A method for noise floor estimation and threshold calculation for dense spectrum signals with low signal-to-noise ratio

Abstract

The present invention relates to the field of signal detection technology, and more particularly to a noise floor estimation and threshold calculation method for a low signal-to-noise ratio (SNR) spectrally dense signal. The method determines the noise power spectrum density by statistically analyzing a histogram of the minimum to mean interval and finding the histogram maximum value. This method avoids the problem of misdetecting the power spectrum of a broadband signal as the noise floor when the signal width accounts for a large proportion of the entire spectrum width, addresses the shortcoming that most noise floor estimation algorithms cannot detect dense signals on the spectrum, and improves the accuracy of noise floor estimation.

Description

Technical Field

[0001] The present invention relates to the field of signal detection technology, and in particular to a method for estimating the noise floor and calculating a threshold of a low signal-to-noise ratio (SNR) spectrum-dense signal. Background Art

[0002] In communication electronic warfare systems, the first step in signal processing is to determine the existence of the signal, which is signal detection.

[0003] The paper "Threshold Algorithms for Wideband Reconnaissance Signal Detection" points out that Fourier receivers have been proven to be optimal for the presence of a large number of signals. Furthermore, the noise threshold detection method presented in the paper requires knowledge of the noise subband or a certain probability distribution, but such prior information is generally unavailable for blind detection. The N-sigma threshold algorithm mentioned in the paper is suitable for situations with a small number of signals and very narrow bandwidths, but is not suitable when the signal spectrum occupies a large proportion of the entire spectrum.

[0004] The CME algorithm, described in another paper, "Research on Interference Detection Algorithms in Communication Countermeasures," divides frequencies into two sets: a noise set and an interfered set. It iteratively extracts frequencies above a threshold from the noise set, adds them to the interfered set, and then recalculates the threshold to remove them until no frequencies are removed. This method iteratively removes signals from the power spectral density, gradually modifying the noise floor estimate. However, this method suffers from poor noise floor estimation accuracy.

[0005] To address the above-mentioned shortcomings, the present application provides a noise floor estimation and threshold calculation method for dense signals in low signal-to-noise ratio spectrum, which solves the shortcomings of most noise floor estimation algorithms that cannot detect dense signals on the spectrum, improves the noise floor estimation accuracy, etc. Summary of the Invention

[0006] The purpose of the present invention is to avoid the problem of misdetecting the power spectrum of a broadband signal as the noise floor when the signal width occupies a large proportion of the entire spectrum width. It solves the shortcomings of most noise floor estimation algorithms that cannot detect dense signals on the spectrum and the poor noise floor estimation accuracy.

[0007] To achieve the above object, the present invention provides the following technical solutions:

[0008] A method for estimating the noise floor and calculating the threshold of a dense signal with a low signal-to-noise ratio spectrum comprises the following steps:

[0009] Step 1: Use the periodogram method to perform power spectrum estimation on the baseband IQ signal, and the power spectrum estimation result is psd;

[0010] Step 2: Calculate the mean of the power spectrum psd, recorded as psd_mean; find the minimum value of psd, recorded as psd_min, and normalize the values of psd falling into the interval [psd_min,psd_mean] to the interval [0,1];

[0011] Step 3: For the above normalized [0,1] interval, count the number of items K that fall into each interval according to the number of divided intervals P. i ,i=1,2,3...P,get the histogram of normalized power spectrum;

[0012] Step 4: Find the interval corresponding to the maximum value of the power spectrum histogram, denormalize the interval to obtain the noise floor estimate S noise ;

[0013] Step 5: Based on the noise power spectrum estimate obtained in the previous step, calculate th1 and th2 as follows:

[0014] th1=K1*Snoise+C1*(psd_mean-psd_min);

[0015] th2=K2*Snoise+C2*(psd_mean-psd_min);

[0016] Among them, K1 and K2 are threshold multiplication factors, for example, K1=6, K2=12, C1 and C2 are DC bias coefficients much smaller than 1. The role of C1 and C2 is to avoid the problem of too small threshold calculation when the signal-to-noise ratio is extremely high.

[0017] A further improvement of the technical solution of the present invention is that in step 1, the IQ signal is subjected to an N_fft-point FFT, the modulus square is taken, and the modulus square is averaged N_mean times, N_mean ≥ 64, and FFT(k) represents the k-th point FFT result of the signal; the power spectrum estimate psd(k) of the signal is obtained:

[0018]

[0019] Wherein, the letters represent the meanings: psd(k) represents the estimated power spectral density value at the kth point, N_fft represents the number of Fourier transform points, N_mean represents the number of Fourier transform smoothing times, i represents the i-th Fourier transform, and |·| represents the modulus.

[0020] The further improvement of the technical solution of the present invention is that: in step 2, the minimum value psd_min and the mean value psd_mean of the power spectrum psd are calculated, and the interval from the minimum value to the mean value [psd_min, psd_mean] is normalized to the interval [0,1], and the normalized interval is recorded as psd [0,1] ,psd [0,1] It can be expressed as follows:

[0021]

[0022] In the formula, the letters express the meaning: psd [0,1] (k), represents the normalized value of the power spectrum at point k, the subscript [0,1] represents normalization, psd_min is the minimum value of psd, and psd_mean is the mean value of psd. Nk is the number of psds in the interval [psd_min,psd_mean], which means that only the power spectra smaller than psd_mean are normalized. A further improvement of the technical solution of the present invention is that: in step 3, psd [0,1] Divide into P intervals and calculate psd [0,1] The number Lj falling into each interval, the detailed steps are:

[0023] 1. Initialize the number of items falling into each interval to 0, and record Lj=0, j=1, 2, 3…, P;

[0024] 2. Calculate each PSD [0,1] (k) The interval number j belongs to = floor(P×psd [0,1] (k))+1, the number of times corresponding to interval number j is incremented by 1, i.e. L j =L j +1, where floor(·) means rounding down.

[0025] A further improvement of the technical solution of the present invention is that: in step 4, find L j The j corresponding to the maximum value is recorded as j max , for j max Do inverse normalization to get the noise average power spectral density S noise :

[0026]

[0027] In the formula, the letters mean: S noise represents the estimated noise power spectral density, j max Represents the normalized power spectrum psd [0,1] The interval number corresponding to the maximum value of the histogram, P represents the number of intervals in which the power spectrum histogram is divided, psd_mean is the mean of the power spectrum density, and psd_min is the minimum value of the power spectrum density.

[0028] The theoretical basis is that psd obeys the chi-square distribution with N_mean degrees of freedom. When the number of smoothing times N_mean is large, psd tends to normal distribution, and the probability density function of the normal distribution reaches its maximum at the expected point.

[0029] A further improvement of the technical solution of the present invention is that in step 5, the upper and lower thresholds are calculated, and the lower and upper threshold multiples are set to K1 and K2, and C1 and C2 are fixed DC biases. The calculation method of the lower threshold th1 and the upper threshold th2 is:

[0030] th1=K1·S noise +C1·(psd_mean-psd_min)

[0031] th2=K2·S noise + C2 (psd_mean - psd_min)

[0032] Where, the letters in the formula represent the following meanings: th1 represents the calculated lower threshold, th2 represents the calculated upper threshold, C1 represents the lower threshold DC offset parameter, and C2 represents the upper threshold DC offset parameter. psd_min represents the minimum power spectral density, and psd_mean represents the mean power spectral density. K1 represents the lower threshold noise multiple parameter, and K2 represents the upper threshold noise multiple parameter.

[0033] Compared with the prior art, the noise floor estimation and threshold calculation method for a low signal-to-noise ratio spectrum dense signal provided by the present invention has the following beneficial effects:

[0034] 1. The present invention provides a noise floor estimation and threshold calculation method for dense signals with low signal-to-noise ratio (SNR). This method avoids the problem of misdetecting the power spectrum of a broadband signal as the noise floor when the signal width accounts for a large proportion of the entire spectrum width. It solves the shortcoming that most noise floor estimation algorithms cannot detect dense signals on the spectrum, and improves the accuracy of noise floor estimation. BRIEF DESCRIPTION OF THE DRAWINGS

[0035] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0036] Figure 1 The present invention provides an overall signal processing block diagram of a noise floor estimation and threshold calculation method for a low signal-to-noise ratio spectrum dense signal.

[0037] Figure 2 The present invention provides a statistical histogram of the power spectrum of a method for estimating the noise floor and calculating the threshold of a dense signal with a low signal-to-noise ratio spectrum.

[0038] Figure 3 The present invention provides a method for estimating the noise floor and calculating the threshold of a low signal-to-noise ratio spectrum dense signal, as well as a calculation result diagram of the thresholds th1 and th2. DETAILED DESCRIPTION

[0039] The technical solutions of the present invention will be clearly and completely described below through specific implementation methods. Obviously, the described embodiments are only some of the embodiments of the present invention, not all of them. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making any creative efforts are within the scope of protection of the present invention.

[0040] like Figure 1-3 As shown, this embodiment provides a method for estimating the noise floor and calculating the threshold of a low signal-to-noise ratio spectrum dense signal, including the following steps:

[0041] Step 1: Use the periodogram method to perform power spectrum estimation on the baseband IQ signal, and the power spectrum estimation result is psd;

[0042] Step 2: Calculate the mean of the power spectrum psd, recorded as psd_mean; find the minimum value of psd, recorded as psd_min, and normalize the values of psd falling into the interval [psd_min,psd_mean] to the interval [0,1];

[0043] Step 3: For the above normalized [0,1] interval, count the number of items K that fall into each interval according to the number of divided intervals P. i ,i=1,2,3...P,get the histogram of normalized power spectrum;

[0044] Step 4: Find the interval corresponding to the maximum value of the power spectrum histogram, denormalize the interval to obtain the noise floor estimate S noise ;

[0045] Step 5: Based on the noise power spectrum estimate obtained in the previous step, calculate th1 and th2 as follows:

[0046] th1=K1*Snoise+C1*(psd_mean-psd_min);

[0047] th2=K2*Snoise+C2*(psd_mean-psd_min);

[0048] Among them, K1 and K2 are threshold multiplication factors, for example, K1=6, K2=12, C1 and C2 are DC bias coefficients much smaller than 1. The role of C1 and C2 is to avoid the problem of too small threshold calculation when the signal-to-noise ratio is extremely high.

[0049] This embodiment provides a method for estimating the noise floor and calculating the threshold of a dense signal with a low signal-to-noise ratio spectrum:

[0050] Step 1: Perform N_fft-point FFT on the IQ signal, square the modulus, and average the modulus squares N_mean times, where N_mean ≥ 64. FFT(k) represents the k-th point FFT result of the signal. The power spectrum estimate psd(k) of the signal is obtained:

[0051]

[0052] Wherein, the letters represent the meanings: psd(k) represents the power spectral density value at the k-th point, N_fft represents the number of Fourier transform points, N_mean represents the number of Fourier transform smoothing times, i represents the i-th Fourier transform, and |·| represents the modulus.

[0053] Step 2: Find the minimum value psd_min and mean psd_mean of the power spectrum psd, and normalize the interval from the minimum value to the mean [psd_min,psd_mean] to the interval [0,1]. The normalized interval is recorded as psd [0,1] ,psd [0,1] It can be expressed as follows:

[0054]

[0055] In the formula, the letters express the meaning: psd [0,1] (k) represents the normalized power spectrum value at point k. The subscript [0,1] indicates normalization. psd_min is the minimum psd value, and psd_mean is the mean psd value. Nk is the number of psd values in the interval [psd_min,psd_mean], meaning that only power spectrum values less than psd_mean are normalized.

[0056] Step 3, edit the psd [0,1] Divide into P intervals and calculate psd [0,1] The number Lj falling into each interval, the detailed steps are:

[0057] 1. Initialize the number of items falling into each interval to 0, and record Lj=0, j=1, 2, 3…, P;

[0058] 2. Calculate each PSD [0,1] (k) The interval number j belongs to = floor(P×psd [0,1] (k))+1, the number of times corresponding to interval number j is incremented by 1, i.e. L j =L j +1, where floor(·) means rounding down.

[0059] Step 4, find L j The j corresponding to the maximum value is recorded as j max , for j max Do inverse normalization to get the noise average power spectral density Snoise :

[0060]

[0061] In the formula, the letters mean: S noise represents the estimated noise power spectral density, j max Represents the normalized power spectrum psd [0,1] The interval number corresponding to the maximum value of the histogram, P represents the number of intervals in which the power spectrum histogram is divided, psd_mean is the mean of the power spectrum density, and psd_min is the minimum value of the power spectrum density.

[0062] The theoretical basis is that psd obeys the chi-square distribution with N_mean degrees of freedom. When the number of smoothing times N_mean is large, psd tends to normal distribution, and the probability density function of the normal distribution reaches its maximum at the expected point.

[0063] Step 5, calculate the high threshold and low threshold, set the low threshold and high threshold multiples to K1 and K2, C1 and C2 are fixed DC bias, the calculation method of the low threshold th1 and the high threshold th2 is:

[0064] th1=K1·S noise +C1·(psd_mean-psd_min)

[0065] th2=K2·S noise + C2 (psd_mean - psd_min)

[0066] Wherein, the letters in the formula represent the following meanings: th1 represents the calculated low threshold, th2 represents the calculated high threshold, C1 is the low threshold DC bias parameter, C2 is the high threshold DC bias parameter, psd_min is the minimum power spectral density, psd_mean is the mean power spectral density, K1 is the low threshold noise multiple parameter, and K2 is the high threshold noise multiple parameter.

[0067] The embodiments described above are merely descriptions of preferred implementations of the present invention and are not intended to limit the scope of the present invention. Without departing from the spirit of the present invention, various modifications and improvements made to the technical solutions of the present invention by ordinary technicians in this field should fall within the scope of protection determined by the claims of the present invention.

Claims

1. A method for estimating the noise floor and calculating the threshold of a dense signal with a low signal-to-noise ratio spectrum, characterized by: The steps include: Step 1: Use the periodogram method to perform power spectrum estimation on the baseband IQ signal, and the power spectrum estimation result is psd; Step 2: Calculate the mean of the power spectrum psd, recorded as psd_mean; find the minimum value of psd, recorded as psd_min, and normalize the values of psd falling into the interval [psd_min,psd_mean] to the interval [0,1]; Step 3: For the above normalized [0,1] interval, count the number of items K that fall into each interval according to the number of divided intervals P. i ,i=1,2,3...P,get the histogram of normalized power spectrum; Step 4: Find the interval corresponding to the maximum value of the power spectrum histogram, denormalize the interval to obtain the noise floor estimate S noise ; Step 5: Based on the noise power spectrum estimate obtained in the previous step, calculate th1 and th2 as follows: th1=K1*Snoise+C1*(psd_mean-psd_min); th2=K2*Snoise+C2*(psd_mean-psd_min); Among them, K1 and K2 are threshold multiplication factors, C1 and C2 are DC bias coefficients much smaller than 1. The role of C1 and C2 is to avoid the problem of the threshold being too small when the signal-to-noise ratio is extremely high.

2. The method for estimating the noise floor and calculating the threshold of a dense spectrum signal with a low signal-to-noise ratio according to claim 1, characterized in that: In step 1, perform N_fft-point FFT on the IQ signal, square the modulus, and average the modulus squares N_mean times, where N_mean ≥ 64. FFT(k) represents the k-th FFT result of the signal. The power spectrum estimate psd(k) of the signal is obtained: Wherein, the letters represent the meanings: psd(k) represents the estimated power spectral density value at the kth point, N_fft represents the number of Fourier transform points, N_mean represents the number of Fourier transform smoothing times, i represents the i-th Fourier transform, and |·| represents the modulus.

3. The method for estimating the noise floor and calculating the threshold of a dense signal with a low signal-to-noise ratio spectrum according to claim 1, characterized in that: In step 2, find the minimum value psd_min and mean value psd_mean of the power spectrum psd(k), normalize the interval from the minimum value to the mean value [psd_min,psd_mean] of the power spectrum to the interval [0,1], and record the normalized power spectrum as psd [0,1] ,psd [0,1] It can be expressed as follows: In the formula, the letters express the meaning: psd [0,1] (k) represents the normalized value of the power spectrum at the kth point, the subscript [0,1] indicates normalization, psd_min is the minimum value of psd, psd_mean is the mean value of psd, Nk is the number of psds in the interval [psd_min,psd_mean], meaning that only power spectra smaller than psd_mean are normalized, and N_fft represents the number of Fourier transform points.

4. The method for estimating the noise floor and calculating the threshold of a dense signal with a low signal-to-noise ratio spectrum according to claim 1, characterized in that: In step 3, psd [0,1] Divide into P intervals and calculate psd [0,1] The number Lj falling into each interval, the detailed steps are:

1. Initialize the number of items falling into each interval to 0, and record Lj=0, j=1, 2, 3…, P; 2. Calculate each PSD [0,1] (k) The interval number j belongs to = floor(P×psd [0,1] (k))+1, the number of times corresponding to interval number j is incremented by 1, i.e. L j =L j +1, where floor(·) means rounding down.

5. The method for estimating the noise floor and calculating the threshold of a dense signal with a low signal-to-noise ratio spectrum according to claim 1, characterized in that: In step 4, find L j The j corresponding to the maximum value is recorded as j max , for j max Do inverse normalization to get the noise average power spectral density S noise : In the formula, the letters mean: S noise represents the estimated noise power spectral density, j max Represents the normalized power spectrum psd [0,1] (k) The interval number corresponding to the maximum value of the histogram, P represents the number of intervals divided by the power spectrum histogram, psd_mean is the mean value of power spectrum density, psd_min is the minimum value of power spectrum density, The theoretical basis is that psd obeys the chi-square distribution with N_mean degrees of freedom. When the number of smoothing times N_mean is large, psd tends to normal distribution, and the probability density function of the normal distribution reaches its maximum at the expected point.

6. The method for estimating the noise floor and calculating the threshold of a dense signal with a low signal-to-noise ratio spectrum according to claim 1, characterized in that: In step 5, the upper and lower thresholds are calculated. The lower and upper threshold multipliers are assumed to be K1 and K2. C1 and C2 are fixed DC bias parameters. The DC bias increases the threshold when the signal-to-noise ratio is high to prevent the threshold from being too low. C1 and C2 are numbers much smaller than 1. The calculation method for the lower and upper thresholds th1 and th2 is as follows: th1=K1·S noise +C1·(psd_mean-psd_min) th2=K2·S noise +C2·(psd_mean-psd_min) Wherein, the letters in the formula represent the following meanings: th1 represents the calculated low threshold, th2 represents the calculated high threshold, C1 is the low threshold DC bias parameter, C2 is the high threshold DC bias parameter, psd_min is the minimum value of the power spectral density, psd_mean is the mean value of the power spectral density, K1 is the low threshold noise multiple parameter, and K2 is the high threshold noise multiple parameter.

Images