Non-cooperative signal separation method, device and equipment based on dual reference parameter fusion DSQM, and medium

Abstract

The invention discloses a non-cooperative signal separation method, device and equipment based on dual-reference parameter fusion DSQM and a medium, and relates to the technical field of digital signal processing, and the method comprises the steps: receiving an input signal, and carrying out symbol extension; estimating an original input noise standard deviation, a weak signal and a signal-to-noise ratio, and calculating a right shift number and a dynamic weighted fusion coefficient; performing disturbance injection and phase compensation on the extended signal, generating an actual separation input signal, and estimating a noise standard deviation of the actual separation input signal; calculating a corrected signal-to-noise ratio adaptive shift number k by adopting a dynamic weighted fusion algorithm according to the original input weak signal and the actual separation input signal; calculating a left shift number m; moving the actual separation input signal leftwards for m bits for amplification; sequentially performing n-bit right shift and n-bit left shift to extract strong signal components; separating weak signal components through time domain subtraction; respectively shifting right for m bits to recover the magnitude; and then symmetric saturation truncation and asymmetric truncation are respectively executed, and bit width output is recovered. According to the invention, the signal separation quality is improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of digital signal processing, and in particular to a DSQM non-cooperative signal separation method, device and equipment based on double-reference parameter fusion and a medium. BACKGROUND

[0002] In the electromagnetic sensing system such as external radiation source radar, electronic reconnaissance, anti-jamming communication, the received non-cooperative signal (processing mixed signals without prior signal parameters or cooperative reference signals) usually contains strong direct wave (directly from the transmitting source) and weak echo (target reflection or scattering signal), and the dynamic range difference between the two is more than 60dB. The strong direct wave accounts for more than 90% of the signal energy, and easily covers the weak echo, resulting in loss of target information; the weak echo contains target characteristics (such as distance, speed and angle), but the amplitude is extremely low and is easily submerged by noise and quantization error. Realizing high-precision and faithful separation of strong direct wave and weak echo signals is the basis for subsequent tasks such as synchronous demodulation and reference signal generation.

[0003] In related technologies, the separation technology mainly includes reference antenna method, transform domain separation method, blind source separation method and time domain adaptive filter separation method: The reference antenna method obtains the strong direct wave signal through an independent reference channel, and combines adaptive filtering for cancellation. This method depends on high-isolation antennas, and has high hardware cost; The transform domain separation method uses frequency domain / time-frequency domain (such as FFT / STFT) to suppress the strong signal frequency band. Since directly suppressing the strong signal will cause distortion of the weak signal phase, the phase consistency of this method is difficult to guarantee, and the distortion of the weak signal is serious; The blind source separation method assumes that the strong / weak signals are statistically independent, and realizes signal separation based on ICA / PCA matrix decomposition. The assumption condition of this method is harsh (the strong / weak signals may be related due to multipath effect), and the actual scene performance is unstable, and the iteration convergence is slow, the delay is large, and the real-time performance is insufficient.

[0004] The time domain adaptive filter separation method includes LMS / NLMS algorithm and RLS algorithm, and its principle is to use an adaptive filter to suppress the strong signal and extract the weak signal by adjusting the filter coefficients. However, the filter mechanism of this method cannot adapt to the sudden change of dynamic range with fixed parameter design; the real-time parameter adjustment algorithm has high computational complexity, and is difficult to deploy in low-power hardware, and cannot balance real-time performance and accuracy. SUMMARY

[0005] The present application provides a DSQM non-cooperative signal separation method, device and equipment based on double-reference parameter fusion, which solves the problem of how to improve the signal separation quality.

[0006] To achieve the above purpose, the technical scheme adopted by the present application is as follows: In a first aspect, a DSQM non-cooperative signal separation method based on double-reference parameter fusion is provided, comprising: receiving an input signal, performing symbol extension on the input signal, and filling symbol bits in high bits; estimating original input noise standard deviation, weak signal, and signal-to-noise ratio by statistically calculating extreme values of the undisturbed input signal through a sliding window, and calculating right shift number n and dynamic weighted fusion coefficient; performing disturbance injection and phase compensation on the extended signal to generate an actual separated input signal; estimating noise standard deviation of the actual separated input signal by statistically calculating extreme values of the actual separated input signal through a sliding window; calculating fusion-corrected signal minimum value and noise standard deviation by using a dynamic weighted fusion algorithm according to original input weak signal minimum value, noise standard deviation, and signal-to-noise ratio, and calculating adaptive shift number k of the corrected signal-to-noise ratio in combination with actual separated input signal minimum value and its noise standard deviation; calculating left shift number m according to the right shift number n and the adaptive shift number k of the corrected signal-to-noise ratio; amplifying the actual separated input signal by left shifting m bits; extracting strong signal components by sequentially performing right shift n bits and left shift n bits on the amplified signal; separating weak signal components by time domain subtraction; restoring magnitudes of the strong and weak signal components by right shifting m bits; performing symmetric saturation truncation and asymmetric truncation on the restored strong and weak signal components, respectively, and outputting restored bit width.

[0007] In a second aspect, a DSQM non-cooperative signal separation device based on double-reference parameter fusion is provided, comprising: a signal input and symbol extension module configured to receive an input signal, perform symbol extension on the input signal, and fill symbol bits in high bits; a right shift number and weight coefficient calculation module configured to estimate original input noise standard deviation, weak signal, and signal-to-noise ratio by statistically calculating extreme values of the undisturbed input signal through a sliding window, and calculate right shift number n and dynamic weighted fusion coefficient; an actual separated signal acquisition module configured to perform disturbance injection and phase compensation on the extended signal to generate an actual separated input signal; an actual separated signal calculation module configured to estimate noise standard deviation of the actual separated input signal by statistically calculating extreme values of the actual separated input signal through a sliding window; a signal-to-noise ratio adaptive shift number correction module configured to calculate fusion-corrected signal minimum value and noise standard deviation by using a dynamic weighted fusion algorithm according to original input weak signal minimum value, noise standard deviation, and signal-to-noise ratio, and calculate adaptive shift number k of the corrected signal-to-noise ratio in combination with actual separated input signal minimum value and its noise standard deviation. A left shift number calculation module is configured to calculate a left shift number m according to the right shift number n and the corrected SNR adaptive shift number k. A left shift amplification module is configured to amplify the actual separated input signal by shifting left m bits. A strong signal extraction module is configured to extract a strong signal component by sequentially performing right shift n bits and left shift n bits on the amplified signal. A weak signal separation module is configured to separate a weak signal component by time domain subtraction. A strong and weak signal recovery magnitude module is configured to recover the magnitude of the strong signal component and the weak signal component by right shifting m bits. A truncation output module is configured to perform symmetric saturation truncation and asymmetric truncation on the recovered strong and weak signals, respectively, and output the bit width.

[0008] In a third aspect, an electronic device is provided, which comprises a memory, a processor, and a computer program stored in the memory and executable on the processor, and the computer program, when executed by the processor, implements the steps of the DSQM non-cooperative signal separation method based on dual-reference parameter fusion according to the first aspect.

[0009] In a fourth aspect, a readable storage medium is provided, which stores a program or instructions, and the program or instructions, when executed by a processor, implement the steps of the DSQM non-cooperative signal separation method based on dual-reference parameter fusion according to the first aspect.

[0010] The DSQM non-cooperative signal separation method based on dual-reference parameter fusion has the following beneficial effects: The present application adopts a dual-reference parameter separation calculation and dynamic weighted fusion method to improve the anti-disturbance ability and parameter precision, since the shift parameter n is sensitive to disturbance and k needs to be consistent with the actual demand. The statistical quantities of the original signal and the actual signal are dynamically fused through a sigmoid function, the weight is adaptively adjusted according to the signal-to-noise ratio, and the signal-to-noise ratio drives the weight distribution, thereby solving the shift parameter distortion problem when the DSQM method is combined with injected disturbance suppression, and improving the signal separation quality and precision.

[0011] The device, electronic device, and readable storage medium corresponding to the DSQM non-cooperative signal separation method based on dual-reference parameter fusion can achieve the same technical effects, and thus will not be described here again to avoid repetition. BRIEF DESCRIPTION OF DRAWINGS

[0012] Figure 1 A schematic flowchart of a DSQM non-cooperative signal separation method based on dual-reference parameter fusion is provided for the embodiments of the present application. Figure 2Another schematic flow chart of a DSQM non-cooperative signal separation method based on double reference parameter fusion provided by the embodiment of the present application is shown in FIG. 6. Figure 3 Another schematic flow chart of a DSQM non-cooperative signal separation method based on double reference parameter fusion provided by the embodiment of the present application is shown in FIG. 6. Figure 4 A structural schematic diagram of a DSQM non-cooperative signal separation device based on double reference parameter fusion provided by the embodiment of the present application is shown in FIG. 7. Figure 5 A structural schematic diagram of an electronic device provided by the embodiment of the present application is shown in FIG. 8. DETAILED DESCRIPTION

[0013] For further illustrating the technical means and effects taken by the present application to achieve the predetermined purposes, the technical solutions in the embodiments of the present application are described clearly. Obviously, the described embodiments are some of the embodiments of the present application, but not all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art belong to the scope of protection of the present application.

[0014] The terms "first", "second", etc. in the specification and claims of the present application are used to distinguish similar objects, and are not used to describe a specific order or sequence. It should be understood that the terms used in this way can be interchanged under appropriate circumstances, so that the embodiments of the present application can be implemented in an order other than that illustrated or described herein, and the objects distinguished by "first", "second", etc. are generally a category, not limited to the number of objects, for example, the first object can be one or more. In addition, "and / or" in the specification means at least one of the connected objects, and the character " / ", generally represents that the front and rear associated objects are in an "or" relationship.

[0015] The description of the method flow in the specification and the steps of the flow chart in the drawings of the present application do not necessarily strictly execute the step labels, and the method steps can change the execution order. Moreover, some steps can be omitted, a plurality of steps can be combined into one step for execution, and / or one step can be divided into a plurality of steps for execution.

[0016] In the present specification, a DSQM non-cooperative signal separation method based on double reference parameter fusion, a DSQM non-cooperative signal separation device based on double reference parameter fusion, a computer device, and a computer readable storage medium are provided, which are described in detail one by one below in combination with the drawings and preferred embodiments.

[0017] The application is based on dynamic shift quantization modulation (DSQM) to realize real-time separation of non-cooperative strong and weak signals. Through symbol extension, dynamic parameter calculation, left shift amplification, strong signal extraction, subtraction separation and truncation recovery, high-precision phase-preserving separation of strong and weak signals is realized in pure time domain. The "dynamic shift quantization modulation (DSQM)" refers to a modulation mode for realizing signal energy separation through dynamic shift transformation and quantization bit width expansion transformation. However, when optimizing the DSQM separation performance through perturbation injection, noise shaping, phase pre-compensation and other techniques, the perturbation and compensation operations change the statistical characteristics of the signals actually participating in the separation, and then the shift parameter value calculation may have errors. Based on this, the present application aims to solve the shift parameter distortion problem caused by the perturbation and compensation operations through a hybrid reference parameter correction mechanism, and has high precision and low complexity.

[0018] Please refer to Figures 1-2 The embodiment of the present application provides a DSQM non-cooperative signal separation method based on double reference parameter fusion, as shown in Figures 1-2 , comprising: Step S1, receiving an input signal, performing symbol extension on the input signal, and filling symbol bits in high bits.

[0019] This step performs symbol extension on the input signal to generate a high-bit-width extended signal. The input is a low-bit-width B in original signal data, and the output is a high-bit-width B ext extended signal data. The bit width of the input signal is expanded from Bin to Bext, and the symbol bits are filled in the high bits (for example, 16-bit input is expanded to 32-bit, and symbol bits are filled in the high bits). With signed extension, specifically, the value -328 (16 bits) is expanded to 0xFFFFFEB8 (32 bits). Based on this, the original symbol and precision can be retained, and overflow during shifting can be avoided. Subsequent left shift amplification, strong signal extraction, and weak signal separation are all operated on the expanded bit width.

[0020] Step S2, estimating the original weak signal minimum value, noise standard deviation and signal-to-noise ratio of the original input signal through sliding window statistical extreme value, and calculating the right shift number n and the weighted fusion coefficient.

[0021] In this step, the input signal without perturbation can be the original input signal. In some possible embodiments, it can also be the extended signal data obtained after symbol extension.

[0022] Further, step S2 includes: Step S21, calculating the signal maximum value and minimum value through sliding window statistics, calculating the noise standard deviation of the original input signal through the interquartile range method, and determining the original input weak signal.

[0023] Based on the sliding window statistical maximum Amax_org and minimum Amin_win, the original input signal noise standard deviation σnoise_org is calculated by the piecewise sorting network combined with the interquartile range method, and the initial weak signal amplitude Amin_org = max(Amin_win, 3σnoise_org) is tracked based on the 3σ cut-off principle.

[0024] Step S22, the original input signal-to-noise ratio is calculated according to the original input weak signal and the original input noise standard deviation.

[0025] The original input signal-to-noise ratio SNR_org is calculated based on the original input weak signal amplitude Amin_org and the original input noise standard deviation σnoise_org: SNR_org = MSB(Amin_org) - MSB(σnoise_org).

[0026] Step S23, the right shift number n is calculated according to the signal maximum and the original input weak signal.

[0027] In this step, the right shift number n is the strong signal compression bit number; the right shift of n bits should ensure that the strong signal is fully compressed while retaining the information of the weak signal; the right shift number n is calculated according to the following formula: n = floor(log2(Amax_org / Amin_org)); In the FPGA implementation, it is calculated by looking up table (LUT) or shift approximation. An example is to extract the MSB of the signal maximum Amax_org and the original input weak signal Amin_org by priority encoder, and calculate the right shift number by difference: n = MSB(Amax_org) - MSB(Amin_org).

[0028] Step S24, based on the original input signal-to-noise ratio, the dynamic weighted fusion coefficient is generated by using the sigmoid function dynamic fusion principle of the statistics of the original signal and the actual signal; the weight is adaptively adjusted with the original input signal-to-noise ratio; that is, low signal-to-noise ratio depends on the original statistics, and high signal-to-noise ratio focuses on the actual signal.

[0029] Specifically, the dynamic weighted fusion coefficient function is: w = 1 / (1 + exp(-a·(SNR_org - b)), physical meaning: signal-to-noise ratio drives weight allocation, low SNR depends on original statistics (w ≈ 1), high SNR focuses on actual signal (w ≈ 0). Wherein, a reflects the weight change slope, b reflects the SNR good and bad demarcation point, which can be selected and configured according to the actual situation; typical a = 1, b = 12 dB, indicating that the weight change slope is -45°, and the SNR < 12 dB starts to deteriorate. SNR_org weight w physical meaning 6dB 0.95 low SNR, rely on original statistics 12dB 0.5 medium SNR, balanced fusion 18dB 0.05 high SNR, focus on actual signal In implementation, the hardware look-up table is used to realize the SNR_org query weight coefficient w table with a and b as entries, so as to obtain the weight coefficient in real time; 256 groups of (a, b, w) parameters are pre-stored, and 20 LUTs are occupied; (optionally) low SNR forced mode: when SNR_org < 10 dB, w = 0.8, and the original signal statistics are preferentially relied on.

[0030] In step S3, the perturbation injection and phase compensation are performed on the expanded signal to generate the actual separated input signal.

[0031] In this step, the perturbation injection and phase compensation on the expanded signal are operations for enhancing the subsequent DSQM separation performance, and the purpose is to enhance the noise suppression and improve the separation phase precision.

[0032] In some possible embodiments, the perturbation injection includes: injecting a first perturbation d1 after the symbol expansion, and the amplitude of the first perturbation d1 is |d1| = clamp(2^(m-n), 1, Amin / 8, and the bit width is low log2(Amin)+3. Wherein, m is the left shift amplification bit number, which can be quickly estimated and calculated in the previous time sequence or cycle. Amin is the fast statistical signal minimum value. Exemplarily: the maximum value of the signal in the sliding window is statistically obtained, and a set of dynamic parameters are generated according to the maximum value, including the right shift number n, the SNR adaptive shift number k and the left shift amplification bit number m; n = floor(log2(Amax / max(Amin, ε))) Wherein, Amax is the maximum value of the signal, that is, the strong signal amplitude, Amin is the minimum value, that is, the weak signal amplitude + noise, and ε is a configurable greater than 0 small amount, which avoids the division by 0 operation, for example, 1, that is, n = floor(log2(Amax / max(Amin, 1))); k = ceil(log2(SNR+1)) = ceil(log2(SNR dB / 3)); SNR dB = 20log 10 (Amax / Amin); m = clamp(n+k, 3, B), Wherein, B = B ext -B in (overflow protection bit number), B in represents the low bit width original signal data, and B ext represents the high bit width expanded signal data.

[0033] In some possible implementations, the phase compensation is pre-compensating the phase angle θ(n) of the signal according to the n value through the CORDIC algorithm, where: θ(n)=0.15n° (n≤10) or θ(n)=0.18n° (n>10). The specific hardware implementation is through the CORDIC pre-compensation module: inserted after the symbol extension, supporting segmented dynamic compensation; 3-stage pipelined CORDIC: precision ±0.02°, resource occupation 60 LUTs; dynamic parameter binding: the n value comes from the previous dynamic parameter calculation module.

[0034] Step S4, the actual separation input signal extreme value is counted through a sliding window, and the noise standard deviation thereof is estimated.

[0035] Specifically, the noise standard deviation σnoise_fact of the actual separation input signal is calculated through the interquartile range method of the segmented sorting network based on the actual separation input signal minimum value Amin_fact counted through the sliding window.

[0036] Step S5, according to the original input weak signal minimum value, the noise standard deviation, and the signal-to-noise ratio, the dynamic weighted fusion algorithm is used to calculate the fused and corrected signal minimum value and noise standard deviation, and to calculate the adaptive shift number k of the corrected signal-to-noise ratio, in combination with the actual separation input signal minimum value and the noise standard deviation thereof.

[0037] This step calculates the fused and corrected signal minimum value and noise standard deviation, and calculates the adaptive shift number k of the corrected signal-to-noise ratio, according to the original input weak signal minimum value Amin_org, the noise standard deviation w·σnoise_org, and the signal-to-noise ratio SNR_org, in combination with the actual separation input signal minimum value Amin_fact and the noise standard deviation σnoise_fact of the actual separation input signal minimum value, using the dynamic weighted fusion algorithm; The dynamic weighted fusion is performed through the following formula: Amin=w·Amin_org+(1−w)·max(Amin_fact,3σnoise_fact) σnoise=w·σnoise_org+(1−w)·σnoise_fact, wherein the weight w is the dynamic weighted fusion coefficient calculated as follows: w=1 / (1+exp(−a·(SNR_org−b)), a=1, b=12 dB; The step forms a mechanism for signal-to-noise ratio driving weight distribution, which relies on original statistics (w≈1) at low SNR and focuses on actual signal (w≈0) at high SNR; a reflects the weight change slope, and b reflects the SNR advantage and disadvantage demarcation point, which can be selected and configured according to actual conditions; typically a=1, b=12dB, indicating that the weight change slope is-45°, and SNR<12dB starts to deteriorate, as shown in the following table. In specific implementation, an SNR_org query weight coefficient w table is established with a and b as entries, to provide real-time acquisition of weight coefficients, and 256 groups of (a, b, w) parameters are pre-stored, occupying 20 LUTs. In some possible implementations, a low signal-to-noise ratio forced mode can also be configured: when SNR_org<10dB, w=0.8, to preferentially rely on original signal statistics. SNR_org weight w physical meaning 6dB 0.95 low SNR, rely on original statistics 12dB 0.5 medium SNR, balanced fusion 18dB 0.05 high SNR, focus on actual signal The corrected signal-to-noise ratio adaptive shift number k is: k=ceil((MSB(Amin)-MSB(σnoise)) / 3)+1, where "+1" is an MSB() approximate calculation SNR error compensation term.

[0038] Step S6, calculating the left shift number m according to the right shift number n and the corrected signal-to-noise ratio adaptive shift number k.

[0039] The left shift number m is calculated by the following formula: m=clamp(n+k,3,B), where B=B ext -B in (B overflow protection bits), B in represents low-bit-width original signal data, and B ext represents high-bit-width extended signal data.

[0040] Further, based on the left shift number m, the right shift number n, and the signal-to-noise ratio adaptive coefficient k, a DSQM (Dynamic Shift Quantization Modulation) strong and weak signal separation step is performed. It includes: Step S7, amplifying the actual separated input signal by left shifting m bits; Specifically, the signal is left shifted by m bits with the sign bit unchanged, and the bit width is not expanded (still Bext), the low bits are zero-padded, and the amplified signal=(extended signal<<m), and the output amplified signal data. The left shifted amplified signal is amplified by 2^m times, so that the weak signal is greater than the quantization noise and the processability is enhanced. The left shift number m is determined by dynamic calculation to ensure that the weak signal is sufficiently amplified but does not overflow.

[0041] Step S8, sequentially performing right shift n bits and left shift n bits on the amplified signal to extract the strong signal component.

[0042] Based on the amplified signal data, the strong signal component is approximately extracted by sequentially right and left shifting, and the phase is preserved (which can ensure that the error is less than 0.1%). The amplified signal data is right shifted by n bits and then left shifted by n bits, the low n bits are discarded, and the phase information remains unchanged due to linear shifting; the amplified strong signal = (amplified signal >> n) << n), and the amplified strong signal data is output. The strong signal compression bit number n is determined by dynamic calculation to ensure that the strong signal is compressed to the extent of the weak signal.

[0043] In step S9, the weak signal component is separated by time domain subtraction of the amplified signal and the strong signal component.

[0044] Based on the time domain difference operation to separate the weak signal, the amplified weak signal data is output by inputting the amplified signal data and the amplified strong signal data, and the amplified weak signal = (amplified signal) - (amplified strong signal data). Since subtraction is a linear operation, the phase error can be ensured to be less than 1° In step S10, the separated strong signal component and weak signal component are right shifted by m bits to restore the order of magnitude.

[0045] In this step, the right shift by m bits offsets the amplification effect of the left shift by m bits, and the restored order of magnitude strong signal data and the restored order of magnitude weak signal data are output by inputting the amplified strong signal data and the amplified weak signal data; the restored order of magnitude strong signal = (amplified strong signal data >> m), and the restored order of magnitude weak signal = (amplified weak signal data >> m).

[0046] In step S11, symmetric saturation truncation and asymmetric truncation are respectively performed on the restored order of magnitude strong signal and weak signal, and the restored bit width is output.

[0047] For the restored order of magnitude strong signal data, the symmetric truncation is the original bit width. The symmetric saturation truncation is, for example, 16-bit output range of -32767~32767. The restored bit width strong signal data is output.

[0048] For the restored order of magnitude weak signal data, the asymmetric truncation is the original bit width. The asymmetric truncation preserves the negative details, such as sign extension truncation. The restored bit width weak signal data is output.

[0049] Referring to Figure 3 Based on the above scheme, the application adopts a double-reference parameter separation calculation and dynamic weighted fusion method to improve the anti-disturbance ability and parameter precision, in which n is sensitive to disturbance and k needs to be fitted to actual demand. The statistical quantities of the original signal and the actual signal are dynamically fused through a sigmoid function, the weight is adaptively adjusted according to the signal-to-noise ratio, the signal-to-noise ratio drives the weight distribution, the original statistics are relied on at low SNR (w≈1), and the actual signal is focused on at high SNR (w≈0). According to the data display, the n error of the application is less than or equal to 5% (d1=200), the k error is less than or equal to 1 dB, and the phase error is less than or equal to 0.5°.

[0050] By way of example, the following provides specific application examples for illustration: Example 1: External source radar signal separation (1) Input signal: Strong direct wave: 16 bits, Amax_org = 32767; Weak echo: Amin_org = 500 (SNR_org = 12 dB), disturbance d1 = 200.

[0051] (2) Dynamic parameters: Traditional method: Amin_win = 700 after disturbance → n = 6 (error 40%); Invention: original Amin_org = 500 → n = 8, weighted k = 5 → m = 13; (3) Effect comparison: index conventional method the present invention improvement effect n value error 40% 3% -37% weak signal SNR improvement 30dB 45dB +15dB resource occupation 600 LUTs 620 LUTs +3.3% Example 2: 5G communication anti-jamming (1) Input signal: Strong interference: Amax_org = 30000, adjacent frequency noise -30dBm; Weak user signal: Amin_org = 300 (SNR_org = 10 dB).

[0052] (2) Dynamic parameters: Traditional method: fixed compensation k = 4 → m = 10, error rate 1e-3; Invention: dynamically corrected k = 5 → m = 11, error rate 1e-6; Effect comparison: index conventional method the present invention improvement effect bit error rate (BER) 1e-3 1e-6 reduce by 3 orders of magnitude processing delay 100ns 85ns -15% Example 3: Medical ultrasound imaging (1) Input signal: Strong tissue reflection: 24 bits, Amax_org = 8,388,607; Weak blood flow signal: Amin_org = 1,000 (SNR_org = 60 dB).

[0053] (2) Dynamic parameters: Traditional method: n = 11 (error 15%), k = 4 → m = 15; Invention: n = 13 (error 0%), k = 4 → m = 17; Effect comparison: index conventional method the present invention improvement effect quantization noise suppression 40dB 66dB +26dB image contrast 30:1 60:1 improve by 100% In specific implementation, the application multiplexes hardware modules to reduce resource overhead and optimize timing, and has the following advantages: 1. High precision: n error ≤5% (d1 = 200), k error ≤1dB, phase error ≤0.5°; 2. Low latency: 85 ns end-to-end latency, supporting real-time processing; 3. Strong compatibility: seamlessly adapting to the previous DSQM architecture, extending the dynamic range to 120 dB; 4. Dynamic adaptability: supporting 80~120 dB dynamic range, adapting to burst signal scenarios.

[0054] Referring to Figure 4 Corresponding to the above-mentioned dual-reference parameter fusion DSQM non-cooperative signal separation method embodiment, the embodiment of the application provides a dual-reference parameter fusion DSQM non-cooperative signal separation device, comprising: A signal input and symbol extension module 1001 is configured to perform symbol extension on an input signal, and fill symbol bits in high bits; A right shift number and weight coefficient calculation module 1002 is configured to estimate the original input noise standard deviation, weak signal and signal-to-noise ratio by calculating the extreme value of the input signal without disturbance through a sliding window, and calculate the right shift number n and dynamic weighted fusion coefficient; An actual separation signal acquisition module 1003 is configured to perform disturbance injection and phase compensation on the extended signal and generate an actual separation input signal; An actual separation signal calculation module 1004 is configured to estimate the noise standard deviation of the actual separation input signal by calculating the extreme value of the actual separation input signal through a sliding window; A signal-to-noise ratio adaptive shift number correction module 1005 is configured to calculate the fusion-corrected signal minimum value and noise standard deviation by using a dynamic weighted fusion algorithm according to the original input weak signal minimum value, noise standard deviation and signal-to-noise ratio, combining the actual separation input signal minimum value and its noise standard deviation, and calculate the corrected signal-to-noise ratio adaptive shift number k; A left shift number calculation module 1006 is configured to calculate the left shift number m according to the right shift number n and the corrected signal-to-noise ratio adaptive shift number k; A left shift amplification module 1007 is configured to left shift the actual separation input signal by m bits for amplification; A strong signal extraction module 1008 is configured to sequentially perform right shift n bits and left shift n bits on the amplified signal to extract a strong signal component; A weak signal separation module 1009 is configured to separate a weak signal component through time domain subtraction; A strong and weak signal recovery magnitude module 1010 is configured to right shift the strong signal component and the weak signal component by m bits to recover the magnitude; A truncation output module 1011 is configured to perform symmetric saturation truncation and asymmetric truncation on the recovered strong and weak signals, respectively, and output the recovered bit width.

[0055] The above-mentioned dual-reference parameter fusion DSQM non-cooperative signal separation device realizes the steps and various processes of the above-mentioned dual-reference parameter fusion DSQM non-cooperative signal separation method embodiment, and can achieve the same technical effects. To avoid repetition, details are not repeated here.

[0056] Referring to Figure 5 Corresponding to the above-mentioned dual-reference parameter fusion DSQM non-cooperative signal separation method embodiment, the embodiment of the present application provides an electronic device, which comprises a memory, a processor and a computer program stored in the memory and executable on the processor. The computer program is executed by the processor to realize the steps and various processes of the above-mentioned dual-reference parameter fusion DSQM non-cooperative signal separation method embodiment, and can achieve the same technical effects. To avoid repetition, details are not repeated here.

[0057] The memory 1012 can be used to store software programs and various data. The memory 1012 can mainly include a first storage area storing programs or instructions and a second storage area storing data, wherein the first storage area can store an operating system, application programs or instructions required by at least one function (such as a sound playing function, an image playing function, etc.), etc. In addition, the memory 1012 can include a volatile memory or a non-volatile memory, or the memory 1012 can include both volatile and non-volatile memories. The non-volatile memory can be a Read-Only Memory (ROM), a Programmable ROM (PROM), an Erasable Programmable ROM (EPROM), an Electrically EPROM (EEPROM) or a flash memory. The volatile memory can be a Random Access Memory (RAM), a Static RAM (SRAM), a Dynamic RAM (DRAM), a Synchronous DRAM (SDRAM), a Double Data Rate SDRAM (DDR SDRAM), an Enhanced SDRAM (ESDRAM), a Synchlink DRAM (SLDRAM) and a Direct Rambus RAM (DRRAM). The memory 1012 in the embodiment of the present application includes but is not limited to these and any other suitable types of memory.

[0058] The processor 1013 can include one or more processing units; alternatively, the processor 1013 integrates an application processor and a modem processor, wherein the application processor mainly processes operations related to operating systems, user interfaces, and application programs, and the modem processor mainly processes wireless communication signals, such as a baseband processor. It can be understood that the above-mentioned modem processor can also not be integrated into the processor 1013.

[0059] Corresponding to the above-mentioned DSQM non-cooperative signal separation method embodiment based on double reference parameter fusion, the embodiment of the present application also provides a readable storage medium, the readable storage medium stores a program or instructions, the program or instructions are executed by a processor to realize the steps of the above-mentioned DSQM non-cooperative signal separation method embodiment based on double reference parameter fusion and the various processes of the embodiment, and the same technical effects can be achieved. To avoid repetition, this will not be repeated here.

[0060] The processor is the processor in the electronic device described in the above-mentioned embodiments of the present application. The readable storage medium includes a computer readable storage medium, such as a computer readable only memory ROM, a random access memory RAM, a magnetic disk or an optical disk, etc.

[0061] It should be noted that in this paper, the term "includes", "contains" or any other variant thereof is intended to cover non-exclusive inclusion, so that the process, method, article or device including a series of elements not only includes those elements, but also includes other elements not explicitly listed, or includes elements inherent to such process, method, article or device. Without more limitations, the element defined by the statement "includes a" does not exclude the presence of another identical element in the process, method, article or device including the element. In addition, it should be pointed out that the scope of the method and device in the present application is not limited to the order of the functions shown or discussed, but also includes the functions performed in a substantially simultaneous manner or in the opposite order, for example, the described method can be performed in an order different from the described order, and various steps can also be added, omitted or combined. In addition, the features described with reference to some examples can be combined in other examples.

[0062] Through the description of the above embodiments, those skilled in the art can clearly understand that the above-mentioned example methods can be realized by means of software and a necessary general hardware platform, and of course, can also be realized by hardware, but in many cases, the former is a better embodiment. Based on such understanding, the technical solutions of the present application can be embodied in the form of a computer software product in essence or in the form of a part of the prior art that contributes to the present application. The computer software product is stored in a storage medium (such as a ROM / RAM, a magnetic disc, an optical disc), and includes a plurality of instructions for causing a terminal (which can be a mobile phone, a computer, a server, or a network device, etc.) to execute the methods described in the various embodiments of the present application.

[0063] It can be understood that the embodiments of the present application are described above in combination with the drawings, but the present application is not limited to the above-mentioned specific embodiments, and the above-mentioned specific embodiments are only illustrative and not limiting. Those skilled in the art know that various changes or equivalent replacements can be made to these features and embodiments without departing from the spirit and scope of the present application. In addition, those skilled in the art can modify these features and embodiments to adapt to specific conditions and materials under the inspiration or teaching of the present application without departing from the spirit and scope of the present application. Therefore, the present application is not limited by the specific embodiments disclosed herein, and all embodiments falling within the scope of the claims of the present application are within the scope of protection of the present application.

Claims

1. A non-cooperative signal separation method based on dual-reference parameter fusion DSQM, characterized in that, include: Receive the input signal, perform sign extension on the input signal, and fill the high-order bits with the sign bit; For an undisturbed input signal, extreme values ​​are statistically analyzed using a sliding window to estimate the standard deviation of the original input noise, weak signals, and signal-to-noise ratio, and the right-shifted number n and dynamic weighted fusion coefficients are calculated. Perturbation injection and phase compensation are performed on the extended signal to generate the actual separated input signal; The noise standard deviation is estimated by statistically analyzing the extreme values ​​of the actual separated input signal using a sliding window. Based on the original weak signal minimum, noise standard deviation and signal-to-noise ratio, and combined with the actual separated input signal minimum and noise standard deviation, a dynamic weighted fusion algorithm is used to calculate the fused corrected signal minimum and noise standard deviation, and to calculate the adaptive shift k of the corrected signal-to-noise ratio; The left shift number m is calculated based on the right shift number n and the corrected signal-to-noise ratio adaptive shift number k. The actual separated input signal is shifted left by m bits and amplified. The amplified signal is sequentially shifted right by n bits and left by n bits to extract the strong signal components. Time-domain subtraction separates weak signal components; Shift the strong signal component and the weak signal component to the right by m bits to restore their magnitude; Symmetric saturation truncation and asymmetric truncation are performed on strong and weak signals of the recovery magnitude, respectively, to restore the bit width output.

2. The method for separating non-cooperative signals based on dual-reference parameter fusion DSQM according to claim 1, characterized in that, For the undisturbed input signal, the extreme values ​​are statistically analyzed using a sliding window to estimate the standard deviation of the original input noise, weak signals, and signal-to-noise ratio. The right-shifted bits n and weighted fusion coefficients are then calculated, including: The original weak input signal is determined by statistically analyzing the maximum and minimum values ​​of the signal using a sliding window and calculating the noise standard deviation of the original input signal using the interquartile range method. Calculate the original input signal-to-noise ratio based on the original weak input signal and the original input noise standard deviation; Calculate the right shift number n based on the maximum signal value and the original weak input signal; Based on the statistical principle of dynamically fusing the original signal and the actual signal using the sigmoid function according to the original input signal-to-noise ratio, dynamic weighted fusion coefficients are generated; the weights are adaptively adjusted according to the original input signal-to-noise ratio.

3. The method for separating non-cooperative signals based on dual-reference parameter fusion DSQM according to claim 2, characterized in that, The number of right shifts, n, is calculated using the following formula: n=floor(log2(Amax_org / Amin_org)), Where Amax_org is the signal maximum and Amin_org is the original weak input signal.

4. The DSQM non-cooperative signal separation method based on dual-reference parameter fusion according to claim 2, characterized in that, The dynamic weighted fusion coefficient function is: w=1 / (1+exp(−a·(SNR_org−b)), Among them, 'a' reflects the slope of the weight change, and 'b' reflects the dividing point between good and bad SNR. The configuration can be selected according to the actual situation.

5. The method for separating non-cooperative signals based on dual-reference parameter fusion DSQM according to claim 2, characterized in that, The step of calculating the fused corrected signal minimum and noise standard deviation using a dynamic weighted fusion algorithm based on the original weak input signal minimum, noise standard deviation, and signal-to-noise ratio, combined with the actual separated input signal minimum and noise standard deviation, and calculating the adaptive shift number k of the corrected signal-to-noise ratio, includes: Dynamic weighted fusion is achieved using the following formula: Amin=w·Amin_org+(1−w)·max(Amin_fact,3σnoise_fact) σnoise=w·σnoise_org+(1−w)·σnoise_fact, Wherein, the weight w is the dynamic weighted fusion coefficient; Amin_org is the minimum value of the original weak input signal, w·σnoise_org is the standard deviation of the noise of the original weak input signal, SNR_org is the signal-to-noise ratio of the original weak input signal, Amin_fact is the minimum value of the actual separated input signal, and σnoise_fact is the standard deviation of the noise of the minimum value of the actual separated input signal.

6. The method for separating non-cooperative signals based on dual-reference parameter fusion DSQM according to claim 5, characterized in that, The corrected signal-to-noise ratio adaptive shift number k is: k=ceil((MSB(Amin)−MSB(σnoise)) / 3)+1.

7. The method for separating non-cooperative signals based on dual-reference parameter fusion DSQM according to claim 6, characterized in that, The number of left shifts, m, is calculated using the following formula: m = clamp(n + k, 3, B), Where B=B ext -B in B in B represents the low-bit width raw signal data. ext This indicates high-bit-width extended signal data.

8. A non-cooperative signal separation device based on dual-reference parameter fusion DSQM, characterized in that, include: The signal input and sign extension module is used to receive input signals, extend the input signals by sign, and fill the high-order bits with sign bits; The right shift bit and weighting coefficient calculation module is used to estimate the original input noise standard deviation, weak signal and signal-to-noise ratio by statistically analyzing the extreme values ​​through a sliding window for the undisturbed input signal, and to calculate the right shift bit n and dynamic weighted fusion coefficient; The actual separation signal acquisition module is used to perform perturbation injection and phase compensation on the extended signal and generate the actual separation input signal; The actual separated signal calculation module is used to calculate the extreme values ​​of the actual separated input signal through a sliding window and estimate its noise standard deviation. The signal-to-noise ratio adaptive shift correction module is used to calculate the fused corrected signal minimum and noise standard deviation based on the original weak input signal minimum, noise standard deviation and signal-to-noise ratio, combined with the actual separated input signal minimum and noise standard deviation, using a dynamic weighted fusion algorithm, and to calculate the corrected signal-to-noise ratio adaptive shift k. The left shift calculation module is used to calculate the left shift number m based on the right shift number n and the corrected signal-to-noise ratio adaptive shift number k. The left-shift amplification module is used to shift the actual separated input signal to the left by m bits and amplify it. The strong signal extraction module is used to extract the strong signal component by sequentially shifting the amplified signal right by n bits and left by n bits. A weak signal separation module is used for time-domain subtraction to separate weak signal components. The strong and weak signal recovery level module is used to right-shift the strong signal component and the weak signal component by m bits to recover the level. The truncation output module is used to perform symmetrical saturation truncation and asymmetrical truncation on strong and weak signals of the recovery magnitude, respectively, to restore the bit width output.

9. An electronic device, characterized in that, The electronic device includes: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the computer program, when executed by the processor, implements the steps of the DSQM non-cooperative signal separation method based on dual reference parameter fusion as described in any one of claims 1 to 6.

10. A readable storage medium, characterized in that, The readable storage medium stores a program or instructions that, when executed by a processor, implement the steps of the DSQM non-cooperative signal separation method based on dual reference parameter fusion as described in any one of claims 1 to 6.

Images