Algorithm and device for quickly setting initial weight of environment sound self-adaptive noise reduction filter for auscultation, medium and equipment

Abstract

The invention discloses a rapid setting algorithm and device for an initial weight of an environment sound adaptive noise reduction filter for auscultation, a medium and equipment. According to the algorithm, pulse responses of a main channel and an auxiliary channel are collected at the same time in a quiet environment; deconvolution is carried out on the main channel impulse response by taking the auxiliary channel impulse response as a kernel, and an obtained result is the impulse response of the adaptive filter expectation model; and finally, determining the order and the initial weight of the adaptive filter according to the pulse response obtained by deconvolution. The algorithm is small in calculation amount, the initial weight and order of the adaptive filter can be determined on line, and the algorithm is suitable for assembly line batch production.

Description

Technical Field

[0001] The present invention relates to the field of medical measurement and signal processing technology, and in particular to a fast setting algorithm, device, medium and equipment for initial weights of an ambient sound adaptive noise reduction filter for auscultation. Background Art

[0002] Heart sounds and respiratory sounds are weak and easily interfered by environmental noise during auscultation. Dual-channel adaptive filtering is one of the commonly used environmental noise reduction technologies. Its application in body sound auscultation can effectively suppress the interference of environmental noise.

[0003] Setting a reasonable filter order and initial weight can greatly shorten the convergence process of the adaptive filter, which is the key to determining whether the noise reduction algorithm is practical. To address this problem, the existing method is to identify the equivalent transfer function based on the cross-correlation function between the main and auxiliary channel environmental noise, and then determine the order and initial weight of the adaptive filter. However, this method requires the construction of autocorrelation matrix and cross-correlation matrix, and requires matrix inversion and singular value decomposition. The implementation of the algorithm is relatively complex and the amount of calculation is large; it is not suitable for setting the parameters of the factory stethoscope online on the production line. Summary of the invention

[0004] In order to overcome the shortcomings and deficiencies in the prior art, the purpose of the present invention is to provide an algorithm for quickly setting the initial weights of an ambient sound adaptive noise reduction filter for auscultation; the algorithm can be used to conveniently and quickly set the filter order and initial weights online on the production line.

[0005] In order to achieve the above purpose, Figure 2 As shown, the present invention uses the secondary channel environmental noise as the input signal and the main channel environmental noise signal as the output signal to identify the order and impulse response of the desired model of the dual-channel adaptive filter. Specifically, it is implemented by the following technical scheme: a fast setting algorithm for the initial weights of the environmental sound adaptive noise reduction filter for auscultation, characterized in that: an artificial mouth or a speaker plays a pulse signal as the environmental noise source signal, a dual-channel digital stethoscope is used to collect the environmental noise signal, the environmental noise signal collected by the main channel is N′1(k), and the environmental noise signal collected by the secondary channel is N′2(k), and the signals collected by the main and secondary channels are preprocessed by bandpass filtering to suppress low-frequency and high-frequency components, and the main channel filtered signal N1(k) and the secondary channel filtered signal N2(k) are obtained respectively; the impulse response sequence G of the desired model of the adaptive filter is estimated by deconvolution; the order and initial weights of the dual-channel adaptive filter can be determined according to G.

[0006] Specifically, the steps include:

[0007] Step S1, in a quiet environment, use an artificial mouth or a speaker to play a pulse signal as environmental noise, and record the pulse start time as time zero, that is, k=0.

[0008] In step S2, the main channel environmental noise signal N′1(k) and the secondary channel environmental noise signal N′2(k) collected by the main channel and the secondary channel respectively under the excitation of the pulse signal of the dual-channel digital stethoscope are recorded.

[0009] In step S3, the main channel environmental noise signal N′1(k) and the secondary channel environmental noise signal N′2(k) are preprocessed in the same manner using a bandpass filter to suppress low-frequency and high-frequency components, and the main channel filtered signal N1(k) and the secondary channel filtered signal N2(k) are obtained respectively. Preferably, the order of the bandpass filter is not less than 2, and the cutoff frequencies are f L ∈[20,80]Hz and f H ∈[1500,2000]Hz.

[0010] Step S4: After a period of time, the amplitudes of signals N1(k) and N2(k) will converge to a certain threshold and will not exceed the threshold over time. The threshold corresponding to N1(k) is δ1, and the threshold corresponding to N2(k) is δ2; and the time when the amplitude of N1(k) converges to the threshold δ1 is t1, and the time when the amplitude of N2(k) converges to the threshold δ2 is t2.

[0011] Preferably, δ1=c1max{|N1(k)|}, δ2=c2max{|N2(k)|}, and c1,c2∈[0.01,0.1].

[0012] The main channel impulse response is denoted as vector p1 = [N1(0) ... N1(t1)], and the sub-channel impulse response is denoted as vector p2 = [N2(0) ... N2(t2)].

[0013] Step S5, with p2 as the kernel, deconvolute p1, and the result is the impulse response sequence G = [g0, g1, ...] of the expected model of the adaptive filter; when j is greater than a certain value, g j The magnitude of will converge to a certain threshold and will no longer exceed the threshold over time.

[0014] Let this threshold be δ g , and g j The amplitude converges to the threshold δ g The time is t g , that is, when j>t g When |g j |<δ g , then the order of the adaptive filter can be taken as t g +1, the initial value of the adaptive filter is

[0015] Preferably, take δ g =c g max{|g j |}, c g ∈[0.01,0.1].

[0016] The technical principle of the algorithm of the present invention is:

[0017] like Figure 2 As shown, the present invention uses the secondary channel environmental noise as the input signal and the main channel environmental noise signal as the output signal to identify the order and impulse response of the expected model of the dual-channel adaptive filter of the digital stethoscope.

[0018] An artificial mouth or a speaker plays a pulse signal as the ambient noise source signal, and a dual-channel digital stethoscope is used to collect the ambient noise signal. The ambient noise signal collected by the main channel is N′1(k), and the ambient noise signal collected by the secondary channel is N′2(k).

[0019] Depend on Figure 2 It can be seen that the impulse response sequence of the adaptive filter expectation model is It can be obtained by deconvolving the main channel impulse response with the secondary channel impulse response. Since the transfer function of the ideal adaptive filter should be consistent with the transfer function of the adaptive filter expected model, the order of the adaptive filter can be taken as t g +1, and its initial value can be taken as

[0020] If the environmental noise signals N′1(k) and N′2(k) collected by the main and auxiliary channels of the digital stethoscope are directly used to calculate the impulse response sequence G, the obtained impulse response sequence may contain low-frequency oscillation components and their harmonics generated by the vibration of the stethoscope diaphragm, which can easily lead to misjudgment of the desired model order and initial weights of the adaptive filter. Therefore, before calculating the order and initial weights, the environmental noise signals N′1(k) and N′2(k) collected by the main and auxiliary channels of the digital stethoscope need to be preprocessed with the same bandpass filter to suppress the low-frequency oscillation components and their high-frequency noise. Preferably, the order of the bandpass filter is not less than 2, and the cut-off frequency is f L ∈[20,80]Hz and f H ∈[1500,2000]Hz.

[0021] An environmental sound adaptive noise reduction device for auscultation realizes the rapid setting algorithm of the initial weights of the environmental sound adaptive noise reduction filter for auscultation.

[0022] A storage medium, wherein the storage medium stores a computer program, which, when executed by a processor, causes the processor to execute the above-mentioned fast setting algorithm for initial weights of an ambient sound adaptive noise reduction filter for auscultation.

[0023] A computing device includes a processor and a memory for storing a program executable by the processor. When the processor executes the program stored in the memory, the above-mentioned fast setting algorithm for initial weights of an ambient sound adaptive noise reduction filter for auscultation is implemented.

[0024] Compared with the prior art, the present invention has the following advantages and beneficial effects:

[0025] 1. The present invention only needs to perform deconvolution operation on the main and auxiliary channel signals, avoiding the operation of obtaining the signal correlation matrix and inverting the high-order matrix. The calculation amount is small and the weights and filter orders can be set online, which can be directly applied to assembly line batch production.

[0026] 2. The present invention can also be applied to wearable body sound collection equipment; when the use environment changes, the equipment can use the method of the present invention to automatically set the initial weight online to improve the noise reduction performance. BRIEF DESCRIPTION OF THE DRAWINGS

[0027] Figure 1 is a flow chart of the algorithm of the present invention;

[0028] Figure 2 It is a schematic diagram of the calculation principle of the transfer function G(s) of the desired model of the dual-channel adaptive filter;

[0029] Figure 3 The waveform diagram of the output signals of the main and auxiliary channels measured when a pulse signal is played by a speaker in a quiet environment in the first embodiment of the present invention;

[0030] Figure 4 It is the impulse response waveform of the expected model of the adaptive filter calculated by the method of the present invention in the first embodiment of the present invention. DETAILED DESCRIPTION

[0031] The present invention is further described in detail below in conjunction with the accompanying drawings and specific implementation methods.

[0032] Embodiment 1

[0033] The present embodiment is an algorithm for quickly setting the initial weights of an ambient sound adaptive noise reduction filter for auscultation. The algorithm addresses the problem that existing methods have large computational complexity and are not suitable for online setting of initial parameters of factory stethoscopes on the production line. Based on the relationship between the pulse responses of the main and auxiliary channels of the stethoscope, the algorithm quickly estimates the order and initial weights of the adaptive filter.

[0034] To implement the above-mentioned fast setting algorithm of initial weights of the ambient sound adaptive noise reduction filter for auscultation, this embodiment includes the following steps:

[0035] Step 1: In a quiet environment, use an artificial mouth or a speaker to play a pulse signal as environmental noise, and record the pulse start time as time zero, that is, k=0.

[0036] Step 2: Record the main channel environmental noise signal N′1(k) and the sub-channel environmental noise signal N′2(k) collected by the main channel and the sub-channel respectively under the excitation of the pulse signal of the dual-channel digital stethoscope.

[0037] Step 3: Use a bandpass filter to perform the same preprocessing on the main channel ambient noise signal N′1(k) and the secondary channel ambient noise signal N′2(k) to suppress the low-frequency and high-frequency components, and obtain the main channel filtered signal N1(k) and the secondary channel filtered signal N2(k) respectively. The order of the bandpass filter is 6, and the cutoff frequency is f L =50Hz and f H =1500Hz.

[0038] Step 4. After a period of time, the amplitudes of signals N1(k) and N2(k) will converge to a certain threshold and will no longer exceed the threshold over time. The threshold is determined according to the maximum amplitude of the signal, and the thresholds are set to δ1 = 0.1max{|N1(k)|} and δ2 = 0.1max{|N2(k)|}. Then the amplitude of N1(k) converges to the threshold δ1 after the 240th sampling period, and the amplitude of N2(k) converges to the threshold δ2 after the 400th sampling period. The main channel impulse response vector p1 = [N1(0)…N1(239)], and the secondary channel impulse response vector p2 = [N2(0)…N2(399)].

[0039] Step 5: Deconvolve p1 with p2 as the kernel, and the result is the impulse response sequence G = [g0, g1, ...] of the expected model of the adaptive filter; take δ g =0.1max{|g j |}, determine g j The amplitude converges to δ g The time t g =111, that is, when j>111, |g j |<δ g .

[0040] Then the order of the adaptive filter is 112, and its initial value is [g0,g1,…,g 111 ].

[0041] An environmental sound adaptive noise reduction device for auscultation realizes the rapid setting algorithm of the initial weights of the environmental sound adaptive noise reduction filter for auscultation.

[0042] A storage medium, wherein the storage medium stores a computer program, which, when executed by a processor, causes the processor to execute the above-mentioned fast setting algorithm for initial weights of an ambient sound adaptive noise reduction filter for auscultation.

[0043] A computing device includes a processor and a memory for storing a program executable by the processor. When the processor executes the program stored in the memory, the above-mentioned fast setting algorithm for initial weights of an ambient sound adaptive noise reduction filter for auscultation is implemented.

[0044] The above embodiments are preferred implementation modes of the present invention, but the implementation modes of the present invention are not limited to the above embodiments. Any other changes, modifications, substitutions, combinations, and simplifications that do not deviate from the spirit and principles of the present invention should be equivalent replacement methods and are included in the protection scope of the present invention.

Claims

1. A fast algorithm for setting the initial weights of an adaptive noise reduction filter for ambient sound used in auscultation, characterized in that The steps include: Step S1: In a quiet environment, use an artificial mouth or a speaker to play a pulse signal as environmental noise, and record the pulse start time as time zero, that is, k = 0; Step S2, recording the main channel environmental noise signal N′1(k) and the auxiliary channel environmental noise signal N′2(k) collected by the main channel and the auxiliary channel respectively under the excitation of the pulse signal of the dual-channel digital stethoscope; Step S3: Use a bandpass filter to perform the same preprocessing on the main channel environmental noise signal N′1(k) and the sub-channel environmental noise signal N′2(k) to suppress low-frequency and high-frequency components, and obtain the main channel filtered signal N1(k) and the sub-channel filtered signal N2(k) respectively; Step S4, after a period of time, the amplitudes of signals N1(k) and N2(k) will converge to a certain threshold and will no longer exceed the threshold over time. The threshold corresponding to N1(k) is δ1, and the threshold corresponding to N2(k) is δ2; and the moment when the amplitude of N1(k) converges to the threshold δ1 is t1, and the moment when the amplitude of N2(k) converges to δ2 is t2. The main channel impulse response vector p1 = [N1(0)…N1(t1)], and the sub-channel impulse response vector p2 = [N2(0)…N2(t2)]; Step S5, with p2 as the kernel, deconvolute p1, and the result is the impulse response sequence G = [g0, g1, ...] of the expected model of the adaptive filter; when j is greater than a certain value, g j The amplitude of will converge to a certain threshold and will no longer exceed the threshold over time; Let this threshold be δ g , and g j The amplitude converges to the threshold δ g The time is t g , that is, when j>t g When |g j |<δ g , then the order of the adaptive filter can be taken as t g +1, take its initial value Preferably, take δ g =c g max{|g j |}, c g ∈[0.01,0.1].

2. The fast setting algorithm of initial weights of the ambient sound adaptive noise reduction filter for auscultation according to claim 1, characterized in that: In step S3, the main channel ambient noise signal N′1(k) and the secondary channel ambient noise signal N′2(k) are preprocessed in the same manner using a bandpass filter to suppress low-frequency and high-frequency components, respectively, to obtain a main channel filtered signal N1(k) and a secondary channel filtered signal N2(k), respectively.

3. The fast setting algorithm of initial weights of the ambient sound adaptive noise reduction filter for auscultation according to claim 2, characterized in that: The preferred value of the order of the bandpass filter is not less than 2, and the preferred values ​​of the cut-off frequency are f L ∈[20,80]Hz and f H ∈[1500,2000]Hz.

4. The fast setting algorithm of initial weights of the ambient sound adaptive noise reduction filter for auscultation according to claim 1, characterized in that: In step S5, the order and initial weight of the adaptive filter are determined as follows: Step S5, with the secondary channel impulse response vector p2 as the core, the main channel impulse response vector p1 is deconvolved, and the result is the impulse response sequence G = [g0, g1, ...] of the expected model of the adaptive filter; when j is greater than a certain value, |g j |will converge to a certain threshold and will not exceed that threshold over time; Let this threshold be δ g , while |g j |Converge to threshold δ g The time is t g , that is, when j>t g When |g j |<δ g , then the order of the adaptive filter can be taken as t g +1, take its initial value Preferably, take δ g =c g max{g j }, c g ∈[0.01,0.1].

5. The fast setting algorithm of initial weights of the ambient sound adaptive noise reduction filter for auscultation according to claim 4, characterized in that: The main channel impulse response vector p1=[N1(0)…N1(t1)], wherein t1 is the moment when |N1(k)| converges to the threshold δ1, that is, when k>t1, |N1(k)|<δ1; preferably, the threshold δ1=c1max{|N1(k)|}, and c1∈[0.01,0.1].

6. The fast setting algorithm of initial weights of the ambient sound adaptive noise reduction filter for auscultation according to claim 4, characterized in that: The secondary channel impulse response vector p2=[N2(0)…N2(t2)], wherein t2 is the moment when |N2(k)| converges to the threshold δ2, that is, when k>t2, |N2(k)|<δ2; preferably, the threshold δ2=c2max{|N2(k)|}, and c2∈[0.01,0.1].

7. The fast setting algorithm of initial weights of the ambient sound adaptive noise reduction filter for auscultation according to claim 5, characterized in that: The main channel impulse response vector p1=[N1(0)…N1(t1)] is used to determine the threshold δ1 at the time t1, and its preferred value is δ1=c1max{|N1(k)|}, and c1∈[0.01,0.1].

8. The fast setting algorithm of initial weights of the ambient sound adaptive noise reduction filter for auscultation according to claim 6, characterized in that: The secondary channel impulse response vector p2=[N2(0)…N2(t2)] is used to determine the threshold δ2 at the time t2, and its preferred value is δ2=c2max{|N2(k)|}, and c2∈[0.01,0.1].

9. A storage medium, characterized in that: The storage medium stores a computer program, which, when executed by a processor, enables the processor to execute the fast setting algorithm for initial weights of an ambient sound adaptive noise reduction filter for auscultation according to any one of claims 1 to 8.

10. A computing device comprising a processor and a memory for storing a program executable by the processor, characterized in that: When the processor executes the program stored in the memory, the fast setting algorithm of the initial weights of the ambient sound adaptive noise reduction filter for auscultation according to any one of claims 1 to 8 is implemented.

Images