An improved wavelet denoising method based on the Sagnac fiber optic acoustic sensing system

Abstract

The present invention discloses an improved wavelet denoising method based on a Sagnac fiber acoustic sensing system. The method includes: acquiring an original sound signal; decomposing the original sound signal by using a wavelet denoising method to obtain a decomposed sound signal; filtering the decomposed sound signal by using an improved wavelet threshold denoising algorithm, and then performing wavelet reconstruction to obtain a filtered noisy sound signal; and performing low-pass filtering on the filtered noisy sound signal to obtain a noise-reduced sound signal. The present invention can improve the quality of the acquired sound signal.

Description

Technical Field

[0001] The present invention relates to the field of sound noise reduction, and particularly to an improved wavelet denoising method based on a Sagnac fiber acoustic sensing system. Background Art

[0002] The fiber optic acoustic sensing system is a new type of sound signal sensor constructed based on the principle of fiber optic vibration sensing. It can effectively solve many deficiencies of traditional acoustic sensors, such as the problem of being difficult to use in extremely harsh environments such as strong magnetic interference, strong corrosion, and high humidity. Therefore, fiber optic acoustic sensors can be widely used in important fields such as aviation detection, energy transmission, and perimeter security.

[0003] In practical applications, the sound signals collected by the fiber optic acoustic sensing system will inevitably be affected by noise interference, which will greatly affect the system performance. Therefore, it is necessary to filter out the noise. To improve the quality of the collected sound signals, many experts and scholars have proposed various solutions. Some scholars designed a micro-bent fiber optic microphone sensor using the principle of fiber optic micro-bending loss, but its frequency characteristics and dynamic range are relatively narrow; some scholars designed a fiber optic acoustic sensor based on a Sagnac fiber interferometer using a 3×3 fiber optic coupler, but the signal demodulation of this system is complex and the signal quality after demodulation is poor; some other scholars designed a Sagnac / Φ-OTDR hybrid fiber optic acoustic sensor, but it cannot achieve the adaptation of voice and positioning. Summary of the Invention

[0004] Based on the above facts, it is necessary to provide an improved wavelet denoising method based on a linear Sagnac fiber acoustic sensing system to improve the quality of sound signals for the above problems.

[0005] To achieve the above object, the present invention provides the following technical solutions:

[0006] An improved wavelet denoising method based on a Sagnac fiber acoustic sensing system, comprising:

[0007] Obtain the original sound signal;

[0008] Decompose the original sound signal using a wavelet denoising method to obtain a decomposed sound signal containing noise signals;

[0009] Filter the decomposed sound signal using an improved wavelet threshold denoising algorithm, and then perform wavelet reconstruction to obtain a filtered noisy sound signal;

[0010] Perform a further low-pass filtering process on the filtered noisy sound signal to obtain a noise-reduced sound signal.

[0011] Further, the decomposed noisy voice signal is filtered by using an improved wavelet threshold denoising algorithm, and then wavelet reconstruction is performed to obtain the filtered noisy voice signal, which specifically includes:

[0012] Determine the wavelet basis function to perform wavelet decomposition on the signal, and obtain high-frequency wavelet coefficients and low-frequency wavelet coefficients of different layers according to the wavelet basis function and the decomposition level;

[0013] Use the improved threshold function and the improved threshold to remove the high-frequency wavelet coefficients in the decomposed signal, while retaining the low-frequency wavelet coefficients;

[0014] Reconstruct the retained wavelet coefficients to obtain the denoised signal.

[0015] Further, the expression of the improved threshold function is:

[0016]

[0017] Where: The parameter α ∈ [0, ∞] is the approaching speed adjustment factor, and the smaller α is, the slower the approaching speed is.

[0018] Further, the expression of the improved threshold is:

[0019]

[0020] In the formula: i is the wavelet decomposition scale;

[0021] In the above formula, the threshold T of the wavelet coefficient i Shows a slow downward trend as the decomposition scale i increases.

[0022] Compared with the prior art, the improved wavelet denoising method based on the linear Sagnac fiber acoustic sensing system of the present invention obtains the original voice signal; uses the wavelet denoising method to decompose the original voice signal to obtain the decomposed voice signal; uses the improved wavelet threshold denoising algorithm to filter the decomposed voice signal, and then performs wavelet reconstruction to obtain the filtered noisy voice signal; performs low-pass filtering on the filtered noisy voice signal to obtain the denoised voice signal. By integrating the improved wavelet threshold denoising algorithm and the low-pass filtering method, the present invention proposes a signal filtering scheme with higher accuracy and adaptability to achieve voice denoising and improve voice quality. Description of the Drawings

[0023] Figure 1 It is a flowchart of the improved wavelet denoising method based on the Sagnac fiber acoustic sensing system provided by the present invention;

[0024] Figure 2Flow chart of the practical application of the improved wavelet denoising method based on the sagnac fiber optic acoustic sensing system provided by the present invention;

[0025] Figure 3 Schematic diagram of the structure of the fiber optic acoustic sensor system based on the linear Sagnac principle provided by the present invention;

[0026] Figure 4 Flow chart of the wavelet denoising method of the present invention;

[0027] Figure 5 Comparison chart of the soft threshold function, hard threshold function, threshold function proposed by the prior art I and the improved threshold function of the present invention;

[0028] Figure 6 Waveform diagram of the analog signal 1 of the present invention processed by different denoising algorithms;

[0029] Figure 7 Waveform diagram of the analog signal 2 of the present invention processed by different denoising algorithms;

[0030] Figure 8 Waveform diagram of the measured signal of the present invention processed by different denoising algorithms. Detailed implementation manners

[0031] To facilitate the understanding of the present invention, the present invention will be described more comprehensively below with reference to the embodiments. However, the present invention can be implemented in many different forms and is not limited to the embodiments described herein. On the contrary, these embodiments are provided to make the understanding of the disclosure of the present invention more thorough and comprehensive.

[0032] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by those of ordinary skill in the technical field to which the present invention belongs. The terms used in the specification of the present invention are only for the purpose of describing specific embodiments and are not intended to limit the present invention. The term "and / or" used herein includes any and all combinations of one or more of the related listed items.

[0033] Please refer to Figure 1 , the present invention provides an improved wavelet denoising method based on the sagnac fiber optic acoustic sensing system, including:

[0034] Step 201: Obtain the original sound signal. In practice, the original sound signal is collected by a fiber optic acoustic sensor system based on the linear Sagnac principle, as specifically shown in Figure 3 .

[0035] Step 202: In the fiber optic sensing system, the interference optical signal is received by a photodetector and converted into an electrical signal, and then the signal is collected by an acquisition card. The collected electrical signal can obtain the original sound signal after demodulation, but this signal is a noisy signal. In practical applications, the signal after wavelet decomposition includes a noisy sound signal and a noise signal. As Figure 4 shown, the low-frequency wavelet coefficients (A i , i = 1, 2, 3) correspond to the noisy sound signal, while the high-frequency wavelet coefficients (D i , i = 1, 2, 3) correspond to the noise signal.

[0036] Step 203: The decomposed sound signal is filtered using an improved wavelet threshold denoising algorithm, and then wavelet reconstruction is performed to obtain the filtered noisy sound signal.

[0037] Specifically, it includes:

[0038] The above-mentioned decomposed high-frequency and low-frequency wavelet coefficients are processed using an improved threshold function and an improved threshold, that is, the high-frequency wavelet coefficients are removed, while the low-frequency wavelet coefficients are retained. Finally, the retained wavelet coefficients are reconstructed to obtain the denoised signal. According to the different amplitudes of the high-frequency and low-frequency wavelet coefficients during signal decomposition, by selecting an appropriate threshold, it can have a better improvement effect on signal noise removal.

[0039] The construction of the threshold function is an important step in the wavelet threshold denoising process. An appropriate threshold function can greatly improve the denoising effect. Since the hard threshold function is discontinuous, the signal may mutate, and the soft threshold function has a constant deviation problem. Therefore, in order to reduce the influence of noise on the signal, the threshold function needs to be improved accordingly.

[0040] Selecting an appropriate threshold is also a very important step in the wavelet threshold denoising process. Different threshold rules affect the noise filtering effect. If the threshold is selected too small, the noise removal is incomplete, resulting in poor noise removal effect; if the threshold is selected too large, the useful information in the signal will be filtered out, which may also lead to an unsatisfactory denoising effect.

[0041] Wavelet reconstruction is the last step in the wavelet threshold denoising process. This step is to reconstruct the wavelet coefficients obtained after wavelet decomposition of the signal, so as to obtain a relatively pure signal.

[0042] The expression of the improved threshold function is:

[0043]

[0044] Where: 0 ≤ β ≤ 1. The parameter α ∈ [0, ∞] is the approaching speed adjustment factor. The smaller α is, the slower the approaching speed is.

[0045] The threshold setting of the high-frequency wavelet coefficients adopts a new threshold rule. It can effectively improve the accuracy and adaptability of signal processing, and its expression is as follows:

[0046]

[0047] In the formula: i is the wavelet decomposition scale.

[0048] In the above formula, the wavelet coefficient threshold T i shows a slow downward trend as the decomposition scale i gradually increases, which exactly conforms to the characteristic that the amplitude of the noise wavelet coefficients gradually decreases as the decomposition scale increases.

[0049] Step 204: Perform low-pass filtering on the filtered noisy sound signal to obtain the finally noise-reduced sound signal.

[0050] The present invention provides an improved wavelet sound noise reduction method based on a linear Sagnac fiber acoustic sensing system, including: obtaining an original sound signal through a linear Sagnac fiber acoustic sensing system; decomposing the original sound signal using a wavelet denoising method; filtering the decomposed sound signal using an improved wavelet threshold denoising algorithm, and then performing wavelet reconstruction to obtain a filtered noisy sound signal; passing the filtered noisy sound signal through low-pass filtering to obtain a noise-reduced sound signal. The present invention realizes sound denoising by integrating an improved wavelet threshold denoising algorithm and a low-pass filtering method, and it can accurately restore the sound signal in a strong noise environment and improve the quality of the sound signal.

[0051] As Figure 2 shown, the present invention provides an improved wavelet denoising method based on a sagnac fiber acoustic sensing system, and its specific steps in practical applications:

[0052] Step 101: Obtain the original sound signal: Build a system according to the linear Sagnac fiber sensing principle to collect the original sound signal. The structure of the fiber acoustic sensing system is as Figure 3 shown, where Laser is the light source, PD is the photodetector, DAQ is the data acquisition card, and PC is the computer; 1, 2, and 3 all represent the 3 input ends of the 3×3 coupler a; 4, 5, and 6 represent the 3 output ends of the 3×3 coupler a; b represents the delay fiber; 7 and 8 respectively represent the two input ends of the 2×1 coupler c, 9 represents the output end of the 2×1 coupler, 10 represents the position of the disturbance intrusion point; d represents the 1×2 coupler; 11 represents the series-connected fiber at the output end of the 1×2 coupler d. The linear Sagnac fiber acoustic sensing system has the following four optical paths:

[0053] (a) Optical path 1: 1 - 6 - delay optical fiber - 8 - 9 - 10 - 11 - 10 - 9 - 7 - 4 - 3. (Clockwise optical path, CW)

[0054] (b) Optical path 2: 1 - 4 - 7 - 9 - 10 - 11 - 10 - 9 - 7 - 4 - 3

[0055] (c) Optical path 3: 1 - 4 - 7 - 9 - 10 - 11 - 10 - 9 - 8 - b - 6 - 3. (Counterclockwise optical path, CCW)

[0056] (d) Optical path 4: 1 - 6 - b - 8 - 9 - 10 - 11 - 10 - 9 - 8 - b - 6 - 3.

[0057] In the above optical paths, the optical paths of Optical path 1 and Optical path 3 are the same, which satisfy the interference light conditions and can stably interfere with each other at the 3×3 coupler. Optical path 2 and Optical path 4 can be regarded as DC signal lights that are easy to eliminate.

[0058] Step 102: Perform wavelet transform on the collected original sound signal s(t) according to the following formula:

[0059]

[0060] In the formula, φ(t) represents the wavelet basis function, α represents the scaling function, and τ represents the translation distance.

[0061] Under the action of the wavelet basis function, the original sound signal will be transformed into a two - dimensional space including the time domain and the scale domain. At the same time, the signal will perform frequency - domain and time - domain conversions, which is conducive to the specific analysis of the signal.

[0062] The original sound signal will be decomposed into high - frequency wavelet coefficients and low - frequency wavelet coefficients under wavelet transform. Among them, the noise mainly exists in the high - frequency wavelet coefficients and needs to be subjected to corresponding denoising processing; while the useful signal mainly exists in the low - frequency wavelet coefficients and needs to be retained.

[0063] Generally speaking, all signals can be decomposed infinitely, but the actual number of decomposition layers needs to be analyzed according to specific problems. For example, when decomposing a sound signal into three layers, its expression can be represented as:

[0064] S = A i + D i

[0065] In the formula, s is the original sound signal, the decomposed low - frequency signal is represented by A i and the high - frequency signal is represented by D i where i = 1, 2, 3 represents the number of decomposition layers. From the above formula and Figure 3It can be seen that under the action of wavelet transform, the sound signal is decomposed into wavelet coefficients. And for each decomposition, one more high-frequency coefficient and one low-frequency coefficient will be generated.

[0066] Step 103: Filter the decomposed sound signal using the improved wavelet threshold denoising algorithm: First, decompose the sound signal according to the selected wavelet basis function and the decomposition level to obtain high-frequency wavelet coefficients and low-frequency wavelet coefficients of different levels; Secondly, use the improved threshold function and threshold to remove the high-frequency wavelet coefficients in the decomposed signal, while retaining the low-frequency wavelet coefficients; Then reconstruct the retained wavelet coefficients to obtain the filtered original sound signal.

[0067] The improved threshold is expressed as follows:

[0068]

[0069] In the formula: i is the wavelet decomposition scale.

[0070] In the above formula, the threshold T of the wavelet coefficient i shows a slow downward trend as the decomposition scale i gradually increases, which exactly conforms to the characteristic that the amplitude of the noise wavelet coefficient gradually decreases as the decomposition scale increases.

[0071] The expression of the improved threshold function is as follows:

[0072]

[0073] Where: 0 ≤ β ≤ 1. The parameter α ∈ [0, ∞] is the approaching speed adjustment factor. The smaller α is, the slower the approaching speed is.

[0074] Step 104: Perform low-pass filtering on the filtered noisy sound signal to obtain the finally denoised sound signal.

[0075] Comparing the soft threshold function, the hard threshold function, the existing threshold function with the improved threshold function of the present invention, it can be known that the improved threshold function of the present invention has the following properties:

[0076] (1) This function is continuous at the threshold;

[0077] (2) By changing the value of β, it can be flexibly transformed into soft and hard threshold functions. For example, when the β value is 0, it is transformed into a soft threshold function, and when the β value is 1, it is a hard threshold function;

[0078] (3) The improved threshold function has an oblique asymptote y = x. The improved threshold function inherits the advantages of the traditional threshold function well, and at the same time, it also solves the problem of constant deviation, that is, the disadvantages of slow convergence speed and discontinuity at the threshold. It can also prevent the occurrence of the pseudo-Gibson phenomenon. It has a better noise removal effect on noise and is more practical. The comparison of the four threshold functions is as Figure 5 shown.

[0079] For related content, reference can be made to the prior art I (Cheng Shu, Ma Weijiao, Niu Yingjie, etc. Measurement data processing of an improved wavelet threshold function integrating weighted average [J]. Chinese Science Bulletin, 2018, 013(015): 1811-1816.).

[0080] The parameters in the noise reduction method of the present invention are optimized using the coordinate descent method. The results show that the sym5 wavelet basis function and 5-layer decomposition have the best denoising effect.

[0081] To verify the effect of the method proposed in the present invention, several common filtering methods are compared and analyzed. The results are shown in Table 1 and Figure 6 shown. It can be seen that by using the method proposed in the present invention, the numerical values of the signal-to-noise ratio (SNR), root mean square error (RMSE), and subjective speech quality (PESQ) indicators of the sound signal after denoising are the best, and the waveform burrs after filtering are also the fewest, which shows the effectiveness of this method.

[0082] Table 1 Evaluation indicators of simulated signal 1 processed by different denoising algorithms

[0083]

[0084] Method 1 is: using an ordinary threshold and a hard threshold function to denoise the signal;

[0085] Method 2 is: using an ordinary threshold and a soft threshold function to denoise the signal;

[0086] Method 3 is: using the threshold and its threshold function in the prior art II (Gong Jing. A wavelet denoising algorithm optimized by a new controllable threshold function and threshold operator [J]. Journal of Beijing University of Civil Engineering and Architecture, 2020, 36(2): 67-73) to denoise the signal;

[0087] Method 4 is: using the improved threshold function and threshold of the present invention to denoise the signal.

[0088] Change the signal and repeat the above steps to obtain the results shown in Table 2 and Figure 7 shown:

[0089] Table 2 Evaluation indicators of simulated signal 2 processed by different denoising algorithms

[0090]

[0091] By analyzing the signal waveform diagram and related indicators, it can be concluded that the method proposed in the present invention has a better denoising effect.

[0092] To further verify the effect of the method proposed in the present invention, a section of measured sound signal was selected for denoising processing, and the above-mentioned different methods were simultaneously used for comparative analysis. The results are shown in Table 3 and Figure 8 as follows. It can be seen from the comparison that the filtering method proposed in the present invention has the best effect.

[0093] Table 3 Evaluation indicators of the measured signal after being processed by different denoising algorithms

[0094]

[0095] Compared with the prior art, the advantages of the present invention are as follows: According to the characteristics of the sound signal collected by the fiber optic acoustic sensing system, a wavelet denoising method is adopted to filter the sound signal, and then an improved wavelet threshold denoising algorithm and a low-pass filtering method are fused to form an improved wavelet sound denoising method. This method can effectively improve the quality of the collected sound signal. As shown in Table 3, its SNR is increased by 2.778, RMSE is decreased by 0.0744, and PESQ is increased by 0.417.

[0096] The above only illustrates the specific embodiments of the present invention, but it cannot be used as the protection scope of the present invention. Any equivalent changes or modifications made according to the design spirit of the present invention should be considered to fall within the protection scope of the present invention.

Claims

1. An improved wavelet denoising method based on the Sagnac fiber acoustic sensing system, characterized in that Including: Obtaining an original sound signal; Determining a wavelet basis function to perform wavelet decomposition on the original sound signal, and obtaining high-frequency wavelet coefficients and low-frequency wavelet coefficients of different layers according to the wavelet basis function and the decomposition level; Using an improved threshold function and an improved threshold to remove the high-frequency wavelet coefficients in the decomposed signal, while retaining the low-frequency wavelet coefficients; Reconstructing the retained wavelet coefficients to obtain a filtered noisy sound signal; Performing low-pass filtering on the filtered noisy sound signal to obtain a noise-reduced sound signal; The expression of the improved threshold function is: ； Wherein: , , the parameter is the approaching speed adjustment factor. The smaller α is, the slower the approaching speed is.

2. The improved wavelet denoising method based on the sagnac fiber acoustic sensing system according to claim 1, wherein, The expression of the improved threshold is: ； Wherein: is the wavelet decomposition scale; N is the length of the Nth high-frequency wavelet coefficient; In the above formula, the threshold of the wavelet coefficient decreases slowly as the decomposition scale increases.

Images