Terahertz time-domain spectral signal multi-channel noise reduction method and device

Abstract

The invention provides a terahertz time-domain spectral signal multichannel noise reduction method and device, and relates to the technical field of terahertz time-domain spectroscopy, and the method comprises the steps: obtaining a terahertz time-domain spectral signal; denoising the terahertz time-domain spectral signal to obtain a denoised signal; performing low-pass filtering on the de-noised signal through an adaptive cut-off frequency filter to obtain a first filtered signal; performing high-pass filtering on the de-noised signal through a transmission filtering model to obtain a second filtering signal; performing median filtering on the de-noised signal according to an adaptive median filter to obtain a third filtering signal; and performing weighted fusion on the first filtering signal, the second filtering signal and the third filtering signal to obtain a terahertz time-domain spectral signal after multi-channel noise reduction. According to the invention, the problem that the traditional single-channel noise reduction technology is difficult to meet the high signal-to-noise ratio is solved.

Description

Technical Field

[0001] The present invention relates to the technical field of terahertz time-domain spectroscopy, and in particular to a method and device for multi-channel noise reduction of terahertz time-domain spectroscopy signals. Background Art

[0002] With the development of terahertz science, terahertz time-domain spectroscopy (THz-TDS) technology has shown great potential in the fields of materials science, biomedicine, security inspection, etc. due to its non-invasiveness, high penetration and unique absorption characteristics for many substances. In the study of terahertz wave skin tissue detection, noise control and signal processing are crucial links. In order to ensure the accuracy and reliability of the test results, researchers use signal averaging technology to measure the same sample multiple times and take the average value, which effectively reduces the influence of random noise and improves the signal-to-noise ratio. In order to make the signal average effective, this single-channel noise reduction method requires the measurement system to remain highly stable during multiple measurements. Any drift or change in the system may cause distortion of the averaged signal. This traditional single-channel noise reduction technology is difficult to meet the requirements of high signal-to-noise ratio, especially in complex environments, and the difficulty of signal processing is further increased. Therefore, developing an effective multi-channel noise reduction method to improve the quality of THz-TDS signals has become an urgent problem to be solved. Summary of the invention

[0003] The purpose of the present invention is to provide a method and device for multi-channel noise reduction of terahertz time-domain spectroscopy signals to improve the above-mentioned problems. In order to achieve the above-mentioned purpose, the technical solution adopted by the present invention is as follows:

[0004] In a first aspect, the present application provides a terahertz time-domain spectroscopy signal multi-channel noise reduction method, comprising:

[0005] Acquire a terahertz time-domain spectroscopy signal, wherein the terahertz time-domain spectroscopy signal is acquired by a sensor;

[0006] De-noising the terahertz time-domain spectrum signal to obtain a de-noised signal;

[0007] Constructing an adaptive cutoff frequency filter, and performing low-pass filtering on the denoised signal through the adaptive cutoff frequency filter to obtain a first filtered signal;

[0008] Constructing a transmission filter model, and performing high-pass filtering on the denoised signal through the transmission filter model to obtain a second filtered signal;

[0009] Constructing an adaptive median filter, and performing median filtering on the denoised signal according to the adaptive median filter to obtain a third filtered signal;

[0010] The first filtered signal, the second filtered signal and the third filtered signal are weightedly fused to obtain a terahertz time-domain spectrum signal after multi-channel noise reduction.

[0011] In a second aspect, the present application also provides a terahertz time-domain spectroscopy signal multi-channel noise reduction device, comprising:

[0012] An acquisition module, used for acquiring a terahertz time-domain spectrum signal, wherein the terahertz time-domain spectrum signal is acquired through a sensor;

[0013] A denoising module, used for denoising the terahertz time-domain spectrum signal to obtain a denoised signal;

[0014] A first filtering module, used to construct an adaptive cutoff frequency filter, and to perform low-pass filtering on the denoised signal through the adaptive cutoff frequency filter to obtain a first filtered signal;

[0015] A second filtering module is used to construct a transmission filtering model, and perform high-pass filtering on the denoised signal through the transmission filtering model to obtain a second filtered signal;

[0016] A third filtering module is used to construct an adaptive median filter, and perform median filtering on the denoised signal according to the adaptive median filter to obtain a third filtered signal;

[0017] A fusion module is used to perform weighted fusion on the first filtered signal, the second filtered signal and the third filtered signal to obtain a terahertz time-domain spectrum signal after multi-channel noise reduction.

[0018] The beneficial effects of the present invention are:

[0019] The present invention designs different filters for multiple channels, uses an adaptive cutoff frequency filter based on signal strength and environmental noise level, a transmission filter model based on terahertz transmission characteristics, and an adaptive median filter that adaptively adjusts the filter window for multi-channel filtering, and fuses the noise-reduced signals to obtain a composite signal with higher purity. The purity of terahertz time-domain spectroscopy signals can be significantly improved, the imaging quality can be improved, and the detection capability of weak signals can be enhanced, and a more reliable basis can be provided for subsequent data analysis.

[0020] Other features and advantages of the present invention will be set forth in the following description, and in part will become apparent from the description, or may be understood by practicing embodiments of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings required for use in the embodiments are briefly introduced below. It should be understood that the following drawings only show certain embodiments of the present invention and therefore should not be regarded as limiting the scope. For ordinary technicians in this field, other related drawings can be obtained based on these drawings without creative work.

[0022] Figure 1 This is a schematic diagram of the process of the multi-channel noise reduction method for terahertz time-domain spectroscopy signals described in an embodiment of the present invention;

[0023] Figure 2 Schematic diagram of the structure of the multi-channel noise reduction device for terahertz time-domain spectroscopy signal described in an embodiment of the present invention.

[0024] Markings in the figure: 901, acquisition module; 902, denoising module; 903, first filtering module; 904, second filtering module; 905, third filtering module; 906, fusion module; 9031, first transform unit; 9032, first setting unit; 9033, first filtering unit; 9034, second transform unit; 9041, third transform unit; 9042, second setting unit; 9043, first calculation unit; 9044, second filtering unit; 9041, fourth transform unit; 9051, second calculation unit; 9052, third setting unit; 9053, fourth setting unit; 9054, third filtering unit. DETAILED DESCRIPTION

[0025] In order to make the purpose, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, rather than all of the embodiments. The components of the embodiments of the present invention generally described and shown in the drawings here can be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of the present invention provided in the drawings is not intended to limit the scope of the claimed invention, but merely represents selected embodiments of the present invention. Based on the embodiments in the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.

[0026] It should be noted that similar reference numerals and letters represent similar items in the following drawings, so once an item is defined in one drawing, it does not need to be further defined and explained in the subsequent drawings. At the same time, in the description of the present invention, the terms "first", "second", etc. are only used to distinguish the description and cannot be understood as indicating or implying relative importance.

[0027] Embodiment 1:

[0028] This embodiment provides a multi-channel noise reduction method for terahertz time-domain spectroscopy signals.

[0029] See also Figure 1 , the figure shows that the method includes step S100, step S200, step S300, step S400, step S500, and step S600.

[0030] Step S100: Acquire a terahertz time-domain spectrum signal, wherein the terahertz time-domain spectrum signal is acquired by a sensor;

[0031] In this embodiment, the sampling formula of the terahertz time-domain spectroscopy signal is:

[0032]

[0033] In the formula, S sum represents the terahertz time-domain spectroscopy signal, N represents the number of sampling points, Δt represents the time interval between two adjacent sampling points, S(t i ) represents the i-th time point t i The signal amplitude obtained by the sensor.

[0034] Step S200: denoising the terahertz time-domain spectrum signal to obtain a denoised signal;

[0035] In this embodiment, advanced digital signal processing technology is used to process the background noise of the original terahertz time-domain spectrum signal, wherein the calculation formula of the denoised signal is:

[0036] S t =S sum -S background

[0037] In the formula, S t represents the denoised signal, S sum represents the terahertz time-domain spectroscopy signal, S background Represents background noise.

[0038] Step S300: constructing an adaptive cutoff frequency filter, and performing low-pass filtering on the denoised signal through the adaptive cutoff frequency filter to obtain a first filtered signal;

[0039] The step S300 includes:

[0040] Step S301: performing Fourier transform on the denoised signal to obtain a frequency domain signal;

[0041] In this embodiment, the time domain signal is converted to the frequency domain, wherein the formula for Fourier transform is:

[0042]

[0043] In the formula, S f represents the frequency domain signal, S t represents the denoised signal, j represents an imaginary number, f represents frequency, and t represents time.

[0044] Step S302: setting a first cutoff frequency, and calculating an adaptive cutoff frequency response according to the first cutoff frequency and the frequency domain signal;

[0045] In this embodiment, the intensity of the terahertz time-domain spectroscopy signal is monitored in real time, especially in the low-frequency part. If the signal intensity is low, it may be caused by noise or signal attenuation. At this time, the first cutoff frequency can be appropriately reduced to retain more signal components. The formula for the adaptive cutoff frequency response is:

[0046]

[0047] In the formula, H low (f) represents the adaptive cutoff frequency response about frequency f, S f represents the frequency domain signal, |·| represents the absolute value, f c_low Indicates the first cutoff frequency.

[0048] Step S303: low-pass filtering the frequency domain signal by using the adaptive cutoff frequency response to obtain a first filtered spectrum signal;

[0049] In this embodiment, the calculation formula of the first filtered spectrum signal is:

[0050] S filtered_low (f) = S f *H low (f)

[0051] In the formula, S filtered_low (f) represents the first filtered spectrum signal about frequency f, S f represents the frequency domain signal, H low (f) represents the adaptive cutoff frequency response with respect to frequency f.

[0052] Step S304: performing inverse Fourier transform on the first filtered spectrum signal to obtain a first filtered signal.

[0053] In this embodiment, the formula for performing inverse Fourier transform on the first filtered spectrum signal is:

[0054]

[0055] In the formula, S t_low represents the first filtered signal, S filtered_low(f) represents the first filtered spectrum signal about frequency f, j represents an imaginary number, f represents frequency, and t represents time.

[0056] Step S400: constructing a transmission filter model, and performing high-pass filtering on the denoised signal through the transmission filter model to obtain a second filtered signal;

[0057] In this embodiment, the terahertz time-domain spectroscopy signal will be affected by the medium during transmission. Therefore, a transmission filtering model considering the terahertz time-domain spectroscopy signal in different media is established, and special filter parameters are designed, taking into account factors such as absorption and scattering.

[0058] The step S400 includes:

[0059] Step S401: performing Fourier transform on the denoised signal to obtain a frequency domain signal;

[0060] Step S402: setting a second cutoff frequency, and calculating a high-pass frequency response according to the second cutoff frequency and the frequency domain signal;

[0061] Step S403: calculating the propagation frequency response according to the propagation distance of the terahertz time-domain spectroscopy signal and the high-pass frequency response;

[0062] In this embodiment, the calculation formula of the propagation frequency response is:

[0063]

[0064] H prop (f) = H high (f)e -αz

[0065] In the formula, H high (f) represents the high-pass frequency response about frequency f, S f represents the frequency domain signal, |·| represents the absolute value, f c_high represents the second cutoff frequency, α represents the attenuation coefficient of the medium, H prop (f) represents the propagation frequency response about frequency f, and z represents the propagation distance of the THz time-domain spectroscopy signal.

[0066] Step S404: performing high-pass filtering on the frequency domain signal according to the propagation frequency response to obtain a second filtered spectrum signal;

[0067] In this embodiment, the calculation formula of the second filtered spectrum signal is:

[0068] S filtered_high (f) = S f *H prop (f)

[0069] In the formula, S filtered_high (f) represents the first filtered spectrum signal about frequency f, S f represents the frequency domain signal, H prop (f) represents the propagation frequency response about frequency f.

[0070] Step S405: performing inverse Fourier transform on the second filtered spectrum signal to obtain a second filtered signal.

[0071] In this embodiment, the formula for performing inverse Fourier transform on the second filtered spectrum signal is:

[0072]

[0073] In the formula, S t_high represents the second filtered signal, S filtered_high (f) represents the second filtered spectrum signal about frequency f, j represents an imaginary number, f represents frequency, and t represents time.

[0074] Step S500: constructing an adaptive median filter, and performing median filtering on the denoised signal according to the adaptive median filter to obtain a third filtered signal;

[0075] In this embodiment, since the frequency of the terahertz time-domain spectroscopy signal is high and the signal changes quickly, the noise level in different areas may be different. Therefore, the size of the filter window is adaptively adjusted according to the local characteristics of the signal. The local characteristics of the signal are detected by calculating the local standard deviation or gradient. If the local standard deviation or gradient is large, it means that the signal changes quickly or there is more noise, and a smaller window is used. Otherwise, a larger window is used.

[0076] At the same time, a weighted median filter is introduced to assign different weights to each signal point to better protect the details of the signal. For each signal point in the signal, the values ​​of several signal points around it are taken, and the number in the middle after the values ​​are sorted is found, and the median value is used to replace the original signal value.

[0077] The step S500 includes:

[0078] Step S501: obtaining the local characteristics of each signal point in the denoised signal by calculating the local standard deviation of the denoised signal;

[0079] Step S502: setting the size of the filtering window of each signal point according to the local characteristics;

[0080] Step S503: assigning different window weights to filter windows of different sizes;

[0081] Step S504: performing median filtering on all signal points in the denoised signal using the filtering window and the corresponding window weight to obtain a third filtered signal.

[0082] The step S504 includes:

[0083] Step A100: selecting filtering signal points around each signal point according to the size of the filtering window of the signal point;

[0084] Step A200: sorting the signal point and the filtered signal point according to signal strength to obtain a signal point subsequence;

[0085] In this embodiment, the size of the filtering window is 2k+1, that is, 2k+1 signal points are taken out to form a subsequence, and the subsequence includes the signal point at the center of the filtering window and the filtered signal points around the signal point, where k represents the index parameter of the signal point at the center of the filtering window.

[0086] Step A300: selecting a median of a signal point subsequence, and adjusting the median by using a window weight of a filter window of the signal point to obtain an adjusted median, and using the adjusted median to replace the signal point;

[0087] Step A400: After replacing each signal point in the denoised signal, a third filtered signal is obtained.

[0088] In this embodiment, the calculation formula for adjusting the median is:

[0089]

[0090] S t_mid =W(b)med{S[bk], S[b-k+1],…, S[b],…, S[b+k-1], S[b+k]}

[0091] In the formula, S t_mid represents the third filtered signal, W(b) represents the window weight of the b-th signal point, k represents the index parameter of the central signal point of the filtering window, σ represents the standard deviation of the window weight, med{·} represents the median operation, S[c] represents the c-th signal point of the denoised signal, c=bk,b-k+1,…,b+k.

[0092] Step S600: performing weighted fusion on the first filtered signal, the second filtered signal and the third filtered signal to obtain a terahertz time-domain spectrum signal after multi-channel noise reduction.

[0093] In this embodiment, corresponding weights are assigned according to the characteristics of signals in different channels, and noise reduction is achieved through weighted averaging, that is, the multi-channel filtered signals after noise reduction are synthesized to obtain the multi-channel terahertz time-domain spectrum signal after noise reduction.

[0094] The step S600 includes:

[0095] Step S601: setting weight parameters of the first filtered signal, the second filtered signal and the third filtered signal;

[0096] Step S602: weighted fusion is performed on the first filtered signal, the second filtered signal and the third filtered signal by using the weight parameter to obtain a terahertz time-domain spectrum signal after multi-channel noise reduction.

[0097] In this embodiment, the calculation formula of the terahertz time-domain spectrum signal after multi-channel noise reduction is:

[0098]

[0099] In the formula, S t_multi represents the terahertz time-domain spectral signal after multi-channel noise reduction, W low represents the first weight parameter, W high represents the second weight parameter, W mid represents the third weight parameter, S t_low represents the first filtered signal, S t_high represents the second filtered signal, S t_mid represents the third filtered signal.

[0100] In this embodiment, a training method for a multi-channel denoising method for a terahertz time-domain spectral signal is also provided, including setting initial weight parameters of an adaptive cutoff frequency filter, a transmission filter model and an adaptive median filter, preparing subcutaneous tissue samples, and testing 1,000 groups of signals, of which 700 groups are training sets and 300 groups are test sets. At the same time, the multi-channel denoising method for a terahertz time-domain spectral signal designed by the present invention is integrated into a denoising model.

[0101] Update the denoising model and set the first weight parameter W low , the second weight parameter W high and the third weight parameter W mid As the update parameter. Calculate the mean square error between the terahertz time-domain spectral signal after multi-channel denoising and the original terahertz time-domain spectral signal, and use the minimum mean square error as the loss function to obtain the final weight parameter. At the same time, in the unsupervised training process, the signal-to-noise ratio is used to measure the denoising performance, and the signal-to-noise ratio changes in the iterative process are compared. The calculation formula of the signal-to-noise ratio is:

[0102]

[0103] In the formula, SNR represents signal-to-noise ratio, P signal Represents the signal power, P noiseRepresents the noise power.

[0104] Before training begins, use the noisy signal to calculate the initial signal-to-noise ratio, perform forward propagation, input the noisy signal into the denoising model, obtain the denoised signal, calculate the power of the denoised signal and the power of the noise, use the above formula to calculate the signal-to-noise ratio of the current denoising model, perform back propagation according to the loss function, and update the first weight parameter W low , the second weight parameter W high and the third weight parameter W mid After each iteration, a new signal-to-noise ratio is calculated. If the signal-to-noise ratio does not improve significantly after multiple consecutive iterations, the training is stopped, the final weight parameters are obtained, and the multi-channel denoising of the terahertz time-domain spectroscopy signal is performed through the trained denoising model.

[0105] In summary, the noise reduction method designed by the present invention can improve the signal-to-noise ratio of the terahertz time-domain spectroscopy signal, and can significantly improve the purity of the terahertz time-domain spectroscopy signal, improve the imaging quality, thereby enhancing its performance in various applications, and is particularly suitable for the field of biomedical research that requires high-precision detection. At the same time, it can enhance the detection capability of weak signals, provide a more reliable basis for subsequent data analysis, and promote the in-depth application of terahertz technology.

[0106] Embodiment 2:

[0107] like Figure 2 As shown, this embodiment provides a terahertz time-domain spectroscopy signal multi-channel noise reduction device, the device comprising:

[0108] An acquisition module, used for acquiring a terahertz time-domain spectrum signal, wherein the terahertz time-domain spectrum signal is acquired through a sensor;

[0109] A denoising module, used for denoising the terahertz time-domain spectrum signal to obtain a denoised signal;

[0110] A first filtering module, used to construct an adaptive cutoff frequency filter, and to perform low-pass filtering on the denoised signal through the adaptive cutoff frequency filter to obtain a first filtered signal;

[0111] A second filtering module is used to construct a transmission filtering model, and perform high-pass filtering on the denoised signal through the transmission filtering model to obtain a second filtered signal;

[0112] A third filtering module is used to construct an adaptive median filter, and perform median filtering on the denoised signal according to the adaptive median filter to obtain a third filtered signal;

[0113] A fusion module is used to perform weighted fusion on the first filtered signal, the second filtered signal and the third filtered signal to obtain a terahertz time-domain spectrum signal after multi-channel noise reduction.

[0114] The first filtering module comprises:

[0115] A first transform unit, configured to perform Fourier transform on the denoised signal to obtain a frequency domain signal;

[0116] A first setting unit, configured to set a first cutoff frequency, and calculate an adaptive cutoff frequency response according to the first cutoff frequency and the frequency domain signal;

[0117] A first filtering unit, configured to perform low-pass filtering on the frequency domain signal through the adaptive cutoff frequency response to obtain a first filtered spectrum signal;

[0118] The second transform unit is used to perform inverse Fourier transform on the first filtered spectrum signal to obtain a first filtered signal.

[0119] The second filtering module comprises:

[0120] A third transform unit, configured to perform Fourier transform on the denoised signal to obtain a frequency domain signal;

[0121] A second setting unit, used for setting a second cutoff frequency, and calculating a high-pass frequency response according to the second cutoff frequency and the frequency domain signal;

[0122] A first calculation unit, configured to calculate a propagation frequency response according to a propagation distance of a terahertz time-domain spectroscopy signal and the high-pass frequency response;

[0123] A second filtering unit, configured to perform high-pass filtering on the frequency domain signal through the propagation frequency response to obtain a second filtered spectrum signal;

[0124] The fourth transform unit is used to perform inverse Fourier transform on the second filtered spectrum signal to obtain a second filtered signal.

[0125] The third filtering module comprises:

[0126] A second calculation unit, configured to obtain a local characteristic of each signal point in the denoised signal by calculating a local standard deviation of the denoised signal;

[0127] A third setting unit, used to set the size of the filter window of each signal point according to the local characteristics;

[0128] A fourth setting unit, used for assigning different window weights to filter windows of different sizes;

[0129] The third filtering unit is used to perform median filtering on all signal points in the denoised signal through the filtering window and the corresponding window weight to obtain a third filtered signal.

[0130] It should be noted that, regarding the device in the above embodiment, the specific manner in which each module performs the operation has been described in detail in the embodiment of the method, and will not be elaborated here.

[0131] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. For those skilled in the art, the present invention may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

[0132] The above description is only a specific implementation mode of the present invention, but the protection scope of the present invention is not limited thereto. Any technician familiar with the technical field can easily think of changes or substitutions within the technical scope disclosed by the present invention, which should be covered by the protection scope of the present invention.

Claims

1. A multi-channel noise reduction method for terahertz time-domain spectroscopy signals, characterized in that: include: Acquire a terahertz time-domain spectroscopy signal, wherein the terahertz time-domain spectroscopy signal is acquired by a sensor; De-noising the terahertz time-domain spectrum signal to obtain a de-noised signal; Constructing an adaptive cutoff frequency filter, and performing low-pass filtering on the denoised signal through the adaptive cutoff frequency filter to obtain a first filtered signal; Constructing a transmission filter model, and performing high-pass filtering on the denoised signal through the transmission filter model to obtain a second filtered signal; Constructing an adaptive median filter, and performing median filtering on the denoised signal according to the adaptive median filter to obtain a third filtered signal; The first filtered signal, the second filtered signal and the third filtered signal are weightedly fused to obtain a terahertz time-domain spectrum signal after multi-channel noise reduction.

2. The multi-channel denoising method for terahertz time-domain spectroscopy signal according to claim 1, characterized in that , the denoising signal is low-pass filtered by the adaptive cutoff frequency filter to obtain a first filtered signal, comprising: Performing Fourier transform on the denoised signal to obtain a frequency domain signal; Setting a first cutoff frequency, and calculating an adaptive cutoff frequency response according to the first cutoff frequency and the frequency domain signal; Performing low-pass filtering on the frequency domain signal through the adaptive cutoff frequency response to obtain a first filtered spectrum signal; Perform an inverse Fourier transform on the first filtered spectrum signal to obtain a first filtered signal.

3. The multi-channel denoising method for terahertz time-domain spectroscopy signal according to claim 1, characterized in that , the denoised signal is subjected to high-pass filtering through the transmission filter model to obtain a second filtered signal, comprising: Performing Fourier transform on the denoised signal to obtain a frequency domain signal; Setting a second cutoff frequency, and calculating a high-pass frequency response according to the second cutoff frequency and the frequency domain signal; Calculate the propagation frequency response by the propagation distance of the terahertz time-domain spectroscopy signal and the high-pass frequency response; Performing high-pass filtering on the frequency domain signal through the propagation frequency response to obtain a second filtered spectrum signal; Perform an inverse Fourier transform on the second filtered spectrum signal to obtain a second filtered signal.

4. The multi-channel denoising method for terahertz time-domain spectroscopy signal according to claim 1, characterized in that , performing median filtering on the denoised signal according to the adaptive median filter to obtain a third filtered signal, comprising: Obtaining the local characteristics of each signal point in the denoised signal by calculating the local standard deviation of the denoised signal; According to the local characteristics, setting the size of the filtering window of each signal point; Assign different window weights to filter windows of different sizes; Median filtering is performed on all signal points in the denoised signal using the filter window and the corresponding window weight to obtain a third filtered signal.

5. The multi-channel noise reduction method for terahertz time-domain spectroscopy signal according to claim 4 is characterized in that , performing median filtering on each signal point in the denoised signal through the filtering window and the corresponding window weight to obtain a third filtered signal, including: According to the size of the filtering window of each signal point, filter signal points around the signal point are selected; Sort the signal point and the filtered signal point according to signal strength to obtain a signal point subsequence; Selecting a median value of a signal point subsequence, and adjusting the median value by using a window weight of a filter window of the signal point to obtain an adjusted median value, and replacing the signal point with the adjusted median value; After replacing each signal point in the denoised signal, a third filtered signal is obtained.

6. The multi-channel noise reduction method for terahertz time-domain spectroscopy signal according to claim 1, characterized in that The weighted fusion of the first filtered signal, the second filtered signal and the third filtered signal to obtain a multi-channel denoised terahertz time domain spectrum signal includes: Setting weight parameters of the first filtered signal, the second filtered signal and the third filtered signal; The first filtered signal, the second filtered signal and the third filtered signal are weightedly fused by using the weight parameter to obtain a terahertz time-domain spectrum signal after multi-channel noise reduction.

7. A multi-channel noise reduction device for terahertz time-domain spectroscopy signal, characterized in that: include: An acquisition module, used for acquiring a terahertz time-domain spectrum signal, wherein the terahertz time-domain spectrum signal is acquired through a sensor; A denoising module, used for denoising the terahertz time-domain spectrum signal to obtain a denoised signal; A first filtering module, used to construct an adaptive cutoff frequency filter, and to perform low-pass filtering on the denoised signal through the adaptive cutoff frequency filter to obtain a first filtered signal; A second filtering module is used to construct a transmission filtering model, and perform high-pass filtering on the denoised signal through the transmission filtering model to obtain a second filtered signal; A third filtering module is used to construct an adaptive median filter, and perform median filtering on the denoised signal according to the adaptive median filter to obtain a third filtered signal; A fusion module is used to perform weighted fusion on the first filtered signal, the second filtered signal and the third filtered signal to obtain a terahertz time-domain spectrum signal after multi-channel noise reduction.

8. The multi-channel noise reduction device for terahertz time-domain spectroscopy signals according to claim 7, characterized in that: The first filtering module comprises: A first transform unit, configured to perform Fourier transform on the denoised signal to obtain a frequency domain signal; A first setting unit, configured to set a first cutoff frequency, and calculate an adaptive cutoff frequency response according to the first cutoff frequency and the frequency domain signal; A first filtering unit, configured to perform low-pass filtering on the frequency domain signal through the adaptive cutoff frequency response to obtain a first filtered spectrum signal; The second transform unit is used to perform inverse Fourier transform on the first filtered spectrum signal to obtain a first filtered signal.

9. The multi-channel noise reduction device for terahertz time-domain spectroscopy signals according to claim 7, characterized in that: The second filtering module comprises: A third transform unit, configured to perform Fourier transform on the denoised signal to obtain a frequency domain signal; A second setting unit, used for setting a second cutoff frequency, and calculating a high-pass frequency response according to the second cutoff frequency and the frequency domain signal; A first calculation unit, configured to calculate a propagation frequency response according to a propagation distance of a terahertz time-domain spectroscopy signal and the high-pass frequency response; A second filtering unit, configured to perform high-pass filtering on the frequency domain signal through the propagation frequency response to obtain a second filtered spectrum signal; The fourth transform unit is used to perform inverse Fourier transform on the second filtered spectrum signal to obtain a second filtered signal.

10. The multi-channel noise reduction device for terahertz time-domain spectroscopy signals according to claim 7, characterized in that: The third filtering module comprises: A second calculation unit, configured to obtain a local characteristic of each signal point in the denoised signal by calculating a local standard deviation of the denoised signal; A third setting unit, used to set the size of the filtering window of each signal point according to the local characteristics; A fourth setting unit, used for assigning different window weights to filter windows of different sizes; The third filtering unit is used to perform median filtering on all signal points in the denoised signal through the filtering window and the corresponding window weight to obtain a third filtered signal.

Images