Noise suppression and gain control method for high-frequency signals

Abstract

The invention provides a noise suppression and gain control method for a high-frequency signal, and relates to the technical field of adaptive control, and the method comprises the steps: collecting a high-frequency target signal and an environment reference signal through a dual-path parallel input channel; performing filtering processing on the high-frequency target signal by using an adaptive band-pass filter; carrying out weighted correlation detection on the filtered high-frequency target signal and the environment reference signal, calculating a correlation degree through a dynamically updated weight coefficient, and identifying effective signal components; performing gain control on the signal by adopting a Newton iteration method based on a second derivative, wherein the gain control comprises two stages of step-by-step coarse tuning and continuous fine tuning; signal-to-noise ratio monitoring is carried out at the output stage, and a monitoring result is fed back to the adaptive band-pass filter and the gain control link, so that bidirectional coupling closed-loop control is formed. According to the invention, the high peak noise in the high-frequency signal can be effectively suppressed, and the signal gain is kept in a proper range.

Description

Technical Field

[0001] The present invention relates to the technical field of adaptive control, and particularly to a noise suppression and gain control method for high-frequency signals. Background Art

[0002] In a high-frequency signal processing system, a signal gain amplifier is a key device for realizing effective signal reception and processing. High-frequency signals have characteristics such as high frequency, wide bandwidth, and strong time-varying characteristics, and are easily interfered by various noises during transmission and processing, especially the influence of impulsive noise and burst interference. These noises will not only reduce the signal-to-noise ratio of the signal, but also cause signal distortion and information loss. Therefore, effective noise suppression measures must be taken.

[0003] Signal gain control technology is an important means to improve the quality of signal processing, mainly by adjusting the gain of the amplifier to keep the amplitude of the output signal within an appropriate range. Traditional automatic gain control (AGC) methods usually adopt the methods of envelope detection and feedback control, and adjust the gain coefficient according to the amplitude characteristics of the output signal. At the same time, in order to ensure the linear amplification of the signal, the working characteristics of the amplifier need to be considered to avoid entering the non-linear region.

[0004] However, the existing signal processing technologies have some deficiencies in processing high-frequency signals: a single gain control strategy is difficult to adapt to complex and changeable signal environments; the traditional noise suppression methods have limited effects on high-peak impulsive noise; at the same time, gain control and noise suppression are often independent processes, lacking an effective cooperation mechanism, which makes it difficult for the system to achieve optimal performance in the face of complex signal environments. Summary of the Invention

[0005] In view of this, the present invention proposes a noise suppression and gain control method for high-frequency signals. By collecting signals through a dual-channel parallel input channel, combining adaptive filtering with non-linear penalty, dynamic weighted correlation detection, and Newton iteration gain control based on the second derivative, it can effectively suppress high-peak noise in high-frequency signals, while keeping the signal gain within an appropriate range, and improving the adaptability of the system to complex signal environments through a two-way coupled closed-loop control mechanism, thereby improving the overall performance of high-frequency signal processing.

[0006] The technical solution of the present invention is implemented as follows:

[0007] The present invention provides a noise suppression and gain control method for high-frequency signals, including:

[0008] S1. Collect the high-frequency target signal and the environmental reference signal respectively through a dual-channel parallel input channel;

[0009] S2. Filter the high-frequency target signal using an adaptive band-pass filter with a non-linear penalty term, where the non-linear penalty term is calculated based on the signal kurtosis and is used to suppress high-kurtosis noise;

[0010] S3. Perform weighted correlation detection on the filtered high-frequency target signal and the environmental reference signal, calculate the degree of correlation through dynamically updated weight coefficients, and identify effective signal components;

[0011] S4. Use the Newton iteration method based on the second derivative to perform gain control on the signal, including two stages: step-by-step coarse tuning and continuous fine tuning;

[0012] S5. Monitor the signal-to-noise ratio at the output stage and feedback the monitoring result to the adaptive band-pass filter and the gain control link to form a two-way coupled closed-loop control.

[0013] Based on the above solution, preferably, the adaptive band-pass filter uses the recursive least squares algorithm for parameter update, and its objective function is:

[0014]

[0015] where w(n) is the impulse response coefficient vector of the filter, w * (n) is the optimal impulse response coefficient vector of the filter at the nth moment, x(i) is the ith sampling value of the input signal sequence, d(i) is the ith sampling value of the desired output signal sequence, λ is the forgetting factor, n is the length of the signal sequence, w T represents the transpose of w, κ(x(i)) represents the kurtosis measure of the input signal, Ψ(·) is the non-linear mapping function of this kurtosis term, Ψ(κ(x(i))) represents the non-linear penalty term, and α is the weighting factor used to control the weight of the non-linear penalty term in the overall optimization.

[0016] Based on the above solution, preferably, the non-linear mapping function Ψ(·) uses a soft threshold or a Gaussian-type penalty function; when the value of κ(x(i)) increases, the non-linear penalty term increases accordingly, prompting the filter weight update to adjust in the direction of suppressing high-kurtosis components.

[0017] Based on the above solution, preferably, the weighted correlation detection uses the following formula to calculate the weighted correlation coefficient:

[0018]

[0019] In the formula, ρ ω is the weighted correlation coefficient, x(i) is the ith sampling value of the main channel input signal sequence, y(i) is the ith sampling value of the reference channel signal sequence, ω iis the dynamic weight corresponding to the i-th sampling point, and i is the time series index.

[0020] On the basis of the above scheme, preferably, the dynamic weight ω i is updated as follows:

[0021] ω i = ω i-1 ·f(κ(x(i)), κ(y(i)), SNR) × g(BW, k ref )

[0022] where κ(x(i)) represents the kurtosis measure of x(i), κ(y(i)) represents the kurtosis measure of y(i), κ ref is the reference channel kurtosis, f(·) is a heuristic function that comprehensively considers the current pulse level, signal-to-noise ratio, and the detection situation of several past frames, and g(BW, k ref ) is a quadratic correction function based on the instantaneous bandwidth BW and the reference channel kurtosis k ref .

[0023] On the basis of the above scheme, preferably, when the instantaneous bandwidth BW temporarily widens and the reference channel kurtosis k ref is relatively high, the value of g(BW, κ ref ) is less than 1 to reduce the weight and weaken the impact of burst sampling; when the instantaneous bandwidth BW narrows and the reference channel kurtosis κ ref is relatively low, the value of g(BW, κ ref ) is close to or slightly greater than 1 to increase the weight and enhance the ability to extract weak target signals.

[0024] On the basis of the above scheme, preferably, the Newton iteration method based on the second derivative includes:

[0025] Define the objective function Φ(G) = (SNR target - SNR(G)) 2 , where G is the current gain, SNR target is the target signal-to-noise ratio, and SNR(G) is the actually measured signal-to-noise ratio when the gain value is G;

[0026] Update the gain using the following iterative formula:

[0027]

[0028] where η is the step size scaling factor used to control the convergence speed and stability; L(G) is the linearity compensation function; μ is the compensation weight factor.

[0029] On the basis of the above scheme, preferably, the linearity compensation function L(G) is defined as:

[0030] L(G) = max(0, P out (G) - P IIP3-thre ) 2

[0031] where P out (G) represents the estimated output power adjusted with the gain G; P IIP3-thre is the safety power interval reserved according to the third-order intercept point of the amplifier; when the output power approaches the nonlinear region of the amplifier, L(G) and its derivative increase, suppressing the continuous increase of the gain.

[0032] Based on the above scheme, preferably, the closed-loop control of the bidirectional coupling includes:

[0033] When high pulse noise is detected, increase the weight factor of the nonlinear penalty term to preferentially suppress the spikes;

[0034] When the output signal is detected to be saturated or too low, correct the gain iteratively and simultaneously adjust the bandwidth and center frequency of the filter.

[0035] Based on the above scheme, preferably, a monitor integrating the Kalman filter and the linear prediction model is set at the output stage to predict the signal-to-noise ratio and peak fluctuation at the next moment according to the output of historical frames and the current state; when it is predicted that the signal has an obvious sudden increase trend, pre-adjust the step attenuator or reduce the gain of the variable gain amplifier to prevent instantaneous saturation.

[0036] The present invention has the following beneficial effects compared with the prior art:

[0037] (1) Through the dual-channel parallel input structure and the closed-loop control mechanism of bidirectional coupling, the present invention organically combines the adaptive filtering of nonlinear penalty, the dynamic weighted correlation detection, and the gain control based on the second derivative, realizing the adaptive processing of high-frequency signals, effectively improving the anti-interference ability and signal processing quality of the system in a complex electromagnetic environment;

[0038] (2) By adopting the nonlinear penalty term based on the signal kurtosis and dynamically updating the filter parameters through the recursive least squares algorithm, the system can quickly identify and suppress high-kurtosis noise, while maintaining a low distortion degree for the effective signal components, improving the selectivity and adaptive performance of the filter;

[0039] (3) By introducing the dynamic weight update strategy and combining the quadratic correction mechanism of the instantaneous bandwidth and the reference channel kurtosis, the correlation detection has stronger adaptability to signal state changes, can effectively suppress the influence of burst sampling, and can enhance the ability to extract weak target signals;

[0040] (4) The Newton iteration method based on the second derivative, combined with the linearity compensation function, realizes the fast convergence and precise adjustment of gain control. At the same time, by monitoring the relationship between the output power and the third-order intercept point of the amplifier, it effectively avoids the amplifier entering the non-linear region and ensures the linearity of signal processing.

[0041] (5) The monitoring mechanism that integrates Kalman filtering and linear prediction is adopted at the output stage, which can predict the signal change trend in advance and perform gain pre-adjustment, effectively preventing signal instantaneous saturation and improving the system's fast response ability and processing stability to burst signals. Description of the Drawings

[0042] In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required in the description of the embodiments or the prior art. Obviously, the following drawings are only some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can also be obtained based on these drawings.

[0043] Figure 1 It is the flowchart of the method of the present invention;

[0044] Figure 2 It is the technical implementation diagram of the present invention. Detailed Embodiments

[0045] The following will describe clearly and completely the technical solutions in the embodiments of the present invention in combination with the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, rather than all embodiments. All other embodiments obtained by those of ordinary skill in the art without creative efforts based on the embodiments of the present invention belong to the scope of protection of the present invention.

[0046] As Figure 1 shown, the present invention provides a noise suppression and gain control method for high-frequency signals, including:

[0047] S1. Collect the high-frequency target signal and the environmental reference signal through the dual-channel parallel input channels respectively;

[0048] S2. Filter the high-frequency target signal by using an adaptive band-pass filter with a non-linear penalty term, where the non-linear penalty term is calculated based on the signal kurtosis and is used to suppress high-kurtosis noise;

[0049] S3. Perform weighted correlation detection on the filtered high-frequency target signal and the environmental reference signal, calculate the degree of correlation through the dynamically updated weight coefficient, and identify the effective signal components;

[0050] S4. Use the Newton iteration method based on the second derivative to perform gain control on the signal, including two stages: step-by-step coarse tuning and continuous fine tuning;

[0051] S5. Monitor the signal-to-noise ratio at the output stage and feedback the monitoring result to the adaptive band-pass filter and the gain control link to form a two-way coupled closed-loop control.

[0052] Please refer to Figure 2 , the present invention provides a noise suppression and gain control method for high-frequency signals. The high-frequency target signal and the environmental reference signal are respectively collected through the dual-channel parallel input channels. The adaptive band-pass filter with a non-linear penalty term is used to filter the target signal, and the non-linear penalty term is calculated based on the kurtosis of the signal to suppress the high-kurtosis noise; then, the weighted correlation detection is performed on the filtered signal and the environmental reference signal, and the correlation degree is calculated through the dynamically updated weight coefficient to identify the effective signal components; the Newton iteration method based on the second derivative is used for gain control, including two stages: step-by-step coarse tuning and continuous fine tuning; finally, the signal-to-noise ratio is monitored at the output stage, and the monitoring result is feedback to the adaptive band-pass filter and the gain control link to form a two-way coupled closed-loop control, realizing the adaptive processing and noise suppression of high-frequency signals.

[0053] Specifically, in an embodiment of the present invention, step S1 includes:

[0054] First, the high-frequency target signal and the environmental reference signal are respectively collected through the dual-channel parallel input channels. Among them, the main channel is used to receive the high-frequency target signal to be processed and input the signal to the adaptive filtering and subsequent signal processing module; the reference channel is used to collect the environmental reference signal with a similar bandwidth to the main channel signal. The reference signal mainly contains environmental noise information and is used to provide auxiliary information for subsequent correlation detection and noise reduction processing.

[0055] Specifically, the main channel includes a high-frequency front-end circuit, which is composed of an antenna, a low-noise amplifier, a mixer, and an analog-to-digital converter. After the antenna receives the high-frequency target signal, it is pre-amplified by the low-noise amplifier, then the signal is frequency-converted to the intermediate frequency by the mixer, and finally the analog signal is converted into a digital signal sequence x(i) by the analog-to-digital converter, where i is the time series index.

[0056] The structure of the reference channel is similar to that of the main channel, but the directivity and gain characteristics of its antenna are optimized to better collect environmental noise information. The signal collected by the reference channel also undergoes amplification, frequency conversion, and digital processing to obtain the reference signal sequence y(i). The sampling clocks of the two channels are kept synchronized to ensure the timing alignment of the signals during subsequent correlation detection.

[0057] Through the above dual-channel parallel input structure, the system can simultaneously obtain real-time information of the target signal and the environmental reference signal, providing the necessary input data for subsequent processing such as adaptive filtering, correlation detection, and noise suppression. This parallel acquisition method has stronger anti-interference ability compared with the single-channel structure and can better handle complex high-frequency signal environments.

[0058] Specifically, in one embodiment of the present invention, step S2 includes:

[0059] First, perform adaptive band-pass filtering on the high-frequency target signal collected from the main channel. The present invention uses an adaptive band-pass filter with a non-linear penalty term, and its core is to update the filter parameters in real time through the recursive least squares (RLS) algorithm. The objective function of this filter is:

[0060]

[0061] where w(n) is the impulse response coefficient vector of the filter, w * (n) is the optimal impulse response coefficient vector of the filter at the nth moment, x(i) is the ith sampling value of the input signal sequence, d(i) is the ith sampling value of the desired output signal sequence, λ is the forgetting factor, n is the length of the signal sequence, and w T represents the transpose of w.

[0062] This objective function introduces a non-linear penalty term α·Ψ(k(x(i))). Among them, k(x(i)) represents the kurtosis measure of the input signal, which is used to characterize the spike or impulsive noise components in the signal; Ψ(·) is a non-linear mapping function of this kurtosis term, and a soft threshold function or a Gaussian-type penalty function can be used; α is a weighting factor used to control the weight of the penalty term in the overall optimization.

[0063] Specifically, when it is detected that the kurtosis κ(x(i)) value of the input signal x(i) increases, it indicates that there is obvious impulsive noise or spike noise. At this time, the penalty term Ψ(κ(x(i))) increases accordingly, prompting the filter weight update to adjust in the direction of suppressing high-kurtosis components.

[0064] Specifically, the non-linear mapping function Ψ(·) selects different forms according to the specific scenario:

[0065] When using a soft threshold function, it is set as:

[0066] Ψ(κ) = max(0, κ - κ th )

[0067] where κ th is the preset kurtosis threshold.

[0068] When using a Gaussian-type penalty function, it is set as:

[0069] Ψ(κ) = 1 - exp(-κ 2 / 2σ 2 )

[0070] where σ is an adjustable shape parameter.

[0071] The selection of the weighting factor α needs to balance the filtering performance and the computational complexity. An overly large α value will lead to excessive suppression of high kurtosis components, possibly losing effective signals; an overly small α value will reduce the suppression effect on impulse noise. Specifically, the α value is dynamically adjusted according to the complexity of the signal environment to make the system have better adaptability.

[0072] Through the above adaptive band-pass filter, the system can effectively identify and suppress high kurtosis noise while maintaining the frequency band selectivity for the target signal, improving the anti-interference ability and signal processing quality in a complex high-frequency environment.

[0073] Specifically, in an embodiment of the present invention, step S3 includes:

[0074] First, weighted correlation detection is performed on the high-frequency target signal and the environmental reference signal after adaptive band-pass filtering. This detection uses a weighted correlation coefficient calculation method, and its calculation formula is:

[0075]

[0076] In the formula, ρ ω is the weighted correlation coefficient, x(i) is the i-th sampling value of the main channel input signal sequence, y(i) is the i-th sampling value of the reference channel signal sequence, ω i is the dynamic weight value corresponding to the i-th sampling point, and i is the time series index.

[0077] The update strategy of the dynamic weight value ω i is:

[0078] ω i = ω i-1 ·f(k(x(i)), k(y(i)), SNR) × g(BW, κ ref )

[0079] In the formula, κ(x(i)) represents the kurtosis measure of x(i), κ(y(i)) represents the kurtosis measure of y(i), k ref is the kurtosis of the reference channel, f(·) is a heuristic function that comprehensively considers the current pulse level, signal-to-noise ratio, and the detection situation of several past frames, and g(BW, κ ref ) is a quadratic correction function based on the instantaneous bandwidth BW and the kurtosis κ ref of the reference channel.

[0080] Specifically, when the system detects that the instantaneous bandwidth BW temporarily widens and the kurtosis κ of the reference channel ref is relatively high, it indicates that the distribution range of the high-kurtosis noise becomes larger. At this time, the value of g(BW, κ ref ) is less than 1, and the influence of burst sampling on the overall correlation degree is weakened by reducing the weight value. On the contrary, when the instantaneous bandwidth BW narrows and the kurtosis κ of the reference channel ref is relatively low, it indicates that the noise environment is relatively stable. At this time, the value of g(BW, κ ref ) is close to or slightly greater than 1, and the weight value is appropriately increased to enhance the ability to extract weak target signals.

[0081] This correlation detection method based on dynamic weights has the following advantages: By introducing dynamic weights, the system can adaptively adjust the contribution degree of each sampling point to the correlation coefficient according to signal characteristics; Combining the quadratic correction mechanism of instantaneous bandwidth and reference channel kurtosis improves the system's ability to identify burst noise; While maintaining the simplicity and high efficiency of the algorithm, it significantly enhances the ability of correlation detection to extract weak target signals.

[0082] Through the above weighted correlation detection process, the system can effectively identify the effective signal components in high-frequency target signals, providing a reliable basis for subsequent gain control.

[0083] Specifically, in an embodiment of the present invention, step S4 includes:

[0084] First of all, the present invention adopts a hierarchical gain adjustment strategy, dividing the gain control into two stages: step-by-step coarse adjustment and continuous fine adjustment. In the coarse adjustment stage, a step attenuator is used to limit the magnitude of the received high-frequency signal, roughly adjusting the signal to an operable range to avoid amplifier saturation or overdrive.

[0085] In the fine adjustment stage, a variable gain amplifier (VGA) is used for continuous adjustment to meet more refined gain requirements. Define the objective function:

[0086] Φ(G)=(SNR target -SNR(G)) 2

[0087] where G is the current gain, SNR target is the target signal-to-noise ratio, and SNR(G) is the actually measured signal-to-noise ratio when the gain value is G;

[0088] The present invention uses the Newton iteration method based on the second derivative for gain update, and its iteration formula is:

[0089]

[0090] Among them, η is the step size scaling factor used to control the convergence speed and stability; L(G) is the linearity compensation function; μ is the compensation weight factor. To ensure fast response in high-frequency scenarios, an adaptive step size adjustment strategy can be adopted: when is large, η is appropriately reduced to prevent oscillation; when the gradient becomes small, η is increased to accelerate convergence.

[0091] Considering that a high-speed amplifier can maintain good linearity only within a specific power range, the present invention introduces the linearity compensation function L(G):

[0092] L(G) = max(0, P out (G) - P IIP3-thre ) 2

[0093] In the formula, P out (G) represents the estimated output power adjusted with the gain G; P IIP3-thre is the safety power range reserved according to the third-order intercept point of the amplifier; when the output power approaches the nonlinear region of the amplifier, L(G) and its derivative increase, suppressing the continuous increase of the gain, thereby avoiding intermodulation distortion caused by excessive power.

[0094] In specific implementation, the system first performs coarse adjustment through a step attenuator to quickly control the signal level within the safe region. Then, it starts the VGA gain adjustment process based on the second-order derivative Newton iteration, and continuously updates the gain value through the above iteration formula. In each iteration, the system calculates the current output power and corrects the gain update through the linearity compensation function to ensure that the output signal not only meets the target signal-to-noise ratio requirement but also does not cause the amplifier to enter the nonlinear region due to excessive output power.

[0095] This gain control method based on the second-order derivative has a faster convergence speed and a smaller overshoot compared to the traditional first-order gradient descent method, and is particularly suitable for high-frequency systems with high requirements for transient response. At the same time, by introducing the linearity compensation and adaptive step size strategies, the nonlinear distortion problem of high-frequency amplifiers under large-signal conditions is effectively solved, and the stability and reliability of the system in a complex high-frequency environment are improved.

[0096] Specifically, in an embodiment of the present invention, step S5 includes:

[0097] First, a monitor integrating a Kalman filter and a linear prediction model is set at the output stage of the system. This monitor predicts the signal-to-noise ratio and peak fluctuation at the next moment based on the output of historical frames and the current state. Specifically, the monitor adopts the following prediction model:

[0098] The current state vector x(k) includes the signal-to-noise ratio SNR(k) and the output power P out(k) Two components. The Kalman filter tracks and predicts these two quantities in real time through the state equation and the observation equation. At the same time, based on the historical data of the most recent N frames, a linear prediction model is used to estimate the change trend of the signal. When it is predicted that the signal has an obvious sudden increase trend (for example, the predicted value exceeds the preset threshold), the system will pre-adjust the step attenuator or reduce the gain of the variable gain amplifier to prevent instantaneous saturation.

[0099] Secondly, the monitoring results are fed back to the adaptive band-pass filter and the gain control link through a two-way coupled closed-loop control mechanism. Specifically, it includes:

[0100] 1. When high pulse noise is detected (judged by the kurtosis measure κ(x(i))), the system will increase the weight factor α of the non-linear penalty term, so that the filter preferentially suppresses these spike noises. The adjustment of the weight factor α follows the following principle:

[0101] When κ(x(i)) > κ th-high then, α = α0 × (1 + β × (κ(x(i)) - κ th-high ));

[0102] When κ(x(i)) ≤ κ th-high then, α gradually returns to the default value α0.

[0103] Among them, κ th-high is the preset high kurtosis threshold, and β is the adjustment coefficient.

[0104] 2. When it is detected that the output signal is saturated or too low, the system will make the following adjustments simultaneously:

[0105] Through gain iteration correction: adjust the step size scaling factor η and the compensation weight factor μ;

[0106] Update the filter parameters: adjust the bandwidth and the center frequency to better adapt to the current signal characteristics;

[0107] Specifically, when the output power is close to saturation, increase the weight of the linearity compensation function L(G) and reduce the gain increase speed; when the output power is too low, appropriately widen the bandwidth of the band-pass filter to improve the signal passing rate.

[0108] 3. In terms of prediction and monitoring, the system combines the advantages of Kalman filtering and linear prediction: Kalman filtering provides the optimal estimate of the current state; linear prediction predicts the short-term trend based on historical data; after the results of both are weighted and fused, they are used to guide the pre-adjustment of system parameters.

[0109] Through the above two-way coupled closed-loop control mechanism, the system can respond in a timely manner to the changes in the signal state, while maintaining the signal quality, effectively preventing saturation distortion or signal attenuation during the processing process.

[0110] In a specific embodiment of the present invention, the execution process of the control method of the present invention is as follows: The system first adopts a dual-channel parallel input structure. The main channel collects high-frequency target signals to be processed through a high-frequency front-end circuit (antenna, low-noise amplifier, mixer, and analog-to-digital converter), and the reference channel collects environmental noise signals. The target signal is first processed by an adaptive band-pass filter with a non-linear penalty term. This filter suppresses high-kurtosis noise by calculating the kurtosis of the signal in real time and introducing corresponding penalty terms. The filtered signal is subjected to weighted correlation detection with the reference channel signal, and the effective signal components are identified through dynamically updated weight coefficients. Then the signal enters the gain control link. First, a step attenuator is used for coarse adjustment to avoid saturation, and then a variable gain amplifier (VGA) performs fine gain adjustment based on the Newton iteration method of the second derivative. The system sets a monitor that combines Kalman filtering and linear prediction at the output stage, continuously monitors the signal-to-noise ratio and output power, and feeds the monitoring results back to the filter and gain control module through a two-way coupled closed-loop: when high pulse noise is detected, the weight of the non-linear penalty term of the filter is automatically increased; when the output signal is abnormal, the gain iteration parameters and filter characteristics are adjusted simultaneously; when it is predicted that the signal may mutate, the gain is adjusted in advance to prevent distortion, thereby realizing the adaptive processing and precise control of high-frequency signals.

[0111] In a specific embodiment, the hardware system of the present invention is implemented as follows:

[0112] 1. Dual-channel signal acquisition unit:

[0113] Main channel: A broadband low-noise amplifier (LNA) is used as the RF front-end, with a gain range of 0 - 30 dB and a noise figure < 2 dB; the mixer down-converts the RF signal to an intermediate frequency, and the local oscillator frequency is programmable; a 14-bit high-speed ADC is used for digitization, with a sampling rate ≥ 100 MSPS to ensure effective sampling of GHz-level signals. Reference channel: The same architecture but optimized antennas and front-end circuits are used, focusing on collecting environmental noise information; a 14-bit ADC is also configured to maintain synchronous sampling with the main channel. Signal conditioning circuit: It includes a programmable gain amplifier (PGA) and a band-pass filter for signal preprocessing and anti-aliasing.

[0114] 2. Core processing unit:

[0115] An integrated digital circuit is used as the main controller, with a clock frequency ≥ 200 MHz; an integrated hardware multiplier array and DSP module are supported for real-time adaptive filtering and correlation operations; on-chip registers are used for signal caching and intermediate result storage; a parallel processing architecture is designed to process dual-channel signal data simultaneously; hardware acceleration of non-linear penalty RLS algorithms, weighted correlation detection, and Newton iteration is realized.

[0116] 3. Gain control output unit:

[0117] Digital step attenuator: attenuation range 0 - 31.5 dB, step 0.5 dB, for coarse signal adjustment; High-precision DAC: 14-bit resolution, update rate ≥ 200 MSPS, driving a variable gain amplifier; Wideband variable gain amplifier (VGA): gain range -10 dB to +20 dB, bandwidth ≥ 500 MHz; Output protection circuit: overvoltage protection, overcurrent protection and temperature monitoring.

[0118] In the specific implementation process, the system works according to the following process:

[0119] (1) Signal acquisition stage:

[0120] The main channel receives the high-frequency target signal through the LNA, is down-converted by the mixer and sampled by the ADC; The reference channel synchronously acquires the ambient noise signal and keeps the sampling clock synchronized; The two signals are sent to the integrated digital circuit for digital processing after preprocessing.

[0121] (2) Signal processing stage:

[0122] The integrated digital circuit implements an adaptive band-pass filtering algorithm with a non-linear penalty term; The hardware multiplier array completes the weighted correlation detection operation; The DSP module executes Newton iteration gain control based on the second derivative.

[0123] (3) Gain control output stage:

[0124] First, coarse adjustment is performed through the digital step attenuator; Then, the DAC drives the VGA to achieve fine gain adjustment; The signal-to-noise ratio and power level of the output signal are monitored in real time.

[0125] (4) Feedback control stage:

[0126] Collect the signal status information at the output end; The integrated digital circuit executes the Kalman filter and linear prediction algorithm; The filter parameters and gain control parameters are dynamically adjusted according to the prediction results.

[0127] The hardware system of the present invention adopts a modular design, and each functional unit is interconnected through a high-speed digital interface, which is convenient for system integration and maintenance. By reasonably selecting device indicators and optimizing the system architecture, it is ensured that a low noise figure and distortion can still be maintained in the GHz frequency band, meeting the performance requirements of high-frequency signal processing.

[0128] The above is only the preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A noise suppression and gain control method for high-frequency signals, characterized in that, Including: S1. Collecting high-frequency target signals and environmental reference signals respectively through a dual-channel parallel input channel; S2. Filtering the high-frequency target signals by using an adaptive band-pass filter with a non-linear penalty term, where the non-linear penalty term is calculated based on the signal kurtosis and is used to suppress high-kurtosis noise; S3. Performing weighted correlation detection on the filtered high-frequency target signals and the environmental reference signals, calculating the degree of correlation through a dynamically updated weight coefficient, and identifying effective signal components; S4. Performing gain control on the signals by using a Newton iteration method based on the second derivative, including two stages of step-by-step coarse tuning and continuous fine tuning; S5. Monitoring the signal-to-noise ratio at the output stage and feeding back the monitoring result to the adaptive band-pass filter and the gain control link to form a two-way coupled closed-loop control.

2. A noise suppression and gain control method for high-frequency signals according to claim 1, characterized in that, The adaptive band-pass filter updates its parameters by using a recursive least squares algorithm, and its objective function is: where \(w(n)\) is the impulse response coefficient vector of the filter, \(w\) * (n) is the optimal impulse response coefficient vector of the filter at the \(n\)-th moment, \(x(i)\) is the \(i\)-th sampling value of the input signal sequence, \(d(i)\) is the \(i\)-th sampling value of the desired output signal sequence, \(\lambda\) is the forgetting factor, \(n\) is the length of the signal sequence, \(w\) T represents the transpose of \(w\), \(\kappa(x(i))\) represents the kurtosis measure of the input signal, \(\Psi(\cdot)\) is the non-linear mapping function for this kurtosis term, \(\Psi(\kappa(x(i)))\) represents the non-linear penalty term, and \(\alpha\) is the weighting factor used to control the weight of the non-linear penalty term in the overall optimization.

3. A noise suppression and gain control method for high-frequency signals according to claim 2, characterized in that The non-linear mapping function Ψ(·) adopts a soft threshold or a Gaussian-type penalty function; when the value of κ(x(i)) increases, the non-linear penalty term increases accordingly, prompting the filter weight update to adjust in the direction of suppressing high-kurtosis components.

4. A noise suppression and gain control method for high-frequency signals according to claim 1, characterized in that The weighted correlation detection calculates the weighted correlation coefficient by using the following formula: where ρ ω is the weighted correlation coefficient, x(i) is the i-th sampling value of the main channel input signal sequence, y(i) is the i-th sampling value of the reference channel signal sequence, ω i is the dynamic weight corresponding to the i-th sampling point, and i is the time series index.

5. A noise suppression and gain control method for high-frequency signals according to claim 4, characterized in that, The dynamic weight ω i is updated as follows: ω i = ω i-1 ·f(κ(x(i)), κ(y(i)), SNR) × g(BW, κ ref ) where κ(x(i)) represents the kurtosis measure of x(i), κ(y(i)) represents the kurtosis measure of y(i), and κ ref is the kurtosis of the reference channel, f(·) is a heuristic function that comprehensively considers the current pulse level, signal-to-noise ratio, and the detection conditions of several past frames, and g(BW, κ ref ) is a quadratic correction function based on the instantaneous bandwidth BW and the kurtosis κ of the reference channel ref .

6. A noise suppression and gain control method for high-frequency signals according to claim 5, characterized in that When the instantaneous bandwidth BW temporarily widens and the kurtosis κ of the reference channel ref is relatively high, the value of g(BW, κ ref ) is less than 1, and the weight is reduced to weaken the impact of burst sampling; when the instantaneous bandwidth BW narrows and the kurtosis κ of the reference channel ref is relatively low, the value of g(BW, κ ref ) is close to or slightly greater than 1, and the weight is increased to enhance the ability to extract weak target signals.

7. A noise suppression and gain control method for high-frequency signals according to claim 1, characterized in that The Newton iteration method based on the second derivative includes: Define the objective function Φ(G) = (SNR target - SNR(G)) 2 , where G is the current gain, SNR target is the target signal-to-noise ratio, and SNR(G) is the actually measured signal-to-noise ratio when the gain value is G; Updating the gain by using the following iteration formula: Where η is a step size scaling factor used to control the convergence speed and stability; L(G) is a linearity compensation function; μ is a compensation weight factor.

8. A noise suppression and gain control method for high-frequency signals according to claim 7, characterized in that The linearity compensation function L(G) is defined as: L(G) = max(0, P out (G) - P IIP3-thre ) 2 where P out (G) represents the estimated output power adjusted with the gain G; P IIP3-thre is the safety power margin reserved according to the third-order intercept point of the amplifier; when the output power approaches the nonlinear region of the amplifier, L(G) and its derivative increase, suppressing the continuous increase of the gain.

9. A noise suppression and gain control method for high-frequency signals according to claim 1, characterized in that The two-way coupled closed-loop control includes: When high pulse noise is detected, increasing the weight factor of the non-linear penalty term to preferentially suppress spikes; When it is detected that the output signal is saturated or too low, correcting through gain iteration and simultaneously adjusting the bandwidth and center frequency of the filter.

10. A noise suppression and gain control method for high-frequency signals according to claim 1, characterized in that, Setting a monitor that combines a Kalman filter and a linear prediction model at the output stage, predicting the signal-to-noise ratio and peak fluctuation at the next moment based on the output of historical frames and the current state; when it is predicted that the signal has an obvious sudden increase trend, pre-adjusting the step attenuator or reducing the gain of the variable gain amplifier to prevent instantaneous saturation.

Images