CN-121978417-A - Denoising method and device for very low-frequency-strobe electric broadband signals

CN121978417ACN 121978417 ACN121978417 ACN 121978417ACN-121978417-A

Abstract

The application discloses a denoising method and device for very low frequency electric broadband signals, which relate to the technical field of very low frequency electromagnetic wave detection and application, and the method comprises the steps of obtaining normalized signals based on original sampling data, and carrying out framing processing to extract the signal frequency spectrum of each frame; and (3) denoising sequentially by taking each frame as a target frame, calculating a posterior signal-to-noise ratio and an priori signal-to-noise ratio based on a noise reference, calculating a full-band average log-likelihood ratio of the target frame to obtain a noise reference of the next frame, obtaining a spectral gain coefficient based on the posterior signal-to-noise ratio, the priori signal-to-noise ratio and a preset distortion index to correct a signal spectrum of the target frame to obtain a denoised signal spectrum, and restoring to obtain a denoised time domain frame of the target frame, and obtaining a denoised very low strobe electrical signal based on the denoised time domain frame of each frame. The application is suitable for ground VLF lightning monitoring stations and weak lightning event capturing under the condition of low signal to noise ratio, and can provide reliable data basis for lightning event detection, classification and positioning front-end feature extraction.

Inventors

XU WEI
LIU SHICHENG
ZHANG BOWEN
GU XUDONG
NI BINBIN
HUANG GONGPING
WANG SHIWEI
LIN HONGWEI
Chen Zhishe
GUO SIMING

Assignees

武汉大学

Dates

Publication Date: 20260505
Application Date: 20260407

Claims (10)

1. A method for denoising very low strobe electrical broadband signals, comprising: Acquiring original sampling data, and processing to obtain a normalized signal; based on normalized signal framing processing, obtaining a plurality of time domain frame signals, and extracting a signal frequency spectrum of each frame; obtaining an initial background noise power spectrum based on signal spectrums of a plurality of continuous frames at the beginning of a first frame, and taking the initial background noise power spectrum as a noise reference of the first frame; Calculating a posterior signal-to-noise ratio and a priori signal-to-noise ratio of the target frame based on a noise reference by taking the first frame as the target frame; Calculating a full-band average log-likelihood ratio based on the posterior signal-to-noise ratio and the prior signal-to-noise ratio, and obtaining a noise reference of the next frame based on the full-band average log-likelihood ratio of the target frame; Obtaining a spectral gain coefficient based on the posterior signal-to-noise ratio, the priori signal-to-noise ratio and a preset distortion index; Correcting the signal spectrum of the target frame based on the spectral gain coefficient to obtain a denoising signal spectrum; obtaining a denoising time domain frame of the target frame based on the denoising signal spectrum reduction; turning to the step of calculating the posterior signal-to-noise ratio and the priori signal-to-noise ratio of the target frame based on the noise reference until a denoising time domain frame of each frame is obtained; And obtaining the denoised very low stroboscopic electrical signal based on the denoised time domain frame of each frame.
2. The denoising method for very low-frequency electrical broadband signals according to claim 1, wherein the obtaining the original sampled data and processing to obtain the normalized signal comprise: Receiving original sampling data, and carrying out mean value removal processing on the original sampling data to obtain a signal after mean value removal; And carrying out normalization processing on the signal after the mean value is removed to obtain a normalized signal.
3. The denoising method for very low-frequency electrical broadband signals according to claim 1, wherein the frame processing based on normalized signal to obtain a plurality of time-domain frame signals, extracting the signal spectrum of each frame comprises: framing the normalized signal in the time domain to obtain a plurality of time domain frame signals; And applying a hanning window with a preset window length to each frame of signal, performing fast Fourier transform, and extracting to obtain a corresponding signal frequency spectrum.
4. The method for denoising very low-frequency electrical broadband signal according to claim 1, wherein the obtaining the initial background noise power spectrum based on the signal spectrum of the plurality of consecutive frames from the first frame comprises: Extracting a plurality of continuous frames from the plurality of time domain frames, wherein the continuous frames start from a first frame, and acquiring signal spectrums of the continuous frames; for each frequency point, extracting an amplitude value of each frame at the corresponding frequency point based on signal spectrums of a plurality of continuous frames; sequencing the amplitude values of each frame at each frequency point according to the sequence from small to large, and extracting the amplitude values at the quantile of a preset value to obtain an initial background noise amplitude spectrum of each frequency point; And processing the initial background noise amplitude spectrum and the scaling factor based on each frequency point to obtain an initial background noise power spectrum.
5. The method for denoising very low-frequency electrical broadband signal according to claim 4, wherein the calculating the a-priori signal-to-noise ratio and the a-priori signal-to-noise ratio of the target frame based on the noise reference comprises: The posterior signal-to-noise ratio of the target frame is calculated based on the noise reference, and the formula is expressed as: ; calculating the prior signal-to-noise ratio of the target frame, and the formula is as follows: ; where n represents a frame number, k represents a frequency point number, Is the signal spectrum of the nth frame; a posterior signal to noise ratio of the nth frame; is a preset upper threshold value for preventing value overflow, and min represents fetch A minimum value within @, n=1, Representing an initial background noise power spectrum; In the time-course of which the first and second contact surfaces, The background noise power spectrum of the n-1 th frame is the noise reference of the n-th frame; A priori signal to noise ratio for the nth frame; the denoised signal spectrum for frame n-1, max represents the number [ Maximum value in ]; is a smoothing factor.
6. The method for denoising very low frequency electrical broadband signal according to claim 5, wherein the calculating the full-band average log likelihood ratio based on a posterior signal-to-noise ratio and an a priori signal-to-noise ratio comprises: calculating the log-likelihood ratio of the target frame at each frequency point based on the posterior signal-to-noise ratio and the prior signal-to-noise ratio of the target frame; calculating a full-band average log-likelihood ratio based on an average value of the log-likelihood ratios of each frequency point; The obtaining the noise reference of the next frame based on the full-band average log likelihood ratio of the target frame comprises the following steps: Comparing the average log likelihood ratio of the full frequency band with the value of the noise updating judgment threshold; Under the condition that the full-band average log likelihood ratio is smaller than or equal to a noise updating judgment threshold, judging the target frame as a noise frame, and updating the background noise power spectrum of the target frame based on a noise reference and the signal spectrum of the target frame, wherein the formula is as follows: ; under the condition that the full-band average log likelihood ratio is larger than the noise updating judgment threshold, judging the target frame as a signal frame, determining a noise reference as a background noise power spectrum of the target frame, and expressing as follows: ; taking the background noise power spectrum of the target frame as a noise reference of the next frame; Wherein, the The smoothing factor is updated for the noise, For the background noise power spectrum of the nth frame, Is the background noise power spectrum for the n-1 th frame.
7. The method for denoising very low-frequency electrical broadband signal according to claim 6, wherein obtaining the spectral gain coefficient based on the posterior signal-to-noise ratio, the prior signal-to-noise ratio and the preset distortion index comprises: And calculating an auxiliary variable based on the posterior signal-to-noise ratio and the prior signal-to-noise ratio of the target frame, wherein the formula is as follows: ; obtaining a spectral gain coefficient based on the auxiliary variable and a preset distortion index, wherein the formula is expressed as follows: ; Correcting the signal spectrum of the target frame based on the spectral gain coefficient to obtain a denoising signal spectrum, wherein the formula is as follows: ; Wherein, the Is an auxiliary variable; Is a confluent super-geometric function; is a gamma function; in order to be a distortion index, ; Is a spectral gain coefficient; The denoising signal spectrum at the frequency point k of the nth frame.
8. A very low strobe electrical broadband signal-oriented denoising apparatus, comprising: The data acquisition unit is used for acquiring original sampling data and processing the original sampling data to obtain normalized signals; the signal processing unit is used for obtaining a plurality of time domain frame signals based on normalized signal framing processing and extracting a signal frequency spectrum of each frame; The signal processing unit is further used for obtaining an initial background noise power spectrum based on signal spectrums of a plurality of continuous frames from the first frame, and the initial background noise power spectrum is used as a noise reference of the first frame; The signal denoising unit is used for calculating the posterior signal-to-noise ratio and the prior signal-to-noise ratio of the target frame based on the noise reference by taking the first frame as the target frame; The signal denoising unit is also used for calculating a full-band average log-likelihood ratio based on the posterior signal-to-noise ratio and the prior signal-to-noise ratio and obtaining a noise reference of the next frame based on the full-band average log-likelihood ratio of the target frame; the signal denoising unit is also used for obtaining a spectrum gain coefficient based on the posterior signal-to-noise ratio, the priori signal-to-noise ratio and a preset distortion index; the signal denoising unit is also used for correcting the signal spectrum of the target frame based on the spectral gain coefficient to obtain a denoised signal spectrum; The signal denoising unit is also used for obtaining a denoising time domain frame of the target frame based on denoising signal frequency spectrum restoration; the signal denoising unit is further used for taking the next frame as a new target frame, and turning to a step of calculating the posterior signal-to-noise ratio and the priori signal-to-noise ratio of the target frame based on a noise reference until a denoising time domain frame of each frame is obtained; the signal denoising unit is also used for obtaining a denoised very low stroboscopic electrical signal based on the denoised time domain frame of each frame.
9. An electronic device comprising a memory, a processor and a computer program stored on the memory and executable on the processor, characterized in that the processor implements the steps of the method according to any one of claims 1 to 7 when the program is executed by the processor.
10. A non-transitory computer readable storage medium, on which a computer program is stored, characterized in that the computer program, when being executed by a processor, implements the steps of the method according to any one of claims 1 to 7.

Description

Denoising method and device for very low-frequency-strobe electric broadband signals Technical Field The application relates to the technical field of very low frequency electromagnetic wave detection and application, in particular to a denoising method and device for very low frequency electric broadband signals. Background Very low frequency (Very Low Frequency, VLF) waves refer to electromagnetic waves having a frequency in the range of 3-30: 30 kHz, which may originate from natural radiation generated by global thunderstorm activity or from continuous transmissions from an artificial very low frequency station. The propagation of very low frequencies is constrained by Earth-ionosphere waveguides (Earth-Ionosphere Waveguide, EIWG) with very low attenuation rates (about 2-3 dB/Mm) within the waveguides, which can be adapted for wide range electromagnetic environment monitoring and event detection, providing an important information basis for related detection applications. Lightning discharges radiate broadband electromagnetic pulses (Sferics) whose energy peaks are concentrated mainly in the very low frequency band. The method can realize continuous tracking and event identification of lightning activity, provide data support for lightning positioning and activity evaluation, is highly sensitive to changes of the state of a propagation channel and an ionized layer when broadband electromagnetic pulse propagates in the ionized layer waveguide, and can be used as a natural probe for researching the problems of ionized layer burst disturbance, earthquake precursor electromagnetic abnormality, magnetic layer wave particle interaction and the like. The above applications generally rely on stable extraction and fidelity analysis of lightning pulse waveforms, particularly in complex contexts where it is desirable to maintain detectability of weak events and to accurately characterize key features (arrival times, peaks, pulse widths, and broadband patterns, etc.) to support subsequent identification, localization, and mechanism analysis. However, very low frequency monitoring devices are often deployed in ground environments, where the collected data is inevitably superimposed on the electromagnetic noise of the environment and the thermal noise of the system, and often affected by multiple types of strong interference, resulting in masking or distortion of lightning pulses in the time-frequency domain. The power line harmonic interference (Power Line Harmonic Interference, PLHI) is derived from a 50 Hz/60 Hz power supply system and a nonlinear load, higher harmonics can extend to thousands of hertz or even tens of kilohertz except a fundamental frequency, sharp and dense comb-shaped spectrum peaks are displayed on a frequency spectrum, and the comb-shaped spectrum peaks are displayed as dense horizontal spectrum lines on a time frequency domain, and the comb-shaped spectrum peaks are obviously overlapped with broadband energy of broadband electromagnetic pulses in a very low frequency range, so that noise bottoms are easily raised and weak lightning events are shielded. The narrow-band interference (Narrowband Interference, NBI) mainly comes from a high-power VLF communication station (such as NAA, NWC, NLK) with frequencies of 10-30 kHz, has the characteristics of concentrated energy, long duration, stable frequency and the like, forms a high-amplitude peak in a frequency domain, and presents a high-amplitude sine component in a time domain, and when the broadband spectrum of a lightning pulse overlaps with the frequency points, obvious spectrum masking and waveform distortion are easy to generate. In the related technology, a trap is usually constructed to inhibit PLHI/NBI by adopting spectral subtraction or interference frequency points, but a large number of comb-shaped traps easily introduce phase distortion and weaken effective broadband information of lightning pulses, and meanwhile, the fundamental frequency of a power grid is not strictly fixed at 50/60 Hz, drift exists, and the fixed frequency point traps easily have mismatch, so that interference residues or insufficient inhibition are caused. On the other hand, conventional wiener filtering generally constructs a linear optimal filter based on mean square error minimization, but under the coexistence condition of non-stationary background noise and structured interference presented by very low-frequency electric data, if adaptive estimation and constraint on noise and interference statistical characteristics are lacking, it is often difficult to achieve significant signal-to-noise ratio (SNR) improvement and waveform feature fidelity at the same time. Therefore, a noise mitigation method for very low-frequency electrical broadband data is needed, which can solve the above-mentioned technical drawbacks and provide a reliable basis for the identification, analysis and further utilization of lightning signals. Disclosure of Invention The application pr