Digital signal filtering processing method and system for blood flow graph

Abstract

The present invention relates to the field of signal processing technology, and specifically to a digital signal filtering processing method and system for a blood flow graph, comprising: preliminarily filtering out basic impedance drift; filtering out high-frequency noise; increasing frequency doubling; finding heartbeat characteristic points to divide heartbeat cycles and obtain blood flow graph wave groups; filtering out basic impedance again; normalizing characteristic blood flow graph cycles: mainly including processing descending branch waveforms; matrix singular value decomposition to capture characteristic REG waveforms; the system comprehensively integrates the core advantages of bioimpedance detection technology and brain impedance blood flow graph technology, thereby improving the accuracy of data acquisition and detection results; creatively designs high-precision analog front-end hardware circuits into a size suitable for wear, which is more portable and expands application scenarios; the collected bioimpedance data has higher requirements for digital signal processing algorithms, and the adopted algorithms can effectively reduce system noise, correct the phase of the data, and accurately realize data reconstruction.

Description

Technical Field

[0001] The present invention relates to the technical field of signal processing, and in particular to a digital signal filtering processing method and system for a blood flow graph. Background Art

[0002] Bioelectrical impedance detection technology is based on the basic principle that the human body is a large bioelectric conductor. Each tissue has a certain impedance. When a local tissue of the human body is diseased, the impedance of that part must be different from other parts. Therefore, the pathological changes of human tissue can be diagnosed through impedance detection. A known voltage is injected into a specific part of the human body to measure the current caused on the body surface, or a known current is injected to measure the voltage caused on the body surface. The impedance presented by each tissue inside the human body under the action of the electric field is calculated using the measured current or voltage value according to a certain reconstruction algorithm.

[0003] The analog front-end circuit is the most critical part of the headband hardware design. In order to reduce the power consumption of the system, it is necessary to use a high-precision, low-power analog signal processing chip. High precision and low power consumption are a pair of contradictory indicators, and it is necessary to find a balance between the two. Therefore, it is also necessary to use specially designed analog filters and digital filters that combine software and hardware to ensure the overall impedance measurement accuracy of the system.

[0004] Cerebral blood flow graph, also known as encephalogram, or REG for short, is a biophysical examination method. It is one of the non-invasive and safe examination methods. It has been used in clinical practice in China since the 1960s, but due to the limitations of electronic circuit technology, there has been no better method for digital signal filtering processing of blood flow graphs.

[0005] The problems in the prior art are that the drift noise and Gaussian noise of the basic impedance signal are not filtered out sufficiently, and the characteristic waveform is not captured accurately enough. Summary of the invention

[0006] The purpose of the present invention is to provide a digital signal filtering processing method for a blood flow diagram to solve the problems raised in the above background technology.

[0007] The technical solution of the present invention is: a method for filtering and processing a digital signal of a blood flow graph, comprising the following steps:

[0008] Step 1: Preliminary filtering of basic impedance drift;

[0009] Step 2: Filter out high-frequency noise;

[0010] Step 3: Increase the frequency multiplier;

[0011] Step 4: Find the heartbeat characteristic points to divide the heartbeat cycle and obtain the blood flow wave group;

[0012] Step 5: Filter out the basic impedance again;

[0013] Step 6: Normalization of characteristic blood flow graph cycle: mainly includes processing the descending branch waveform;

[0014] Step 7: Matrix singular value decomposition is used to capture the characteristic REG waveform.

[0015] Preferably, in step 1, the basic impedance drift is mostly low-frequency signal fluctuation, and Butterworth 8th-order high-pass filtering is used to obtain the basic impedance drift curve, wherein Fs=50 Hz and Fc=2 Hz.

[0016] Preferably, in step 2, Butterworth 6th-order low-pass filtering is used, wherein Fs=50; Fpass=5; Fstop=20; Apass=1; Astop=100.

[0017] Preferably, the filtering processing in step one and step two both adopts double palindrome data length filtering: the input data length is at least 8.84s*50hz=442 points of the amplitude response length; wherein 8.84s is the filtering response time; and positive and negative filtering are adopted twice: the phase drift caused by filtering is eliminated.

[0018] Preferably, the specific method of step three is: multiplying the original data sampling frequency of 50 Hz to 200 Hz, and using cubic spline interpolation to supplement the midpoint frequency doubling.

[0019] Preferably, the specific method of step four is: perform FFT transformation on the frequency-multiplied curve to obtain the heartbeat frequency, then take the maximum value point within a certain range window as the heartbeat feature point and locate the feature point, and the window width is determined by the heartbeat frequency calculation.

[0020] Preferably, the specific method of step five is: taking the heartbeat characteristic point as the reference point and performing cubic spline interpolation to make a basic impedance curve, so that: blood flow curve = -(impedance curve - basic impedance curve).

[0021] Preferably, in step six, the specific method for normalizing the period of the characteristic blood flow graph is: obtaining the average period T of the wave group; and linearly interpolating the descending branch band to normalize all waveform periods in the wave group to T.

[0022] Preferably, the specific method of step seven is: assuming that there are n bands in the wave group and m discrete points of the waveform within the period T, the wave group can constitute a matrix M (n×m), and M is subjected to singular value decomposition to obtain N (1×m), where N is the typical feature mapping of the blood flow signal under the influence of the pulse in different heartbeat bands.

[0023] The digital signal filtering and processing system of the blood flow graph includes two parts: a filtering circuit and a filtering algorithm. The filtering circuit is responsible for collecting more accurate and clear impedance waveforms, and the filtering algorithm is responsible for filtering out impurities in the collected waveforms.

[0024] The present invention improves the digital signal filtering processing method of the blood flow diagram provided herein, and has the following improvements and advantages compared with the prior art:

[0025] First, the present invention adopts a method of filtering out the basic impedance twice (the first time is Butterworth low-pass filtering, and the second time is feature point interpolation). The two different basic impedance drift filtering methods make the blood flow map signal under the influence of the pulse more accurate;

[0026] Secondly, the normalization method formulated by the present invention for the blood flow map characteristics (the ascending branch remains unchanged) and the capture of typical waveforms by matrix singular value decomposition are more advantageous than the traditional accumulation method. BRIEF DESCRIPTION OF THE DRAWINGS

[0027] The present invention will be further explained below in conjunction with the accompanying drawings and embodiments:

[0028] Figure 1 is a filtering flow chart of the present invention;

[0029] Figure 2 It is a frequency response curve diagram of Butterworth 8th order high pass filter of the present invention;

[0030] Figure 3 is an amplitude response curve diagram of the Butterworth 8th-order high-pass filter of the present invention;

[0031] Figure 4 It is a frequency response curve diagram of Butterworth 6th order low pass filter of the present invention;

[0032] Figure 5 is an amplitude response curve diagram of the Butterworth 6th-order low-pass filter of the present invention;

[0033] Figure 6 This is a waveform diagram before the high-frequency noise is filtered out by the present invention;

[0034] Figure 7 It is a waveform diagram after filtering out high-frequency noise in the present invention;

[0035] Figure 8 It is a waveform diagram of the present invention when taking heartbeat characteristic points;

[0036] Fig. 9 is the basic impedance curve of the present invention;

[0037] Fig.10 is the blood flow diagram curve of the present invention;

[0038] Fig.11 This is the waveform diagram of the descending branch processed by the present invention. DETAILED DESCRIPTION

[0039] The present invention is described in detail below, and the technical solutions in the embodiments of the present invention are clearly and completely described. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0040] The present invention improves the digital signal filtering processing method of the blood flow diagram provided herein, and the technical solution of the present invention is:

[0041] like Figure 1-Figure 5 As shown, the digital signal filtering processing method of the blood flow graph comprises the following steps:

[0042] Step 1: Preliminary filtering of basic impedance drift: Basic impedance drift is mostly low-frequency signal fluctuations. The basic impedance drift is obtained by using Butterworth 8th-order high-pass filtering, where Fs = 50 Hz and Fc = 2 Hz.

[0043] Step 2, filter out high-frequency noise: use Butterworth 6th-order low-pass filtering, where Fs=50; Fpass=5; Fstop=20; Apass=1; Astop=100 (the filtering processing in steps 1 and 2 both use double palindrome data length filtering: the input data length is at least 8.84s*50hz=442 points of the amplitude response length; 8.84s is the filter response time; and use positive and negative filtering twice: eliminate the phase drift caused by filtering);

[0044] Step 3: Increase the frequency: multiply the original data sampling frequency from 50Hz to 200hz, and use cubic spline interpolation to supplement the frequency of the intermediate points;

[0045] Step 4: Find the characteristic points of the heartbeat to divide the heartbeat cycle and obtain the wave group of the blood flow graph: perform FFT transformation on the frequency-multiplied curve to obtain the heartbeat frequency, then take the maximum value point within a certain range window as the heartbeat characteristic point and locate the characteristic point. The window width is determined by the heartbeat frequency calculation;

[0046] Step 5: Filter out the basic impedance again: Use the heartbeat feature point as the reference point and use cubic spline interpolation to make a basic impedance curve, so that: blood flow curve = - (impedance curve - basic impedance curve), the formula is:

[0047] ω=-(λ-λ b );

[0048] Among them, impedance curve λ refers to the above-mentioned frequency-doubled curve;

[0049] Basic impedance curve λ b:The basic impedance is the sum of the inherent impedance of human tissue and the transmission impedance of the circuit. It will drift slightly during the acquisition process, so it is necessary to capture (cubic spline interpolation of feature points) and remove the basic impedance drift curve in the signal to obtain the effective blood flow curve;

[0050] Blood flow graph curve ω: blood flow Figure 1 Generally expressed as a blood flow volume curve, the impedance signal needs to be negated to flip, so there is a "-";

[0051] Step 6, cycle normalization of characteristic blood flow graph: mainly includes processing descending branch waveforms (obtaining all blood flow graph cycle bands. Due to the slight changes in heart rate during acquisition, the partial cycles of each blood flow graph are inconsistent, so cycle normalization processing is required. Since the rapid rising segment changes less with the heart rate, the cycle changes are mainly concentrated in the descending branch, so the cycle normalization processing mainly processes the descending branch waveform): The specific method of cycle normalization of characteristic blood flow graph is: obtaining the average value T of the wave group cycle; linearly interpolating the descending branch band to normalize all waveform cycles in the wave group to T;

[0052] Step 7: Matrix singular value decomposition to capture the characteristic REG waveform: Assume that there are n bands in the wave group and m discrete points of the waveform in period T, then the wave group can form a matrix M (n×m). Perform singular value decomposition on M to obtain N (1×m), where N is the typical feature mapping of the blood flow signal under the influence of pulse in different heartbeat bands.

[0053] The digital signal filtering and processing system of the blood flow graph includes two parts: a filtering circuit and a filtering algorithm. The filtering circuit is responsible for collecting more accurate and clear impedance waveforms, and the filtering algorithm is responsible for filtering out impurities in the collected waveforms. The two complement each other to make the obtained blood flow graph waveform more accurate and reliable.

[0054] Specifically, the input brain electrical impedance signal is first processed by the input buffer (BUF), then the signal is gain amplified (PGA), then processed by the anti-aliasing filter (AAF), and then buffered (BUF) again, and then the analog signal is digitized (AD), and the digitized digital signal is processed by discrete Fourier transform (DFT) (to remove interference caused by impurity components to the image), and finally the entire signal is filtered through hardware digital filters.

[0055] The acquisition of brain electrical impedance signals has high requirements for electrode conductivity, accuracy, noise and polarization potential stability. It needs to have good conductivity, high accuracy, low noise and stable polarization potential. This circuit design removes noise from the acquired brain electrical impedance signals by filtering the signals and effectively extracts blood flow map signals. The algorithm mainly uses a specialized Butterworth low-pass algorithm, fast Fourier transform capture cycle, and matrix singular value decomposition SVD to obtain typical waveform templates.

[0056] This system integrates the core advantages of bioimpedance detection technology and brain impedance blood flow mapping technology to improve the accuracy of data acquisition and detection results; creatively designs high-precision analog front-end hardware circuits into a wearable size, which is more portable and expands the application scenarios; the collected bioimpedance data has higher requirements for digital signal processing algorithms, and the adopted algorithm can effectively reduce system noise and correct the phase of the data to accurately achieve data reconstruction.

[0057] The above description enables those skilled in the art to implement or use the present invention. Various modifications to these embodiments will be apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention will not be limited to the embodiments shown herein, but will conform to the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A method for filtering and processing a digital signal of a blood flow graph, characterized in that: The following steps are involved: Step 1: Preliminarily filter out the basic impedance drift and obtain the basic impedance curve by Butterworth high-pass filtering; Step 2: Filter out high-frequency noise; Step 3: Increase the frequency multiplier; Step 4: Find the heartbeat characteristic points to divide the heartbeat cycle and obtain the blood flow wave group; Step 5: Filter out the basic impedance again and obtain the basic impedance curve by cubic spline interpolation; Step 6: Normalizing the characteristic blood flow graph cycle: including processing the descending branch waveform to normalize the characteristic blood flow graph cycle; Step 7: Matrix singular value decomposition is used to capture the characteristic REG waveform.

2. The method for filtering the digital signal of a blood flow graph according to claim 1, characterized in that: In step 1, the basic impedance drift is mostly low-frequency signal fluctuation, and the basic impedance drift curve is obtained by using Butterworth 8th-order high-pass filtering, where Fs=50 Hz and Fc=2 Hz.

3. The method for filtering and processing the digital signal of the blood flow diagram according to claim 1, characterized in that: In step 2, a Butterworth 6th-order low-pass filter is used, wherein Fs=50; Fpass=5; Fstop=20; Apass=1; Astop=100.

4. The method for filtering and processing the digital signal of a blood flow graph according to claim 1, characterized in that: The Butterworth filtering in step 1 and step 2 uses two positive and negative data filtering, and the input data length is at least 8.84s*50hz=442 points of the amplitude response length; wherein 8.84s is the filter response time.

5. The method for filtering and processing digital signals of a blood flow graph according to claim 1, characterized in that: The specific method of step three is: multiply the original data sampling frequency of 50Hz to 200hz, and use cubic spline interpolation to supplement the midpoint frequency doubling.

6. The method for filtering and processing digital signals of a blood flow graph according to claim 1, characterized in that: The specific method of step four is: perform FFT transformation on the frequency-multiplied curve to obtain the heartbeat frequency, then take the maximum value point in a certain range window as the heartbeat feature point and locate the feature point, and the window width is determined by the heartbeat frequency calculation.

7. The method for filtering and processing digital signals of a blood flow graph according to claim 1, characterized in that: The specific method of step five is: taking the heartbeat characteristic point as the reference point and performing cubic spline interpolation to make a basic impedance curve, so that: blood flow curve = - (impedance curve - basic impedance curve).

8. The method for filtering and processing digital signals of a blood flow graph according to claim 1, characterized in that: In step six, the specific method for normalizing the characteristic blood flow image period is: obtaining the average value T of the wave group period; linearly interpolating the descending branch band to normalize all waveform periods in the wave group to T.

9. The method for filtering and processing digital signals of a blood flow graph according to claim 1, characterized in that: The specific method of step seven is: suppose there are n bands in the wave group, and the number of discrete waveform points in period T is m, then the wave group can form a matrix M (n×m), and M is subjected to singular value decomposition to obtain N (1×m), where N is the typical feature mapping of the blood flow signal under the influence of the pulse in different heartbeat bands.

10. A system using the method for filtering and processing the digital signal of a blood flow image as claimed in claim 1, characterized in that: It includes two parts: a filtering circuit and a filtering algorithm. The filtering circuit is responsible for collecting more accurate and clear impedance waveforms, and the filtering algorithm is responsible for filtering out impurities in the collected waveforms.

Images