embodiment 3:

[0107] Referring to Figure 2 The present application provides a high-precision signal conditioning method, comprising the following steps:

[0108] S1, signal acquisition and conversion: converting the input signal through the signal conversion module, including current-to-frequency (I / F) conversion, voltage-to-frequency (V / F) conversion or capacitance-to-voltage (C / V) conversion;

[0109] S2, signal preprocessing: inputting the converted signal into the filtering processing module to remove high-frequency noise or interference signals;

[0110] S3, signal amplification and isolation: using the isolation amplification module to electrically isolate and amplify the preprocessed signal to enhance the signal strength and prevent interference;

[0111] S4, feature extraction and noise reduction: use VAE module to extract the latent features of the signal, generate the latent distribution parameters mu and logsigma2 of the signal through the encoder; generate feature representation z = mu + sigma * e through latent space sampling, and reconstruct the signal through the decoder to filter out noise;

[0112] S5, time series modeling and prediction: input the feature representation z generated by the VAE module into the LSTM module, based on the dynamic modeling formula of time series: h t = sigma (W o · [h t-1 ,z t ]+b o )·tang(C t )Use the hidden state h t to predict the future value of the signal

[0113] S6, anomaly detection and feedback optimization: according to the deviation between the predicted value t and the actual value y If delta t > delta 阈值 , it is determined that the signal is abnormal; adjust the filter frequency range or the zero offset compensation of the signal conversion module according to the feedback result;

[0114] S7, signal output: output the optimized signal for external devices to use.

[0115] Example 4: high-precision conditioning and anomaly detection of industrial vibration signals, background: the key equipment of a factory produces vibration signals during operation, and equipment failure usually manifests as abnormal changes in vibration signal frequency and amplitude. In order to detect the running state of the equipment and predict potential failures, it is necessary to perform high-precision conditioning, noise reduction, modeling and anomaly detection on the vibration signals.

[0116] Input signal: the collected vibration signal is as follows:

[0117] x = [0.3, 0.28, 0.35, 0.4, 0.42, 0.41, 0.38, 0.6, 0.95, 1.2, 0.85];

[0118] The signal is mixed with high-frequency noise and periodic interference.

[0119] The processing process includes:

[0120] 1, signal acquisition and conversion (signal conversion module): the input signal is converted through the signal conversion module. The converted signal is standardized to adapt to the subsequent processing.

[0121] 2、Signal preprocessing: Through the filter module, a low-pass filter (cut-off frequency of 200 Hz) is used to remove high-frequency noise, and the filtered signal is:

[0122] x filtered =[0.3, 0.28, 0.35, 0.4, 0.42, 0.41, 0.38, 0.6, 0.8, 0.9, 0.75];

[0123] 3、Signal amplification and isolation (isolation amplification module): The filtered signal is isolated and gain-adjusted by the isolation amplification module to enhance signal strength and reduce external interference.

[0124] 4、Feature extraction and noise reduction (VAE module): The encoder generates latent distribution parameters, and the encoder calculates the mean and log variance of the latent space:

[0125] μ = f μ (x filtered ) = [0.31, 0.29, 0.36, 0.41, 0.43, 0.42, 0.39, 0.61, 0.81, 0.91, 0.76];

[0126]

[0127] Latent space sampling: According to the formula z = μ + σ * ∈, where sampling gets:

[0128] z = [0.31, 0.29, 0.36, 0.41, 0.43, 0.42, 0.39, 0.61, 0.81, 0.91, 0.76];

[0129] Decoder reconstructs signal: The decoder reconstructs the signal according to the latent feature z:

[0130] Through reconstruction, the signal noise has been significantly reduced.

[0131] 5、Dynamic modeling and prediction (LSTM module): Input latent features, input the latent features z extracted by VAE to the LSTM module, and the LSTM updates the state according to the following formula:

[0132] Forget gate: f t = σ(W f · [h t-1 , z t ] + b f );

[0133] Input gate: i t = σ(W i · [h t-1 , zt ]+b i ),

[0134] Cell state:

[0135] Output gate: o t =σ(W o ·[h t-1 ,z t ]+b o ),h t =o t ·tanh(C t );

[0136] Future signal prediction: using hidden state h t Predicting future signal values

[0137] The predicted future signals are:

[0138]

[0139] 6. Anomaly detection and feedback optimization: Based on predicted values and actual value y t deviation Determine if there is an anomaly:

[0140] δ t = [0.01, 0.01, 0.01, 0.06, 0.07, 0.08, 0.09, 0.05, 0.03, 0.02, 0.03]; Set the threshold δ 阈值 =0.2. The result shows:

[0141] δ t <δ 阈值 No abnormal signals were detected; the frequency range of the filter was optimized based on feedback to improve signal processing accuracy.

[0142] 7. Signal Output: Based on the prediction bias of the LSTM, adjust the filter parameters (increase the cutoff frequency to 220Hz) and output the final optimized signal:

[0143] x output = [0.31, 0.29, 0.36, 0.41, 0.43, 0.42, 0.39, 0.61, 0.81, 0.91, 0.76].

[0144] The high-precision signal conditioning system of the application achieves the following technical effects: 1. Improved noise reduction capability: the VAE module significantly reduces noise interference and preserves key signal characteristics. 2. The time series modeling LSTM accurately captures the time sequence characteristics of the signal, and the prediction result is highly accurate. 3. The abnormality detection result is accurate, with no false positives or false negatives. 4. Self-adaptive optimization: the system adjusts the filtering parameters in real time to improve the quality of signal output.

[0145] This example demonstrates the excellent performance of the application in complex dynamic signal conditioning, providing an efficient solution for operation monitoring and fault warning of industrial equipment.

[0146] The above-described embodiments are merely used to illustrate the technical solutions of the application, rather than limit them; although the application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that they can still modify the technical solutions recorded in the foregoing embodiments, or make equivalent replacements for some technical features; and these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the application.

Claims

1. A high-precision signal conditioning system, comprising a signal conversion module, a filtering processing module, an isolation amplification module, and a VAE-LSTM combined algorithm module, characterized in that: the signal conversion module comprises a current-frequency conversion circuit, a voltage-frequency conversion circuit, and a capacitance-voltage conversion circuit; different forms of signals are converted into a standard form for subsequent processing; the filtering processing module comprises a low-pass filter and a band-pass filter; noise and interference in the signal are removed to improve the quality and stability of the signal; the isolation amplification module isolates and amplifies the signal; the safety, stability, and anti-interference ability of the system are improved; the VAE-LSTM combined algorithm module comprises a VAE module and an LSTM module; the VAE module is used to extract the latent features of the signal and filter out noise; the LSTM module is used to model and predict the time series of the latent features extracted by the VAE module, combining the advantages of variational autoencoder VAE and long short-term memory network LSTM, to extract deep features, filter noise, and model and predict time series of the signal; the VAE module calculates the latent feature z through sampling of the latent space: the VAE-LSTM combined algorithm module dynamically adjusts the parameters of the filtering module through the following feedback adjustment mechanism: based on the signal prediction result of the LSTM, the frequency range of the filter is optimized; based on the signal reconstructed by the VAE, the zero offset compensation of the signal conversion module is adjusted; through the feedback mechanism, the dynamic optimization of signal conditioning is realized: the frequency range of the filter is adjusted according to the prediction result of the LSTM; the zero offset compensation of the signal conversion module is adjusted based on the signal features reconstructed by the VAE; the filtering and signal conversion modules of the system can adapt to dynamic environments to ensure the quality of signal output; the parameters are automatically adjusted according to the real-time changes of the signal environment to improve the intelligence and stability of signal processing. The VAE module generates the latent distribution parameters of the signal through the encoder, and the mean and log variance are respectively: The LSTM module updates the time series state of the signal through the following steps: The latent space dimension z of the VAE module is compressed to 10%-30% relative to the dimension of the input signal, and the calculation of the latent space dimension z is performed through the following formula: The LSTM module judges abnormal signals based on the deviation of the prediction error from the actual signal using a threshold, and the deviation calculation formula is: The following steps are included: where ∈ is a standard normal distributed noise sampled randomly; σ = exp(0.5 log σ 2 ); z denotes the latent feature computed by sampling in latent space; N(0, I) denotes a multivariate standard normal distribution; where 0 denotes the mean vector, indicating that the mean of the distribution is 0 in each dimension; I is an identity matrix, denoting the covariance matrix, meaning that each dimension is independent of each other, and the variance is 1; S1, signal acquisition and conversion: the input signal is converted by the signal conversion module, including current-frequency I / F conversion, voltage-frequency V / F conversion, or capacitance-voltage C / V conversion; S2, signal preprocessing: the converted signal is input into the filtering processing module to remove high-frequency noise or interference signals; S3, signal amplification and isolation: the preprocessed signal is electrically isolated and amplified by the isolation amplification module to enhance the signal strength and prevent interference; S4, feature extraction and noise reduction: the VAE module extracts the latent features of the signal, generates the latent distribution parameters of the signal through the encoder; the feature representation is generated through sampling of the latent space, and the signal is reconstructed through the decoder to filter out noise; 2. The high precision signal conditioning system of claim 1, wherein: ​ μ = f μ (x; θ e ), where μ and logσ 2 are the mean and log-variance of the latent distribution, respectively, θ e are trainable parameters of the encoder; x represents the signal input to the VAE module encoder; f u (·) is a function determined by the encoder, taking as input the signal x and the trainable parameters of the encoder θ e outputs the mean u of the latent distribution; f σ2 (·) is a function determined by the encoder, taking as input the signal x and the trainable parameters of the encoder θ e outputs the log-variance of the latent distribution.

3. The high precision signal conditioning system of claim 1, wherein: ​ forget gate: f t = σ(W f ·[h t-1 ,z t ]+b f ); where f t is the output of the forget gate; W f is the weight matrix of the forget gate; b f is the bias vector of the forget gate; h t-1 is the hidden state of the previous time step; z t is the input feature of the current time step; is the sigmoid activation function; t denotes the time step. Input gate: i t = σ(W i · [h t-1 , z t ] + b i ), where i t is the output of the input gate; is the candidate cell state value at the current time step; W i , W c are the weight matrices for the input gate and the candidate cell state value, respectively; b i , b c are the bias vectors for the input gate and the candidate cell state value, respectively; is the hyperbolic tangent activation function with a value range of [-1, 1]. Cell state: where C t is the cell state at the current time step; C t-1 is the cell state at the previous time step; Output gate: o t = σ(W o · [h t-1 , z t ] + b o ), h t = o t · tanh(C t ) ; where o t is the output of the output gate; h t is the hidden state of the current time step; b o is the bias vector of the output gate.

4. The high precision signal conditioning system of claim 3, wherein: The LSTM module utilizes the hidden state h t The output value of the prediction signal where W y is a weight matrix of the prediction layer, b y is a bias parameter; t represents a time step; h t is the hidden state of the LSTM module at time step t.

5. The high precision signal conditioning system of claim 2, wherein: ​ 6. The high precision signal conditioning system of claim 1, wherein: The LSTM module adopts a sliding window technique to dynamically model the time series signal, and the window length is T w Each update is according to the following formula: where, is the predicted output, y t is the actual value; t is the time step index; L LSTM is the loss function value based on the LSTM model, ||·|| represents the norm operation.

7. The high precision signal conditioning system of claim 1, wherein: ​ If δ t > δ 阈值 , then determine an abnormal signal; where δ t represents the deviation value between the prediction error and the actual signal at time step t, and δ 阈值 represents a pre-set fixed value as the standard limit for determining whether the signal is abnormal.

8. A high-precision signal conditioning method based on the high-precision signal conditioning system of any one of claims 1-7, characterized in that, ​ ​ ​ ​ ​ S5, time series modeling and prediction: input the feature representation generated by the VAE module into the LSTM module, use the hidden state to predict the future value of the signal based on the dynamic modeling formula of the time series; S6, anomaly detection and feedback optimization: determine the abnormal signal according to the deviation between the predicted value and the actual value; adjust the filter frequency range or the zero offset compensation of the signal conversion module according to the feedback result; S7, signal output: output the optimized signal for external devices to use.

Images