A diaphragm electromyogram signal denoising method

Abstract

The present application relates to a kind of diaphragm myoelectric signal noise reduction method, comprising the following steps, S1, by means of acquisition device obtains diaphragm myoelectric signal and reference electrocardio noise signal;S2, diaphragm myoelectric signal and reference electrocardio noise signal are carried out wavelet transform, obtain first wavelet coefficient sequence;S3, first wavelet transform coefficient sequence is handled using independent component analysis method for first noise reduction, obtain second wavelet coefficient sequence;S4, second wavelet coefficient sequence is handled using wavelet entropy method for second noise reduction, obtain third wavelet coefficient sequence;S5, third wavelet coefficient sequence is carried out wavelet inverse transform, obtain the diaphragm myoelectric signal of noise reduction.The present application uses independent component analysis one-time noise reduction superposition wavelet entropy secondary noise reduction double noise reduction technology, can effectively inhibit and reduce the influence of electrocardio noise on diaphragm myoelectric signal acquisition, improve the effectiveness and accuracy of diaphragm myoelectric signal acquisition.

Description

Technical Field

[0001] This invention relates to a method for noise reduction of diaphragmatic electromyographic signals. Background Technology

[0002] The diaphragm is the main respiratory muscle in the human body, accounting for 80% of the breathing process. Diaphragmatic electromyography (EMG) is a non-stationary bioelectrical signal generated by the movement of the diaphragm. Diaphragmatic EMG can be used to assess the diaphragmatic fatigue state in patients with chronic obstructive pulmonary disease (COPD). Combined with other respiratory signals, it can be used to calculate the airway resistance and other physiological characteristics of patients with obstructive sleep apnea, providing a scientific basis for clinical diagnosis and assessment.

[0003] There are three methods for collecting diaphragmatic electromyography (EMG): needle electrodes, esophageal electrodes, and surface electrodes. Needle electrodes and esophageal electrodes have low noise interference but are invasive and have low patient tolerance, limiting their clinical application. While surface electrodes experience greater noise interference, their non-invasive nature and ease of collection make them a promising candidate for widespread clinical use.

[0004] In diaphragmatic electromyography (EMG) signals acquired using surface electrodes, the interference from cardiac noise is significant. The amplitude of cardiac noise is several times higher than that of diaphragmatic EMG, and the spectra of cardiac noise and diaphragmatic EMG signals overlap considerably. Since the spectral characteristics of diaphragmatic EMG are important indicators for clinical assessment of diaphragmatic function, cardiac noise is an interference that must be eliminated.

[0005] In existing technologies, most methods employ Independent Component Analysis (ICA) for noise reduction of multi-channel diaphragmatic electromyography (EMG) signals. However, due to the strong amplitude of central electrical noise in diaphragmatic EMG signals and its overlap with the diaphragmatic EMG spectrum, and the differences in the data acquisition channels and the distance of electrode positions from the heart in practical applications, the amplitude and waveform of the ECG noise collected in each channel vary. Using ICA alone to remove ECG noise results in uneven noise reduction effects across multiple channels and poor overall noise reduction performance.

[0006] To address this issue, this invention, based on the independent component analysis method for initial noise reduction, utilizes wavelet entropy to perform secondary noise reduction on channels where the initial noise reduction effect is unsatisfactory. This effectively solves the problem of inconsistent removal of ECG noise across channels and poor noise reduction effect in some channels by the independent component analysis method.

[0007] Meanwhile, to enhance the noise separation effect during the independent component analysis (ICA) process, this invention incorporates a reference electrocardiogram (ECG) noise signal into the input of the ICA method. This allows for more effective noise information extraction during ICA, thereby enhancing the noise reduction effect. The reference ECG noise signal can be acquired using a surface electrode located on the left side of the body near the heart using a data acquisition device. Summary of the Invention

[0008] (a) Technical problems to be solved

[0009] In view of the above-mentioned shortcomings and deficiencies of the prior art, the present invention provides a method for noise reduction of diaphragmatic electromyographic signals.

[0010] (II) Technical Solution

[0011] To achieve the above objectives, the main technical solutions adopted by the present invention include:

[0012] In a first aspect, embodiments of the present invention provide a method for noise reduction of diaphragmatic electromyographic signals, comprising the following steps:

[0013] S1, acquires diaphragmatic electromyographic signals and reference electrocardiographic noise signals with the aid of a data acquisition device;

[0014] S2, perform wavelet transform on the diaphragm electromyography signal and the reference electrocardiogram noise signal to obtain the first wavelet coefficient sequence;

[0015] S3, perform first noise reduction processing on the first wavelet transform coefficient sequence using independent component analysis to obtain the second wavelet coefficient sequence;

[0016] S4, the second wavelet coefficient sequence is subjected to a second noise reduction process using the wavelet entropy method to obtain a third wavelet coefficient sequence;

[0017] S5, perform inverse wavelet transform on the third wavelet coefficient sequence to obtain the denoised diaphragm electromyographic signal.

[0018] Optionally, S2 includes:

[0019] The diaphragm electromyography signal and the reference electrocardiogram noise signal are subjected to 5-level Haar wavelet transforms respectively, and the wavelet coefficients of each transformed signal are merged into the first wavelet coefficient sequence.

[0020] Optionally, S3 includes:

[0021] S31, the first wavelet coefficient sequence Wx = [Wx1; Wx2; ...; Wx n As input to the Independent Component Analysis (ICA) algorithm, the source signal matrix Y = [y1; y2; ...; y3] is obtained through ICA analysis. n], and the unmixing matrix W and the mixing matrix A,

[0022] Where Wx and Y are n×N matrices, W and A are both n×n matrices, N is the length of the wavelet coefficient vector, and n is the total number of signals, including diaphragmatic electromyography signals and reference electrocardiogram noise signals from several channels.

[0023] S32, the obtained source signal sequences y i Wavelet transform coefficient sequence Wx corresponding to reference ECG noise 噪声 Perform correlation calculations to obtain the absolute value θ of the correlation coefficient. i The calculation formula is as follows:

[0024]

[0025] S33, based on the preset correlation threshold T1, process each source signal sequence y according to the following formula. i :

[0026]

[0027] S34, process y i Insert the data into the i-th row of Y to obtain the processed source data matrix Y1;

[0028] S35, perform independent component analysis inverse transform on Y1 using the mixing matrix A to obtain the second wavelet transform coefficient sequence Wx1 of each channel signal after the first noise reduction. The calculation formula is as follows:

[0029] Wx1 = A*Y1.

[0030] Optionally, the first wavelet transform coefficient sequence is subjected to independent component analysis and a first noise reduction process using the FASTICA algorithm.

[0031] Optionally, S4 includes:

[0032] S41, based on a preset noise index, select channels exceeding the preset noise index from the second wavelet transform coefficient sequence to form the second noise reduction processing sequence Wx2, and the other channels to form Wx11;

[0033] S42, the second noise reduction sequence Wx2 is subjected to second noise reduction using the wavelet entropy method to obtain Wx22;

[0034] S43, merges the second denoised Wx22 and Wx11 into the third wavelet transform coefficients.

[0035] Optionally, S42 includes:

[0036] The wavelet coefficient sequence Cxn of each channel and each scale of the second noise reduction sequence Wx2 is subjected to wavelet transform.i Perform the following processing:

[0037] A1 is divided into M equal intervals based on its sequence length L, with the length of each interval being β = floor(L / M). The entropy value H of the wavelet coefficients in each interval is calculated.

[0038] like If the interval is not marked as a low wavelet coefficient interval, then it is marked as a high wavelet coefficient interval.

[0039] A2, for Cxn i Points in all high wavelet coefficient intervals within the wavelet coefficient interval are denoised using the following formula:

[0040]

[0041] Among them, Hxn i (k, w) represents the w-th wavelet coefficient value within the k-th high wavelet coefficient interval, q is the preset wavelet entropy high adjustment threshold, and H1 is Cxn. i The average of the absolute values ​​of wavelet coefficients across all high wavelet coefficient intervals;

[0042] A3, for Cxn i Points in all lower wavelet coefficient intervals within the wavelet coefficient interval are denoised using the following formula:

[0043]

[0044] Among them, Lxn i (k, w) is Cxn i The w-th wavelet coefficient value is located in the k-th low wavelet coefficient interval within the wavelet coefficient interval, where p is the preset wavelet entropy low adjustment threshold, and L1 is Cxn. i The average of the absolute values ​​of wavelet coefficients in all low wavelet coefficient intervals;

[0045] A5, the processed Cxn i Wavelet coefficients replace the original Cxn i Wavelet coefficients.

[0046] In a second aspect, the present invention provides a computer device including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps of the diaphragm electromyography signal noise reduction method described above.

[0047] Thirdly, the present invention provides a diaphragm electromyography (EMG) signal acquisition device, comprising an acquisition unit, a processor, and an output unit. The acquisition unit is used to acquire the diaphragm EMG signal and the reference electrocardiogram (ECG) noise signal. The processor is used to execute any of the above-described diaphragm EMG signal noise reduction methods. The output unit is used to output a noise-reduced diaphragm EMG signal free of ECG noise.

[0048] Optionally, the collector uses surface electrodes to acquire electromyographic signals of the diaphragm.

[0049] Optionally, the collector includes four sets of surface electrodes, symmetrically placed on the left and right sides of the human body near the diaphragm, for collecting diaphragmatic electromyographic signals, and one set of surface electrodes, placed on the left side of the human body near the heart, for collecting reference electrocardiogram noise signals.

[0050] (III) Beneficial Effects

[0051] Compared with existing technologies, this invention employs a dual denoising technique that combines independent component analysis (ICA) for primary denoising with wavelet entropy for secondary denoising. This technique overcomes the problem of uneven denoising effects across multiple channels and poor overall denoising performance when using ICA alone. It can effectively suppress and reduce the impact of ECG noise on diaphragmatic electromyography (EMG) signal acquisition, thereby improving the effectiveness and accuracy of diaphragmatic EMG signal acquisition.

[0052] In addition, the present invention incorporates a reference electrocardiogram noise signal into the input of the independent component analysis method, thereby enabling more effective extraction of noise information and enhancing the noise reduction effect in independent component analysis. Attached Figure Description

[0053] Figure 1 This is a flowchart of the noise reduction method of the present invention;

[0054] Figure 2 The data collected in each channel are from an embodiment of the present invention;

[0055] Figure 3 This is a wavelet coefficient sequence after wavelet transform of each data channel according to an embodiment of the present invention;

[0056] Figure 4 This is a source signal decomposed by independent component analysis according to an embodiment of the present invention;

[0057] Figure 5 These are the multi-channel wavelet coefficients after noise reduction via independent component analysis, according to an embodiment of the present invention.

[0058] Figure 6 This is multi-channel data after noise reduction through independent component analysis, according to an embodiment of the present invention.

[0059] Figure 7This is a comparison image of wavelet entropy denoising before and after one embodiment of the present invention;

[0060] Figure 8 This is data obtained by removing ECG noise from multichannel diaphragmatic electromyography according to an embodiment of the present invention. Detailed Implementation

[0061] To better understand the above technical solutions, exemplary embodiments of the present invention will be described in more detail below with reference to the accompanying drawings. Although exemplary embodiments of the present invention are shown in the drawings, it should be understood that the present invention can be implemented in various forms and should not be limited to the embodiments set forth herein. Rather, these embodiments are provided so that the present invention can be understood more clearly and thoroughly, and that the scope of the present invention can be fully conveyed to those skilled in the art.

[0062] Example 1

[0063] like Figure 1 As shown, this embodiment provides a method for noise reduction of diaphragmatic electromyographic signals, including the following steps:

[0064] S1, acquires diaphragmatic electromyographic signals and reference electrocardiographic noise signals with the aid of a data acquisition device;

[0065] S2, perform wavelet transform on the diaphragm electromyography signal and the reference electrocardiogram noise signal to obtain the first wavelet coefficient sequence;

[0066] S3, perform first noise reduction processing on the first wavelet transform coefficient sequence using independent component analysis to obtain the second wavelet coefficient sequence;

[0067] S4, the second wavelet coefficient sequence is subjected to a second noise reduction process using the wavelet entropy method to obtain a third wavelet coefficient sequence;

[0068] S5, perform inverse wavelet transform on the third wavelet coefficient sequence to obtain the denoised diaphragm electromyographic signal.

[0069] To better illustrate step S2, the specific steps for wavelet transforming each signal in S2 (including multiple channels of diaphragmatic electromyography signals and one channel of reference ECG noise signal) are explained below:

[0070] S21, let these multiple data points X(t) = [x1(t); x2(t); ...; x n Each channel in [t] undergoes a 5-level Haar wavelet transform to obtain the wavelet transform coefficients Cx for each channel. i (j, k).

[0071] Among them, Cx i(j, k) is the k-th coefficient of the j-th scale of the i-th data, j = 1, 2, ..., 5, k is the translation factor, and x takes the value a or d, where a represents the low-frequency component and d represents the high-frequency component.

[0072] The calculation formula is as follows:

[0073]

[0074] S22, the five-level discrete wavelet coefficients of each data point obtained from the decomposition are concatenated to obtain the wavelet coefficient sequence of each data point, and the concatenation form is Wx. i =[Ca5 i Cd5 i Cd4 i Cd3 i Cd2 i Cd1 i ], where Wx i Let Ca5 represent the wavelet coefficient sequence of the i-th channel, which is a 1×N row vector. i It is the low-frequency wavelet coefficient of the 5th scale in the i-th channel, Cdn i These are the high-frequency wavelet coefficients of the i-th channel at the n-th scale, where n = 1, 2, ..., 5.

[0075] To explain the working steps of S3 in detail, S3 is broken down into sub-steps S31 to S35 below, including:

[0076] S31, the first wavelet coefficient sequence Wx = [Wx1; Wx2; ...; Wx n As input to the Independent Component Analysis (ICA) algorithm, the source signal matrix Y = [y1; y2; ...; y3] is obtained through ICA analysis. n ], and the unmixing matrix W and the mixing matrix A,

[0077] Where Wx and Y are n×N matrices, W and A are both n×n matrices, N is the length of the wavelet coefficient vector, and n is the total number of signals, including diaphragmatic electromyography signals and reference electrocardiogram noise signals from several channels.

[0078] S32, the obtained source signal sequences y i Wavelet transform coefficient sequence Wx corresponding to reference ECG noise 噪声 Perform correlation calculations to obtain the absolute value θ of the correlation coefficient. i The calculation formula is as follows:

[0079]

[0080] θ i The value range of is [0, 1]. As a criterion for correlation, its physical meaning is that the larger its absolute value, the more similar it is to the noise source signal.

[0081] S33, based on the preset correlation threshold T1, process each source signal sequence y according to the following formula. i :

[0082]

[0083] S34, process y i Insert the data into the i-th row of Y to obtain the processed source data matrix Y1;

[0084] S35, perform independent component analysis inverse transform on Y1 using the mixing matrix A to obtain the second wavelet transform coefficient sequence Wx1 of each channel signal after the first noise reduction. The calculation formula is as follows:

[0085] Wx1 = A*Y1.

[0086] Independent component analysis (ICA) has various implementation methods. This invention adopts the FASTICA algorithm based on the negative entropy criterion. This algorithm uses a distributed and parallel computing method and has the characteristics of fast running speed and small memory consumption.

[0087] To better illustrate S4, the working method of S4 is broken down into sub-steps S41 to S43, including the following:

[0088] S41, based on a preset noise index, select channels exceeding the preset noise index from the second wavelet transform coefficient sequence to form the second noise reduction processing sequence Wx2, and the other channels to form Wx11;

[0089] S42, the second noise reduction sequence Wx2 is subjected to second noise reduction using the wavelet entropy method to obtain Wx22;

[0090] S43, merges the second denoised Wx22 and Wx11 into the third wavelet transform coefficients.

[0091] S42 includes:

[0092] The wavelet coefficient sequence Cxn of each channel and each scale of the second noise reduction sequence Wx2 is subjected to wavelet transform. i Perform the following processing:

[0093] A1 is divided into M equal intervals based on its sequence length L, with the length of each interval being β = floor(L / M). The entropy value H of the wavelet coefficients in each interval is calculated.

[0094] It should be noted that the entropy value H of the wavelet coefficients can be calculated using the following steps:

[0095] AA1, calculate Cxn iThe sum of the absolute values ​​of wavelet coefficients in each small division of the coefficients, where the formula for calculating the sum of the absolute values ​​of wavelet coefficients in the j-th small division is as follows:

[0096]

[0097] AA2, calculate Cxn i The sum of the absolute values ​​of each small equal part of the wavelet coefficients accounts for Cxn i The proportion of the sum of the absolute values ​​of the wavelet coefficients, i.e., the probability. Where Cxn i The probability of the j-th smallest wavelet coefficient is P i，x，n (j), the calculation formula is as follows:

[0098]

[0099] Where j = 1, 2, ..., M.

[0100] AA3, Calculate Cxn i Each small portion of the wavelet coefficients represents a portion of the total N equal portions Cxn. i The entropy value in the wavelet coefficients, where the coefficient of the j-th smallest equal part accounts for the entire Cxn of the N equal parts. i The entropy value in wavelet coefficients is calculated using the following formula:

[0101] H i，x，n (j)=-P i，x，n (j)*log2(P i，x，n (j));

[0102] Where j = 1, 2, ..., M.

[0103] Based on the calculated entropy value H of the wavelet coefficients for each interval, if If the interval is not marked as a low wavelet coefficient interval, then it is marked as a high wavelet coefficient interval.

[0104] A2, for Cxn i Points in all high wavelet coefficient intervals within the wavelet coefficient interval are denoised using the following formula:

[0105]

[0106] Among them, Hxn i (k, w) represents the w-th wavelet coefficient value within the k-th high wavelet coefficient interval, q is the preset wavelet entropy high adjustment threshold, and H1 is Cxn. i The average of the absolute values ​​of wavelet coefficients across all high wavelet coefficient intervals;

[0107] A3, for Cxn i Points in all lower wavelet coefficient intervals within the wavelet coefficient interval are denoised using the following formula:

[0108]

[0109] Among them, Lxn i (k, w) is Cxn i The w-th wavelet coefficient value is located in the k-th low wavelet coefficient interval within the wavelet coefficient interval, where p is the preset wavelet entropy low adjustment threshold, and L1 is Cxn. i The average of the absolute values ​​of wavelet coefficients in all low wavelet coefficient intervals;

[0110] A5, the processed Cxn i Wavelet coefficients replace the original Cxn i Wavelet coefficients.

[0111] The diaphragm electromyography (EMG) signal denoising method provided in this invention first separates most of the ECG noise from the channels using Independent Component Analysis (ICA). Then, for channels with strong residual ECG noise or minimal reduction in ECG noise, wavelet entropy is used to divide the wavelet coefficients at each scale into high and low wavelet entropy intervals. Different wavelet thresholds are applied to these two intervals for secondary denoising. This double denoising process achieves excellent denoising results for the diaphragm EMG signal, effectively improving the accuracy and effectiveness of diaphragm EMG signal acquisition and testing.

[0112] It should be noted that the denoising method for diaphragmatic electromyography (EMG) signals in this embodiment of the invention is designed based on the characteristics of diaphragmatic EMG signals and their noise, and is not applicable to all human measurement signals, such as EEG signals. This is because surface diaphragmatic EMG signals are electromyographic signals with large amplitudes on the order of microseconds (V), primarily in the frequency range of 20–250 Hz, and their noise primarily originates from electrocardiogram (ECG) signals. In contrast, EEG signals are very weak, with amplitudes on the order of microseconds (µV) and a frequency range of 0–60 Hz. They are non-stationary, nonlinear, and highly random, making them susceptible to various noise interferences, including those from ECG, electrooculogram (EOG), EMG, and other physiological artifacts. The significant differences in signal morphology mean that the focus and specific methods used for denoising the two signals differ, and the denoising methods are not interchangeable.

[0113] Example 2

[0114] This embodiment provides an application example of using the diaphragm electromyography signal noise reduction method of the present invention for diaphragm signal acquisition and noise reduction.

[0115] In this embodiment, the acquisition device uses four sets of surface electrodes to acquire diaphragmatic electromyography (EMG) signals and one set of surface electrodes to acquire reference electrocardiogram (ECG) noise signals. The electrodes for acquiring diaphragmatic EMG signals are symmetrically placed on the left and right sides of the human body near the diaphragm, while the electrode for acquiring reference ECG noise signals is placed on the left side of the human body near the heart.

[0116] The sampling frequency of each channel is 2.5 kHz. The data collected by the electrodes is then subjected to a high-pass filter with a cutoff frequency of 15 Hz, a low-pass filter with a cutoff frequency of 500 Hz, and corresponding spectral interpolation to obtain the initial collected data, such as... Figure 2 As shown, the data consists of five channels. The first four channels, from top to bottom, correspond to the diaphragmatic electromyography (EMG) signals collected by the electrodes: data11, data21, data31, and data41. The horizontal axis of each channel represents the number of sampling time points, and the vertical axis represents the voltage amplitude of the measured signal, in volts (V). Data21 and data31 are signals collected by electrodes on the left side of the body (closer to the heart), and therefore have stronger ECG noise. Data11 and data41 are signals collected on the right side of the body, farther from the heart, and therefore have relatively weaker ECG noise. The fifth channel represents the ECG reference noise.

[0117] These five data points are used as inputs to the noise reduction method of this invention. Figure 2 The data in the dataset are subjected to wavelet transforms to obtain the wavelet transform coefficients for each channel, such as... Figure 3 As shown.

[0118] exist Figure 3 As can be seen, the ECG noise components of data21 and data31 are relatively strong, with very high spikes, while the ECG noise of data11 and data41 is relatively weak.

[0119] Will Figure 3 The data is input into the ICA algorithm to obtain the various source components, such as Figure 4 As shown.

[0120] Will Figure 4 Correlation operations are performed on the wavelet coefficient sequences of each source component signal and the reference noise to identify and remove source components that are highly correlated with ECG noise. The noise-removed source component data are then mixed using a mixing matrix to obtain the denoised wavelet coefficients for each channel, as shown below. Figure 5 .

[0121] To verify the effectiveness of the ICA algorithm, the wavelet coefficients processed by ICA were subjected to inverse wavelet transform, and the denoised time-domain signal was observed. (Refer to...) Figure 6As can be seen, the amplitude of ECG noise in data21 and data31 is significantly reduced, and the signal amplitude has also changed considerably. The amplitude has changed from being dominated by ECG noise to being dominated by diaphragmatic electromyography, resulting in a significant improvement in the signal-to-noise ratio. The amplitude changes in data11 and data41 are relatively small.

[0122] To evaluate the denoising effect of the ICA algorithm, the center frequency and mean square value of the signal can be used as evaluation indicators. The intermediate frequency and mean square value of the signal are calculated by calculating the multi-channel time domain data before ICA denoising and the data after ICA denoising, respectively, and the data in Table 1 and Table 2 are obtained.

[0123] Table 1 Performance indicators of each channel before noise reduction

[0124] data11 data21 data31 data41 <![CDATA[f m (Intermediate frequency Hz) 132.05 100.10 98.59 136.81 RMS (mean square value) 0.0974 0.1954 0.3123 0.1138

[0125] Table 2 Performance indicators of each channel after noise reduction

[0126] data11 data21 data31 data41 <![CDATA[f m (Intermediate frequency Hz) 132.99 132.89 127.80 137.98 RMS (mean square value) 0.0949 0.0746 0.0562 0.1115

[0127] Table 1 shows the mean square (MSP) and intermediate frequency (IF) values ​​of each channel data before ICA denoising, and Table 2 shows the MSP and IF values ​​of each channel data after ICA denoising. Due to the use of ICA denoising, the overall MSP value of the multi-channel data signal decreases after denoising. The MSP value of data31 decreases significantly, with the denoised MSP value being approximately 18% of the original. The MSP value of data21 decreases the second largest, becoming 38.2% of the original. The MSP values ​​of data11 and data41 decrease relatively little, only about 2% of the original. This indicates that the ICA denoising method is effective for data21 and data31, but less effective for data11 and data41, although it does remove a small portion of ECG noise.

[0128] Since the spectrum of ECG signals is concentrated in the 20-50Hz range, the original signal's spectrum drops significantly within this frequency range after applying the ICA denoising algorithm, leading to an increase in its mid-frequency value. Looking at the data in Tables 1 and 2, data21 and data31 still show significant effects, with large changes in mid-frequency values. After denoising, the mid-frequency value of data21 increases to 32.76% of its original value, and that of data31 increases to 29.62%. The mid-frequency value of data11 increases by 0.7%, and that of data41 increases by 0.8%. These data show that data21 and data31 have better ECG denoising effects, with larger decreases in the mean square value and higher increases in the mid-frequency value, indicating excellent denoising performance. In contrast, data11 and data41 show smaller decreases in the mean square value and smaller increases in the mid-frequency value, resulting in weaker denoising effects, although they still remove some ECG noise. The reason why data11 and data41 have weaker noise reduction effects is that data21 and data31 are signals collected by electrodes placed on the left side of the body. These signals have large amplitude and strong correlation of ECG noise, making them easy to separate using the ICA method. However, data11 and data41 are data collected by electrodes located on the right side of the body. These signals have smaller amplitude of ECG noise, and compared to the data collected on the left side, the ECG noise differs slightly in amplitude and waveform. This results in poorer separation and less noise removal by ICA on these two data sources.

[0129] To address this issue, wavelet entropy thresholding was used to classify the scale coefficients of the two data streams, data11 and data41, after ICA processing. Each scale coefficient was divided into a high wavelet coefficient range and a low wavelet coefficient range. Different thresholds were applied to wavelet coefficients of different amplitudes for threshold denoising to obtain the processed results.

[0130] like Figure 7 As shown in the figure, this is a comparison of the denoising results of data11 before and after using the wavelet entropy thresholding denoising method. The waveform at the bottom of the figure is the data of data11 after ICA processing, that is, the data before wavelet entropy thresholding denoising. The waveform at the top of the figure is the result of data11 after ICA processing and then wavelet entropy thresholding denoising. It can be seen that the ECG noise in data11 is significantly reduced after wavelet entropy denoising, and the overall effect is good.

[0131] Table 3 Performance indicators of the signal after wavelet entropy threshold denoising

[0132] data11 data41 <![CDATA[f m (Intermediate frequency Hz) 154.67 158.86 RMS (mean square value) 0.0699 0.0809

[0133] Table 3 shows the performance metrics of data11 and data41 before and after wavelet entropy thresholding. After wavelet entropy thresholding denoising, the intermediate frequency (IF) of data11 and data41 processed by the ICA method is significantly improved. The IF of data11 is increased to 16.3% of its original value (compared to the data in Table 2), and the IF of data41 is increased to 15.13% of its original value. Furthermore, the mean square values ​​of data11 and data41 also decrease significantly, with the mean square value of data11 decreasing to 26.3% of its original value and the mean square value of data41 decreasing to 27.4% of its original value. This indicates that wavelet entropy thresholding denoising can effectively denoise ECG noise.

[0134] The data of the four-channel diaphragm electromyography signal after the above-mentioned double noise reduction process is as follows: Figure 8 As shown, from Figure 8 As can be seen, the four diaphragm electromyography data achieved a good noise reduction effect after ICA and wavelet entropy processing.

[0135] Example 3

[0136] This embodiment provides a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements the steps of the diaphragm electromyography signal noise reduction method as described in any of the above embodiments.

[0137] Example 4

[0138] This embodiment provides a diaphragm electromyography (EMG) signal acquisition device, including a collector, a processor, and an output unit. The collector is used to acquire the diaphragm EMG signal and the reference electrocardiogram (ECG) noise signal. The processor is used to execute any of the above-described diaphragm EMG signal noise reduction methods. The output unit is used to output the noise-reduced diaphragm EMG signal without ECG noise. The collector includes four sets of surface electrodes, symmetrically placed on the left and right sides of the human body near the diaphragm, for acquiring diaphragm EMG signals, and one set of surface electrodes, placed on the left side of the human body near the heart, for acquiring the reference ECG noise signal.

[0139] Since the systems / devices described in the above embodiments of the present invention are systems / devices used to implement the methods of the above embodiments of the present invention, those skilled in the art can understand the specific structure and modifications of the systems / devices based on the methods described in the above embodiments of the present invention, and therefore will not be repeated here. All systems / devices used in the methods of the above embodiments of the present invention fall within the scope of protection of the present invention.

[0140] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0141] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, as well as combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions.

[0142] It should be noted that any reference numerals placed between parentheses in the claims should not be construed as limiting the claims. The word "comprising" does not exclude the presence of components or steps not listed in the claims. The word "a" or "an" preceding a component does not exclude the presence of a plurality of such components. The invention can be implemented by means of hardware comprising several different components and by means of a suitably programmed computer. In claims that enumerate several means, several of these means may be embodied by the same hardware. The use of the terms first, second, third, etc., is merely for convenience of expression and does not indicate any order. These terms can be understood as part of the component names.

[0143] Furthermore, it should be noted that in the description of this specification, the terms "one embodiment," "some embodiments," "embodiment," "example," "specific example," or "some examples," etc., refer to specific features, structures, materials, or characteristics described in connection with that embodiment or example, which are included in at least one embodiment or example of the present invention. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples. Furthermore, without contradiction, those skilled in the art can combine and integrate the different embodiments or examples described in this specification, as well as the features of different embodiments or examples.

[0144] Although preferred embodiments of the invention have been described, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the claims should be interpreted to include both the preferred embodiments and all changes and modifications falling within the scope of the invention.

[0145] Obviously, those skilled in the art can make various modifications and variations to this invention without departing from its spirit and scope. Therefore, if these modifications and variations fall within the scope of the claims of this invention and their equivalents, then this invention should also include these modifications and variations.

Claims

1. A method for noise reduction of diaphragmatic electromyographic signals, characterized in that, include: S1, acquires multiple channels of diaphragmatic electromyography signals and reference electrocardiogram noise signals with the aid of a data acquisition device; S2, perform wavelet transform on the diaphragm electromyography (EMG) signals and the reference ECG noise signals from multiple channels to obtain a first wavelet coefficient sequence, wherein the diaphragm EMG signals and the reference ECG noise signals from multiple channels are subjected to wavelet transform separately to obtain the first wavelet coefficient sequence Wx=[ ; ;...; ], where n is the total number of signals including diaphragmatic electromyography signals and reference electrocardiogram noise signals with several channels; S3, perform first noise reduction processing on the first wavelet transform coefficient sequence using independent component analysis to obtain the second wavelet coefficient sequence; S4, the second wavelet coefficient sequence is subjected to a second noise reduction process using the wavelet entropy method to obtain a third wavelet coefficient sequence; S5, perform inverse wavelet transform on the third wavelet coefficient sequence to obtain the denoised diaphragm electromyographic signal.

2. The method for noise reduction of diaphragmatic electromyographic signals according to claim 1, characterized in that, S2 includes: The diaphragm electromyography signal and the reference electrocardiogram noise signal are subjected to 5-level Haar wavelet transforms respectively, and the wavelet coefficients of each transformed signal are merged into the first wavelet coefficient sequence.

3. The method for noise reduction of diaphragmatic electromyographic signals according to claim 1, characterized in that, S3 includes: S31, the first wavelet coefficient sequence Wx=[ ; ;...; As input to the Independent Component Analysis (ICA) algorithm, the source signal matrix Y is obtained through ICA analysis. ], and the unmixing matrix W and the mixing matrix A, Where Wx and Y are Both W and A are The matrix, where N is the length of the wavelet coefficient vector and n is the total number of signals, including several channels of diaphragmatic electromyography signals and reference electrocardiogram noise signals; S32, the obtained source signal sequences Wavelet transform coefficient sequence corresponding to reference ECG noise Perform correlation calculations to obtain the absolute value of the correlation coefficient. The calculation formula is as follows: Among them, each source signal sequence for ; S33, based on the preset correlation threshold T1, process each source signal sequence according to the following formula. : ； S34, processed Insert the data into the i-th row of Y to obtain the processed source data matrix Y1; S35, perform independent component analysis inverse transform on Y1 using the mixing matrix A to obtain the second wavelet transform coefficient sequence Wx1 of each channel signal after the first noise reduction. The calculation formula is as follows: 。 4. The method for noise reduction of diaphragmatic electromyographic signals according to claim 1, characterized in that: The first wavelet transform coefficient sequence is subjected to independent component analysis and first noise reduction processing using the FASTICA algorithm.

5. The method for noise reduction of diaphragmatic electromyographic signals according to claim 1, characterized in that, S4 includes: S41, based on a preset noise index, select channels exceeding the preset noise index from the second wavelet transform coefficient sequence to form the second noise reduction processing sequence Wx2, and the other channels to form Wx11; S42, the second noise reduction sequence Wx2 is subjected to second noise reduction using the wavelet entropy method to obtain Wx22; S43, merges the second denoised Wx22 and Wx11 into the third wavelet transform coefficients.

6. The method for noise reduction of diaphragmatic electromyographic signals according to claim 5, characterized in that, S42 includes: The wavelet coefficient sequence of each channel and each scale of the second noise reduction sequence Wx2. Perform the following processing: A1 is divided into M equal intervals based on its sequence length L, with each interval having a sequence length of... floor(L / M), calculate the entropy value H of the wavelet coefficients for each interval. If H If the interval is not specified, it is marked as a low wavelet coefficient interval; otherwise, it is marked as a high wavelet coefficient interval. A2, yes Points in all high wavelet coefficient intervals within the wavelet coefficient interval are denoised using the following formula: ； in, Let q be the w-th wavelet coefficient value in the k-th high wavelet coefficient interval, q be the preset wavelet entropy high adjustment threshold, and H1 be... The average of the absolute values ​​of wavelet coefficients across all high wavelet coefficient intervals; A3, yes Points in all lower wavelet coefficient intervals within the wavelet coefficient interval are denoised using the following formula: ； in, The wavelet coefficient value is the w-th wavelet coefficient value in the k-th low wavelet coefficient interval of the wavelet coefficient interval, where p is the preset wavelet entropy low adjustment threshold, and L1 is... The average of the absolute values ​​of wavelet coefficients in all low wavelet coefficient intervals; A5, the processed Wavelet coefficients replace the original Wavelet coefficients.

7. A computer device, characterized in that, It includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps of the diaphragm electromyography signal noise reduction method as described in any one of claims 1 to 6.

8. A device for acquiring electromyographic signals of the diaphragm, characterized in that, Includes a data acquisition unit, a processor, and an output unit: The acquisition device is used to acquire the diaphragm electromyography signal and the reference electrocardiogram noise signal; The processor is used to execute the diaphragm electromyography signal noise reduction method according to any one of claims 1 to 6; The output unit is used to output a noise-reduced diaphragmatic electromyographic signal without ECG noise.

9. A diaphragm electromyography signal acquisition device according to claim 8, characterized in that, The acquisition device uses surface electrodes to obtain diaphragmatic electromyographic signals and reference electrocardiographic noise signals.

10. A diaphragm electromyography signal acquisition device according to claim 9, characterized in that, The data collector includes: Four sets of surface electrodes are symmetrically placed on the left and right sides of the human body near the diaphragm for the acquisition of diaphragmatic electromyographic signals. One set of surface electrodes is placed on the left side of the human body near the heart for reference in acquiring electrocardiogram noise signals.

Images