A giant magnetoresistance sensing signal enhancement method based on multi-channel fusion

Abstract

This application relates to a method for enhancing giant magnetoresistive (GMR) sensor signals based on multi-channel fusion. The method includes: calibrating the complex domain frequency response of each GMR channel to obtain discrete complex sampled values ​​of differential-mode sensitivity and common-mode response transfer function at a preset frequency set; acquiring multiple original digital signal streams, and obtaining complex spectral data for each channel and frequency slot through framing and short-time frequency domain transformation; obtaining estimated values ​​of the common-mode transfer function of each channel relative to a reference channel through frame-by-frame adaptive blind identification; obtaining a compensated multi-channel time-domain signal sequence through frequency domain consistency compensation, inverse transformation, and time-domain reconstruction; and finally, outputting the enhanced GMR sensor signal through statistical optimal weighted fusion and sparse domain enhancement. This method can accurately suppress common-mode interference, align the frequency domain characteristics of multiple channels, and improve the signal-to-noise ratio and detection sensitivity of the GMR sensor signal.

Description

Technical Field

[0001] This invention belongs to the field of magnetic field measurement and weak signal detection technology, and in particular relates to a method for enhancing giant magnetoresistive sensing signals based on multi-channel fusion. Background Technology

[0002] Giant magnetoresistive (GMR) sensors have found wide application in fields such as magnetic anomaly detection, biomagnetic field detection, non-destructive testing, and geological exploration due to their high sensitivity, small size, and good frequency response characteristics. In practical engineering deployments, to suppress the smothering effect of far-field electromagnetic interference on target signals, multiple GMR sensing elements are typically arranged in space as an array with a specific baseline length. The principle of differential or gradient measurement is used to cancel out the approximately identical common-mode background magnetic field at adjacent elements. The effectiveness of this approach relies on a fundamental assumption: that each GMR channel has sufficiently consistent and predictable response characteristics to the same common-mode magnetic field, allowing the common-mode component to be sufficiently eliminated after differential operations.

[0003] However, this fundamental assumption has never truly held true in actual hardware systems. Each giant magnetoresistive (GMR) channel consists of a GMR Wheatstone bridge, a front-end instrumentation amplifier, an anti-aliasing filter, and an analog-to-digital conversion link. Its common-mode rejection capability is quantified by the key indicator, the common-mode rejection ratio (CMRR). Influenced by multiple factors, including the microscopic inhomogeneity of the magnetic properties of the GMR multilayer film material, the process mismatch of the bridge arm resistors, the differences in the distribution of parasitic capacitance and inductance on the printed circuit board, and the discreteness of the transistor parameters in the amplifier input stage, the CMRR of different GMR channels not only differs under static DC conditions but also exhibits unique frequency-dependent characteristics within their respective operating frequency bands. In other words, the CMRR is a complex function that varies with frequency, and the degree of magnitude decay with frequency and the phase lag characteristic of its argument with frequency are different between channels. When broadband electromagnetic interference with rich spectral components exists in the environment, the residual common-mode noise at the output of each channel will exhibit a complex and inconsistent relationship in terms of amplitude and phase.

[0004] Traditional differential fusion algorithms, when dealing with inconsistent residual common-mode noise between channels, not only fail to effectively cancel it, but also, due to phase distortions and amplitude differences among the frequency components, mistakenly combine these noise components that should be suppressed into a spurious gradient signal that appears to have clear time-frequency characteristics. If any weak signal enhancement modules based on energy accumulation, matched filtering, or time-frequency analysis are subsequently applied to the differential result, this spurious gradient signal will be amplified as the real target signal, ultimately leading to a sharp increase in the false alarm rate and a decrease in the signal-to-noise ratio (SNR). Current technologies for improving the common-mode rejection ratio (CMRR) mainly focus on improving the static balancing accuracy of giant magnetoresistive bridges, using instrumentation amplifiers with higher CMRR specifications, and adding high-order common-mode filtering stages in the analog domain. While these methods can improve the CMRR of a single channel at a specific frequency to some extent, they cannot solve the systemic problem of inconsistent complex CMRR frequency response curves across the entire operating frequency band. Furthermore, most existing digital post-compensation methods assume that there is only a fixed multiple of amplitude mismatch between channels, while ignoring phase mismatch, which is also a key factor that leads to the generation of false signals. They also fail to provide a closed-loop processing mechanism that can track changes in the environmental interference spectrum in real time during the measurement process and dynamically adjust the compensation strategy. Summary of the Invention

[0005] Therefore, it is necessary to provide a method for enhancing giant magnetoresistive sensing signals based on multi-channel fusion, which can eliminate false gradient signals caused by channel inconsistency through dynamic identification and adaptive compensation, and enhance the real weak gradient signals through multi-channel optimal fusion.

[0006] In a first aspect, this application provides a method for enhancing giant magnetoresistive sensing signals based on multi-channel fusion, including:

[0007] S1. Perform complex domain frequency response calibration on each giant magnetoresistive channel to obtain the discrete complex sampled values ​​of the differential-mode sensitivity transfer function and common-mode response transfer function of each giant magnetoresistive channel on a preset frequency set.

[0008] S2. Acquire multiple raw digital signal streams collected during the measurement phase of each giant magnetoresistive channel. Based on the discrete complex sampling values, perform frame division and short-time frequency domain transformation on the multiple raw digital signal streams to obtain the complex spectrum data of each giant magnetoresistive channel in each frequency slot.

[0009] S3. Based on the complex spectrum data of each giant magnetoresistive channel in each frequency slot, the common-mode transfer function between channels is adaptively blindly identified frame by frame to obtain the estimated value of the common-mode transfer function of each giant magnetoresistive channel in each frequency slot relative to the reference channel in each frame of complex spectrum data.

[0010] S4. Based on the common-mode transfer function estimate, perform consistency compensation on the complex spectrum data of each giant magnetoresistive channel in the frequency domain to obtain compensated complex spectrum data. Then, perform inverse transformation and time-domain reconstruction on the compensated complex spectrum data to obtain a compensated multi-channel time-domain signal sequence.

[0011] S5. Perform statistical optimal weighted fusion on the compensated multi-channel time-domain signal sequence to obtain a linear preliminary estimation sequence of gradient signals in the multiple original digital signal streams; based on the phase consistency of the multi-channel, perform sparse domain enhancement on the linear preliminary estimation sequence to obtain an enhanced gradient signal time series; the enhanced gradient signal time series is used as the enhanced giant magnetoresistive sensing signal output.

[0012] The aforementioned multi-channel fusion-based giant magnetoresistive (GMR) sensor signal enhancement method, by performing complex domain frequency response calibration on each GMR channel, obtains discrete complex sampled values ​​of the differential-mode sensitivity transfer function and common-mode response transfer function of each GMR channel on a preset frequency set. This enables precise complex domain quantization characterization of the full-band frequency domain response characteristics of each GMR channel, solving the problems of incomplete channel characteristic characterization and lack of accurate frequency domain reference in subsequent signal processing caused by traditional GMR channel calibration focusing only on single-frequency amplitude characteristics and ignoring phase information and complex domain response characteristics. By performing frame segmentation and short-time frequency domain transformation on multiple original digital signal streams to generate complex spectrum data, and then performing frame-by-frame adaptive blind identification based on the channel frequency domain data to obtain the common-mode transfer function estimate, it can achieve accurate time-frequency domain conversion of the original acquired signal and dynamic adaptive tracking of the common-mode transfer function between channels. This solves the problems of traditional fixed-parameter identification failing to adapt to the time-varying characteristics of signals and lacking precision in common-mode interference suppression. This paper addresses the problem of incomplete interference suppression caused by the inter-channel transfer function. It compensates for the frequency domain consistency of complex spectral data by estimating the common-mode transfer function, followed by inverse transformation and time-domain reconstruction to generate a compensated multi-channel time-domain signal sequence. This achieves precise alignment of the multi-channel frequency domain response and root-cause elimination of common-mode interference, resolving issues such as signal mismatch, residual common-mode interference, and submersion of effective signals caused by inconsistent frequency domain characteristics between channels. Furthermore, it generates a linear preliminary estimation sequence by statistically optimal weighted fusion of the compensated multi-channel time-domain signal sequence, and then enhances the linear preliminary estimation sequence in the sparse domain based on multi-channel phase consistency, generating an enhanced gradient signal time series. This achieves optimal convergence of effective signals from multiple channels and deep enhancement of weak gradient signals, solving the problems of traditional single-channel enhancement methods failing to utilize multi-channel correlation characteristics, insufficient weak gradient signal extraction capabilities, and limited signal enhancement effects. This improves the signal-to-noise ratio and detection sensitivity of giant magnetoresistive sensing signals. Attached Figure Description

[0013] To more clearly illustrate the technical solutions in the embodiments or related technologies of this application, the accompanying drawings used in the description of the embodiments or related technologies will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0014] Figure 1 This is a flowchart of a giant magnetoresistive sensing signal enhancement method based on multi-channel fusion according to the present invention;

[0015] Figure 2 This is a flowchart illustrating the calculation of discrete complex sample values ​​in one optional embodiment of the present invention;

[0016] Figure 3 This is a flowchart illustrating the calculation of the common-mode transfer function estimate in one optional embodiment of the present invention. Detailed Implementation

[0017] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0018] In one embodiment, such as Figure 1 As shown, a method for enhancing giant magnetoresistive (GMR) sensing signals based on multi-channel fusion is provided. This embodiment illustrates the application of this method to a GMR signal processing terminal. It is understood that this method can also be applied to a GMR signal processing server, and further to a GMR signal processing system including both a GMR signal processing terminal and a GMR signal processing server, and is implemented through the interaction between the two. In this embodiment, the method includes the following steps:

[0019] Step S1: Perform complex domain frequency response calibration on each giant magnetoresistive channel to obtain discrete complex sampled values ​​of the differential-mode sensitivity transfer function and common-mode response transfer function of each giant magnetoresistive channel on a preset frequency set.

[0020] Optionally, the giant magnetoresistive (GMR) signal processing terminal can determine a preset frequency set for frequency response calibration based on the operating frequency band range of the GMR sensor array and the frequency requirements of the measurement scenario. The GMR signal processing terminal can sequentially apply standard magnetic field excitation signals at each frequency point in the preset frequency set to each GMR channel, and synchronously acquire the digital output sequence of each GMR channel at the corresponding frequency point through the synchronous acquisition unit. The GMR signal processing terminal can perform complex domain correlation operations on the digital output sequence of each GMR channel at each frequency point and the reference sequence of the standard magnetic field excitation signal to calculate the discrete complex sampled values ​​of the differential-mode sensitivity transfer function and common-mode response transfer function of each GMR channel at each frequency point in the preset frequency set.

[0021] Specifically, the giant magnetoresistive (GMR) channel can be a complete signal acquisition link consisting of a GMR Wheatstone bridge, a front-end signal conditioning circuit, an anti-aliasing filter unit, and an analog-to-digital converter unit; the complex domain frequency response calibration can be a calibration process that performs complex domain quantization characterization of the amplitude-frequency and phase-frequency characteristics of each GMR channel within a preset operating frequency range. The complex domain frequency response calibration can be used to obtain the amplitude gain and phase shift information of the channel response; the differential-mode sensitivity transfer function can be a complex variable function characterizing the response characteristics of the GMR channel to the target gradient magnetic field. The discrete complex sampled values ​​of the differential-mode sensitivity transfer function can be used to quantify the sensitivity gain and phase response of the GMR channel to the differential-mode magnetic field at different frequencies; the common-mode response transfer function can be a complex variable function characterizing the response characteristics of the GMR channel to the background common-mode magnetic field. The discrete complex sampled values ​​of the common-mode response transfer function can be used to quantify the response amplitude and phase shift of the GMR channel to common-mode interference at different frequencies; the preset frequency set can be a discretely distributed set of frequency points covering the operating frequency band of the GMR sensor.

[0022] Step S2: Acquire the multiple raw digital signal streams collected during the measurement phase of each giant magnetoresistive channel. Based on the discrete complex sampling values, perform frame division and short-time frequency domain transformation on the multiple raw digital signal streams to obtain the complex spectrum data of each giant magnetoresistive channel on each frequency slot.

[0023] Optionally, after obtaining authorization, the giant magnetoresistive (GMR) signal processing terminal can synchronously acquire multiple raw digital signal streams acquired by each GMR channel during the measurement phase through a multi-channel synchronous data acquisition interface, and perform sampling synchronization verification and DC component removal on the multiple raw digital signal streams to obtain pre-processed multiple raw digital signal streams. The GMR signal processing terminal can determine the frame length and frame shift parameters for frame processing according to the frequency resolution of a preset frequency set, and perform frame processing on the pre-processed multiple raw digital signal streams according to the frame length and frame shift parameters to obtain short-time signal frames for each GMR channel. The GMR signal processing terminal can perform short-time frequency domain transformation on each short-time signal frame to obtain complex spectrum data. The GMR signal processing terminal can divide the complex spectrum data according to frequency slots and perform frequency alignment between multiple channels to obtain complex spectrum data of each GMR channel in each frequency slot.

[0024] Alternatively, the calculation formula for the short-time frequency domain transform can be:

[0025]

[0026] in, The first short-time signal frame The frame signal is in the first Complex spectral data on each frequency slot; This refers to the time-domain sampling point number of a single short-time signal frame; For the first The sequence number of the short-time signal frame is The discrete amplitude of the giant magnetoresistive signal corresponding to the time-domain sampling points; of Time Domain ; The frame length of the short-time signal frame; It is the imaginary unit.

[0027] Specifically, the multiple raw digital signal streams can be time-domain digital signal sequences synchronously acquired by each giant magnetoresistive channel during the measurement phase and arranged in the order of sampling time; framing can be the process of dividing a continuous time-domain digital signal stream into multiple continuous short-time signal frames according to a preset frame length and frame shift; short-time frequency domain transformation can refer to the frequency domain transformation processing performed on each short-time signal frame, which is used to convert the time-domain signal into complex spectral data in the frequency domain; frequency slots can be discrete frequency units divided in the frequency domain according to fixed frequency intervals after short-time frequency domain transformation, with each frequency slot corresponding to a fixed frequency interval; complex spectral data can be complex data output by short-time frequency domain transformation, which is used to characterize the frequency characteristics of the signal frame in the corresponding frequency slot.

[0028] Step S3: Based on the complex spectrum data of each giant magnetoresistive channel in each frequency slot, perform frame-by-frame adaptive blind identification of the common-mode transfer function between channels to obtain the estimated value of the common-mode transfer function of each giant magnetoresistive channel in each frequency slot relative to the reference channel in each frame of complex spectrum data.

[0029] Optionally, the giant magnetoresistive (GMR) signal processing terminal can select a GMR channel that meets a preset benchmark condition from all GMR channels based on the complex domain frequency of each GMR channel in each frequency slot. The GMR signal processing terminal can use a short-time signal frame as the basic unit and the complex spectrum data of the reference channel as the benchmark to calculate the frequency domain response correlation characteristics of each GMR channel and the reference channel in each frequency slot of each frame of complex spectrum data. Based on the frequency domain response correlation characteristics, the GMR signal processing terminal can perform adaptive blind identification of the common-mode transfer function between channels to obtain the estimated value of the common-mode transfer function of each GMR channel in each frequency slot of each frame of complex spectrum data relative to the reference channel.

[0030] Alternatively, the formula for calculating the common-mode transfer function estimate can be:

[0031]

[0032] in, For the first Frame complex spectrum data in the first The first frequency slot Each giant magnetoresistive channel relative to the reference channel The common-mode transfer function estimate; For the first Frame complex spectrum data in the first The first frequency slot Complex spectral data of a giant magnetoresistive channel; For the first The first frame of complex spectral data Complex spectral data of the reference channel on each frequency slot.

[0033] Specifically, frame-by-frame adaptive blind identification can be a process of adaptively identifying the common-mode response transfer function between channels using short-time signal frames as the processing unit; the reference channel can be a giant magnetoresistive channel selected from the giant magnetoresistive channels as a benchmark for comparing the response characteristics between channels; the common-mode transfer function estimate can be a complex estimate of the common-mode response transfer function of each giant magnetoresistive channel relative to the reference channel obtained through blind identification.

[0034] Step S4: Based on the common-mode transfer function estimate, perform consistency compensation on the complex spectrum data of each giant magnetoresistive channel in the frequency domain to obtain compensated complex spectrum data. Then, perform inverse transformation and time-domain reconstruction on the compensated complex spectrum data to obtain a compensated multi-channel time-domain signal sequence.

[0035] Optionally, the giant magnetoresistive (GMR) signal processing terminal can calculate the frequency domain compensation coefficients on each frequency slot based on the common-mode transfer function estimate and the frequency domain response characteristics of the reference channel. Based on these frequency domain compensation coefficients, the GMR signal processing terminal can perform amplitude and phase correction on the complex spectrum data of each GMR channel to achieve frequency domain consistency compensation and obtain compensated complex spectrum data. The GMR signal processing terminal can perform an inverse transform corresponding to the short-time frequency domain transform on the compensated complex spectrum data, converting the frequency domain data into short-time signal frame time-domain data. The GMR signal processing terminal can perform time-domain overlap and addition reconstruction on the time-domain data of each short-time signal frame to obtain continuous time-domain signals for each GMR channel. Finally, the GMR signal processing terminal can integrate the continuous time-domain signals from all GMR channels to obtain a compensated multi-channel time-domain signal sequence.

[0036] Alternatively, the formula for calculating the compensated complex spectral data can be:

[0037]

[0038] in, For the first Frame complex spectrum data in the first The first frequency slot Compensated complex spectral data for a giant magnetoresistive channel; For the first Frame complex spectrum data in the first The first frequency slot Complex spectral data of a giant magnetoresistive channel; For the first Frame complex spectrum data in the first The first frequency slot Each giant magnetoresistive channel relative to the reference channel The common-mode transfer function estimate.

[0039] Specifically, frequency domain consistency compensation can be a process of performing amplitude and phase correction on the complex spectral data of each giant magnetoresistive channel in the frequency domain based on the common-mode transfer function estimate, so that the frequency domain response characteristics of each channel to the common-mode magnetic field are consistent; the compensated complex spectral data can be the complex spectral data of each channel after frequency domain consistency compensation, in which the amplitude-frequency characteristics and phase-frequency characteristics are aligned; the inverse transform can be the inverse frequency domain transform corresponding to the short-time frequency domain transform; the time domain reconstruction can be the process of restoring the short-time signal frames obtained by the inverse transform into a continuous time domain signal sequence according to the frame length and frame shift parameters at the time of framing, through the overlapping and addition method; the compensated multi-channel time domain signal sequence can be the continuous time domain signal sequence of each giant magnetoresistive channel after frequency domain consistency compensation and time domain reconstruction.

[0040] Step S5: Perform statistical optimal weighted fusion on the compensated multi-channel time-domain signal sequence to obtain a linear preliminary estimation sequence of gradient signals in the multiple original digital signal streams; based on the phase consistency of the multi-channel, perform sparse domain enhancement on the linear preliminary estimation sequence to obtain an enhanced gradient signal time series.

[0041] Optionally, the giant magnetoresistive (GMR) signal processing terminal can perform second-order statistical characteristic analysis on the compensated multi-channel time-domain signal sequence, calculating the covariance matrix of the compensated multi-channel time-domain signal sequence and the relative sensitivity characteristics of each GMR channel to the gradient field; the GMR signal processing terminal can solve for the statistically optimal weighting vector of the compensated multi-channel time-domain signal sequence based on common-mode suppression constraints and unbiased gradient signal estimation constraints; the GMR signal processing terminal can perform weighted fusion on the compensated multi-channel time-domain signal sequence based on the optimal weighting vector to obtain a linear preliminary estimation sequence of the gradient signal; the GMR signal processing terminal can process the compensated multi-channel time-domain signal... The sparse domain transformation is performed on the sequence and the linear preliminary estimation sequence to obtain the sparse domain coefficient sequence. The giant magnetoresistive signal processing terminal can calculate the phase concentration measure of each sparse domain coefficient position based on the multi-channel phase consistency characteristics. Based on the phase concentration measure, the giant magnetoresistive signal processing terminal can selectively enhance and suppress noise on the sparse domain coefficients of the linear preliminary estimation sequence to obtain the enhanced sparse domain coefficient sequence. The giant magnetoresistive signal processing terminal can perform an inverse sparse domain transformation on the enhanced sparse domain coefficient sequence to obtain the enhanced gradient signal time series, and output the enhanced gradient signal time series as the enhanced giant magnetoresistive sensing signal.

[0042] Specifically, statistical optimal weighted fusion can be a multi-channel fusion method that calculates the optimal weighting coefficients based on the second-order statistical properties of multi-channel signals under preset constraints, and then performs weighted fusion of multi-channel time-domain signals; the linear preliminary estimation sequence can be the preliminary estimation time-domain sequence of gradient signals in multiple original digital signal streams obtained after statistical optimal weighted fusion; multi-channel phase consistency can refer to the consistent phase characteristics of gradient signals in the sparse transform domain of each giant magnetoresistive channel; sparse domain enhancement can be a process of selectively enhancing and suppressing noise of the sparse coefficients of the signal in the sparse domain of wavelet transform based on multi-channel phase consistency; and enhanced gradient signal time series can be the time-domain signal sequence obtained after sparse domain enhancement.

[0043] In the aforementioned method for enhancing giant magnetoresistive (GMR) sensor signals based on multi-channel fusion, the discrete complex sampled values ​​of differential-mode sensitivity and common-mode response transfer function at a preset frequency set are obtained by calibrating the complex domain frequency response of each GMR channel. Multiple original digital signal streams are acquired, and the complex spectrum data of each frequency slot of each channel is obtained through framing and short-time frequency domain transformation. The estimated value of the common-mode transfer function of each channel relative to the reference channel is obtained through frame-by-frame adaptive blind identification. The compensated multi-channel time-domain signal sequence is obtained through frequency domain consistency compensation, inverse transformation, and time-domain reconstruction. Finally, the enhanced GMR sensor signal is output through statistical optimal weighted fusion and sparse domain enhancement. This method can accurately suppress common-mode interference, align the frequency domain characteristics of multiple channels, and improve the signal-to-noise ratio and detection sensitivity of the GMR sensor signal.

[0044] In one embodiment, such as Figure 2 As shown, the complex domain frequency response calibration is performed on each giant magnetoresistive channel to obtain the discrete complex sampled values ​​of the differential-mode sensitivity transfer function and common-mode response transfer function of each giant magnetoresistive channel on a preset frequency set, which may include:

[0045] Step S21: Place the giant magnetoresistive sensor array in a uniform magnetic field generator, drive the uniform magnetic field generator to generate a differential mode magnetic field with constant amplitude and frequency step-sweep, and synchronously collect the differential mode digital output sequence of each giant magnetoresistive channel at each frequency point.

[0046] Optionally, the giant magnetoresistive (GMR) signal processing terminal can fix the GMR sensor array to be calibrated within the uniform magnetic field region of the uniform magnetic field generator, thus completing the electrical connection between the GMR sensor array and the uniform magnetic field generator. The GMR signal processing terminal can generate an excitation control signal with frequency step sweep according to a preset frequency set, driving the uniform magnetic field generator to generate a differential-mode magnetic field with constant amplitude and frequency step-by-step variation according to the preset frequency set. After the excitation magnetic field at each frequency point stabilizes, the GMR signal processing terminal can synchronously acquire the differential-mode digital output sequence of each GMR channel under differential-mode magnetic field excitation at each frequency point.

[0047] Specifically, the uniform magnetic field generator can be a Helmholtz coil device capable of generating a spatially uniform standard magnetic field with controllable amplitude and frequency within a calibrated area; the differential-mode magnetic field can be a standard magnetic field that generates gradient differences at different sensitive units of a giant magnetoresistive sensor array, and the differential-mode magnetic field can be used to simulate the target gradient magnetic field in actual measurement; the frequency stepping sweep can be a sweep method that sequentially switches the frequency of the excitation magnetic field from low frequency to high frequency according to the frequency point sequence of a preset frequency set, while keeping the amplitude of the excitation magnetic field constant at each frequency point; the differential-mode digital output sequence can be a time-domain digital signal sequence output by the giant magnetoresistive channel under differential-mode magnetic field excitation.

[0048] Step S22: Perform complex division between the differential-mode digital output sequence of each giant magnetoresistive channel at each frequency point and the reference excitation signal to obtain the discrete complex sampled values ​​of the differential-mode sensitivity transfer function of each giant magnetoresistive channel at each frequency point.

[0049] Optionally, the giant magnetoresistive (GMR) signal processing terminal can perform frequency domain transformation on the reference excitation signal at each frequency point and the differential-mode digital output sequence of each GMR channel to obtain the complex spectral values ​​of the reference excitation signal and the differential-mode digital output sequence at the corresponding excitation frequency point; the GMR signal processing terminal can perform complex division operation on the complex spectral values ​​of the differential-mode digital output sequence and the complex spectral values ​​of the reference excitation signal to obtain the complex values ​​of the differential-mode sensitivity transfer function of each GMR channel at each frequency point; the GMR signal processing terminal can traverse all frequency points and all GMR channels, integrate the complex values ​​of the differential-mode sensitivity transfer function of each GMR channel at each frequency point, and obtain the discrete complex sampled values ​​of the differential-mode sensitivity transfer function of each GMR channel at each frequency point.

[0050] Alternatively, the formula for calculating discrete complex sample values ​​can be:

[0051]

[0052] in, For the first A giant magnetoresistive channel at frequency point Discrete complex sampled values ​​of the differential-mode sensitivity transfer function at the location; For the first A giant magnetoresistive channel at frequency point Differential digital output sequence at the location; Frequency point Reference excitation signal at the location; This is for Fast Fourier Transform (FFT) operations.

[0053] Specifically, the reference excitation signal can be a standard sinusoidal excitation signal that drives a uniform magnetic field generator to produce a differential-mode magnetic field; complex division can be a division operation performed on two complex number sequences in the frequency domain, and complex division can be used to calculate the amplitude gain and phase shift of the channel response.

[0054] Step S23: Drive the uniform magnetic field generator to generate a common-mode magnetic field with constant amplitude and frequency step-sweep, and synchronously collect the common-mode digital output sequence of each giant magnetoresistive channel at each frequency point.

[0055] Optionally, the giant magnetoresistive signal processing terminal can keep the giant magnetoresistive sensor array in a fixed position in the uniform magnetic field generator, generate a common-mode excitation control signal with frequency step sweep, and drive the uniform magnetic field generator to generate a common-mode magnetic field with constant amplitude and frequency step by step according to a preset frequency set; after the excitation magnetic field at each frequency point is stable, the giant magnetoresistive signal processing terminal can synchronously collect the common-mode digital output sequence of each giant magnetoresistive channel under the common-mode magnetic field excitation at each frequency point.

[0056] Specifically, the common-mode magnetic field can be a uniform background magnetic field with completely consistent amplitude, phase and frequency at all sensitive units of the giant magnetoresistive sensor array. The common-mode magnetic field can be used to simulate the environmental common-mode interference magnetic field in actual measurement. The common-mode digital output sequence can be a time-domain digital signal sequence output by the giant magnetoresistive channel under common-mode magnetic field excitation.

[0057] Step S24: Perform complex division between the common-mode digital output sequence of each giant magnetoresistive channel at each frequency point and the reference excitation signal to obtain the discrete complex sampled values ​​of the common-mode response transfer function of each giant magnetoresistive channel at each frequency point.

[0058] Optionally, the giant magnetoresistive (GMR) signal processing terminal can perform frequency domain transformation on the reference excitation signal at each frequency point and the common-mode digital output sequence of each GMR channel to obtain the complex spectral values ​​of the reference excitation signal and the common-mode digital output sequence at the corresponding excitation frequency point; the GMR signal processing terminal can perform complex division operation on the complex spectral values ​​of the common-mode digital output sequence and the complex spectral values ​​of the reference excitation signal to obtain the complex values ​​of the common-mode response transfer function of each GMR channel at each frequency point; the GMR signal processing terminal can traverse all frequency points and all GMR channels, integrate the complex values ​​of the common-mode response transfer function of each GMR channel at each frequency point, and obtain the discrete complex sampled values ​​of the common-mode response transfer function of each GMR channel at each frequency point.

[0059] Optionally, the formula for calculating the discrete complex sampled values ​​of the common-mode response transfer function can be:

[0060]

[0061] in, For the first A giant magnetoresistive channel at frequency point Discrete complex sampled values ​​of the common-mode response transfer function at the location; For the first A giant magnetoresistive channel at frequency point Common-mode digital output sequence at the location; Frequency point Reference excitation signal at the location; This is for Fast Fourier Transform (FFT) operations.

[0062] In one embodiment, such as Figure 3 As shown, based on the complex spectrum data of each giant magnetoresistive channel in each frequency slot, the common-mode transfer function between channels is adaptively blindly identified frame by frame to obtain the estimated value of the common-mode transfer function of each giant magnetoresistive channel in each frequency slot relative to the reference channel in each frame of complex spectrum data, which may include:

[0063] Step S31: Based on the discrete complex sampled values ​​of the differential-mode sensitivity transfer function of each giant magnetoresistive channel, calculate the flatness index of the differential-mode sensitivity amplitude-frequency characteristic of each giant magnetoresistive channel; select the giant magnetoresistive channel with the optimal differential-mode sensitivity amplitude-frequency characteristic flatness index as the reference channel.

[0064] Optionally, the giant magnetoresistive (GMR) signal processing terminal can retrieve the discrete complex sampled values ​​of the differential-mode sensitivity transfer function of each GMR channel, extract the modulus of the discrete complex sampled values ​​at each frequency point, and obtain the differential-mode sensitivity amplitude-frequency characteristic curve of each GMR channel. Based on the differential-mode sensitivity amplitude-frequency characteristic curve, the GMR signal processing terminal can calculate the fluctuation variance of the amplitude-frequency response within the working frequency band and obtain the differential-mode sensitivity amplitude-frequency characteristic flatness index of each GMR channel. The GMR signal processing terminal can sort the differential-mode sensitivity amplitude-frequency characteristic flatness indices of all GMR channels and select the GMR channel with the smallest differential-mode sensitivity amplitude-frequency characteristic flatness index as the reference channel.

[0065] Optionally, the formula for calculating the flatness index of the differential mode sensitivity amplitude-frequency response can be:

[0066]

[0067] in, For the first The flatness index of differential-mode sensitivity amplitude-frequency response of a giant magnetoresistive channel; The total number of frequency points in the preset frequency set; For the first A giant magnetoresistive channel at frequency point The magnitude of the discrete complex sampled values ​​of the differential-mode sensitivity transfer function at the given location; For the first The average value of the differential mode sensitivity amplitude-frequency response of each giant magnetoresistive channel within the operating frequency band.

[0068] Specifically, the differential-mode sensitivity amplitude-frequency characteristic flatness index can be used as an indicator to quantify the degree of amplitude response fluctuation of the differential-mode sensitivity of the giant magnetoresistive channel within the operating frequency band. The smaller the index value, the flatter the amplitude-frequency characteristic of the giant magnetoresistive channel, the better the linearity of the frequency response, and the more suitable it is as a reference channel.

[0069] Step S32: In each frame of complex spectrum data, for the giant magnetoresistive channel other than the reference channel, calculate the original complex ratio of the complex spectrum of the giant magnetoresistive channel in each frequency slot of the current frame to the complex spectrum of the reference channel in the corresponding frequency slot.

[0070] Optionally, the giant magnetoresistive (GMR) signal processing terminal can traverse all complex spectral data frame by frame, and traverse all GMR channels in the current frame except for the reference channel; the GMR signal processing terminal can traverse all frequency slots of the GMR channels except for the reference channel, and extract the complex spectral data of the GMR channel in each frequency slot and the complex spectral data of the reference channel in each frequency slot; the GMR signal processing terminal can perform complex division operations on the complex spectral data of each GMR and the complex spectral data of the reference channel to obtain the original complex ratio of each GMR channel in each frequency slot.

[0071] Specifically, the original complex ratio can be the ratio of the complex spectral data of the giant magnetoresistive channel (excluding the reference channel) and the reference channel on the same frequency slot.

[0072] Step S33: For each frequency slot, determine whether the power spectral density of the reference channel in the frequency slot of the current frame exceeds the noise threshold; if it does, store the original complex ratio value into the complex ratio circular buffer corresponding to the frequency slot to obtain the complex ratio value sequence.

[0073] Optionally, the giant magnetoresistive (GMR) signal processing terminal can allocate a fixed-length complex ratio circular buffer for each frequency slot. The complex ratio circular buffer can be used to store the original complex ratio. The GMR signal processing terminal can calculate the power spectral density of the reference channel in each frequency slot of the current frame. The GMR signal processing terminal can compare the power spectral density with a noise threshold. If the power spectral density exceeds the noise threshold, the original complex ratio of the GMR channel in the current frequency slot of the current frame is stored in the complex ratio circular buffer corresponding to the current frequency slot. If the power spectral density does not exceed the noise threshold, the original complex ratio of the GMR channel in the current frequency slot of the current frame is discarded, and the complex ratio circular buffer is not updated. The GMR signal processing terminal can traverse all frequency slots, update all complex ratio circular buffers, and obtain the complex ratio sequence for each frequency slot.

[0074] Alternatively, the formula for calculating the power spectral density can be:

[0075]

[0076] in, For the first The first frame Power spectral density of the reference channel on each frequency slot; For the first The first frame Complex spectral data of the reference channel on each frequency slot.

[0077] Specifically, the power spectral density can be the signal power of the reference channel in the corresponding frequency slot, and the power spectral density can be used to characterize the signal strength in the frequency slot; the noise threshold can be a critical value of the power spectral density used to distinguish between effective signals and background noise; the complex ratio circular buffer can be a fixed-length circular buffer allocated separately for each frequency slot to store historical original complex ratios; the complex ratio sequence can be a set of original complex ratios arranged in chronological order stored in the complex ratio circular buffer.

[0078] Step S34: Extract the last number of valid sample original complex ratios from the complex ratio sequence as the median filtering input sequence. Perform median filtering on the real and imaginary parts of the median filtering input sequence respectively. Combine the median of the real part with the median of the imaginary part to form a complex number, which is used as the common mode transfer function estimate of the current frame.

[0079] Optionally, the giant magnetoresistive (GMR) signal processing terminal can extract the number of original complex ratios that were last stored as valid samples from the complex ratio sequence of each frequency slot to form a median-filtered input sequence. The GMR signal processing terminal can split each original complex ratio in the median-filtered input sequence into real and imaginary parts to obtain real and imaginary part sequences. The GMR signal processing terminal can perform median filtering on the real and imaginary parts respectively to obtain the real median and imaginary median. The GMR signal processing terminal can use the real median as the real part and the imaginary median as the imaginary part, combine them into a new complex number, and use it as the common-mode transfer function estimate of the current frame.

[0080] Specifically, the effective sample size can be the number of the latest original complex ratios extracted from the complex ratio sequence for median filtering; median filtering can be a nonlinear filtering method that sorts the values ​​of the input sequence and then takes the median.

[0081] In one embodiment, the giant magnetoresistive sensing signal enhancement method based on multi-channel fusion may further include:

[0082] Step S41: Calculate the adaptive forgetting factor based on the rate of change of the complex spectral data of the current frame and the previous frame.

[0083] Optionally, the giant magnetoresistive signal processing terminal can calculate the difference in modulus between the current frame and the previous frame of complex spectral data, and statistically analyze the average modulus difference rate across the entire frequency band to obtain the rate of change of the complex spectral data; the giant magnetoresistive signal processing terminal can set upper and lower limits for the forgetting factor, and perform linear interpolation on the rate of change of the complex spectral data based on the upper and lower limits of the forgetting factor to calculate the adaptive forgetting factor.

[0084] Alternatively, the formula for calculating the adaptive forgetting factor can be:

[0085]

[0086] in, For the first Adaptive forgetting factor corresponding to frame complex spectral data; and These are the upper and lower limits of the forgetting factor, respectively; For the first The rate of change of the complex spectral data of a frame relative to the complex spectral data of the previous frame; The maximum rate of change threshold; It is a minimum value function.

[0087] Specifically, the adaptive forgetting factor can be a coefficient used to dynamically adjust the number of effective samples, and its value range can be (0,1]; the rate of change of complex spectral data can be the degree of difference between the complex spectral data of the current frame and the previous frame.

[0088] Step S42: Multiply the adaptive forgetting factor and the total number of original complex ratios in the complex ratio sequence, and then round down to obtain the effective sample number.

[0089] Optionally, the giant magnetoresistive signal processing terminal can obtain the total number of original complex ratios stored in the adaptive forgetting factor and the complex ratio ring buffer; the giant magnetoresistive signal processing terminal can multiply the adaptive forgetting factor and the total number of original complex ratios, and round down the result to obtain the number of effective samples.

[0090] Alternatively, the formula for calculating the number of effective samples can be:

[0091]

[0092] in, For the first The number of valid samples corresponding to a frame of complex spectral data; This is the lower limit of the minimum sample size; This is a floor function; For the first Adaptive forgetting factor corresponding to frame complex spectral data; This represents the total number of original complex ratios in the complex ratio circular buffer.

[0093] In one embodiment, based on the common-mode transfer function estimate, the complex spectral data of each giant magnetoresistive channel is subjected to consistency compensation in the frequency domain to obtain compensated complex spectral data, which may include:

[0094] Step S51: Based on the discrete complex sampled values ​​of the common-mode response transfer function of each giant magnetoresistive channel, calculate the fluctuation index of the common-mode response amplitude-frequency characteristic of each giant magnetoresistive channel; select the giant magnetoresistive channel with the optimal fluctuation index of the common-mode response amplitude-frequency characteristic as the common-mode reference template channel; and obtain the estimated value of the common-mode transfer function of the common-mode reference template channel relative to the reference channel.

[0095] Optionally, the giant magnetoresistive (GMR) signal processing terminal can retrieve the discrete complex sampled values ​​of the common-mode response transfer function of each GMR channel, extract the modulus of the discrete complex sampled values ​​at each frequency point, and obtain the common-mode response amplitude-frequency characteristic curve of each GMR channel. Based on the common-mode response amplitude-frequency characteristic curve, the GMR signal processing terminal can calculate the coefficient of variation of the amplitude-frequency response within the working frequency band to obtain the common-mode response amplitude-frequency characteristic fluctuation index of each GMR channel. The GMR signal processing terminal can sort the common-mode response amplitude-frequency characteristic fluctuation indices of all GMR channels and select the GMR channel with the smallest common-mode response amplitude-frequency characteristic fluctuation index as the common-mode reference template channel. The GMR signal processing terminal can retrieve the common-mode transfer function estimate of the current frame and calculate the common-mode transfer function estimate of the common-mode reference template channel relative to the reference channel.

[0096] Optionally, the formula for calculating the common-mode response amplitude-frequency characteristic fluctuation index can be:

[0097]

[0098] in, For the first The degree of fluctuation in the common-mode response amplitude-frequency characteristic of a giant magnetoresistive channel; The total number of frequency points in the preset frequency set; For the first A giant magnetoresistive channel at frequency point The magnitude of the discrete complex sampled values ​​of the common-mode response transfer function at point; For the first The average value of the common-mode response amplitude-frequency response of each giant magnetoresistive channel within the operating frequency band.

[0099] Specifically, the common-mode response amplitude-frequency characteristic fluctuation index can be an index that quantifies the degree of fluctuation of the amplitude of the common-mode response of the giant magnetoresistive channel with frequency within the operating frequency band; the common-mode reference template channel can be the giant magnetoresistive channel with the most stable common-mode response characteristics selected from the giant magnetoresistive channels.

[0100] Step S52: For the giant magnetoresistive (GMR) channels other than the reference channel, subtract the product of the complex spectrum data of the reference channel in the corresponding frequency slot and the estimated value of the common-mode transfer function of the reference channel from the complex spectrum data of the GMR channel in each frequency slot of the current frame; add the product of the complex spectrum data of the reference channel in the corresponding frequency slot and the estimated value of the common-mode transfer function of the common-mode reference template channel to obtain the compensated complex spectrum data.

[0101] Optionally, the giant magnetoresistive (GMR) signal processing terminal can traverse all GMR channels in the current frame except for the reference channel, and traverse all frequency slots of each GMR channel to extract the original complex spectrum data of each GMR channel in each frequency slot, the complex spectrum data of the reference channel in each frequency slot, the estimated common-mode transfer function (CMF) of each GMR channel relative to the reference channel, and the estimated common-mode transfer function of the common-mode reference template channel relative to the reference channel. The GMR signal processing terminal can set compensation operation rules and, according to the compensation operation rules, perform complex domain operations on the original complex spectrum data of each GMR channel in each frequency slot, the complex spectrum data of the reference channel in each frequency slot, the estimated common-mode transfer function of each GMR channel relative to the reference channel, and the estimated common-mode transfer function of the common-mode reference template channel relative to the reference channel to obtain the compensated complex spectrum data of each GMR channel in each frequency slot.

[0102] In one embodiment, performing statistical optimal weighted fusion on the compensated multi-channel time-domain signal sequence to obtain a linear preliminary estimate sequence of gradient signals in the multiple original digital signal streams may include:

[0103] Step S61: Organize the compensated multi-channel time-domain signal sequence at each sampling time into a signal column vector, and perform an arithmetic average of the signal column vectors at the sampling time and the previous sampling time to obtain the residual common-mode component estimate.

[0104] Optionally, the giant magnetoresistive signal processing terminal can traverse the compensated multi-channel time-domain signal sequence according to the sampling time, arrange the signal sample values ​​of all giant magnetoresistive channels at each sampling time in the order of the giant magnetoresistive channels, and obtain the signal column vector; the giant magnetoresistive signal processing terminal can select all the signal column vectors of each sampling time and the previous sampling time of each sampling time, and perform an arithmetic mean operation on all the signal column vectors to obtain the residual common mode component estimate value corresponding to each sampling time.

[0105] Specifically, the signal column vector can be a column vector formed by arranging the time-domain signal sampled values ​​of all giant magnetoresistive channels at the same sampling time in the order of the giant magnetoresistive channels; the residual common-mode component estimate can be an estimate of the residual common-mode interference that exists in each giant magnetoresistive channel, obtained by the arithmetic mean of the multi-channel signals.

[0106] Step S62: At the sampling time when the complex spectrum data is below the spectrum threshold, the residual common-mode component estimate is subtracted from the signal column vector to obtain the residual vector sample.

[0107] Optionally, the giant magnetoresistive signal processing terminal can calculate the total spectral energy of the signal frame across the entire frequency band at each sampling time, and compare the total spectral energy with the spectral threshold. If the total spectral energy is lower than the spectral threshold, the residual common-mode component estimate is subtracted from the signal column vector at that sampling time to obtain the residual vector sample.

[0108] Specifically, the spectral threshold can be a pre-set critical value of spectral energy used to distinguish between signal segments and noise segments; the residual vector sample can be a vector obtained by removing the residual common-mode component estimate from the signal column vector, and the residual vector sample can be used to characterize the noise and residual interference characteristics of each channel.

[0109] Step S63: Covariance estimation of the residual vector samples is performed by exponential moving average to obtain the covariance matrix of each giant magnetoresistive channel.

[0110] Optionally, the giant magnetoresistive signal processing terminal can set a smoothing coefficient and use the exponential moving average method to estimate the covariance of the residual vector samples to obtain the covariance matrix of each giant magnetoresistive channel.

[0111] Alternatively, the formula for calculating the covariance matrix can be:

[0112]

[0113] in, Sampling time The covariance matrix; is the exponential smoothing coefficient, and its value range can be (0,1); Let be the covariance matrix of the previous sampling time. Sampling time The corresponding residual vector samples; This is a vector transpose operation.

[0114] Specifically, the exponential moving average can be a weighted average method for time-series samples, which can adaptively track the time-varying characteristics of noise statistics; the covariance matrix can be a square matrix that characterizes the noise correlation between multi-channel signals.

[0115] Step S64: Based on the relative sensitivity vector and covariance matrix of each giant magnetoresistive channel to the gradient field intensity, the optimal weighting vector is obtained under the constraints that the sum of the weighting coefficients is zero and the inner product of the weighting coefficients and the relative sensitivity vector is one.

[0116] Optionally, the giant magnetoresistive (GMR) signal processing terminal can obtain the average differential-mode sensitivity of each GMR channel within the operating frequency band to obtain the relative sensitivity vector. The GMR signal processing terminal can construct a constrained optimization problem with the minimum noise variance of the fused signal as the optimization objective and the constraint that the sum of the weighting coefficients is zero and the inner product of the weighting coefficients and the relative sensitivity vector is one. The GMR signal processing terminal can solve this constrained optimization problem using the Lagrange multiplier method to obtain the optimal weighting vector.

[0117] Specifically, the relative sensitivity vector can be a vector composed of the response sensitivities of each giant magnetoresistive channel to the gradient field intensity; the optimal weighting vector can be a weighting coefficient vector that satisfies the constraints and minimizes the noise variance of the fused signal.

[0118] Step S65: Perform an inner product operation on the optimal weighted vector and the signal column vectors at each sampling time to obtain a linear preliminary estimation sequence of the gradient signal.

[0119] Optionally, the giant magnetoresistive signal processing terminal can perform an inner product operation on the optimal weighted vector and the signal column vector at each sampling time to obtain the gradient signal estimate at each sampling time; the giant magnetoresistive signal processing terminal can arrange all gradient signal estimates in ascending order of sampling time to obtain a linear preliminary estimate sequence of gradient signals.

[0120] In one embodiment, the expression for the optimal weighted vector can be:

[0121]

[0122] in, The optimal weighted vector. Let covariance matrix be the variance matrix. Relative sensitivity vector In a completely one vector The projection vector on the orthogonal complement space, This represents the vector transpose operation. This represents the matrix inversion operation.

[0123] In one embodiment, sparse domain enhancement of the linear preliminary estimation sequence based on multi-channel phase consistency to obtain an enhanced gradient signal time series may include:

[0124] Step S71: Perform discrete wavelet transform on the compensated multi-channel time-domain signal sequence to obtain the channel-independent wavelet coefficient sequence.

[0125] Optionally, the giant magnetoresistive signal processing terminal can set the wavelet basis function and the number of decomposition levels for discrete wavelet transform, and perform discrete wavelet transform on the compensated multi-channel time-domain signal sequence according to the wavelet basis function and the number of decomposition levels to obtain the complex wavelet coefficients of each giant magnetoresistive channel at each decomposition scale and at each wavelet coefficient position; the giant magnetoresistive signal processing terminal can integrate the complex wavelet coefficients of each giant magnetoresistive channel at each decomposition scale and at each wavelet coefficient position to obtain the channel-independent wavelet coefficient sequence of each giant magnetoresistive channel.

[0126] Specifically, discrete wavelet transform can be a sparse domain transform method that decomposes a time-domain signal into wavelet coefficients of different scales; channel-independent wavelet coefficient sequence can be a sequence of wavelet coefficients of each scale and position obtained after the time-domain signal of each giant magnetoresistive channel has undergone discrete wavelet transform.

[0127] Step S72: Perform discrete wavelet transform on the linear preliminary estimation sequence to obtain the wavelet coefficient sequence of the fused signal.

[0128] Optionally, the giant magnetoresistive signal processing terminal can perform discrete wavelet transform on the linear preliminary estimation sequence to obtain the complex wavelet coefficients of the linear preliminary estimation sequence at each decomposition scale and at each wavelet coefficient position; the giant magnetoresistive signal processing terminal can integrate the complex wavelet coefficients of the linear preliminary estimation sequence at each decomposition scale and at each wavelet coefficient position to obtain the fused signal wavelet coefficient sequence.

[0129] Step S73: For each wavelet coefficient position, calculate the magnitude of the sum of the unit vectors of the independent wavelet coefficients of the channel independent wavelet coefficient sequence at the wavelet coefficient position in the complex plane to obtain the phase concentration measure.

[0130] Optionally, the giant magnetoresistive signal processing terminal can extract the independent wavelet coefficients of all giant magnetoresistive channels at each wavelet coefficient position; the giant magnetoresistive signal processing terminal can divide each independent wavelet coefficient by the modulus of each wavelet coefficient position to obtain the unit vector of each wavelet coefficient position on the complex plane; the giant magnetoresistive signal processing terminal can sum the unit vectors of all giant magnetoresistive channels to obtain the summation vector, and calculate the modulus of the summation vector to obtain the phase concentration measure.

[0131] Specifically, the phase concentration measure can be an index that quantifies the degree of phase consistency of wavelet coefficients in each channel at the same wavelet coefficient position. The larger the modulus of the phase concentration measure, the more concentrated the phase of the wavelet coefficients in each giant magnetoresistive channel is.

[0132] Step S74: Compare the phase concentration measure and the phase consistency threshold. For wavelet coefficient positions where the phase concentration measure exceeds the phase consistency threshold, multiply the fused signal wavelet coefficients at the wavelet coefficient positions by the enhancement gain factor. For wavelet coefficient positions where the phase concentration measure does not exceed the phase consistency threshold, set the fused signal wavelet coefficients at the wavelet coefficient positions to zero to obtain the enhanced fused signal wavelet coefficient sequence.

[0133] Optionally, the giant magnetoresistive signal processing terminal can set a phase consistency threshold and an enhancement gain factor. The phase concentration measure of each wavelet coefficient position is compared with the phase consistency threshold. If the phase concentration measure exceeds the phase consistency threshold, the wavelet coefficient position is determined to be a valid signal position, and the fused signal wavelet coefficients of the fused signal wavelet coefficient sequence at that position are multiplied by the enhancement gain factor. If the phase concentration measure does not exceed the phase consistency threshold, the wavelet coefficient position is determined to be a noise position, and the fused signal wavelet coefficients of the fused signal wavelet coefficient sequence at that position are set to zero to obtain the enhanced fused signal wavelet coefficient sequence.

[0134] Specifically, the phase consistency threshold can be a critical value for phase concentration measurement used to distinguish between effective signals and noise; the enhancement gain factor can be a gain coefficient used to amplify the wavelet coefficients of the effective signal; and the enhanced fused signal wavelet coefficient sequence can be a wavelet coefficient sequence obtained after selective enhancement and noise suppression.

[0135] Step S75: Perform inverse discrete wavelet transform on the wavelet coefficient sequence of the enhanced fusion signal to obtain the time sequence of the enhanced gradient signal.

[0136] Optionally, the giant magnetoresistive signal processing terminal can perform inverse discrete wavelet transform on the wavelet coefficient sequence of the enhanced fusion signal to convert the wavelet coefficients in the sparse domain back to the time domain signal; the giant magnetoresistive signal processing terminal can use the time domain signal sequence obtained by the inverse transform as the time sequence of the enhanced gradient signal to complete the enhancement of the giant magnetoresistive sensing signal.

[0137] It should be understood that although the steps in the flowcharts of the above embodiments are shown sequentially according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the flowcharts of the above embodiments may include multiple steps or multiple stages. These steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these steps or stages is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the steps or stages of other steps.

[0138] The above embodiments merely illustrate several implementation methods of the embodiments of this application, and their descriptions are relatively specific and detailed, but they should not be construed as limiting the scope of the patent application. It should be noted that those skilled in the art can make several modifications and improvements without departing from the concept of the embodiments of this application, and these all fall within the protection scope of the embodiments of this application.

Claims

1. A method for enhancing giant magnetoresistive sensing signals based on multi-channel fusion, characterized in that, The method includes: S1. Perform complex domain frequency response calibration on each giant magnetoresistive channel to obtain discrete complex sampled values ​​of the differential-mode sensitivity transfer function and common-mode response transfer function of each giant magnetoresistive channel on a preset frequency set; S2. Acquire multiple raw digital signal streams collected by each of the giant magnetoresistive channels during the measurement phase. Based on the discrete complex sampling values, perform frame division and short-time frequency domain transformation on the multiple raw digital signal streams to obtain the complex spectrum data of each of the giant magnetoresistive channels in each frequency slot. S3. Based on the complex spectrum data of each of the giant magnetoresistive channels in each of the frequency slots, perform frame-by-frame adaptive blind identification of the common-mode transfer function between channels to obtain the estimated value of the common-mode transfer function of each of the giant magnetoresistive channels in each of the frequency slots relative to the reference channel in each frame of complex spectrum data. S4. Based on the estimated value of the common-mode transfer function, perform consistency compensation on the complex spectrum data of each giant magnetoresistive channel in the frequency domain to obtain compensated complex spectrum data, and perform inverse transformation and time-domain reconstruction on the compensated complex spectrum data to obtain a compensated multi-channel time-domain signal sequence. S5. Perform statistical optimal weighted fusion on the compensated multi-channel time-domain signal sequence to obtain a linear preliminary estimation sequence of gradient signals in the multi-channel original digital signal stream; based on multi-channel phase consistency, perform sparse domain enhancement on the linear preliminary estimation sequence to obtain an enhanced gradient signal time sequence; the enhanced gradient signal time sequence is used as the enhanced giant magnetoresistive sensing signal output.

2. The method according to claim 1, characterized in that, The step of calibrating the complex domain frequency response of each giant magnetoresistive channel to obtain discrete complex sampled values ​​of the differential-mode sensitivity transfer function and common-mode response transfer function of each giant magnetoresistive channel on a preset frequency set includes: S21. Place the giant magnetoresistive sensor array in a uniform magnetic field generator, drive the uniform magnetic field generator to generate a differential mode magnetic field with constant amplitude and frequency step-sweep, and synchronously collect the differential mode digital output sequence of each giant magnetoresistive channel at each frequency point. S22. Perform a complex division between the differential-mode digital output sequence of each of the giant magnetoresistive channels at each frequency point and the reference excitation signal to obtain the discrete complex sampled value of the differential-mode sensitivity transfer function of each of the giant magnetoresistive channels at each frequency point. S23. Drive the uniform magnetic field generator to generate a common-mode magnetic field with constant amplitude and frequency step-sweep, and synchronously collect the common-mode digital output sequence of each giant magnetoresistive channel at each frequency point. S24. Perform complex division between the common-mode digital output sequence of each of the giant magnetoresistive channels at each frequency point and the reference excitation signal to obtain the discrete complex sampled value of the common-mode response transfer function of each of the giant magnetoresistive channels at each frequency point.

3. The method according to claim 1, characterized in that, The step of performing frame-by-frame adaptive blind identification of the common-mode transfer function (CMJ) between channels based on the complex spectral data of each of the giant magnetoresistive (GMR) channels in each of the frequency slots in each frame of complex spectral data to obtain the estimated value of the CMM relative to the reference channel for each of the GMR channels in each of the frequency slots in each frame of complex spectral data includes: S31. Based on the discrete complex sampled values ​​of the differential-mode sensitivity transfer function of each of the giant magnetoresistive channels, calculate the differential-mode sensitivity amplitude-frequency characteristic flatness index of each of the giant magnetoresistive channels; select the giant magnetoresistive channel with the optimal differential-mode sensitivity amplitude-frequency characteristic flatness index as the reference channel. S32. In each frame of complex spectrum data, for the giant magnetoresistive channel other than the reference channel, calculate the original complex ratio of the complex spectrum of the giant magnetoresistive channel in each frequency slot of the current frame to the complex spectrum of the reference channel in the corresponding frequency slot. S33. For each frequency slot, determine whether the power spectral density of the reference channel in the frequency slot of the current frame exceeds the noise threshold; if it does, store the original complex ratio value into the complex ratio value circular buffer corresponding to the frequency slot to obtain the complex ratio value sequence. S34. Extract the number of the last stored valid samples and the original complex ratios from the complex ratio sequence as the median filtering input sequence. Perform median filtering on the real and imaginary parts of the median filtering input sequence respectively. Combine the median of the real part with the median of the imaginary part into a complex number, which is used as the common mode transfer function estimate of the current frame.

4. The method according to claim 3, characterized in that, The method further includes: S41. Calculate the adaptive forgetting factor based on the rate of change of the complex spectrum data of the current frame and the previous frame; S42. Multiply the adaptive forgetting factor and the total number of the original complex ratios in the complex ratio sequence, and then round down to obtain the number of effective samples.

5. The method according to claim 1, characterized in that, The step of performing consistency compensation on the complex spectral data of each of the giant magnetoresistive channels in the frequency domain based on the common-mode transfer function estimate to obtain compensated complex spectral data includes: S51. Based on the discrete complex sampled values ​​of the common-mode response transfer function of each of the giant magnetoresistive channels, calculate the common-mode response amplitude-frequency characteristic fluctuation index of each of the giant magnetoresistive channels; select the giant magnetoresistive channel with the optimal common-mode response amplitude-frequency characteristic fluctuation index as the common-mode reference template channel; and obtain the estimated value of the common-mode transfer function of the common-mode reference template channel relative to the reference channel. S52. For the giant magnetoresistive channels other than the reference channel, subtract the product of the complex spectrum data of the reference channel in the corresponding frequency slot and the estimated value of the common-mode transfer function of the reference channel from the complex spectrum data of the giant magnetoresistive channel in each frequency slot of the current frame; add the product of the complex spectrum data of the reference channel in the corresponding frequency slot and the estimated value of the common-mode transfer function of the common-mode reference template channel to obtain the compensated complex spectrum data.

6. The method according to claim 1, characterized in that, The step of performing statistical optimal weighted fusion on the compensated multi-channel time-domain signal sequence to obtain a linear preliminary estimation sequence of gradient signals in the multiple original digital signal streams includes: S61. Organize the compensated multi-channel time-domain signal sequence at each sampling time into a signal column vector, and perform an arithmetic average on the signal column vectors at the sampling time and the sampling time preceding the sampling time to obtain the residual common-mode component estimate. S62. At the sampling time when the complex spectrum data is lower than the spectrum threshold, the residual common mode component estimate is subtracted from the signal column vector to obtain the residual vector sample. S63. The covariance of the residual vector samples is estimated by exponential moving average to obtain the covariance matrix of each giant magnetoresistive channel; S64. Based on the relative sensitivity vector of each giant magnetoresistive channel to the gradient field intensity and the covariance matrix, under the constraint that the sum of the weighting coefficients is zero and the inner product of the weighting coefficients and the relative sensitivity vector is one, the optimal weighting vector is obtained; the gradient field intensity is the amplitude of the gradient signal at each sampling time. S65. Perform an inner product operation on the optimal weighted vector and the signal column vector at each sampling time to obtain the linear preliminary estimation sequence of the gradient signal.

7. The method according to claim 6, characterized in that, The expression for the optimal weighted vector is: in, Let be the optimal weighted vector. Let be the covariance matrix. The relative sensitivity vector In a completely one vector The projection vector on the orthogonal complement space, This represents the vector transpose operation. This represents the matrix inversion operation.

8. The method according to claim 1, characterized in that, The step of enhancing the linear preliminary estimation sequence based on multi-channel phase consistency to obtain an enhanced gradient signal time series includes: S71. Perform discrete wavelet transform on the compensated multi-channel time-domain signal sequence to obtain a channel-independent wavelet coefficient sequence; S72. Perform discrete wavelet transform on the linear preliminary estimation sequence to obtain the wavelet coefficient sequence of the fused signal; S73. For each wavelet coefficient position, calculate the magnitude of the sum of the unit vectors of the independent wavelet coefficients of the channel independent wavelet coefficient sequence at the wavelet coefficient position in the complex plane to obtain the phase concentration measure. S74. Compare the phase concentration measure and the phase consistency threshold. For wavelet coefficient positions where the phase concentration measure exceeds the phase consistency threshold, multiply the fused signal wavelet coefficients of the fused signal wavelet coefficient sequence at the wavelet coefficient positions by an enhancement gain factor. For wavelet coefficient positions where the phase concentration measure does not exceed the phase consistency threshold, set the fused signal wavelet coefficients of the fused signal wavelet coefficient sequence at the wavelet coefficient positions to zero to obtain an enhanced fused signal wavelet coefficient sequence. S75. Perform inverse discrete wavelet transform on the wavelet coefficient sequence of the enhanced fusion signal to obtain the time sequence of the enhanced gradient signal.

Images