Dynamic de-noising circuit for multi-dimensional signals and working method of dynamic de-noising circuit

Abstract

The embodiment of the invention discloses a dynamic de-noising circuit for a multi-dimensional signal and a working method thereof, and the circuit comprises a multi-dimensional signal selection unit which is used for selecting a multi-dimensional signal for dynamic de-noising; the signal data buffer dual port is used for storing a multi-dimensional signal; the multiplication and addition denoising unit is used for carrying out Gaussian denoising processing on the stored multi-dimensional signals; the denoising coefficient storage single port is used for storing a denoising coefficient for the multiplication and addition denoising unit to retrieve, and the denoising coefficient is an optimal denoising coefficient determined by learning multi-dimensional signal features under different conditions through the adaptive denoising coefficient adjustment unit; and the self-adaptive denoising coefficient adjusting unit is used for adjusting the denoising coefficient according to a denoising result. By implementing the circuit disclosed by the embodiment of the invention, the aim of reducing the scale and the power consumption of the denoising circuit can be fulfilled, so that the aim of miniaturization and low power consumption of multi-dimensional signal detection equipment is fulfilled, the circuit can achieve the optimal denoising effect in different working environments, and the accuracy and the reliability of detection are improved.

Description

Technical Field

[0001] The invention relates to the technical field of live detection of electric power systems, and in particular to a dynamic denoising circuit for multi-dimensional signals and a working method thereof. Background Art

[0002] In the field of live detection of power systems, infrared thermal imaging temperature measurement detection technology and partial discharge detection technology including transient ground voltage detection, ultrasonic detection, high frequency detection, ultra-high frequency detection and other detection methods have been widely used. Infrared thermal imaging temperature measurement detection technology can detect and find equipment failures with abnormal temperatures in power systems, and partial discharge detection technology can detect partial discharge failures in transformers, combination electrical appliances, switch cabinets and other equipment.

[0003] At present, infrared thermal imaging temperature measurement and detection equipment and partial discharge detection equipment are independent of each other, and the equipment is large and inconvenient to carry, resulting in heavy detection tasks and long detection cycles, which affects the implementation of operation and maintenance tasks and the efficiency of operation and maintenance management. Therefore, the integration of infrared thermal imaging detection, transient ground voltage detection, ultrasonic detection and other conventional detection technologies has become an urgent need for power system operation and maintenance detection. In addition, if the processing of infrared, transient ground voltage, ultrasonic and other signals is simply a simple mechanical superposition of multiple devices to achieve the integration of multiple detection functions, it only formally provides a multi-dimensional signal detection device, but cannot solve the problems of miniaturization, low power consumption, and long battery life of the equipment; and the denoising coefficients used in the automatic denoising method of the current denoising circuit are all denoised according to the pre-set coefficients. In this way, the circuit cannot achieve the optimal denoising effect under different working environments.

[0004] Therefore, it is necessary to design a new circuit to reduce the scale and power consumption of the denoising circuit, so that the multi-dimensional signal detection equipment can achieve the purpose of miniaturization and low power consumption, and automatically adjust the denoising coefficient according to the characteristics of the input signal to ensure that the circuit can achieve the optimal denoising effect under different working environments, thereby improving the accuracy and reliability of detection. Summary of the invention

[0005] The technical problem to be solved by the present invention is to provide a dynamic denoising circuit for multi-dimensional signals and a working method thereof.

[0006] To solve the above technical problems, the object of the present invention is achieved through the following technical solutions: Provide a dynamic denoising circuit for a multidimensional signal, characterized in that it includes: a denoising mode control unit, a multidimensional signal selection unit, a signal data buffer dual port, an adaptive denoising coefficient adjustment unit and a multiplication-addition denoising unit; wherein the multidimensional signal selection unit is used to select a multidimensional signal for dynamic denoising; the signal data buffer dual port is used to store the multidimensional signal; the multiplication-addition denoising unit is used to perform Gaussian denoising on the multidimensional signal stored in the signal data buffer dual port; the denoising mode control unit is connected to the multidimensional signal selection unit, the signal data buffer dual port and the multiplication-addition denoising unit respectively;

[0007] It also includes a denoising coefficient storage single port; wherein the denoising coefficient storage single port is used to store the denoising coefficient for the multiplication and addition denoising unit to retrieve, and the denoising coefficient is the optimal denoising coefficient determined by the adaptive denoising coefficient adjustment unit learning the multidimensional signal characteristics of different situations; the adaptive denoising coefficient adjustment unit is used to adjust the denoising coefficient in the denoising coefficient storage single port according to the denoising result.

[0008] A further technical solution thereof is: it also includes an enhanced signal preprocessing unit, and the enhanced signal preprocessing unit is used for filtering, amplifying and correcting the original signal.

[0009] Its further technical solution is: the enhanced signal preprocessing unit includes a programmable gain amplifier, a programmable filter, an automatic correction circuit, a real-time signal quality monitoring module and a microcontroller, and the programmable gain amplifier, programmable filter, automatic correction circuit, and real-time signal quality monitoring module are respectively connected to the microcontroller; the microcontroller is connected to the multidimensional signal selection unit.

[0010] A further technical solution is: the multiplication-addition denoising unit comprises a multiplication-addition circuit and a filtering circuit connected in sequence.

[0011] In addition, the technical problem to be solved by the present invention is to provide a working method of a dynamic denoising circuit for a multi-dimensional signal, comprising:

[0012] The multi-dimensional signal selection unit selects the multi-dimensional signal for dynamic denoising;

[0013] The signal data buffer dual ports store the multi-dimensional signal;

[0014] The multiplication-addition denoising unit retrieves the denoising coefficient from the denoising coefficient storage single port, and performs Gaussian denoising on the multidimensional signal stored in the signal data buffer dual ports;

[0015] The adaptive denoising coefficient adjustment unit adjusts the denoising coefficient in the denoising coefficient storage port according to the denoising result.

[0016] A further technical solution thereof is: determining whether the multi-dimensional signal is a calibrated infrared signal;

[0017] If the multi-dimensional signal is a corrected infrared signal, two lines of buffered infrared signals are read out in sequence from two groups of signal data buffer dual ports;

[0018] The Gaussian denoising filter coefficient is read out from the denoising coefficient storage single port to obtain the denoising coefficient;

[0019] The multiplication-addition denoising unit uses 9 multipliers and 8 adders and utilizes the denoising coefficient to perform a two-dimensional convolution operation on the infrared signal to obtain a signal after Gaussian denoising processing.

[0020] A further technical solution is: after determining whether the multi-dimensional signal is a calibrated infrared signal, the method further comprises:

[0021] If the multi-dimensional signal is not a calibrated infrared signal, determining whether the multi-dimensional signal is a transient voltage signal;

[0022] If the multi-dimensional signal is a transient voltage signal, the transient voltage signal is sequentially read out from two groups of signal data buffer dual ports;

[0023] The L-order low-pass denoising filter coefficient is read out from the denoising coefficient storage single port to obtain the denoising coefficient;

[0024] The multiplication-addition denoising unit performs a one-dimensional convolution operation on the transient voltage signal using a denoising coefficient to obtain a signal after Gaussian denoising.

[0025] A further technical solution is: after determining whether the multi-dimensional signal is a transient voltage signal, the method further comprises:

[0026] If the multi-dimensional signal is not a transient voltage signal, ultrasonic signals are sequentially read out from two groups of signal data buffer dual ports;

[0027] The L-order bandpass denoising filter coefficient is read out from the denoising coefficient storage single port to obtain the denoising coefficient;

[0028] The multiplication-addition denoising unit performs a one-dimensional convolution operation on the ultrasonic signal using a denoising coefficient to obtain a signal after Gaussian denoising processing.

[0029] A further technical solution is: before the multidimensional signal selection unit selects the multidimensional signal for dynamic denoising, it also includes:

[0030] A denoising mode control unit configures a denoising mode;

[0031] The denoising mode control unit configures the denoising coefficient.

[0032] A further technical solution is that the adaptive denoising coefficient adjustment unit adjusts the denoising coefficient in the denoising coefficient storage port according to the denoising result, including:

[0033] The adaptive denoising coefficient adjustment unit performs a quality assessment on the denoising result, and fine-tunes the denoising coefficient currently in use according to the quality assessment result.

[0034] Compared with the prior art, the beneficial effects of the present invention are as follows: the present invention utilizes multi-dimensional signal selection and signal data buffering, implements Gaussian denoising processing through a multiplication-addition denoising unit, and also uses adaptive denoising coefficient adjustment to determine the optimal denoising coefficient by learning multi-dimensional signal characteristics, and the denoising mode control unit manages the overall denoising process, thereby achieving the purpose of reducing the scale and power consumption of the denoising circuit, so that the multi-dimensional signal detection device can achieve the purpose of miniaturization and low power consumption, and automatically adjusts the denoising coefficient according to the characteristics of the input signal, ensuring that the circuit can achieve the optimal denoising effect under different working environments, thereby improving the accuracy and reliability of detection.

[0035] The present invention is further described below in conjunction with the accompanying drawings and specific embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0036] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without paying any creative work.

[0037] Figure 1 A schematic block diagram of a dynamic denoising circuit for multi-dimensional signals provided in an embodiment of the present invention;

[0038] Figure 2 A schematic flow chart of a working method of a dynamic denoising circuit for a multi-dimensional signal provided by an embodiment of the present invention;

[0039] Figure 3 A schematic diagram of a sub-flow diagram of a working method of a dynamic denoising circuit for a multi-dimensional signal provided by an embodiment of the present invention;

[0040] Figure 4 A schematic flow chart of a working method of a dynamic denoising circuit for multi-dimensional signals provided by another embodiment of the present invention;

[0041] Description of the symbols in the figure:

[0042] 10. De-noising mode control unit; 20. Multi-dimensional signal selection unit; 30. Signal data buffer dual ports; 40. Multiplication and addition denoising unit; 50. De-noising coefficient storage single port; 60. Main control unit; 70. Adaptive denoising coefficient adjustment unit; 80. Enhanced signal preprocessing unit. DETAILED DESCRIPTION

[0043] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0044] It should be understood that when used in this specification and the appended claims, the terms "include" and "comprises" indicate the presence of described features, integers, steps, operations, elements and / or components, but do not exclude the presence or addition of one or more other features, integers, steps, operations, elements, components and / or combinations thereof.

[0045] It should also be understood that the terms used in this specification of the present invention are only for the purpose of describing specific embodiments and are not intended to limit the present invention. As used in the specification of the present invention and the appended claims, unless the context clearly indicates otherwise, the singular forms "a", "an" and "the" are intended to include plural forms.

[0046] It should be further understood that the term "and / or" used in the present description and the appended claims refers to any and all possible combinations of one or more of the associated listed items, and includes these combinations.

[0047] See also Figure 1 , Figure 1A schematic block diagram of a dynamic denoising circuit for a multidimensional signal provided by an embodiment of the present invention. The circuit can detect infrared, transient ground voltage, ultrasonic and other detections of power equipment in a time-sharing manner, and the collected infrared, transient ground voltage, ultrasonic multidimensional signals are time-sharingly called for time-sharing multiplexing of circuit resources in the same signal processing circuit. In the multidimensional signal processing process, firstly, the denoising mode and denoising coefficient are configured, and then one of the infrared, transient ground voltage, ultrasonic and other multidimensional signals is selected to be cached and written into the dual-port 30RAM, and then the cached signal data and denoising coefficient are read, and the multiplication and addition circuit is called as needed according to the denoising mode, and the denoising circuit structure is reconstructed to perform time-sharing denoising processing of the multidimensional signal, and finally the denoised infrared, ultrasonic and transient ground voltage signals are output for further signal processing by subsequent circuits, thereby achieving the purpose of reducing the scale and power consumption of the denoising circuit, making the miniaturization and low-power design of the multidimensional signal detection equipment possible.

[0048] In addition, the optimal denoising coefficient is determined by learning the multi-dimensional signal characteristics of different situations with the help of an adaptive denoising coefficient adjustment unit and stored in a denoising coefficient storage single port. Moreover, the adaptive denoising coefficient adjustment unit will also adjust the corresponding denoising coefficient according to the result after each denoising to ensure that the circuit can achieve the optimal denoising effect under different working environments, thereby improving the accuracy and reliability of detection.

[0049] See also Figure 1 The above-mentioned dynamic denoising circuit for a multidimensional signal comprises: a denoising mode control unit 10, a multidimensional signal selection unit 20, a signal data buffer dual port 30 and a multiplication-addition denoising unit 40; wherein the multidimensional signal selection unit 20 is used to select a multidimensional signal for dynamic denoising; the signal data buffer dual port 30 is used to store the multidimensional signal; the multiplication-addition denoising unit 40 is used to perform Gaussian denoising on the multidimensional signal stored in the signal data buffer dual port 30; the denoising mode control unit 10 is respectively connected to the multidimensional signal selection unit 20, the signal data buffer dual port 30 and the multiplication-addition denoising unit 40;

[0050] It also includes a denoising coefficient storage single port 50; wherein the denoising coefficient storage single port 50 is used to store the denoising coefficient for the multiplication and addition denoising unit to retrieve, and the denoising coefficient is the optimal denoising coefficient determined by learning the multidimensional signal characteristics of different situations by the adaptive denoising coefficient adjustment unit 70; the adaptive denoising coefficient adjustment unit 70 is used to adjust the denoising coefficient in the denoising coefficient storage single port according to the denoising result.

[0051] The denoising coefficient storage single port 50 and the denoising mode control unit 10 are respectively connected to the main control unit 60 for controlling the operation of the denoising mode control unit 10 and the denoising coefficient storage single port 50 .

[0052] The circuit of this embodiment performs dynamic denoising processing on the collected multi-dimensional signals such as infrared thermal radiation, transient ground voltage, and ultrasonic waves through a dynamic denoising circuit, thereby effectively time-division multiplexing the storage circuit and the multiplication and addition circuit for signal processing, greatly reducing the circuit scale and circuit power consumption of the multi-dimensional signal denoising processing, and enabling multiple live detection functions to be integrated on a single miniaturized portable device.

[0053] The adaptive denoising coefficient adjustment unit 70 uses a machine learning algorithm to adjust the denoising coefficient in real time according to the characteristics of the input signal to ensure the best denoising effect under different working environments, thereby improving the accuracy and reliability of detection.

[0054] In one embodiment, the denoising coefficient storage single port 50 includes two single-port 50 RAMs.

[0055] In one embodiment, the signal data buffer dual port 30 includes two dual port 30 RAMs.

[0056] In one embodiment, the multiplication-addition denoising unit 40 includes a multiplication-addition circuit and a filtering circuit connected in sequence.

[0057] In this embodiment, the original infrared array data IR with N×M resolution output by the infrared detector is collected by an ADC converter. ori (x, y), after non-uniformity correction, output the corrected infrared array data IR nuc (x, y), which is the corrected infrared signal. N represents the number of columns of the infrared array, M represents the number of rows of the infrared array, x represents the horizontal coordinate of the infrared array data, and y represents the vertical coordinate of the infrared array data. nuc (x, y) represents the intensity value of the infrared array data after non-uniform correction of the pixel with coordinates (x, y). nuc (x, y) with a 3×3 Gaussian kernel function Perform Gaussian filtering to remove noise and convert IR nuc (x, y) is convolved with the Gaussian kernel function, i.e. Gaussian denoising infrared data The multiplication-addition denoising unit 40 needs to cache two rows of data through two sets of dual-port 30 RAMs with a storage depth of N, and implement Gaussian denoising of infrared data through a circuit of 9 multipliers and 8 adders.

[0058] The transient ground voltage signal is processed by the analog conditioning circuit including the analog filtering circuit, the logarithmic amplifier circuit, and the peak detection circuit, and the most important peak information in the transient ground voltage signal is output. After ADC sampling, the L-order FIR low-pass denoising process is performed in the digital domain to remove the high-frequency noise contained in the transient ground voltage detection data and improve the accuracy of transient ground voltage detection. The input sequence x of the L even-order FIR low-pass filter is tev (n) and the transfer function h tev (n) Convolution, output denoised transient voltage detection data: The denoising filter adopts the semi-parallel filter design concept, firstly divides the denoising coefficients into two groups and Stored in two sets of single-port 50 RAM, and then using two The deep dual-port 30RAM realizes data buffering, and reads the data in the dual-port 30RAM and the denoising coefficient in the single-port 50RAM again through the read address control, and realizes the filtering operation through time-sharing multiplexing of two multipliers and two adders, and finally outputs the transient voltage detection data y after low-pass denoising tev (n).

[0059] After the ultrasonic signal is filtered by the analog amplifier circuit and the analog low-pass filter with a cutoff frequency of 200kHz, the digital ultrasonic signal is obtained by sampling the ADC with a sampling frequency of 400kHz. The digital ultrasonic signal is processed by L-order FIR bandpass filtering to extract the denoised ultrasonic signal with a center frequency of 40kHz and a frequency range of 20kHz to 60kHz. The input sequence x of the L-order FIR bandpass denoising filter is us (n) and the transfer function h us (n) Convolution, output bandpass denoised ultrasonic signal: The denoising filter uses the semi-parallel filter design concept to divide the denoising coefficients into two groups: and Stored in two sets of single-port 50 RAM. The deep dual-port 30RAM realizes data buffering, and reads the corresponding data in the dual-port 30RAM and the bandpass denoising coefficient in the single-port 50RAM through read address control. The filtering operation is realized by time-division multiplexing two multipliers and two adders, and finally the ultrasonic signal y after bandpass filtering is output. us (n).

[0060] In addition, the workflow of the adaptive denoising coefficient adjustment unit is as follows:

[0061] First, select a suitable machine learning algorithm, such as neural network, support vector machine (SVM), decision tree, etc., according to the complexity of signal features and the amount of data; extract key features from the input signal, i.e., multidimensional signal, through spectrum analysis, time domain analysis, energy distribution and other methods. These features will directly affect the subsequent denoising effect; use machine learning algorithms with labeled data sets for training, so that the model can learn the complex relationship between signal features and the optimal denoising coefficient. A large amount of data is required during the training process to ensure the accuracy and generalization ability of the model. This forms a trained model that can be used to extract the optimal denoising coefficient corresponding to the multidimensional signal features in different situations.

[0062] Secondly, a memory or data structure, namely a denoising coefficient storage port 50, is set to store the best denoising coefficients under different conditions. These coefficients are dynamically updated and adjusted according to the output of the machine learning algorithm.

[0063] Next, at runtime, according to the characteristics of the current input signal, i.e., the multi-dimensional information, appropriate denoising coefficients are obtained from the denoising coefficient storage single port 50. These coefficients are obtained based on previous training and learning, and can be adjusted according to real-time requirements.

[0064] Finally, the quality of the denoised signal is evaluated, such as calculating the signal-to-noise ratio and analyzing the energy within a specific frequency range. These evaluations help the system understand the current processing effect and make decisions for further adjustments. Based on the real-time feedback evaluation, decide whether to fine-tune the currently used denoising coefficient. Through minor adjustments, the denoising effect can be improved to adapt to the changing working environment or signal conditions. Specifically, by calculating the power or energy ratio of the signal to the noise, a quantitative indicator of signal quality can be obtained. The common calculation formula is SNR = 10*log10(P signal / P noise ), where P signal is the signal power, P noise is the noise power.

[0065] Analyze the energy distribution of the signal within a specific frequency range using spectrum analysis tools or Fourier transform and other techniques. This helps determine whether there is noise or interference in the frequency domain; feedback the results of the quality assessment to the denoising mode control unit. This can be achieved through the data transmission and processing module to ensure the timeliness and accuracy of the feedback information. The system determines whether the currently used denoising coefficient is still valid based on the results of the real-time quality assessment. If the signal quality is poor or the working environment changes, the system can trigger the fine-tuning process of the denoising coefficient. Using the previously learned multi-dimensional signal features and the data stored in the denoising coefficient storage port of the best denoising coefficient, the system can use an adaptive algorithm to adjust the denoising coefficient. This algorithm can perform more accurate denoising processing on the signal under the current working conditions based on the quality assessment results fed back in real time; fine-tuning usually uses small increments or decrements to avoid excessive adjustments that cause signal distortion or reduce the processing effect. This ensures that the system can adapt to the dynamic changes in the working environment and signal conditions, and maintain a high denoising effect and signal quality.

[0066] Through the implementation of the above steps, the dynamic denoising circuit can intelligently adjust and optimize the denoising effect according to the real-time quality evaluation results to ensure stable and high-quality signal processing under different working conditions and signal environments.

[0067] These steps form a tightly linked closed-loop system: from identifying and learning signal features, to dynamically adjusting the denoising coefficients based on the learning results, to optimizing the denoising effect based on real-time feedback. This design not only improves the performance and stability of the signal processing system, but also ensures that the best denoising effect can be achieved under various working conditions, thereby effectively improving the overall efficiency and responsiveness of the system.

[0068] In one embodiment, the above-mentioned dynamic denoising circuit for multi-dimensional signals further includes an enhanced signal preprocessing unit 80, and the enhanced signal preprocessing unit 80 is used to filter, amplify and correct the original signal.

[0069] Specifically, the enhanced signal preprocessing unit 80 includes a programmable gain amplifier, a programmable filter, an automatic correction circuit, a real-time signal quality monitoring module and a microcontroller, and the programmable gain amplifier, programmable filter, automatic correction circuit, and real-time signal quality monitoring module are respectively connected to the microcontroller; the microcontroller is connected to the multidimensional signal selection unit.

[0070] In this embodiment, the model of the programmable gain amplifier is AD8367, which has a high gain range and precise gain control capability, can provide a wide dynamic range adjustment in the front end of signal processing, and is suitable for the pre-processing of the multi-dimensional signal selection unit.

[0071] The model of the programmable filter is AD9361, which includes a high-performance digital filter that can filter the signal immediately after receiving it to meet different frequency and bandwidth requirements.

[0072] The automatic correction circuit uses a high-precision, low-noise digital-to-analog converter, model AD5791, which can be used to generate accurate correction signals. Combined with a microcontroller or FPGA, it can achieve accurate offset and gain correction of the signal and eliminate system errors.

[0073] The real-time signal quality monitoring module is a high-speed, high-resolution analog-to-digital converter, model AD9467, which can collect and monitor signal quality parameters such as signal-to-noise ratio and distortion in real time. These parameters can be analyzed in real time by the embedded processor to dynamically adjust the parameters of the pre-processing module.

[0074] The microcontroller model is Microchip PIC32MX, which has rich peripherals and communication interfaces and can be used to manage and control the operation of the entire pre-processing module. It can efficiently exchange and coordinate data with various components such as PGA, PF, DAC and ADC.

[0075] The integration and optimization of these components can form an efficient enhanced signal preprocessing unit 80, which can accurately filter, amplify and correct the original signal before the multi-dimensional signal selection unit, thereby significantly improving the effect of subsequent denoising processing and the overall system performance.

[0076] In another embodiment, the above-mentioned dynamic denoising circuit for multi-dimensional signals further includes: a denoising strategy configuration unit, which is connected to the denoising mode control unit 10 and is used to configure corresponding denoising parameters and algorithm strategies according to the signal type, noise characteristics and denoising targets set by the user. At the same time, the denoising mode control unit 10 uses a machine learning model to predict the best denoising solution.

[0077] Specifically, the input signal needs to be classified and the noise characteristics analyzed first. This involves spectrum analysis of sensor data, time domain feature extraction, and noise power spectrum estimation. These analyses can help the system understand the basic characteristics of the signal and the statistical characteristics of the noise, providing a basis for subsequent denoising decisions.

[0078] Next, you can set the denoising expectations through the system interface or preset parameters, such as reducing the noise level while retaining signal details. These goals will serve as the basis for the system to optimize the denoising algorithm and parameter selection.

[0079] Secondly, according to the signal analysis and the goals set by the user, the denoising strategy configuration unit is responsible for selecting the appropriate denoising algorithm and parameter configuration. This includes selecting processing methods for different frequency bands or time domains, setting filter types and parameters, and adjusting the intensity and timeliness of denoising;

[0080] Next, the denoising mode control unit integrates a machine learning model, whose goal is to predict the best denoising solution based on real-time signal characteristics and previous learning experience. This model can be a classifier or regressor based on supervised learning, a clustering or anomaly detection model based on unsupervised learning, or even an intelligent agent based on reinforcement learning.

[0081] The model is usually trained based on historical data sets, including processing results under various signal types and noise scenarios. When running in real time, the model will output the recommended optimal denoising strategy based on the characteristics of the current signal and the prediction target. This may involve selecting a specific algorithm, adjusting parameters, or dynamically adjusting the priority and intensity of denoising.

[0082] Finally, after applying the denoising algorithm, the quality of the processed signal is monitored in real time. This feedback includes changes in the signal-to-noise ratio, measures of signal distortion, or user-defined quality indicators. If the system detects that the denoising is not working well or the signal environment has changed, it can automatically adjust the denoising parameters or algorithms to optimize the processing results.

[0083] Through these steps and techniques, the dynamic denoising system can efficiently cope with complex signal processing scenarios, ensuring excellent processing performance and user satisfaction under various noise interferences and changes in signal characteristics.

[0084] In another embodiment, the above-mentioned dynamic denoising circuit for multi-dimensional signals further includes: a high-performance data cache and processing array, which reads signal data and denoising coefficients according to the instructions of the denoising strategy configuration unit and executes the selected denoising algorithm. During the processing, the denoising effect is monitored in real time, and the denoising parameters are dynamically adjusted as needed.

[0085] In this embodiment, appropriate denoising algorithms and parameters are selected according to the real-time requirements of the system and user settings. The high-performance data cache and processing array includes a software controller that dynamically generates denoising configuration instructions based on a preset algorithm library and parameter table, or an intelligent model (such as a machine learning model), according to real-time signal characteristics and denoising effect requirements.

[0086] The high-performance data cache and processing array contains a denoising algorithm library and execution unit. These algorithms can cover frequency domain processing (such as FFT, filter design), time domain processing (such as convolution, mean filtering) and complex signal processing algorithms (such as wavelet transform, deep learning model). According to the instructions of the denoising strategy configuration unit, the selected denoising algorithm is called by the execution unit and applied to the input signal in the data stream.

[0087] During the signal processing, the adaptive denoising coefficient adjustment unit 70 monitors the denoising effect in real time. This includes calculating the signal-to-noise ratio of the signal, evaluating the spectral characteristics of the denoised signal or other relevant quality indicators. The monitoring results are used to feed back to the denoising strategy configuration unit and dynamically adjust the denoising parameters.

[0088] Based on the feedback from real-time monitoring, the adaptive denoising coefficient adjustment unit 70 can dynamically adjust the denoising parameters of the current application in combination with the high-performance data cache and processing array. This involves adjusting the cutoff frequency of the filter, increasing or decreasing the denoising intensity, modifying the processing window size of the algorithm, etc. The adjustment can be achieved through a software controller to ensure that the system can quickly respond to changes in the signal environment and adjustments to user needs, that is, the adaptive denoising coefficient adjustment unit 70 adjusts the denoising coefficient, while the high-performance data cache and processing array adjust other contents.

[0089] As a result, the high-performance data cache and processing array can achieve effective denoising for various signal types and complex noise environments while ensuring processing speed and efficiency. This system architecture is not only suitable for application scenarios with high real-time processing requirements, but also can flexibly adjust and optimize denoising strategies according to different application requirements.

[0090] The above-mentioned dynamic denoising circuit for multi-dimensional signals selects one of the infrared, transient ground voltage, ultrasonic and other multi-dimensional signals to be cached and written into the signal data buffer dual port 30 through the configured denoising mode and denoising coefficient, then reads the cached signal data and denoising coefficient, calls the multiplication and addition denoising unit 40 as needed according to the denoising mode, and reconstructs the denoising circuit structure to perform time-sharing denoising processing of the multi-dimensional signal, and finally outputs the denoised infrared, ultrasonic and transient ground voltage signals, thereby achieving the purpose of reducing the scale and power consumption of the denoising circuit, and making the multi-dimensional signal detection equipment miniaturized and low-power.

[0091] In one embodiment, see Figure 2 The working method of the above-mentioned dynamic denoising circuit for multi-dimensional signals includes steps S110 to S140.

[0092] S110, the multidimensional signal selection unit 20 selects a multidimensional signal for dynamic denoising;

[0093] S120, the signal data buffer dual port 30 stores the multi-dimensional signal;

[0094] S130, the multiplication-addition denoising unit 40 performs Gaussian denoising processing on the multidimensional signal stored in the signal data buffer dual port 30;

[0095] In one embodiment, see Figure 3 , the above-mentioned step S130 may include steps S130a~S130k.

[0096] S130a, determining whether the multi-dimensional signal is a calibrated infrared signal;

[0097] S130b, if the multi-dimensional signal is a calibrated infrared signal, read out two lines of buffered infrared signals from two groups of signal data buffer dual ports 30 in sequence;

[0098] S130c, reading out the Gaussian denoising filter coefficient from the denoising coefficient storage single port 50 to obtain the denoising coefficient;

[0099] S130d, the multiplication-addition denoising unit 40 performs a two-dimensional convolution operation on the infrared signal using the denoising coefficient to obtain a signal after Gaussian denoising processing.

[0100] Specifically, the multiplication-addition denoising unit 40 performs a two-dimensional convolution operation on the infrared signal through 9 multipliers and 8 adders and utilizes the denoising coefficient to obtain a signal after Gaussian denoising processing.

[0101] S130e, if the multi-dimensional signal is not a calibrated infrared signal, determining whether the multi-dimensional signal is a transient voltage signal;

[0102] S130f, if the multi-dimensional signal is a transient voltage signal, then sequentially read out the transient voltage signals from the two groups of signal data buffer dual ports 30;

[0103] S130g, reading out the L-order low-pass denoising filter coefficient from the denoising coefficient storage single port 50 to obtain the denoising coefficient;

[0104] S130h: The multiplication-addition denoising unit 40 performs a one-dimensional convolution operation on the transient ground voltage signal using the denoising coefficient to obtain a signal after Gaussian denoising.

[0105] Specifically, the multiplication-addition denoising unit 40 performs a one-dimensional convolution operation on the transient voltage signal by time-division multiplexing two multipliers and two adders and using a denoising coefficient to implement a filtering operation to obtain a signal after Gaussian denoising.

[0106] S130i, if the multi-dimensional signal is not a transient voltage signal, then read out ultrasonic signals from the two sets of signal data buffer dual ports 30 in sequence;

[0107] S130j, reading out the L-order bandpass denoising filter coefficient from the denoising coefficient storage single port 50 to obtain the denoising coefficient;

[0108] S130k: The multiplication-addition denoising unit 40 performs a one-dimensional convolution operation on the ultrasonic signal using the denoising coefficient to obtain a signal after Gaussian denoising processing.

[0109] Specifically, the multiplication-addition denoising unit 40 performs a one-dimensional convolution operation on the transient ground voltage signal by time-division multiplexing two multipliers and two adders and using an L-order bandpass denoising filter coefficient to implement a filtering operation to obtain a signal after Gaussian denoising.

[0110] In this embodiment, the denoising mode control unit 10 selects the input signal to be denoised, and selects one of the infrared, transient voltage, and ultrasonic signals to write into the dual-port 30RAM. When the dynamic denoising circuit performs infrared signal denoising, two sets of dual-port 30RAMs with a storage depth of N are configured, and each set of dual-port 30RAMs caches adjacent rows of infrared data; when the dynamic denoising circuit performs transient voltage or ultrasonic signal denoising, two sets of storage depths are configured. The dual-port 30RAM of each group caches half-depth signal processing data.

[0111] The de-noising mode control unit 10 controls the read address generator of the dual-port 30RAM and controls the read port data output of the dual-port 30RAM. When the dynamic de-noising circuit performs infrared signal de-noising, two lines of infrared buffer data are read out from two groups of N-depth dual-port 30RAM in sequence. When the dynamic de-noising circuit performs transient voltage or ultrasonic signal de-noising, two lines of infrared buffer data are read out from two groups of N-depth dual-port 30RAM in sequence. The deep dual-port 300 RAM reads out two sets of transient voltage or ultrasonic buffer data in sequence.

[0112] The denoising mode control unit 10 controls the read address generator of the single-port 50RAM to read out the denoising coefficients of the corresponding denoising modes in sequence. When the dynamic denoising circuit performs infrared signal denoising, the Gaussian denoising filter coefficients are read out from the single-port 50RAM. When the dynamic denoising circuit performs transient voltage signal denoising, the L-order low-pass denoising filter coefficients are read out from the single-port 50RAM. When the dynamic denoising circuit performs ultrasonic signal denoising, the L-order band-pass denoising filter coefficients are read out from the single-port 50RAM.

[0113] The denoising mode control unit 10 configures a multiplication and addition circuit and a filtering network to reconstruct the denoising circuit. When the dynamic denoising circuit performs infrared signal denoising, the multiplication and addition circuit and the filtering network are configured to implement a two-dimensional convolution operation calculation: When the dynamic denoising circuit performs denoising on the transient voltage signal, the multiplication and addition circuit and the filter network are configured to implement one-dimensional convolution operation calculation: When the dynamic denoising circuit performs denoising on ultrasonic signals, the multiplication and addition circuit and the filter network are configured to implement one-dimensional convolution operation calculation:

[0114] Through the dynamic denoising circuit of this embodiment, two groups of MAX (N, L / 2) deep dual-port 30RAM, two groups of L / 2 deep single-port 50RAM, 9 multiplication circuits, and 8 addition circuits can complete the multi-dimensional denoising processing of infrared, transient ground voltage, and ultrasonic signals, while the use of a separate parallel three-way denoising circuit requires 2×N+4×L / 2 groups of dual-port 30RAM, 2×2×L / 2 groups of single-port 50RAM, 13 multiplication circuits, and 12 addition circuits to complete the corresponding denoising filtering processing. Therefore, the present invention effectively reduces the circuit scale and circuit power consumption of the denoising circuit, which is very critical to the integration, miniaturization, and low-power design of multi-dimensional signal detection products.

[0115] S140 , the adaptive denoising coefficient adjustment unit 70 adjusts the denoising coefficient in the denoising coefficient storage single port 50 according to the denoising result.

[0116] In this embodiment, the adaptive denoising coefficient adjustment unit 70 performs quality assessment on the denoising result, and fine-tunes the currently used denoising coefficient according to the quality assessment result.

[0117] It should be noted that technicians in the relevant field can clearly understand that the specific implementation process of the working method of the above-mentioned dynamic denoising circuit for multi-dimensional signals can refer to the corresponding description in the aforementioned device embodiment, and for the convenience and conciseness of the description, it will not be repeated here.

[0118] Figure 4 FIG. 1 is a flow chart of a working method of a dynamic denoising circuit for multi-dimensional signals provided by another embodiment of the present invention. Figure 4 As shown, the working method of a dynamic denoising circuit for a multi-dimensional signal in this embodiment includes steps S210-S260. Steps S230-S260 are similar to steps S110-S140 in the above embodiment and are not described here. The following is a detailed description of the added steps S210-S220 in this embodiment.

[0119] S210, the denoising mode control unit 10 configures a denoising mode;

[0120] S220 , the denoising mode control unit 10 configures a denoising coefficient.

[0121] The denoising mode refers to the operation mode performed by the multiplication-addition denoising unit 40 , and the denoising coefficients include Gaussian denoising filter coefficients, L-order low-pass denoising filter coefficients, and L-order band-pass denoising filter coefficients.

[0122] It should be noted that technicians in the relevant field can clearly understand that the specific implementation process of the working method of the above-mentioned dynamic denoising circuit for multi-dimensional signals can refer to the corresponding description in the aforementioned device embodiment, and for the convenience and conciseness of the description, it will not be repeated here.

[0123] The above is only a specific embodiment of the present invention, but the protection scope of the present invention is not limited thereto. Any technician familiar with the technical field can easily think of various equivalent modifications or replacements within the technical scope disclosed by the present invention, and these modifications or replacements should be included in the protection scope of the present invention. Therefore, the protection scope of the present invention shall be based on the protection scope of the claims.

Claims

1. A dynamic denoising circuit for multidimensional signals, characterized in that: include: A denoising mode control unit, a multidimensional signal selection unit, a signal data buffer dual port, an adaptive denoising coefficient adjustment unit and a multiplication-addition denoising unit; wherein the multidimensional signal selection unit is used to select a multidimensional signal for dynamic denoising; the signal data buffer dual port is used to store the multidimensional signal; the multiplication-addition denoising unit is used to perform Gaussian denoising on the multidimensional signal stored in the signal data buffer dual port; the denoising mode control unit is connected to the multidimensional signal selection unit, the signal data buffer dual port and the multiplication-addition denoising unit respectively; It also includes a denoising coefficient storage single port; wherein the denoising coefficient storage single port is used to store the denoising coefficient for the multiplication and addition denoising unit to retrieve, and the denoising coefficient is the optimal denoising coefficient determined by the adaptive denoising coefficient adjustment unit learning the multidimensional signal characteristics of different situations; the adaptive denoising coefficient adjustment unit is used to adjust the denoising coefficient in the denoising coefficient storage single port according to the denoising result.

2. A dynamic denoising circuit for multidimensional signals according to claim 1, characterized in that: It also includes an enhanced signal preprocessing unit, which is used to filter, amplify and correct the original signal.

3. The dynamic denoising circuit for multidimensional signals according to claim 1, characterized in that: The enhanced signal preprocessing unit includes a programmable gain amplifier, a programmable filter, an automatic correction circuit, a real-time signal quality monitoring module and a microcontroller. The programmable gain amplifier, the programmable filter, the automatic correction circuit, and the real-time signal quality monitoring module are respectively connected to the microcontroller; the microcontroller is connected to the multidimensional signal selection unit.

4. The dynamic denoising circuit for multidimensional signals according to claim 1, characterized in that: The multiplication-addition denoising unit includes a multiplication-addition circuit and a filtering circuit connected in sequence.

5. A working method of a dynamic denoising circuit for a multi-dimensional signal, characterized in that: include: The multi-dimensional signal selection unit selects the multi-dimensional signal for dynamic denoising; The signal data buffer dual ports store the multi-dimensional signal; The multiplication-addition denoising unit retrieves the denoising coefficient from the denoising coefficient storage single port, and performs Gaussian denoising on the multidimensional signal stored in the signal data buffer dual ports; The adaptive denoising coefficient adjustment unit adjusts the denoising coefficient in the denoising coefficient storage port according to the denoising result.

6. The working method of a dynamic denoising circuit for multi-dimensional signals according to claim 5, characterized in that: The multiplication-addition denoising unit performs Gaussian denoising processing on the multidimensional signal stored in the signal data buffer dual port, including: Determining whether the multi-dimensional signal is a calibrated infrared signal; If the multi-dimensional signal is a corrected infrared signal, two lines of buffered infrared signals are read out in sequence from two groups of signal data buffer dual ports; The Gaussian denoising filter coefficient is read out from the denoising coefficient storage single port to obtain the denoising coefficient; The multiplication-addition denoising unit uses 9 multipliers and 8 adders and utilizes the denoising coefficient to perform a two-dimensional convolution operation on the infrared signal to obtain a signal after Gaussian denoising processing.

7. The working method of a dynamic denoising circuit for multi-dimensional signals according to claim 6, characterized in that: After determining whether the multi-dimensional signal is a calibrated infrared signal, the method further includes: If the multi-dimensional signal is not a calibrated infrared signal, determining whether the multi-dimensional signal is a transient voltage signal; If the multi-dimensional signal is a transient voltage signal, the transient voltage signal is sequentially read out from two groups of signal data buffer dual ports; The L-order low-pass denoising filter coefficient is read out from the denoising coefficient storage single port to obtain the denoising coefficient; The multiplication-addition denoising unit performs a one-dimensional convolution operation on the transient voltage signal using a denoising coefficient to obtain a signal after Gaussian denoising.

8. The working method of a dynamic denoising circuit for multi-dimensional signals according to claim 7, characterized in that: After determining whether the multi-dimensional signal is a transient voltage signal, the method further includes: If the multi-dimensional signal is not a transient voltage signal, ultrasonic signals are sequentially read out from two groups of signal data buffer dual ports; The L-order bandpass denoising filter coefficient is read out from the denoising coefficient storage single port to obtain the denoising coefficient; The multiplication-addition denoising unit performs a one-dimensional convolution operation on the ultrasonic signal using a denoising coefficient to obtain a signal after Gaussian denoising processing.

9. The working method of a dynamic denoising circuit for multi-dimensional signals according to claim 5, characterized in that: Before the multidimensional signal selection unit selects the multidimensional signal for dynamic denoising, the method further comprises: A denoising mode control unit configures a denoising mode; The denoising mode control unit configures the denoising coefficient.

10. The working method of a dynamic denoising circuit for multi-dimensional signals according to claim 6, characterized in that: The step of adjusting the denoising coefficient in the denoising coefficient storage port by the adaptive denoising coefficient adjustment unit according to the denoising result includes: The adaptive denoising coefficient adjustment unit performs a quality assessment on the denoising result, and fine-tunes the denoising coefficient currently in use according to the quality assessment result.

Images