Signal amplification and noise reduction method based on triode active load circuit

Abstract

The application relates to the technical field of signal processing, and relates to a signal amplification and noise reduction method based on a triode active load circuit, which comprises the following steps: obtaining an initial electric signal, performing low-noise amplification processing on the initial electric signal to obtain an amplified analog signal, performing analog-digital conversion on the signal to obtain an amplified digital signal, performing discrete wavelet transform on the amplified digital signal to obtain a wavelet coefficient set, calculating the estimated noise level of each wavelet decomposition layer, dividing the wavelet coefficient set into multiple coefficient sub-blocks, calculating an adaptive threshold value according to the estimated noise level, performing threshold shrinkage processing on the wavelet coefficient to obtain primary estimated wavelet coefficients, inputting the primary estimated wavelet coefficients into a Wiener filter unit to perform Wiener filtering processing and obtain final estimated wavelet coefficients, and continuously performing inverse discrete wavelet transform to obtain a target amplified signal. The application can realize high-gain low-noise amplification and adaptive accurate noise reduction of weak electric signals, and significantly improves the signal-to-noise ratio and waveform fidelity of the output signal.

Description

Technical Field

[0001] This invention relates to the field of signal processing technology, and in particular to a signal amplification and noise reduction method based on a transistor active load circuit. Background Technology

[0002] With the rapid development of sensor technology, bioelectric signal detection, structural health monitoring and IoT sensing, the accurate acquisition and processing of weak electrical signals has become a core technical requirement in many application scenarios. Against this backdrop, weak signal amplification and noise reduction technology faces a severe bottleneck challenge in achieving high gain and low noise in large-scale applications. How to effectively amplify weak signals while suppressing noise to the maximum extent is of vital practical significance for improving the sensitivity and reliability of measurement systems.

[0003] Traditional signal amplification and noise reduction schemes often employ a large resistive load connected to the collector of the amplifying transistor to achieve high voltage gain. However, the large resistor itself generates significant thermal noise, preventing further improvement in the signal-to-noise ratio after amplification. In the digital noise reduction stage, traditional methods often use wavelet denoising with a fixed threshold or a single digital filter. Once the threshold and parameters are set, they cannot adapt to the differentiated characteristics of the signal at different frequency levels and locations. Furthermore, there is a lack of systematic coordination between the analog front-end and the digital noise reduction algorithm. Each module is designed and optimized independently, meaning that noise introduced during amplification cannot be effectively compensated for in subsequent digital processing, severely restricting the overall performance improvement of weak signal detection systems. Summary of the Invention

[0004] This invention provides a signal amplification and noise reduction method based on a transistor active load circuit. Its main purpose is to achieve high-gain, low-noise amplification and adaptive, precise noise reduction of weak electrical signals, thereby significantly improving the signal-to-noise ratio and waveform fidelity of the output signal.

[0005] To achieve the above objectives, the present invention provides a signal amplification and noise reduction method based on a transistor active load circuit, comprising:

[0006] The system receives a signal amplification and noise reduction command, and confirms the signal processing system based on the signal amplification and noise reduction command. The signal processing system includes: a load amplification circuit, a wavelet transform unit, a block thresholding unit, a Wiener filter unit, and an inverse wavelet transform unit.

[0007] An initial electrical signal is acquired, and the initial electrical signal is amplified by the load amplifier circuit to obtain an amplified analog signal. The amplified analog signal is then converted from analog to digital to obtain an amplified digital signal.

[0008] The number of wavelet decomposition layers is set to n. The wavelet transform unit performs discrete wavelet transform on the amplified digital signal according to the number of layers to obtain n sets of wavelet coefficients. Each set of wavelet coefficients corresponds to each wavelet decomposition layer, and each set of wavelet coefficients consists of multiple wavelet coefficients.

[0009] Obtain the median absolute deviation of each set of wavelet coefficients, and calculate the estimated noise level of each wavelet decomposition layer based on the median absolute deviation;

[0010] The optimal block length of each set of wavelet coefficients is determined by the block threshold unit based on the estimated noise level. Based on the optimal block length, each set of wavelet coefficients is divided into multiple coefficient sub-blocks. An adaptive threshold is calculated based on the estimated noise level corresponding to each coefficient sub-block. Threshold shrinkage processing is performed on all wavelet coefficients of the corresponding coefficient sub-block according to the adaptive threshold to obtain the initial estimated wavelet coefficients.

[0011] The initial estimated wavelet coefficients are input into the Wiener filter unit. A Wiener filter is constructed based on the initial estimated wavelet coefficients and the estimated noise level of the corresponding wavelet decomposition layer. The Wiener filter is used to perform Wiener filtering on the initial estimated wavelet coefficients to obtain the final estimated wavelet coefficients.

[0012] The final estimated wavelet coefficients are input into the inverse wavelet transform unit to perform a discrete inverse wavelet transform, thereby obtaining the target amplified signal.

[0013] Optionally, the step of further performing analog-to-digital conversion on the amplified analog signal to obtain an amplified digital signal includes:

[0014] The amplified analog signal is input into an analog-to-digital converter, wherein the analog-to-digital converter has a built-in clock source, a sample-and-hold circuit, a quantization encoding circuit, and a digital calibration circuit.

[0015] Set the sampling frequency and quantization bit depth of the amplified analog signal, and use the clock source to generate a sampling clock signal synchronized with the sampling frequency;

[0016] Based on the sampling clock signal, the sample-and-hold circuit performs equally spaced sampling on the amplified analog signal to obtain m discrete sampled voltage values, and each sampled voltage value corresponds to a sampling time.

[0017] Obtain a pseudo-random sequence generated by digital logic, wherein the pseudo-random sequence contains m pseudo-random sequence values, and each pseudo-random sequence value corresponds to a sampling time.

[0018] A jitter signal with a known amplitude is acquired. The known amplitude of the jitter signal is multiplied sequentially by a pseudo-random sequence value to obtain multiple injected jitter values. The injected jitter values ​​are then added to the sampling voltage values ​​of the same sampling time to obtain multiple jitter sampling values. The jitter sampling values ​​are sequentially input into a quantization encoding circuit. The jitter sampling values ​​are uniformly quantized according to the quantization bit depth to obtain the initial quantization encoding value for each sampling time.

[0019] Based on the initial quantization code value and the corresponding pseudo-random sequence value at each sampling time, the cross-correlation value between the initial quantization code value and the pseudo-random sequence value is calculated using the digital calibration circuit, and the nonlinear error parameter of the analog-to-digital converter is extracted based on the cross-correlation value.

[0020] Based on the nonlinear error parameter, real-time digital compensation is performed on the initial quantization code value to obtain the calibrated quantization code value at each sampling time.

[0021] The calibration quantization encoding values ​​of all sampling times are arranged in chronological order to form a time-domain digital signal sequence, which is then used as an amplified digital signal.

[0022] Optionally, setting the sampling frequency and quantization bit depth of the amplified analog signal includes:

[0023] The amplified analog signal is presampled to obtain a digital sample sequence, and the digital sample sequence is subjected to spectral analysis to obtain a signal spectral distribution. The highest effective frequency is then extracted from the signal spectral distribution.

[0024] Set a safety factor, and multiply the highest effective frequency by the safety factor to obtain the sampling frequency;

[0025] The peak voltage and background noise voltage of the amplified analog signal are obtained, and the ratio of the peak voltage to the background noise voltage is calculated to obtain the dynamic range of the signal.

[0026] Set a signal-to-noise ratio (SNR) threshold, and increment the quantization bit depth bit by bit starting from the first quantization bit depth bit. Calculate the theoretical SNR corresponding to each quantization bit depth bit in sequence, and select the first quantization bit depth bit that satisfies the theoretical SNR not lower than the SNR threshold as the theoretical quantization bit depth bit.

[0027] Calculate the minimum quantization bit depth for the dynamic range of the signal, and use the larger of the theoretical quantization bit depth and the minimum quantization bit depth as the initial quantization bit depth;

[0028] Set the bit adjustment step size, and round the initial quantization bit size up to an integer multiple of the bit adjustment step size to obtain the final quantization bit size;

[0029] The sampling frequency and the final quantization bit depth are used as the operating parameters of the analog-to-digital converter to complete the setting of the sampling frequency and quantization bit depth.

[0030] Optionally, the step of calculating the cross-correlation value between the initial quantized encoded value and the pseudo-random sequence value using the digital calibration circuit, and extracting the nonlinear error parameter of the analog-to-digital converter based on the cross-correlation value, includes:

[0031] A set accumulation count is defined, where the accumulation count is a positive integer not greater than m. Based on the accumulation count, the initial quantization encoding value, and the pseudo-random sequence value, the cross-correlation value is calculated according to the following formula:

[0032] ;

[0033] in, For cross-correlation values, To accumulate the number of times, For the first The initial quantized encoded value at each sampling time. For the first The pseudo-random sequence values ​​corresponding to each sampling time;

[0034] The ratio of the cross-correlation value to the known amplitude of the jitter signal is calculated to obtain the estimated value of the nonlinear error parameter;

[0035] The first-order exponential smoothing method is used to iteratively update the estimated value of the nonlinear error parameter, and the updated estimated value of the nonlinear error parameter is used as the nonlinear error parameter of the analog-to-digital converter.

[0036] Optionally, the step of iteratively updating the estimated value of the nonlinear error parameter using a first-order exponential smoothing method, and using the updated estimated value of the nonlinear error parameter as the nonlinear error parameter of the analog-to-digital converter, includes:

[0037] Confirm the current iteration cycle, obtain the estimated value of the nonlinear error parameter for the current iteration cycle, and set a smoothing factor, wherein the smoothing factor has a value range between 0 and 1.

[0038] Determine whether the current iteration cycle is the first iteration. If it is the first iteration, use the current nonlinear error parameter estimate as the updated nonlinear error parameter estimate after the first iteration.

[0039] If it is not the first iteration, obtain the estimated value of the nonlinear error parameter after the previous iteration update, calculate the product of the smoothing factor and the estimated value of the nonlinear error parameter in the current iteration period, and obtain the first weighted value;

[0040] Calculate the difference obtained by subtracting the smoothing factor from 1, and then multiply the difference by the estimated value of the nonlinear error parameter obtained after the previous iteration update to obtain the second weighted value;

[0041] Add the first weighted value to the second weighted value to obtain the estimated value of the nonlinear error parameter after this iteration.

[0042] Repeat the above steps, and use the estimated value of the nonlinear error parameter after each iteration as the nonlinear error parameter of the analog-to-digital converter.

[0043] Optionally, the step of performing real-time digital compensation on the initial quantization code value based on the nonlinear error parameter to obtain the calibrated quantization code value at each sampling time includes:

[0044] Based on the extracted nonlinear error parameters of the analog-to-digital converter, the input range of the analog-to-digital converter is confirmed;

[0045] The input range is divided into multiple continuous and non-overlapping quantization intervals, and the corresponding nonlinear error parameters are calibrated at the two endpoints of each quantization interval.

[0046] Obtain the initial quantization code value at each sampling time, determine the quantization interval to which the initial quantization code value belongs, and mark it as the target quantization interval;

[0047] Read the initial quantization code value of the left endpoint, the nonlinear error parameter corresponding to the left endpoint, the initial quantization code value of the right endpoint, and the nonlinear error parameter corresponding to the right endpoint of the target quantization interval.

[0048] The compensation amount corresponding to the initial quantized encoded value is calculated according to the following formula:

[0049] ;

[0050] in, For compensation amount, This is the initial quantization encoding value. and These are the initial quantization values ​​at the left and right endpoints of the target quantization interval, respectively. and These are the nonlinear error parameters corresponding to the left endpoint and the nonlinear error parameters corresponding to the right endpoint, respectively.

[0051] The calibration quantization value for each sampling time is obtained by subtracting the corresponding compensation amount from the initial quantization code value at each sampling time.

[0052] Optionally, the step of determining the optimal block length for each set of wavelet coefficients based on the estimated noise level using the block threshold unit, and dividing each set of wavelet coefficients into multiple coefficient sub-blocks based on the optimal block length, includes:

[0053] Set the block length search range and search step size for each wavelet decomposition layer, and select candidate block lengths sequentially according to the search step size for the block length search range.

[0054] Based on the length of each candidate block, the set of wavelet coefficients for each corresponding wavelet decomposition layer is divided into multiple candidate sub-blocks;

[0055] Confirm all wavelet coefficients and the estimated noise level of the corresponding wavelet decomposition layer for each candidate sub-block, and calculate the threshold shrinkage mean square risk estimate for each candidate sub-block.

[0056] The threshold shrinkage mean square risk estimates of all candidate sub-blocks under the same candidate block length are summed to obtain the total risk estimate for each candidate block length.

[0057] Traverse all candidate block lengths and select the candidate block length with the smallest total risk estimate as the optimal block length for the corresponding wavelet decomposition layer.

[0058] The wavelet coefficient set of the wavelet decomposition layer is divided into multiple non-overlapping coefficient sub-blocks according to the optimal block length. The number of wavelet coefficients contained in each coefficient sub-block is equal to the optimal block length. If the total number of wavelet coefficients of the wavelet decomposition layer is not divisible by the optimal block length, the remaining undivided wavelet coefficients are merged into the last coefficient sub-block.

[0059] Optionally, the step of constructing a Wiener filter based on the initially estimated wavelet coefficients and the estimated noise level of the corresponding wavelet decomposition layer, and then applying the Wiener filter to perform Wiener filtering on the initially estimated wavelet coefficients to obtain the final estimated wavelet coefficients, includes:

[0060] The filter window size of each wavelet decomposition layer is set according to the estimated noise level of each wavelet decomposition layer.

[0061] The filter window size is used as the sliding window length. Based on the sliding window length, the sliding window is commanded to slide point by point with a sliding step size of 1 until all the initial estimated wavelet coefficients of the wavelet decomposition layer are traversed.

[0062] Confirm the center position of the sliding window, calculate the arithmetic mean of all initially estimated wavelet coefficients within the sliding window, and use the arithmetic mean as the local signal mean at the center position;

[0063] Calculate the variance of all initially estimated wavelet coefficients within the sliding window, and use the variance as the local signal variance at the center position;

[0064] The local signal variance is compared with the estimated noise level of the corresponding wavelet decomposition layer: if the local signal variance is less than or equal to the estimated noise level, the wavelet coefficients at the center position are replaced with the local signal mean; if the local signal variance is greater than the estimated noise level, the difference between the local signal variance and the estimated noise level is calculated, the difference is divided by the local signal variance to obtain the weighting coefficient, the difference between the wavelet coefficients at the center position and the local signal mean is multiplied by the weighting coefficient to obtain the weighted correction difference, and finally the weighted correction difference is added to the local signal mean to obtain the weighted corrected wavelet coefficients.

[0065] The weighted and corrected wavelet coefficients are used as the Wiener-filtered wavelet coefficients at the center position of the corresponding sliding window. After traversing all the center positions of the sliding window, the Wiener-filtered wavelet coefficients of each wavelet decomposition layer are obtained.

[0066] The final estimated wavelet coefficients are obtained by combining the Wiener-filtered wavelet coefficients of all wavelet decomposition layers.

[0067] Optionally, setting the filter window size for each wavelet decomposition layer based on the estimated noise level of each wavelet decomposition layer includes:

[0068] The initial estimated wavelet coefficients of each wavelet decomposition layer are divided into multiple local regions, wherein each local region contains multiple wavelet coefficients;

[0069] Calculate the sum of the absolute values ​​of all wavelet coefficients in each local region, and divide the sum of absolute values ​​by the total number of wavelet coefficients in the local region to obtain the average amplitude of the coefficients in the local region.

[0070] The amplitude signal-to-noise ratio of the local region is obtained by calculating the ratio of the average amplitude of the coefficients to the estimated noise level of the corresponding wavelet decomposition layer.

[0071] Based on the amplitude signal-to-noise ratio, the range of amplitude signal-to-noise ratio values ​​is obtained, and multiple signal-to-noise ratio thresholds are set. Based on the signal-to-noise ratio thresholds, the range of amplitude signal-to-noise ratio values ​​is divided into multiple continuous intervals.

[0072] The amplitude signal-to-noise ratio of each local region is matched with a continuous interval. After the matching is completed, a corresponding filter window size is assigned to each local region, and the center position of each local region is mapped to the value point of the corresponding filter window size.

[0073] Median filtering is applied sequentially to the filter window sizes corresponding to the value points to obtain the filter window sizes of all sliding windows in each wavelet decomposition layer.

[0074] To achieve the above objectives, the present invention also provides a signal amplification and noise reduction system based on a transistor active load circuit, comprising:

[0075] The instruction system confirmation module is used to receive signal amplification and noise reduction instructions and confirm the signal processing system based on the signal amplification and noise reduction instructions. The signal processing system includes: a load amplification circuit, a wavelet transform unit, a block thresholding unit, a Wiener filtering unit, and an inverse wavelet transform unit.

[0076] The low-noise analog-to-digital converter module is used to acquire an initial electrical signal, perform low-noise amplification processing on the initial electrical signal using the load amplifier circuit to obtain an amplified analog signal, and then perform analog-to-digital conversion on the amplified analog signal to obtain an amplified digital signal.

[0077] The wavelet block noise reduction module is used to set the number of wavelet decomposition layers to n. The wavelet transform unit performs discrete wavelet transform on the amplified digital signal according to the number of layers to obtain n sets of wavelet coefficients. Each set of wavelet coefficients corresponds to one wavelet decomposition layer, and each set consists of multiple wavelet coefficients. The absolute median deviation of each set of wavelet coefficients is obtained. Based on the absolute median deviation, the estimated noise level of each wavelet decomposition layer is calculated. The block threshold unit determines the optimal block length of each set of wavelet coefficients based on the estimated noise level. Based on the optimal block length, each set of wavelet coefficients is divided into multiple coefficient sub-blocks. An adaptive threshold is calculated based on the estimated noise level corresponding to each coefficient sub-block. Threshold shrinkage processing is performed on all wavelet coefficients of the corresponding coefficient sub-block according to the adaptive threshold to obtain the initial estimated wavelet coefficients.

[0078] The Wiener reconstruction output module is used to input the initially estimated wavelet coefficients into the Wiener filtering unit, construct a Wiener filter based on the initially estimated wavelet coefficients and the estimated noise level of the corresponding wavelet decomposition layer, perform Wiener filtering on the initially estimated wavelet coefficients using the Wiener filter to obtain the final estimated wavelet coefficients, and input the final estimated wavelet coefficients into the inverse wavelet transform unit to perform discrete inverse wavelet transform to obtain the target amplified signal.

[0079] To address the above problems, the present invention also provides an electronic device, the electronic device comprising:

[0080] Memory, storing at least one instruction;

[0081] The processor executes the instructions stored in the memory to implement the signal amplification and noise reduction method based on the transistor active load circuit described above.

[0082] To address the aforementioned problems, the present invention also provides a computer-readable storage medium storing at least one instruction, which is executed by a processor in an electronic device to implement the signal amplification and noise reduction method based on a transistor active load circuit described above.

[0083] To address the problems described in the background art, this invention first receives a signal amplification and noise reduction command. Based on this command, a signal processing system is confirmed. This signal processing system includes a load amplification circuit, a wavelet transform unit, a block thresholding unit, a Wiener filtering unit, and an inverse wavelet transform unit. It is evident that this invention addresses the lack of systematic coordination between the analog front-end and digital noise reduction algorithms in traditional solutions. Therefore, by unifying the integrated signal processing system through commands, it organically integrates the low-noise analog front-end with a multi-stage digital noise reduction link, providing a unified hardware foundation and data flow framework for achieving end-to-end collaborative noise reduction. Then, an initial electrical signal is acquired, and the load amplification circuit performs low-noise amplification processing on the initial electrical signal to obtain an amplified analog signal. Finally, the amplified analog signal is converted from analog to digital to obtain an amplified digital signal. As can be seen, this invention addresses the problem of significant thermal noise generated by traditional low-noise amplifier circuits that often use large resistive loads to achieve high gain. Therefore, it replaces the traditional resistive load with a transistor active load structure, using a constant-current-source-biased second transistor as the dynamic active load for the first-stage amplifier. This achieves high voltage gain without introducing additional thermal noise, reducing noise introduced by the amplification stage at its source. Simultaneously, a background digital calibration technique based on pseudo-random jitter injection and cross-correlation calculation is introduced into the analog-to-digital conversion to extract and compensate for the converter's nonlinearity error in real time, further improving the quality of the digital domain signal. Secondly, the number of wavelet decomposition layers is set to n. The wavelet transform unit performs discrete wavelet transform on the amplified digital signal according to the number of layers to obtain n sets of wavelet coefficients. Each set of wavelet coefficients corresponds to each wavelet decomposition layer, and each set of wavelet coefficients consists of multiple wavelet coefficients. The median absolute deviation of each set of wavelet coefficients is obtained. Based on the median absolute deviation, the estimated noise level of each wavelet decomposition layer is calculated. The block threshold unit determines the optimal block length of each set of wavelet coefficients according to the estimated noise level. Based on the optimal block length, each set of wavelet coefficients is divided into multiple coefficient sub-blocks. An adaptive threshold is calculated according to the estimated noise level corresponding to each coefficient sub-block. Threshold shrinkage processing is performed on all wavelet coefficients of the sub-block according to the adaptive threshold to obtain the initial estimated wavelet coefficients. As can be seen, this invention takes into account that traditional thresholds are mostly based on a uniform global noise level setting, which cannot adapt to the differentiated characteristics of coefficients at different decomposition layers and locations. Therefore, it adopts a method of independently estimating the noise level layer by layer. It utilizes the characteristic that the absolute deviation of the median is insensitive to outliers to accurately estimate the actual noise intensity of each layer, so that the threshold processing can be differentiated according to the real situation of each layer. On this basis, a block-adaptive threshold strategy is further adopted. By searching for the block length with the least risk, the coefficients of each layer are divided into multiple sub-blocks and the threshold is calculated independently in each sub-block. This allows the threshold to change dynamically with the local statistical characteristics of the signal, achieving a better balance between noise suppression and signal fidelity.Then, the initially estimated wavelet coefficients are input into the Wiener filtering unit. A Wiener filter is constructed based on the initially estimated wavelet coefficients and the estimated noise level of the corresponding wavelet decomposition layer. The Wiener filter is used to perform Wiener filtering on all initially estimated wavelet coefficients to obtain the final estimated wavelet coefficients. The final estimated wavelet coefficients are then input into the inverse wavelet transform unit to perform a discrete wavelet transform to obtain the target amplified signal. It is evident that this invention considers the potential residual noise caused by the hard decision characteristic of threshold shrinkage. Therefore, after threshold shrinkage, Wiener filtering is further introduced to perform soft correction on the continuous gain coefficients of each coefficient, further reducing residual noise and controlling signal distortion. The window size of the Wiener filter is adaptively adjusted according to the local statistical characteristics of the coefficients. Small windows are used to protect details in signal-rich regions, while large windows are used to fully suppress noise in noise-dominant regions. Finally, the target amplified signal is reconstructed through discrete wavelet transform, achieving end-to-end low-noise amplification and adaptive noise reduction from the initial electrical signal to a high signal-to-noise ratio signal. Therefore, this invention can achieve high-gain, low-noise amplification and adaptive, precise noise reduction of weak electrical signals, significantly improving the signal-to-noise ratio and waveform fidelity of the output signal. Attached Figure Description

[0084] Figure 1 This is a schematic flowchart of a signal amplification and noise reduction method based on a transistor active load circuit provided in an embodiment of the present invention.

[0085] Figure 2 This is a functional block diagram of a signal amplification and noise reduction system based on a transistor active load circuit provided in an embodiment of the present invention;

[0086] Figure 3 This is a schematic diagram of the structure of an electronic device that implements the signal amplification and noise reduction method based on a transistor active load circuit, according to an embodiment of the present invention.

[0087] Explanation of reference numerals in the attached figures:

[0088] 10. Electronic device; 11. Processor; 12. Memory; 13. Bus.

[0089] The realization of the objective, functional features and advantages of the present invention will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0090] It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.

[0091] This application provides a signal amplification and noise reduction method based on a transistor active load circuit. The executing entity of this signal amplification and noise reduction method includes, but is not limited to, at least one of the following electronic devices that can be configured to execute the method provided in this application: a server, a terminal, etc. In other words, the signal amplification and noise reduction method based on a transistor active load circuit can be executed by software or hardware installed on a terminal device or a server device, and the software can be a blockchain platform. The server includes, but is not limited to, a single server, a server cluster, a cloud server, or a cloud server cluster.

[0092] Reference Figure 1 The diagram shown is a schematic flowchart of a signal amplification and noise reduction method based on a transistor active load circuit according to an embodiment of the present invention. In this embodiment, the signal amplification and noise reduction method based on a transistor active load circuit includes:

[0093] S1. Receive signal amplification and noise reduction instructions, and confirm the signal processing system based on the signal amplification and noise reduction instructions. The signal processing system includes: a load amplification circuit, a wavelet transform unit, a block thresholding unit, a Wiener filter unit, and an inverse wavelet transform unit.

[0094] Understandably, the signal amplification and noise reduction command is a digital trigger signal issued by a test engineer or operator through a host computer software interface, hardware control panel, or automated test platform. The signal processing system is a hardware and algorithm collaborative platform that executes all steps of this method, specifically including a load amplification circuit, a wavelet transform unit, a block thresholding unit, a Wiener filtering unit, and an inverse wavelet transform unit. The load amplification circuit is used for low-noise pre-amplification of the initial electrical signal; its core employs a transistor active load structure, providing high voltage gain without introducing additional thermal noise. The wavelet transform unit transforms the time-domain digital signal to the wavelet domain, outputting multi-layer wavelet coefficients to provide a frequency-time joint analysis basis for subsequent noise estimation and thresholding. The block thresholding unit adaptively determines the block length and threshold based on the statistical characteristics of each layer of wavelet coefficients, performing initial noise reduction. The Wiener filtering unit performs a second filtering on the initially estimated wavelet coefficients to further suppress residual noise components. The inverse wavelet transform unit reconstructs the Wiener-filtered wavelet coefficients into a time-domain signal, outputting the target amplified signal.

[0095] S2. Obtain the initial electrical signal, perform low-noise amplification processing on the initial electrical signal using the load amplifier circuit to obtain an amplified analog signal, and then perform analog-to-digital conversion on the amplified analog signal to obtain an amplified digital signal.

[0096] Understandably, the initial electrical signal refers to the raw, weak electrical signal output by the sensor, transducer, or front-end acquisition device. The amplitude of this signal is usually in the millivolt or microvolt range and is mixed with environmental noise, circuit background noise, and power frequency interference. It needs to be amplified with low noise before subsequent analog-to-digital conversion and wavelet transform processing can be performed.

[0097] It should be clarified that the load amplifier circuit adopts a transistor active load structure, with direct coupling between the first-stage and second-stage amplifier transistors. A dynamic active load composed of a second transistor is connected to the collector of the first-stage amplifier transistor. This dynamic active load exhibits extremely high AC impedance under constant current source bias, thereby significantly increasing the voltage gain of the first stage without significantly increasing the DC voltage drop. Simultaneously, the thermal noise generated by the active load itself is much lower than that of a large resistance at the same gain, thus achieving low-noise amplification. When the initial electrical signal is coupled to the input terminal of the load amplifier circuit through the input coupling capacitor, after being amplified by the first-stage common-emitter amplifier, the signal output from the collector of the first-stage amplifier transistor is directly coupled to the base of the second-stage common-emitter amplifier transistor. After being amplified by the second stage, the signal is output from the collector of the second-stage amplifier transistor, obtaining the amplified analog signal. During the amplification process, the operating temperature of the amplifier transistors is continuously monitored. When the temperature exceeds a preset threshold, the bias resistor value is adjusted through a temperature compensation circuit to stabilize the static operating point and suppress additional noise introduced by temperature drift.

[0098] Furthermore, the step of performing analog-to-digital conversion on the amplified analog signal to obtain an amplified digital signal includes:

[0099] The amplified analog signal is input into an analog-to-digital converter, wherein the analog-to-digital converter has a built-in clock source, a sample-and-hold circuit, a quantization encoding circuit, and a digital calibration circuit.

[0100] Set the sampling frequency and quantization bit depth of the amplified analog signal, and use the clock source to generate a sampling clock signal synchronized with the sampling frequency;

[0101] Based on the sampling clock signal, the sample-and-hold circuit performs equally spaced sampling on the amplified analog signal to obtain m discrete sampled voltage values, and each sampled voltage value corresponds to a sampling time.

[0102] Obtain a pseudo-random sequence generated by digital logic, wherein the pseudo-random sequence contains m pseudo-random sequence values, and each pseudo-random sequence value corresponds to a sampling time.

[0103] A jitter signal with a known amplitude is acquired. The known amplitude of the jitter signal is multiplied sequentially by a pseudo-random sequence value to obtain multiple injected jitter values. The injected jitter values ​​are then added to the sampling voltage values ​​of the same sampling time to obtain multiple jitter sampling values. The jitter sampling values ​​are sequentially input into a quantization encoding circuit. The jitter sampling values ​​are uniformly quantized according to the quantization bit depth to obtain the initial quantization encoding value for each sampling time.

[0104] Based on the initial quantization code value and the corresponding pseudo-random sequence value at each sampling time, the cross-correlation value between the initial quantization code value and the pseudo-random sequence value is calculated using the digital calibration circuit, and the nonlinear error parameter of the analog-to-digital converter is extracted based on the cross-correlation value.

[0105] Based on the nonlinear error parameter, real-time digital compensation is performed on the initial quantization code value to obtain the calibrated quantization code value at each sampling time.

[0106] The calibration quantization encoding values ​​of all sampling times are arranged in chronological order to form a time-domain digital signal sequence, which is then used as an amplified digital signal.

[0107] It should be understood that the amplified analog signal is an analog voltage signal output after processing by a load amplifier circuit. The amplitude of this signal has been increased to the volt level, and the signal-to-noise ratio is significantly improved compared to the initial electrical signal, but its physical form remains a continuous-time, continuous-amplitude analog quantity. Since subsequent noise reduction algorithms such as wavelet transform and Wiener filtering need to be performed in the digital domain, this amplified analog signal must be converted into a discrete digital signal sequence before further processing can proceed.

[0108] It should be explained that the clock source generates a stable and precisely periodic pulse signal. Its function is to ensure that the analog-to-digital converter (ADC) performs sampling at fixed time intervals, avoiding time-domain distortion of the digital signal caused by uneven sampling time. If the sampling time is uneven, the recovered waveform will not accurately reflect the frequency and phase characteristics of the original signal, thus affecting the accuracy of subsequent wavelet transform. The sample-and-hold circuit captures and holds the instantaneous voltage value of the amplified analog signal at each sampling pulse on an internal capacitor. Because the quantization encoding process of the ADC requires a certain amount of time, if the signal continues to change during the conversion, the quantization result will not accurately represent the true amplitude at the sampling time. After the capture is completed, the sample-and-hold circuit disconnects the input path, keeping the voltage value faced by the quantization encoding circuit constant during the conversion, thereby ensuring the conversion accuracy of each sampling point.

[0109] Understandably, the quantization encoding circuit converts the held analog voltage amplitude into binary digital code. The pseudo-random sequence, generated by digital logic, takes values ​​of +1 or -1 and is used to control the polarity of the jitter injection. Its function is to disrupt the periodic distribution of quantization noise and improve the spurious-free dynamic range of the analog-to-digital converter. The jitter signal is an analog signal with a known amplitude. Multiplying this amplitude by the pseudo-random sequence value yields an injected jitter value that can be positive or negative. This value is then added to the sampled voltage value at the same sampling time, thus introducing a controllable random disturbance before quantization, making the quantization noise no longer correlated with the input signal.

[0110] It should be explained that the digital calibration circuit is based on background calibration technology: since there is no statistical correlation between the input signal and the pseudo-random sequence, but there is a definite correspondence between the injected jitter and the pseudo-random sequence, calculating the cross-correlation value between the initial quantization code value and the pseudo-random sequence value in the digital domain over the number of accumulations can effectively eliminate the influence of the input signal on error extraction and separate the nonlinear error parameter introduced by the non-ideal characteristics of the analog-to-digital converter. After extraction, the nonlinear error parameter is iteratively updated using first-order exponential smoothing and then used to perform real-time digital compensation on subsequent initial quantization code values, thereby obtaining the calibration quantization code value. The calibration quantization code values ​​at all sampling times are arranged chronologically to obtain the amplified digital signal, which more accurately reflects the true amplitude of the original signal and removes the distortion introduced by the non-ideal characteristics of the analog-to-digital converter.

[0111] Furthermore, the setting of the sampling frequency and quantization bit depth of the amplified analog signal includes:

[0112] The amplified analog signal is presampled to obtain a digital sample sequence, and the digital sample sequence is subjected to spectral analysis to obtain a signal spectral distribution. The highest effective frequency is then extracted from the signal spectral distribution.

[0113] Set a safety factor, and multiply the highest effective frequency by the safety factor to obtain the sampling frequency;

[0114] The peak voltage and background noise voltage of the amplified analog signal are obtained, and the ratio of the peak voltage to the background noise voltage is calculated to obtain the dynamic range of the signal.

[0115] Set a signal-to-noise ratio (SNR) threshold, and increment the quantization bit depth bit by bit starting from the first quantization bit depth bit. Calculate the theoretical SNR corresponding to each quantization bit depth bit in sequence, and select the first quantization bit depth bit that satisfies the theoretical SNR not lower than the SNR threshold as the theoretical quantization bit depth bit.

[0116] Calculate the minimum quantization bit depth for the dynamic range of the signal, and use the larger of the theoretical quantization bit depth and the minimum quantization bit depth as the initial quantization bit depth;

[0117] Set the bit adjustment step size, and round the initial quantization bit size up to an integer multiple of the bit adjustment step size to obtain the final quantization bit size;

[0118] The sampling frequency and the final quantization bit depth are used as the operating parameters of the analog-to-digital converter to complete the setting of the sampling frequency and quantization bit depth.

[0119] It should be explained that the sampling frequency setting is based on the Nyquist sampling theorem, which states that the sampling frequency should not be less than twice the highest frequency of the signal. However, in practical engineering, the transition band width and attenuation characteristics of the anti-aliasing filter also need to be considered. Therefore, a safety factor is required to reserve design margin. To determine the highest effective frequency of the amplified analog signal, the system performs short-time pre-sampling of the amplified analog signal at the highest sampling frequency supported by the analog-to-digital converter to obtain a digital sample sequence. Then, spectral analysis is performed on this sequence to obtain the signal spectral distribution. The frequency corresponding to the amplitude dropping to a set percentage of the peak amplitude is extracted as the highest effective frequency. Multiplying this highest effective frequency by the safety factor yields the sampling frequency that meets the engineering requirements.

[0120] Understandably, the quantization bit depth refers to the number of bits used by the analog-to-digital converter (ADC) to convert the analog voltage amplitude into binary digital code. The setting of the quantization bit depth requires consideration of both signal-to-noise ratio (SNR) and dynamic range requirements. Therefore, the system first calculates the theoretical SNR of the ideal ADC bit depth bitwise, starting from the first bit and incrementing sequentially, based on a preset SNR threshold. The first bit depth that meets the requirements is then used as the theoretical quantization bit depth. Next, the minimum quantization bit depth corresponding to the dynamic range is calculated. The larger of these two values ​​is taken as the initial quantization bit depth. Finally, to adapt to commonly used bit widths, the initial quantization bit depth is rounded up to an integer multiple of the set bit depth by adjusting the step size, resulting in the final quantization bit depth. At this point, both the sampling frequency and the quantization bit depth are determined, and the ADC operates with these parameters.

[0121] Furthermore, the step of calculating the cross-correlation value between the initial quantized encoded value and the pseudo-random sequence value using the digital calibration circuit, and extracting the nonlinear error parameter of the analog-to-digital converter based on the cross-correlation value, includes:

[0122] A set accumulation count is defined, where the accumulation count is a positive integer not greater than m. Based on the accumulation count, the initial quantization encoding value, and the pseudo-random sequence value, the cross-correlation value is calculated according to the following formula:

[0123] ;

[0124] in, For cross-correlation values, To accumulate the number of times, For the first The initial quantized encoded value at each sampling time. For the first The pseudo-random sequence values ​​corresponding to each sampling time;

[0125] The ratio of the cross-correlation value to the known amplitude of the jitter signal is calculated to obtain the estimated value of the nonlinear error parameter;

[0126] The first-order exponential smoothing method is used to iteratively update the estimated value of the nonlinear error parameter, and the updated estimated value of the nonlinear error parameter is used as the nonlinear error parameter of the analog-to-digital converter.

[0127] It should be explained that the cross-correlation value reflects the statistical correlation between the initial quantized encoded value and the pseudo-random sequence. Since there is no statistical correlation between the pseudo-random sequence and the input signal, but a definite correspondence exists between the injected jitter and the pseudo-random sequence, multiplying the initial quantized encoded value and the pseudo-random sequence value and then averaging them will statistically cancel out the contributions of the input signal, while the error component caused by the jitter injection will be retained. A larger number of accumulations results in better statistical noise reduction and higher error extraction accuracy, but a slower response speed. Since the number of accumulations must be within the length of the pseudo-random sequence, it is limited to a positive integer not exceeding m. After obtaining the cross-correlation value, the ratio of the cross-correlation value to the amplitude of the jitter signal is calculated to obtain the estimated value of the nonlinear error parameter. To eliminate random fluctuations and obtain stable error parameters, recursive updates using a first-order exponential smoothing method are required.

[0128] Furthermore, the step of iteratively updating the estimated nonlinear error parameter using a first-order exponential smoothing method, and using the updated estimated nonlinear error parameter as the nonlinear error parameter of the analog-to-digital converter, includes:

[0129] Confirm the current iteration cycle, obtain the estimated value of the nonlinear error parameter for the current iteration cycle, and set a smoothing factor, wherein the smoothing factor has a value range between 0 and 1.

[0130] Determine whether the current iteration cycle is the first iteration. If it is the first iteration, use the current nonlinear error parameter estimate as the updated nonlinear error parameter estimate after the first iteration.

[0131] If it is not the first iteration, obtain the estimated value of the nonlinear error parameter after the previous iteration update, calculate the product of the smoothing factor and the estimated value of the nonlinear error parameter in the current iteration period, and obtain the first weighted value;

[0132] Calculate the difference obtained by subtracting the smoothing factor from 1, and then multiply the difference by the estimated value of the nonlinear error parameter obtained after the previous iteration update to obtain the second weighted value;

[0133] Add the first weighted value to the second weighted value to obtain the estimated value of the nonlinear error parameter after this iteration.

[0134] Repeat the above steps, and use the estimated value of the nonlinear error parameter after each iteration as the nonlinear error parameter of the analog-to-digital converter.

[0135] It should be explained that the first-order exponential smoothing method is a recursive weighted filtering algorithm. That is, each time it is updated, the new error parameter is composed of the currently observed estimated value and the historically stored error parameter according to a certain weight. Since the nonlinear error of the analog-to-digital converter may drift slowly with environmental factors such as temperature and power supply voltage, the first-order exponential smoothing method can smooth random fluctuations, suppress the influence of instantaneous noise on error estimation, and at the same time track the slow trend of error change.

[0136] Understandably, the smoothing factor is a control parameter in the algorithm, with a value ranging from 0 to 1. The smoothing factor determines the proportion of the current observation and historical values ​​in the updated result. In this embodiment, the estimated nonlinear error parameter is the original estimate obtained by calculating the cross-correlation ratio, which still contains some random fluctuations. The updated estimated nonlinear error parameter is a stable output result after first-order exponential smoothing recursive filtering. In the first iteration, since there are no historical values, the algorithm directly uses the current estimated nonlinear error parameter as the first update result. Starting from the second iteration, each time a new estimate is obtained, it is updated according to the weighted combination rule. This recursive process continues, performing an update each time a new estimated nonlinear error parameter is obtained. The final updated estimated nonlinear error parameter is used as the nonlinear error parameter of the analog-to-digital converter for subsequent real-time digital compensation.

[0137] Furthermore, the step of performing real-time digital compensation on the initial quantization code value based on the nonlinear error parameter to obtain the calibration quantization code value at each sampling time includes:

[0138] Based on the extracted nonlinear error parameters of the analog-to-digital converter, the input range of the analog-to-digital converter is confirmed;

[0139] The input range is divided into multiple continuous and non-overlapping quantization intervals, and the corresponding nonlinear error parameters are calibrated at the two endpoints of each quantization interval.

[0140] Obtain the initial quantization code value at each sampling time, determine the quantization interval to which the initial quantization code value belongs, and mark it as the target quantization interval;

[0141] Read the initial quantization code value of the left endpoint, the nonlinear error parameter corresponding to the left endpoint, the initial quantization code value of the right endpoint, and the nonlinear error parameter corresponding to the right endpoint of the target quantization interval.

[0142] The compensation amount corresponding to the initial quantized encoded value is calculated according to the following formula:

[0143] ;

[0144] in, For compensation amount, This is the initial quantization encoding value. and These are the initial quantization values ​​at the left and right endpoints of the target quantization interval, respectively. and These are the nonlinear error parameters corresponding to the left endpoint and the nonlinear error parameters corresponding to the right endpoint, respectively.

[0145] The calibration quantization value for each sampling time is obtained by subtracting the corresponding compensation amount from the initial quantization code value at each sampling time.

[0146] It should be explained that the nonlinear error parameter is an estimated output deviation of the analog-to-digital converter (ADC) at each input code value, reflecting the systematic distortion introduced by non-ideal circuit characteristics such as capacitor mismatch, amplifier finite gain, and comparator offset. Since the nonlinear error of the ADC varies with different input code values, using a single compensation value for the entire input range will significantly reduce the compensation effect in areas with large error variations. Therefore, this invention divides the entire input range of the ADC into multiple continuous and non-overlapping quantization intervals. Each interval corresponds to a continuous range of input code values, and the two endpoints of each interval are respectively labeled with corresponding nonlinear error parameter values. This allows for more accurate compensation based on the location of different code values ​​during subsequent compensation.

[0147] Understandably, the quantization interval is a discretized division of the input range, and the code value corresponding to its endpoint is usually the actual quantized code value at that endpoint. The target quantization interval refers to the specific quantization interval into which the initial quantized code value at the current sampling time falls. After determining the interval, it is necessary to read the initial quantized code value at the left endpoint, the nonlinear error parameter corresponding to the left endpoint, the initial quantized code value at the right endpoint, and the nonlinear error parameter corresponding to the right endpoint for subsequent linear interpolation calculations.

[0148] Furthermore, the compensation amount refers to the estimated output deviation caused by the non-ideal characteristics of the analog-to-digital converter at the current input code value. Since the actual error curve typically changes approximately linearly between adjacent endpoints, a linear interpolation method is used to calculate the compensation amount corresponding to any code value within the interval. Because the initial quantization code value at each sampling moment is independently calculated with corresponding compensation amounts and subtracted successively, the final result is a calibrated quantization code value sequence for the entire sampling sequence. This sequence significantly reduces the distortion introduced by the non-ideal characteristics within the analog-to-digital converter, providing a higher-precision data foundation for subsequent wavelet transform and noise reduction processing.

[0149] S3. Set the number of wavelet decomposition layers to n, and use the wavelet transform unit to perform discrete wavelet transform on the amplified digital signal according to the number of layers to obtain n sets of wavelet coefficients. Each set of wavelet coefficients corresponds to each wavelet decomposition layer, and each set of wavelet coefficients consists of multiple wavelet coefficients.

[0150] It should be explained that the discrete wavelet transform is a signal decomposition method relative to the continuous wavelet transform. The discrete wavelet transform decomposes the signal layer by layer through binary decimation. Each layer of decomposition yields a set of wavelet coefficients to match the requirements of subsequent noise estimation and block thresholding processing for the layer coefficients.

[0151] It should be clarified that the number of wavelet decomposition layers, n, refers to the total number of times the amplified digital signal is decomposed layer by layer. Each decomposition separates the amplified digital signal into a low-frequency approximation component and a high-frequency detail component. Each layer of decomposition generates a set of high-frequency detail coefficients, while the nth layer of decomposition also generates an additional set of low-frequency approximation coefficients. In this method, for the sake of consistency, the high-frequency detail coefficients of each layer and the low-frequency approximation coefficients of the last layer are collectively referred to as the wavelet coefficients of that layer. Therefore, a total of n sets of wavelet coefficients are obtained, and each set corresponds to one decomposition layer.

[0152] Understandably, the amplified digital signal is a time-domain discrete digital sequence obtained after analog-to-digital conversion, and the number of sampling points *m* is known. During discrete wavelet transform, the length of the wavelet coefficient set is approximately half the length of the input of the previous layer after each decomposition layer. The wavelet coefficients are the result of the inner product of the signal and the wavelet basis functions, and their magnitude reflects the energy distribution of the signal within the frequency band corresponding to that decomposition layer. In practical applications, noise components are typically distributed in finer high-frequency detail layers, while the energy of the effective signal is more concentrated in low-frequency approximation layers or certain specific high-frequency layers. By decomposing the signal into multiple frequency band levels, subsequent noise estimation and thresholding can be differentiated at different levels, thereby achieving layer-by-layer noise suppression.

[0153] S4. Obtain the median absolute deviation of each set of wavelet coefficients, and calculate the estimated noise level of each wavelet decomposition layer based on the median absolute deviation.

[0154] It should be explained that the median absolute deviation is a robust statistic used to measure the dispersion of a set of data. The median absolute deviation needs to be calculated independently for each set of wavelet coefficients. Since the wavelet coefficients of different decomposition layers have different statistical properties, calculating the estimated noise level of each layer separately allows subsequent block thresholding and Wiener filtering to be differentiated according to the actual noise intensity of each layer.

[0155] Specifically, for each set of wavelet coefficients, the absolute values ​​of all wavelet coefficients in the set are first extracted to obtain an absolute value sequence; then the values ​​in the absolute value sequence are sorted in ascending order; the median of the sorted sequence is then taken as the absolute deviation of the median for the set; finally, the normal distribution quantile constant of 0.6745 is set, the ratio of the absolute deviation of the median to the constant is calculated, and the ratio is used as the estimated noise level of the corresponding wavelet decomposition layer.

[0156] It should be explained that the absolute value sequence is a sequence formed by taking the absolute values ​​of each wavelet coefficient. The sorted sequence is arranged in ascending order of value. The normal distribution quantile constant 0.6745 comes from the standard normal distribution—the Gaussian distribution. The median of the absolute values ​​is approximately equal to 0.6745 times the standard deviation. Therefore, dividing the absolute deviation of the median by this constant yields a robust estimate of the noise standard deviation. The estimated noise level, i.e., the estimate of the noise standard deviation for this layer, is used for subsequent adaptive threshold calculation and Wiener filter construction.

[0157] S5. Using the block threshold unit, determine the optimal block length of each set of wavelet coefficients based on the estimated noise level. Divide each set of wavelet coefficients into multiple coefficient sub-blocks based on the optimal block length. Calculate an adaptive threshold based on the estimated noise level corresponding to each coefficient sub-block. Perform threshold shrinkage processing on all wavelet coefficients of the corresponding coefficient sub-block based on the adaptive threshold to obtain the initial estimated wavelet coefficients.

[0158] It should be explained that, since wavelet coefficients at different locations within the same decomposition layer may have different signal-to-noise ratios, if the same global threshold is used for the entire decomposition layer, the details in signal-rich areas will be excessively compressed, while noise in noise-dominant areas cannot be adequately suppressed. Therefore, this invention adopts a block-based strategy, dividing the coefficients into multiple continuous and non-overlapping sub-blocks and calculating the threshold independently within each sub-block. This allows the threshold to adapt to the statistical characteristics of the local coefficients, thereby achieving more refined differentiated noise reduction.

[0159] Furthermore, the step of using the block threshold unit to determine the optimal block length of each set of wavelet coefficients based on the estimated noise level, and dividing each set of wavelet coefficients into multiple coefficient sub-blocks based on the optimal block length, includes:

[0160] Set the block length search range and search step size for each wavelet decomposition layer, and select candidate block lengths sequentially according to the search step size for the block length search range.

[0161] Based on the length of each candidate block, the set of wavelet coefficients for each corresponding wavelet decomposition layer is divided into multiple candidate sub-blocks;

[0162] Confirm all wavelet coefficients and the estimated noise level of the corresponding wavelet decomposition layer for each candidate sub-block, and calculate the threshold shrinkage mean square risk estimate for each candidate sub-block.

[0163] The threshold shrinkage mean square risk estimates of all candidate sub-blocks under the same candidate block length are summed to obtain the total risk estimate for each candidate block length.

[0164] Traverse all candidate block lengths and select the candidate block length with the smallest total risk estimate as the optimal block length for the corresponding wavelet decomposition layer.

[0165] The wavelet coefficient set of the wavelet decomposition layer is divided into multiple non-overlapping coefficient sub-blocks according to the optimal block length. The number of wavelet coefficients contained in each coefficient sub-block is equal to the optimal block length. If the total number of wavelet coefficients of the wavelet decomposition layer is not divisible by the optimal block length, the remaining undivided wavelet coefficients are merged into the last coefficient sub-block.

[0166] It should be clarified that in the process of determining the optimal block length described above, the search range for the block length is a continuous integer interval. The search step size is the difference between the lengths of two adjacent candidate blocks.

[0167] It should be explained that determining the optimal block length first requires defining the search space. Since the block length cannot be too small or too large, a reasonable search range needs to be set, along with a search step size to control the fineness of the attempts. For a given candidate block length, the wavelet coefficients of the current layer are temporarily divided into multiple continuous and non-overlapping candidate sub-blocks according to that length. Then, a threshold shrinkage mean square risk estimate is calculated for each candidate sub-block. This threshold shrinkage mean square risk estimate is a quantitative indicator that assesses the expected deviation from the ideal noise-free signal if threshold shrinkage is performed on a per-sub-block basis. Its calculation is based on the amplitude distribution of all wavelet coefficients within the sub-block and the estimated noise level of the layer. The risk estimates of all candidate sub-blocks under the same candidate length are summed to obtain the total risk estimate corresponding to that length, which is used to comprehensively evaluate the applicability of the block length to the entire layer.

[0168] Understandably, by performing the above calculations sequentially on each candidate block length within the search range, the total risk estimate corresponding to each length can be obtained. Selecting the length with the lowest total risk as the optimal block length means that this length achieves the best threshold shrinkage effect for the entire layer among all candidate lengths. After determining the optimal block length, the system performs a final partitioning of the wavelet coefficient set of the current layer according to this length, resulting in coefficient sub-blocks. The number of wavelet coefficients contained in each sub-block is equal to the optimal block length. If the total number of coefficients is not divisible by the optimal block length, the coefficients remaining that are less than a complete block are merged into the last sub-block to ensure that all coefficients are included in a sub-block for threshold processing without discarding any data.

[0169] Furthermore, after completing the coefficient sub-block partitioning, the system enters the adaptive threshold calculation and shrinkage processing stage. For each coefficient sub-block, an adaptive threshold matching the sub-block is calculated based on the amplitude distribution of the wavelet coefficients within the sub-block and the estimated noise level of that layer. This adaptive threshold is typically determined based on the mean square error minimization criterion, ensuring that the mean square error of the sub-block is minimized after threshold shrinkage. Subsequently, threshold shrinkage processing is performed on all wavelet coefficients within the sub-block according to this threshold, and the shrunken wavelet coefficients are used as the output of the sub-block. After all sub-blocks have been processed, the results of each sub-block are concatenated in their original order to obtain the initial estimated wavelet coefficients.

[0170] S6. Input the initially estimated wavelet coefficients into the Wiener filtering unit, construct a Wiener filter based on the initially estimated wavelet coefficients and the estimated noise level of the corresponding wavelet decomposition layer, and perform Wiener filtering on the initially estimated wavelet coefficients using the Wiener filter to obtain the final estimated wavelet coefficients.

[0171] It should be explained that Wiener filtering is an optimal linear filtering method based on the minimum mean square error criterion, and is an existing technology. While threshold shrinkage removes most of the noise, the retained coefficients may still contain residual noise components, especially in resolution layers with low signal-to-noise ratios. Therefore, the initially estimated wavelet coefficients need to be re-input into the Wiener filtering unit for further processing. Wiener filtering utilizes the statistical properties of noise to fine-tune each coefficient, further reducing residual noise without introducing additional distortion, resulting in higher quality final estimated wavelet coefficients.

[0172] Furthermore, the step of constructing a Wiener filter based on the initially estimated wavelet coefficients and the estimated noise level of the corresponding wavelet decomposition layer, and then applying the Wiener filter to perform Wiener filtering on the initially estimated wavelet coefficients to obtain the final estimated wavelet coefficients, includes:

[0173] The filter window size of each wavelet decomposition layer is set according to the estimated noise level of each wavelet decomposition layer.

[0174] The filter window size is used as the sliding window length. Based on the sliding window length, the sliding window is commanded to slide point by point with a sliding step size of 1 until all the initial estimated wavelet coefficients of the wavelet decomposition layer are traversed.

[0175] Confirm the center position of the sliding window, calculate the arithmetic mean of all initially estimated wavelet coefficients within the sliding window, and use the arithmetic mean as the local signal mean at the center position;

[0176] Calculate the variance of all initially estimated wavelet coefficients within the sliding window, and use the variance as the local signal variance at the center position;

[0177] The local signal variance is compared with the estimated noise level of the corresponding wavelet decomposition layer: if the local signal variance is less than or equal to the estimated noise level, the wavelet coefficients at the center position are replaced with the local signal mean; if the local signal variance is greater than the estimated noise level, the difference between the local signal variance and the estimated noise level is calculated, the difference is divided by the local signal variance to obtain the weighting coefficient, the difference between the wavelet coefficients at the center position and the local signal mean is multiplied by the weighting coefficient to obtain the weighted correction difference, and finally the weighted correction difference is added to the local signal mean to obtain the weighted corrected wavelet coefficients.

[0178] The weighted and corrected wavelet coefficients are used as the Wiener-filtered wavelet coefficients at the center position of the corresponding sliding window. After traversing all the center positions of the sliding window, the Wiener-filtered wavelet coefficients of each wavelet decomposition layer are obtained.

[0179] The final estimated wavelet coefficients are obtained by combining the Wiener-filtered wavelet coefficients of all wavelet decomposition layers.

[0180] It should be explained that after determining the window size, the system uses this size as the window length and slides the window point by point with a step size of 1, processing each wavelet coefficient sequentially. At each window position, all initially estimated wavelet coefficients within the window are extracted, and their arithmetic mean and variance are calculated respectively, serving as the local signal mean and local signal variance at the center of the window. Since the coefficients within the window have the strongest correlation with the central coefficient, the statistics estimated by this window best reflect the local environment of the central coefficient; therefore, the coefficient at the center of the window is selected as the current processing target.

[0181] Furthermore, after obtaining the local signal mean and local signal variance, the system compares them with the estimated noise level of the layer to determine the processing method for the current coefficient: if the local signal variance is less than or equal to the estimated noise level, it indicates that the signal strength within the window is submerged by noise, and the central coefficient is mainly contributed by noise, so it is directly replaced by the local signal mean; if the local signal variance is greater than the estimated noise level, it indicates that there is an effective signal within the window, and the Wiener filtering criterion is used to correct the coefficient.

[0182] Understandably, after performing the aforementioned sliding window Wiener filtering on each wavelet decomposition layer, the processed coefficients from each layer are combined layer by layer to obtain the final estimated wavelet coefficients. This coefficient sequence not only removes large-amplitude noise through threshold shrinkage but also suppresses residual noise through fine correction by Wiener filtering. At the same time, the distortion of the effective signal components is controlled at a low level, providing high-quality frequency domain data for subsequent inverse wavelet transform reconstruction of the target amplified signal.

[0183] It should be explained that in the implementation of the Wiener filter described above, the first step is to determine the filter window size, i.e., how many wavelet coefficients each sliding window contains. The window size directly affects the filtering effect. If the window is too small, there are insufficient sample points for averaging, resulting in a large variance in the local signal variance estimation; if the window is too large, coefficients with different statistical characteristics may be mixed together, leading to estimation bias. Therefore, the window size at each position needs to be adaptively determined based on the local statistical characteristics of the wavelet coefficients in that layer.

[0184] Furthermore, the step of setting the filter window size for each wavelet decomposition layer based on the estimated noise level of each wavelet decomposition layer includes:

[0185] The initial estimated wavelet coefficients of each wavelet decomposition layer are divided into multiple local regions, wherein each local region contains multiple wavelet coefficients;

[0186] Calculate the sum of the absolute values ​​of all wavelet coefficients in each local region, and divide the sum of absolute values ​​by the total number of wavelet coefficients in the local region to obtain the average amplitude of the coefficients in the local region.

[0187] The amplitude signal-to-noise ratio of the local region is obtained by calculating the ratio of the average amplitude of the coefficients to the estimated noise level of the corresponding wavelet decomposition layer.

[0188] Based on the amplitude signal-to-noise ratio, the range of amplitude signal-to-noise ratio values ​​is obtained, and multiple signal-to-noise ratio thresholds are set. Based on the signal-to-noise ratio thresholds, the range of amplitude signal-to-noise ratio values ​​is divided into multiple continuous intervals.

[0189] The amplitude signal-to-noise ratio of each local region is matched with a continuous interval. After the matching is completed, a corresponding filter window size is assigned to each local region, and the center position of each local region is mapped to the value point of the corresponding filter window size.

[0190] Median filtering is applied sequentially to the filter window sizes corresponding to the value points to obtain the filter window sizes of all sliding windows in each wavelet decomposition layer.

[0191] It should be explained that in the process of setting the adaptive window size, the initial estimated wavelet coefficients of the current wavelet decomposition layer are first divided into multiple non-overlapping local regions. Each local region contains multiple wavelet coefficients, serving as the basic unit for window size estimation. The average absolute value of all wavelet coefficients within each local region is calculated to obtain the average amplitude, which characterizes the signal energy level of that region. The average amplitude is divided by the estimated noise level of that layer to obtain the amplitude signal-to-noise ratio. A higher ratio indicates a higher probability of valid signals in that region, and a smaller window should be used to preserve details; a lower ratio indicates that the region is dominated by noise, and a larger window should be used to ensure sufficient smoothness.

[0192] Understandably, to achieve the aforementioned qualitative judgment, the system sets multiple signal-to-noise ratio (SNR) thresholds, dividing the amplitude SNR values ​​into multiple continuous intervals, each interval corresponding to a preset window size. For example, a maximum window is allocated when the value is below the first threshold, a medium window is allocated when the value is between the first and second thresholds, and a minimum window is allocated when the value is above the second threshold. Each local region is assigned a corresponding window size based on the interval in which its amplitude SNR falls, with the center of the region serving as the value point for that size.

[0193] Furthermore, since the window size of adjacent local regions may differ significantly due to abrupt changes in signal-to-noise ratio, direct use would cause the window size to jump at the boundary, affecting the continuity of the filtering results. Therefore, median filtering smoothing is performed on the window size of all points—taking a certain size value as the center, the median of multiple size values ​​in its neighborhood is taken as the final size at that position, so that the window size transitions smoothly in space, ultimately obtaining the filtered window size corresponding to the center position of each sliding window.

[0194] S7. Input the final estimated wavelet coefficients into the inverse wavelet transform unit, perform discrete inverse wavelet transform, and obtain the target amplified signal.

[0195] It should be explained that the inverse discrete wavelet transform is the inverse operation of the discrete wavelet transform. The process is as follows: starting from the deepest layer, the low-frequency approximation coefficients and high-frequency detail coefficients of each layer are upsampled, then filtered through corresponding reconstruction low-pass and high-pass filters. The two filtering results are then added together to obtain the reconstruction coefficients of the next layer. This process is repeated until the original time-domain signal with the original number of sampling points is recovered. The final estimated wavelet coefficients are wavelet domain coefficients after block thresholding and Wiener filtering, which effectively suppress noise components and retain the main characteristics of the signal. The time-domain signal obtained after reconstruction by the inverse discrete wavelet transform is the target amplified signal. The waveform of this signal is basically consistent with the initial electrical signal, but the noise level is significantly reduced, the signal-to-noise ratio is effectively improved, and the signal amplitude has been amplified to the target range, making it directly usable for subsequent analysis, display, or storage.

[0196] To address the problems described in the background art, this invention first receives a signal amplification and noise reduction command. Based on this command, a signal processing system is confirmed. This signal processing system includes a load amplification circuit, a wavelet transform unit, a block thresholding unit, a Wiener filtering unit, and an inverse wavelet transform unit. It is evident that this invention addresses the lack of systematic coordination between the analog front-end and digital noise reduction algorithms in traditional solutions. Therefore, by unifying the integrated signal processing system through commands, it organically integrates the low-noise analog front-end with a multi-stage digital noise reduction link, providing a unified hardware foundation and data flow framework for achieving end-to-end collaborative noise reduction. Then, an initial electrical signal is acquired, and the load amplification circuit performs low-noise amplification processing on the initial electrical signal to obtain an amplified analog signal. Finally, the amplified analog signal is converted from analog to digital to obtain an amplified digital signal. As can be seen, this invention addresses the problem of significant thermal noise generated by traditional low-noise amplifier circuits that often use large resistive loads to achieve high gain. Therefore, it replaces the traditional resistive load with a transistor active load structure, using a constant-current-source-biased second transistor as the dynamic active load for the first-stage amplifier. This achieves high voltage gain without introducing additional thermal noise, reducing noise introduced by the amplification stage at its source. Simultaneously, a background digital calibration technique based on pseudo-random jitter injection and cross-correlation calculation is introduced into the analog-to-digital conversion to extract and compensate for the converter's nonlinearity error in real time, further improving the quality of the digital domain signal. Secondly, the number of wavelet decomposition layers is set to n. The wavelet transform unit performs discrete wavelet transform on the amplified digital signal according to the number of layers to obtain n sets of wavelet coefficients. Each set of wavelet coefficients corresponds to each wavelet decomposition layer, and each set of wavelet coefficients consists of multiple wavelet coefficients. The median absolute deviation of each set of wavelet coefficients is obtained. Based on the median absolute deviation, the estimated noise level of each wavelet decomposition layer is calculated. The block threshold unit determines the optimal block length of each set of wavelet coefficients according to the estimated noise level. Based on the optimal block length, each set of wavelet coefficients is divided into multiple coefficient sub-blocks. An adaptive threshold is calculated according to the estimated noise level corresponding to each coefficient sub-block. Threshold shrinkage processing is performed on all wavelet coefficients of the sub-block according to the adaptive threshold to obtain the initial estimated wavelet coefficients. As can be seen, this invention takes into account that traditional thresholds are mostly based on a uniform global noise level setting, which cannot adapt to the differentiated characteristics of coefficients at different decomposition layers and locations. Therefore, it adopts a method of independently estimating the noise level layer by layer. It utilizes the characteristic that the absolute deviation of the median is insensitive to outliers to accurately estimate the actual noise intensity of each layer, so that the threshold processing can be differentiated according to the real situation of each layer. On this basis, a block-adaptive threshold strategy is further adopted. By searching for the block length with the least risk, the coefficients of each layer are divided into multiple sub-blocks and the threshold is calculated independently in each sub-block. This allows the threshold to change dynamically with the local statistical characteristics of the signal, achieving a better balance between noise suppression and signal fidelity.Then, the initially estimated wavelet coefficients are input into the Wiener filtering unit. A Wiener filter is constructed based on the initially estimated wavelet coefficients and the estimated noise level of the corresponding wavelet decomposition layer. The Wiener filter is used to perform Wiener filtering on all initially estimated wavelet coefficients to obtain the final estimated wavelet coefficients. The final estimated wavelet coefficients are then input into the inverse wavelet transform unit to perform a discrete wavelet transform to obtain the target amplified signal. It is evident that this invention considers the potential residual noise caused by the hard decision characteristic of threshold shrinkage. Therefore, after threshold shrinkage, Wiener filtering is further introduced to perform soft correction on the continuous gain coefficients of each coefficient, further reducing residual noise and controlling signal distortion. The window size of the Wiener filter is adaptively adjusted according to the local statistical characteristics of the coefficients. Small windows are used to protect details in signal-rich regions, while large windows are used to fully suppress noise in noise-dominant regions. Finally, the target amplified signal is reconstructed through discrete wavelet transform, achieving end-to-end low-noise amplification and adaptive noise reduction from the initial electrical signal to a high signal-to-noise ratio signal. Therefore, this invention can achieve high-gain, low-noise amplification and adaptive, precise noise reduction of weak electrical signals, significantly improving the signal-to-noise ratio and waveform fidelity of the output signal.

[0197] like Figure 2 The diagram shown is a functional block diagram of a signal amplification and noise reduction system based on a transistor active load circuit provided in an embodiment of the present invention.

[0198] The signal amplification and noise reduction system 100 based on a transistor active load circuit described in this invention can be installed in an electronic device. Depending on the functions implemented, the signal amplification and noise reduction system 100 based on a transistor active load circuit may include an instruction system confirmation module 101, a low-noise analog-to-digital conversion module 102, a wavelet block noise reduction module 103, and a Wiener reconstruction output module 104. The module described in this invention can also be called a unit, which refers to a series of computer program segments that can be executed by the processor of an electronic device and can perform a fixed function, and which are stored in the memory of the electronic device.

[0199] The instruction system confirmation module 101 is used to receive signal amplification and noise reduction instructions and confirm the signal processing system based on the signal amplification and noise reduction instructions. The signal processing system includes: a load amplification circuit, a wavelet transform unit, a block thresholding unit, a Wiener filtering unit, and an inverse wavelet transform unit.

[0200] The low-noise analog-to-digital converter module 102 is used to acquire an initial electrical signal, perform low-noise amplification processing on the initial electrical signal using the load amplifier circuit to obtain an amplified analog signal, and then perform analog-to-digital conversion on the amplified analog signal to obtain an amplified digital signal.

[0201] The wavelet block noise reduction module 103 is used to set the number of wavelet decomposition layers to n, and use the wavelet transform unit to perform discrete wavelet transform on the amplified digital signal according to the number of layers to obtain n sets of wavelet coefficients. Each set of wavelet coefficients corresponds to each wavelet decomposition layer, and each set of wavelet coefficients consists of multiple wavelet coefficients. The absolute median deviation of each set of wavelet coefficients is obtained, and the estimated noise level of each wavelet decomposition layer is calculated based on the absolute median deviation. The optimal block length of each set of wavelet coefficients is determined by the block threshold unit according to the estimated noise level. Each set of wavelet coefficients is divided into multiple coefficient sub-blocks based on the optimal block length. An adaptive threshold is calculated according to the estimated noise level corresponding to each coefficient sub-block, and threshold shrinkage processing is performed on all wavelet coefficients of the corresponding coefficient sub-block according to the adaptive threshold to obtain the initial estimated wavelet coefficients.

[0202] The Wiener reconstruction output module 104 is used to input the initially estimated wavelet coefficients into the Wiener filtering unit, construct a Wiener filter based on the initially estimated wavelet coefficients and the estimated noise level of the corresponding wavelet decomposition layer, perform Wiener filtering on the initially estimated wavelet coefficients using the Wiener filter to obtain the final estimated wavelet coefficients, and input the final estimated wavelet coefficients into the inverse wavelet transform unit to perform discrete inverse wavelet transform to obtain the target amplified signal.

[0203] In detail, the modules in the signal amplification and noise reduction system 100 based on a transistor active load circuit described in this embodiment of the invention employ the same methods as described above. Figure 1 The signal amplification and noise reduction method based on the transistor active load circuit described in the article uses the same technical means and can produce the same technical effect, so it will not be repeated here.

[0204] like Figure 3 The diagram shown is a schematic representation of an electronic device that implements a signal amplification and noise reduction method based on a transistor active load circuit, according to an embodiment of the present invention.

[0205] The electronic device 1 may include a processor 10, a memory 11 and a bus 12, and may also include a computer program stored in the memory 11 and executable on the processor 10, such as a signal amplification and noise reduction method program based on a transistor active load circuit.

[0206] The memory 11 includes at least one type of readable storage medium, such as flash memory, portable hard drive, multimedia card, card-type memory (e.g., SD or DX memory), magnetic memory, magnetic disk, optical disk, etc. In some embodiments, the memory 11 can be an internal storage unit of the electronic device 1, such as a portable hard drive. In other embodiments, the memory 11 can be an external storage device of the electronic device 1, such as a plug-in portable hard drive, smart media card (SMC), secure digital card (SD), flash card, etc., equipped on the electronic device 1. Furthermore, the memory 11 includes both internal storage units and external storage devices of the electronic device 1. The memory 11 can be used not only to store application software and various types of data installed on the electronic device 1, such as code for a signal amplification and noise reduction method program based on a transistor active load circuit, but also to temporarily store data that has been output or will be output.

[0207] In some embodiments, the processor 10 may be composed of integrated circuits, such as a single packaged integrated circuit or multiple integrated circuits with the same or different functions, including combinations of one or more central processing units (CPUs), microprocessors, digital processing chips, graphics processors, and various control chips. The processor 10 is the control unit of the electronic device, connecting various components of the entire electronic device through various interfaces and lines. It executes programs or modules stored in the memory 11 (e.g., signal amplification and noise reduction methods based on transistor active load circuits) and calls data stored in the memory 11 to perform various functions of the electronic device 1 and process data.

[0208] The bus 12 can be a peripheral component interconnect (PCI) bus or an extended industry standard architecture (EISA) bus, etc. The bus 12 can be divided into an address bus, a data bus, a control bus, etc. The bus 12 is configured to realize the connection and communication between the memory 11 and at least one processor 10, etc.

[0209] Figure 3 Only electronic devices with components are shown; those skilled in the art will understand that... Figure 3The structure shown does not constitute a limitation on the electronic device 1, and may include fewer or more components than shown, or combine certain components, or have different component arrangements.

[0210] For example, although not shown, the electronic device 1 may also include a power supply (such as a battery) to power the various components. Preferably, the power supply can be logically connected to the at least one processor 10 through a power management device, thereby enabling functions such as charging management, discharging management, and power consumption management. The power supply may also include one or more DC or AC power supplies, recharging devices, power fault detection circuits, power converters or inverters, power status indicators, and other arbitrary components. The electronic device 1 may also include various sensors, Bluetooth modules, Wi-Fi modules, etc., which will not be described in detail here.

[0211] Furthermore, the electronic device 1 may also include a network interface. Optionally, the network interface may include a wired interface and / or a wireless interface (such as a Wi-Fi interface, a Bluetooth interface, etc.), which is typically used to establish communication connections between the electronic device 1 and other electronic devices.

[0212] Optionally, the electronic device 1 may further include a user interface, which may be a display, an input unit (such as a keyboard), and optionally, a standard wired interface or a wireless interface. Optionally, in some embodiments, the display may be an LED display, a liquid crystal display, a touch-sensitive liquid crystal display, or an OLED (Organic Light-Emitting Diode) touchscreen, etc. The display may also be appropriately referred to as a screen or display unit, used to display information processed in the electronic device 1 and to display a visual user interface.

[0213] The signal amplification and noise reduction method program based on a transistor active load circuit stored in the memory 11 of the electronic device 1 is a combination of multiple instructions. When run in the processor 10, it can achieve the following:

[0214] The system receives a signal amplification and noise reduction command, and confirms the signal processing system based on the signal amplification and noise reduction command. The signal processing system includes: a load amplification circuit, a wavelet transform unit, a block thresholding unit, a Wiener filter unit, and an inverse wavelet transform unit.

[0215] An initial electrical signal is acquired, and the initial electrical signal is amplified by the load amplifier circuit to obtain an amplified analog signal. The amplified analog signal is then converted from analog to digital to obtain an amplified digital signal.

[0216] The number of wavelet decomposition layers is set to n. The wavelet transform unit performs discrete wavelet transform on the amplified digital signal according to the number of layers to obtain n sets of wavelet coefficients. Each set of wavelet coefficients corresponds to each wavelet decomposition layer, and each set of wavelet coefficients consists of multiple wavelet coefficients.

[0217] Obtain the median absolute deviation of each set of wavelet coefficients, and calculate the estimated noise level of each wavelet decomposition layer based on the median absolute deviation;

[0218] The optimal block length of each set of wavelet coefficients is determined by the block threshold unit based on the estimated noise level. Based on the optimal block length, each set of wavelet coefficients is divided into multiple coefficient sub-blocks. An adaptive threshold is calculated based on the estimated noise level corresponding to each coefficient sub-block. Threshold shrinkage processing is performed on all wavelet coefficients of the corresponding coefficient sub-block according to the adaptive threshold to obtain the initial estimated wavelet coefficients.

[0219] The initial estimated wavelet coefficients are input into the Wiener filter unit. A Wiener filter is constructed based on the initial estimated wavelet coefficients and the estimated noise level of the corresponding wavelet decomposition layer. The Wiener filter is used to perform Wiener filtering on the initial estimated wavelet coefficients to obtain the final estimated wavelet coefficients.

[0220] The final estimated wavelet coefficients are input into the inverse wavelet transform unit to perform a discrete inverse wavelet transform, thereby obtaining the target amplified signal.

[0221] Specifically, the processor 10's implementation method for the above instructions can be found in [reference needed]. Figures 1 to 3 The descriptions of the relevant steps in the corresponding embodiments are not repeated here.

[0222] Furthermore, if the modules / units integrated in the electronic device 1 are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. The computer-readable storage medium can be volatile or non-volatile. For example, the computer-readable medium may include: any entity or device capable of carrying the computer program code, a recording medium, a USB flash drive, a portable hard drive, a magnetic disk, an optical disk, a computer memory, or a read-only memory (ROM).

[0223] The present invention also provides a computer-readable storage medium storing a computer program, which, when executed by a processor of an electronic device, can perform the following:

[0224] The system receives a signal amplification and noise reduction command, and confirms the signal processing system based on the signal amplification and noise reduction command. The signal processing system includes: a load amplification circuit, a wavelet transform unit, a block thresholding unit, a Wiener filter unit, and an inverse wavelet transform unit.

[0225] An initial electrical signal is acquired, and the initial electrical signal is amplified by the load amplifier circuit to obtain an amplified analog signal. The amplified analog signal is then converted from analog to digital to obtain an amplified digital signal.

[0226] The number of wavelet decomposition layers is set to n. The wavelet transform unit performs discrete wavelet transform on the amplified digital signal according to the number of layers to obtain n sets of wavelet coefficients. Each set of wavelet coefficients corresponds to each wavelet decomposition layer, and each set of wavelet coefficients consists of multiple wavelet coefficients.

[0227] Obtain the median absolute deviation of each set of wavelet coefficients, and calculate the estimated noise level of each wavelet decomposition layer based on the median absolute deviation;

[0228] The optimal block length of each set of wavelet coefficients is determined by the block threshold unit based on the estimated noise level. Based on the optimal block length, each set of wavelet coefficients is divided into multiple coefficient sub-blocks. An adaptive threshold is calculated based on the estimated noise level corresponding to each coefficient sub-block. Threshold shrinkage processing is performed on all wavelet coefficients of the corresponding coefficient sub-block according to the adaptive threshold to obtain the initial estimated wavelet coefficients.

[0229] The initial estimated wavelet coefficients are input into the Wiener filter unit. A Wiener filter is constructed based on the initial estimated wavelet coefficients and the estimated noise level of the corresponding wavelet decomposition layer. The Wiener filter is used to perform Wiener filtering on the initial estimated wavelet coefficients to obtain the final estimated wavelet coefficients.

[0230] The final estimated wavelet coefficients are input into the inverse wavelet transform unit to perform a discrete inverse wavelet transform, thereby obtaining the target amplified signal.

[0231] In the embodiments provided by this invention, it should be understood that the disclosed devices, systems, and methods can be implemented in other ways. For example, the system embodiments described above are merely illustrative, and actual implementations may have other classification methods.

[0232] The modules described as separate components may or may not be physically separate. The components shown as modules may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs.

[0233] Furthermore, the functional modules in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or in the form of hardware plus software functional modules.

[0234] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the present invention can be implemented in other specific forms without departing from the spirit or essential characteristics of the present invention.

[0235] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention.

Claims

1. A signal amplification and noise reduction method based on a transistor active load circuit, characterized in that, The method includes: The system receives a signal amplification and noise reduction command, and confirms the signal processing system based on the signal amplification and noise reduction command. The signal processing system includes: a load amplification circuit, a wavelet transform unit, a block thresholding unit, a Wiener filter unit, and an inverse wavelet transform unit. An initial electrical signal is acquired, and the initial electrical signal is amplified by the load amplifier circuit to obtain an amplified analog signal. The amplified analog signal is then converted from analog to digital to obtain an amplified digital signal. The number of wavelet decomposition layers is set to n. The wavelet transform unit performs discrete wavelet transform on the amplified digital signal according to the number of layers to obtain n sets of wavelet coefficients. Each set of wavelet coefficients corresponds to each wavelet decomposition layer, and each set of wavelet coefficients consists of multiple wavelet coefficients. Obtain the median absolute deviation of each set of wavelet coefficients, and calculate the estimated noise level of each wavelet decomposition layer based on the median absolute deviation; The optimal block length of each set of wavelet coefficients is determined by the block threshold unit based on the estimated noise level. Based on the optimal block length, each set of wavelet coefficients is divided into multiple coefficient sub-blocks. An adaptive threshold is calculated based on the estimated noise level corresponding to each coefficient sub-block. Threshold shrinkage processing is performed on all wavelet coefficients of the corresponding coefficient sub-block according to the adaptive threshold to obtain the initial estimated wavelet coefficients. The initial estimated wavelet coefficients are input into the Wiener filter unit. A Wiener filter is constructed based on the initial estimated wavelet coefficients and the estimated noise level of the corresponding wavelet decomposition layer. The Wiener filter is used to perform Wiener filtering on the initial estimated wavelet coefficients to obtain the final estimated wavelet coefficients. The final estimated wavelet coefficients are input into the inverse wavelet transform unit to perform a discrete inverse wavelet transform, thereby obtaining the target amplified signal.

2. The signal amplification and noise reduction method based on a transistor active load circuit as described in claim 1, characterized in that, The step of further performing analog-to-digital conversion on the amplified analog signal to obtain an amplified digital signal includes: The amplified analog signal is input into an analog-to-digital converter, wherein the analog-to-digital converter has a built-in clock source, a sample-and-hold circuit, a quantization encoding circuit, and a digital calibration circuit. Set the sampling frequency and quantization bit depth of the amplified analog signal, and use the clock source to generate a sampling clock signal synchronized with the sampling frequency; Based on the sampling clock signal, the sample-and-hold circuit performs equally spaced sampling on the amplified analog signal to obtain m discrete sampled voltage values, and each sampled voltage value corresponds to a sampling time. Obtain a pseudo-random sequence generated by digital logic, wherein the pseudo-random sequence contains m pseudo-random sequence values, and each pseudo-random sequence value corresponds to a sampling time. A jitter signal with a known amplitude is acquired. The known amplitude of the jitter signal is multiplied sequentially by a pseudo-random sequence value to obtain multiple injected jitter values. The injected jitter values ​​are then added to the sampling voltage values ​​of the same sampling time to obtain multiple jitter sampling values. The jitter sampling values ​​are sequentially input into a quantization encoding circuit. The jitter sampling values ​​are uniformly quantized according to the quantization bit depth to obtain the initial quantization encoding value for each sampling time. Based on the initial quantization code value and the corresponding pseudo-random sequence value at each sampling time, the cross-correlation value between the initial quantization code value and the pseudo-random sequence value is calculated using the digital calibration circuit, and the nonlinear error parameter of the analog-to-digital converter is extracted based on the cross-correlation value. Based on the nonlinear error parameter, real-time digital compensation is performed on the initial quantization code value to obtain the calibrated quantization code value at each sampling time. The calibration quantization encoding values ​​of all sampling times are arranged in chronological order to form a time-domain digital signal sequence, which is then used as an amplified digital signal.

3. The signal amplification and noise reduction method based on a transistor active load circuit as described in claim 2, characterized in that, The setting of the sampling frequency and quantization bit depth of the amplified analog signal includes: The amplified analog signal is presampled to obtain a digital sample sequence, and the digital sample sequence is subjected to spectral analysis to obtain a signal spectral distribution. The highest effective frequency is then extracted from the signal spectral distribution. Set a safety factor, and multiply the highest effective frequency by the safety factor to obtain the sampling frequency; The peak voltage and background noise voltage of the amplified analog signal are obtained, and the ratio of the peak voltage to the background noise voltage is calculated to obtain the dynamic range of the signal. Set a signal-to-noise ratio (SNR) threshold, and start from the first quantization bit and increment it bit by bit. Calculate the theoretical SNR corresponding to each quantization bit in turn, and select the first quantization bit that satisfies the theoretical SNR not lower than the SNR threshold as the theoretical quantization bit. Calculate the minimum quantization bit depth for the dynamic range of the signal, and use the larger of the theoretical quantization bit depth and the minimum quantization bit depth as the initial quantization bit depth; Set the bit adjustment step size, and round the initial quantization bit size up to an integer multiple of the bit adjustment step size to obtain the final quantization bit size; The sampling frequency and the final quantization bit depth are used as the operating parameters of the analog-to-digital converter to complete the setting of the sampling frequency and quantization bit depth.

4. The signal amplification and noise reduction method based on a transistor active load circuit as described in claim 2, characterized in that, The step of calculating the cross-correlation value between the initial quantized encoded value and the pseudo-random sequence value using the digital calibration circuit, and extracting the nonlinear error parameter of the analog-to-digital converter based on the cross-correlation value, includes: A set accumulation count is defined, where the accumulation count is a positive integer not greater than m. Based on the accumulation count, the initial quantization encoding value, and the pseudo-random sequence value, the cross-correlation value is calculated according to the following formula: ； in, For cross-correlation values, To accumulate the number of times, For the first The initial quantized encoded value at each sampling time. For the first The pseudo-random sequence values ​​corresponding to each sampling time; The ratio of the cross-correlation value to the known amplitude of the jitter signal is calculated to obtain the estimated value of the nonlinear error parameter; The first-order exponential smoothing method is used to iteratively update the estimated value of the nonlinear error parameter, and the updated estimated value of the nonlinear error parameter is used as the nonlinear error parameter of the analog-to-digital converter.

5. The signal amplification and noise reduction method based on a transistor active load circuit as described in claim 4, characterized in that, The step of iteratively updating the estimated nonlinear error parameter using a first-order exponential smoothing method, and using the updated estimated nonlinear error parameter as the nonlinear error parameter of the analog-to-digital converter, includes: Confirm the current iteration cycle, obtain the estimated value of the nonlinear error parameter for the current iteration cycle, and set a smoothing factor, wherein the smoothing factor ranges from 0 to 1. Determine whether the current iteration cycle is the first iteration. If it is the first iteration, use the current nonlinear error parameter estimate as the updated nonlinear error parameter estimate after the first iteration. If it is not the first iteration, obtain the estimated value of the nonlinear error parameter after the previous iteration update, calculate the product of the smoothing factor and the estimated value of the nonlinear error parameter in the current iteration period, and obtain the first weighted value; Calculate the difference obtained by subtracting the smoothing factor from 1, and then multiply the difference by the estimated value of the nonlinear error parameter obtained after the previous iteration update to obtain the second weighted value; Add the first weighted value to the second weighted value to obtain the estimated value of the nonlinear error parameter after this iteration. Repeat the above steps, and use the estimated value of the nonlinear error parameter after each iteration as the nonlinear error parameter of the analog-to-digital converter.

6. The signal amplification and noise reduction method based on a transistor active load circuit as described in claim 2, characterized in that, The step of performing real-time digital compensation on the initial quantization code value based on the nonlinear error parameter to obtain the calibration quantization code value at each sampling time includes: Based on the extracted nonlinear error parameters of the analog-to-digital converter, the input range of the analog-to-digital converter is confirmed; The input range is divided into multiple continuous and non-overlapping quantization intervals, and the corresponding nonlinear error parameters are calibrated at the two endpoints of each quantization interval. Obtain the initial quantization code value at each sampling time, determine the quantization interval to which the initial quantization code value belongs, and mark it as the target quantization interval; Read the initial quantization code value of the left endpoint, the nonlinear error parameter corresponding to the left endpoint, the initial quantization code value of the right endpoint, and the nonlinear error parameter corresponding to the right endpoint of the target quantization interval. The compensation amount corresponding to the initial quantized encoded value is calculated according to the following formula: ； in, For compensation amount, This is the initial quantization encoding value. and These are the initial quantization values ​​at the left and right endpoints of the target quantization interval, respectively. and These are the nonlinear error parameters corresponding to the left endpoint and the nonlinear error parameters corresponding to the right endpoint, respectively. The calibration quantization value for each sampling time is obtained by subtracting the corresponding compensation amount from the initial quantization code value at each sampling time.

7. The signal amplification and noise reduction method based on a transistor active load circuit as described in claim 1, characterized in that, The step involves using the block threshold unit to determine the optimal block length for each set of wavelet coefficients based on the estimated noise level, and dividing each set of wavelet coefficients into multiple coefficient sub-blocks based on the optimal block length, including: Set the block length search range and search step size for each wavelet decomposition layer, and select candidate block lengths sequentially according to the search step size for the block length search range. Based on the length of each candidate block, the set of wavelet coefficients for each wavelet decomposition layer is divided into multiple candidate sub-blocks; Confirm all wavelet coefficients and the estimated noise level of the corresponding wavelet decomposition layer for each candidate sub-block, and calculate the threshold shrinkage mean square risk estimate for each candidate sub-block. The threshold shrinkage mean square risk estimates of all candidate sub-blocks under the same candidate block length are summed to obtain the total risk estimate for each candidate block length. Traverse all candidate block lengths and select the candidate block length with the smallest total risk estimate as the optimal block length for the corresponding wavelet decomposition layer. The wavelet coefficient set of the wavelet decomposition layer is divided into multiple non-overlapping coefficient sub-blocks according to the optimal block length. The number of wavelet coefficients contained in each coefficient sub-block is equal to the optimal block length. If the total number of wavelet coefficients of the wavelet decomposition layer is not divisible by the optimal block length, the remaining undivided wavelet coefficients are merged into the last coefficient sub-block.

8. The signal amplification and noise reduction method based on a transistor active load circuit as described in claim 1, characterized in that, The process involves constructing a Wiener filter based on the initially estimated wavelet coefficients and the estimated noise level of the corresponding wavelet decomposition layer, and then applying the Wiener filter to perform Wiener filtering on the initially estimated wavelet coefficients to obtain the final estimated wavelet coefficients. This includes: The filter window size of each wavelet decomposition layer is set according to the estimated noise level of each wavelet decomposition layer. The filter window size is used as the sliding window length. Based on the sliding window length, the sliding window is commanded to slide point by point with a sliding step size of 1 until all the initial estimated wavelet coefficients of the wavelet decomposition layer are traversed. Confirm the center position of the sliding window, calculate the arithmetic mean of all initially estimated wavelet coefficients within the sliding window, and use the arithmetic mean as the local signal mean at the center position; Calculate the variance of all initially estimated wavelet coefficients within the sliding window, and use the variance as the local signal variance at the center position; The local signal variance is compared with the estimated noise level of the corresponding wavelet decomposition layer: if the local signal variance is less than or equal to the estimated noise level, the wavelet coefficients at the center position are replaced with the local signal mean; if the local signal variance is greater than the estimated noise level, the difference between the local signal variance and the estimated noise level is calculated, the difference is divided by the local signal variance to obtain the weighting coefficient, the difference between the wavelet coefficients at the center position and the local signal mean is multiplied by the weighting coefficient to obtain the weighted correction difference, and finally the weighted correction difference is added to the local signal mean to obtain the weighted corrected wavelet coefficients. The weighted and corrected wavelet coefficients are used as the Wiener-filtered wavelet coefficients at the center position of the corresponding sliding window. After traversing all the center positions of the sliding window, the Wiener-filtered wavelet coefficients of each wavelet decomposition layer are obtained. The final estimated wavelet coefficients are obtained by combining the Wiener-filtered wavelet coefficients of all wavelet decomposition layers.

9. The signal amplification and noise reduction method based on a transistor active load circuit as described in claim 8, characterized in that, The step of setting the filter window size for each wavelet decomposition layer based on the estimated noise level of each wavelet decomposition layer includes: The initial estimated wavelet coefficients of each wavelet decomposition layer are divided into multiple local regions, wherein each local region contains multiple wavelet coefficients; Calculate the sum of the absolute values ​​of all wavelet coefficients in each local region, and divide the sum of absolute values ​​by the total number of wavelet coefficients in the local region to obtain the average amplitude of the coefficients in the local region. The amplitude signal-to-noise ratio of the local region is obtained by calculating the ratio of the average amplitude of the coefficients to the estimated noise level of the corresponding wavelet decomposition layer. Based on the amplitude signal-to-noise ratio, the range of amplitude signal-to-noise ratio values ​​is obtained, and multiple signal-to-noise ratio thresholds are set. Based on the signal-to-noise ratio thresholds, the range of amplitude signal-to-noise ratio values ​​is divided into multiple continuous intervals. The amplitude signal-to-noise ratio of each local region is matched with a continuous interval. After the matching is completed, a corresponding filter window size is assigned to each local region, and the center position of each local region is mapped to the value point of the corresponding filter window size. Median filtering is applied sequentially to the filter window sizes corresponding to the value points to obtain the filter window sizes of all sliding windows in each wavelet decomposition layer.

10. A signal amplification and noise reduction system based on a transistor active load circuit, characterized in that, The system includes: The instruction system confirmation module is used to receive signal amplification and noise reduction instructions and confirm the signal processing system based on the signal amplification and noise reduction instructions. The signal processing system includes: a load amplification circuit, a wavelet transform unit, a block thresholding unit, a Wiener filtering unit, and an inverse wavelet transform unit. The low-noise analog-to-digital converter module is used to acquire an initial electrical signal, perform low-noise amplification processing on the initial electrical signal using the load amplification circuit to obtain an amplified analog signal, and then perform analog-to-digital conversion on the amplified analog signal to obtain an amplified digital signal. The wavelet block noise reduction module is used to set the number of wavelet decomposition layers to n. The wavelet transform unit performs discrete wavelet transform on the amplified digital signal according to the number of layers to obtain n sets of wavelet coefficients. Each set of wavelet coefficients corresponds to one wavelet decomposition layer, and each set consists of multiple wavelet coefficients. The absolute median deviation of each set of wavelet coefficients is obtained. Based on the absolute median deviation, the estimated noise level of each wavelet decomposition layer is calculated. The block threshold unit determines the optimal block length of each set of wavelet coefficients based on the estimated noise level. Based on the optimal block length, each set of wavelet coefficients is divided into multiple coefficient sub-blocks. An adaptive threshold is calculated based on the estimated noise level corresponding to each coefficient sub-block. Threshold shrinkage processing is performed on all wavelet coefficients of the corresponding coefficient sub-block according to the adaptive threshold to obtain the initial estimated wavelet coefficients. The Wiener reconstruction output module is used to input the initially estimated wavelet coefficients into the Wiener filtering unit, construct a Wiener filter based on the initially estimated wavelet coefficients and the estimated noise level of the corresponding wavelet decomposition layer, perform Wiener filtering on the initially estimated wavelet coefficients using the Wiener filter to obtain the final estimated wavelet coefficients, and input the final estimated wavelet coefficients into the inverse wavelet transform unit to perform discrete inverse wavelet transform to obtain the target amplified signal.

Images