CN104825140A - Digital filter method for pulse wave extraction, and digital filter - Google Patents

Abstract

The present invention relates to the technical field of medical electronics, in which medical devices, such as electronic sphygmomanometers and blood oxygen monitors, need to extract and analyze human pulse signals, and particularly relates to a digital filter method and a digital filter that extract human pulse signals. The digital filter method can be applied to various digital medical devices. Provided in the scheme are a digital filter method and a digital filter, which provide a method that is simple and easy to implement and determination criteria for near-zero-hertz interference caused by a variety of interference sources, so that an extracted pulse wave spectrum is as clean as possible. The digital filter implemented by using the digital filter method can be used to separate a direct current component and an alternating current component from a signal output by a sensor, and extract a required pulse signal. The PCB circuit is simplified, interference and costs are reduced, and meanwhile the problem of a long delay in an analog filter is overcome. In the day of downsized high performance MCU chips, the digital filter method is especially applicable to mini wearable devices and is also applicable to general medical electronic devices.

Description

The digital filtering method that pulse wave extracts and digital filter

Technical field

The present invention relates to the medical electronics technical field that the armarium such as electric sphygmomanometer, blood-oxygen monitor needs to extract and analyze human pulse signal, particularly relate to a kind of digital filtering method and the digital filter that extract human pulse signal, the method can be applicable in multiple digital medical equipment.

Background technology

The present invention, for the electronic blood pressure measurement device based on " oscillographic method ", illustrates a kind of digital filtering method and the digital filter that can be used for extracting and analyzing human pulse signal.As shown in Figure 1, this device can realize the automatic measurement of blood pressure for the basic module of electronic blood pressure measurement device and system composition." oscillographic method " is the static pressure signal being obtained tremulous pulse generation in measuring process by pressure transducer, this signal is the mixed signal comprising the larger direct current signal of amplitude and the less AC signal of amplitude, direct current signal can characterize the real-time pressure in measuring process in cuff, AC signal is wherein then the pulse pressure signal that tremulous pulse is passed in cuff, although the pulse intensity of human body varies with each individual, the intensity of pulse signal is in the scope of 1 ~ 3mmHg usually.Meanwhile, according to medical observation, the power spectrum scope of human pulse signal is roughly 0.2Hz ~ 40Hz, and the concentration of energy of the overwhelming majority is at below 10Hz.

Utilize oscillographic method principle Measure blood pressure, the direct current signal in the static pressure signal exported by sensor and AC signal is needed to realize being separated, extract AC signal, and systolic pressure and the diastolic pressure of test individual is determined by the feature of the amplitude envelope line of AC signal, as shown in Figure 2, specifically determine the method for systolic pressure and diastolic pressure, can with reference to No. 201310129242.6 applications for a patent for invention.Current electronic blood pressure measurement device extensively adopts the method for analog filtering to realize, the active high-pass filter that its common way is realized by one group of amplifier and R, C element and active low-pass filter, form a band filter, from the output pressure signal of pressure transducer, separation and extraction goes out pulse signal.The advantage of analog filtering is the calculation resources can saving MCU, but also has obvious deficiency:

1, after determining the high pass cut off frequency of wave filter, how to select suitable circuit parameter cumbersome, there is larger time delay in wave filter, makes measurement result occur error;

2, take more PCB space, be not suitable for wearable device, and irrational wiring may introduce new interference;

If the analog filter of the higher exponent number of 3 employing, circuit will be huger, and hardware cost can be caused to increase;

The present invention proposes a kind of digital filtering method and digital filter, realize the separation of the direct current to pressure transducer output signal, AC compounent, the pulse signal needed for extraction, as shown in Figure 3.Both simplify PCB circuit, and reduced interference, can reduce costs again, overcome the larger time delay that analog filter exists simultaneously.At high-performance MCU(as ARM) today of chip small-sized, digital filtering method is specially adapted to miniature wearable device, is also applicable to general medical treatment electronic equipment.

Summary of the invention

In order to overcome the deficiency adopting analog filter in above-mentioned measuring process, the present invention proposes a kind of method and digital filter of digital filtering, is separated the direct current mixed in sensor output signal, AC compounent, and realizes extraction and the analysis of pulse signal.

The technical solution adopted in the present invention is:

1, because the energy of human pulse signal mainly concentrates on 0.2Hz ~ 10Hz, the rational sample frequency of setting A/D converter fs=100Hz ~ 500Hz(for simplicity, all gets below fs=100Hz).

2, need emphasis consideration to the suppression of 50Hz Hz noise and filtration simultaneously, so devise two digital filters respectively, a cut-off frequency f1the high pass filter of=0.2Hz and a cut-off frequency f2the low pass filter of=10Hz, be connected in series by these two digital filters and can form a band filter, the frequency band that its passband is the most intensive with energy in pulse signal power spectrum is consistent.Here high pass filter is N rank iir filter (N>=2), the reason adopting IIR form is for the cut-off frequency requirement close to 0Hz, only have the FIR on even thousands of rank, hundreds of rank could obtain the performance similar to three rank IIR and effect, most MCU cannot meet the FIR of the even thousands of exponent number of hundreds of to the requirement of calculation resources at all.And low pass filter can be IIR or the FIR filter (N>=2) on N rank, as long as the requirement that this wave filter suppressed sidelobes level attenuation degree coincidence measurement system is filtered signal, so corresponding filter parameter and admissible.

3, cut-off frequency f1the high pass filter parameter designing of=0.2Hz.The present invention proposes the decision method of a kind of cut-off frequency near the IIR high pass filter parameter designing of 0Hz, as follows:

(1) according to list of references, low-frequency disturbance (coming from the factors such as baseline drift, human body human body displacement) may be there is in the pulse wave signal comprised in sensor output signal between 0Hz ~ 1Hz, particularly close to the place of 0 frequency, exist so-called " 0 subharmonic " of larger component, in order to suppress and eliminate this interference, wish high pass filter 0 ~ f1hz interval can decline fast.When calculating the amplitude-frequency response of this digital filter, it is frequency fperiodic function, in computational process, frequency ffrom 0, and by the interval set continuous increase, " nearly 0 frequency " that define this high pass filter is , so use the size of place's amplitude fading weigh this wave filter 0 ~ f1the degree of the interval decline speed of Hz, decay larger, declines faster, decline faster, suppresses and to eliminate the interference effect of nearly 0 frequency better.

(2) in order to eliminate the interference at nearly 0 frequency place, the pulse signal that reduction frequency spectrum is clean, only need at normalized logarithm amplitude frequency curve place's amplitude fading makes restriction, as <=-40db.For the IIR high pass filter meeting this decay characteristics, the present invention is defined and is classified as---the nearly 0 IIR high pass filter frequently suppressed.

Setting place's amplitude fading is-40db, uses IIR(Butterworth) method design low-pass cut-off frequencies f1the high pass filter of=0.2Hz.Calculate and find, when N=2 rank, the amplitude-frequency characteristic of wave filter can meet the demands, and the parameter of this wave filter SOS second order joint is as follows:

a1=1 a2=-2 a3=1 b1=1 b2=-1.982228929792528 b3=0.982385450614125

Limit meets stability requirement.The amplitude-frequency response characteristic of this wave filter as shown in Figure 4.

The system function of this wave filter is:

4, cut-off frequency f2the low pass filter parameter designing of=10Hz.Design the filtering that this wave filter needs mainly to consider the Hz noise of 50Hz, and the width of transition band.The degree needing equally to consider that wave filter suppresses Hz noise and standard, decision method is: by the log magnitude-frequency characteristics of observation filter device, if the relative main lobe of secondary lobe is decayed reach-40db, just thinks that the effect of wave filter can accept.

Wave filter has two kinds of selections: IIR low pass filter, FIR low pass filter.If iir filter, select the transition band width that is suitable, as being 10Hz, find by calculating, the IIR low pass filter on N=6 rank can meet design requirement, and the parameter of three SOS second order joints of this wave filter is as follows:

a11=1 a12=2 a13=1 b11=1 b12=-1.404384890471582 b13=0.735915191196472

a21=1 a22=2 a23=1 b21=1 b22=-1.142980502539901 b23=0.412801598096189

a31=1 a32=2 a33=1 b31=1 b32=-1.032069405319710 b33=0.275707942472944

Limit meets stability requirement.The amplitude-frequency response characteristic of this wave filter as shown in Figure 5.

The system function of this wave filter is:

If FIR filter, transition band width is similarly 10Hz, finding, selecting Hamming window function, when N=30 rank, having had good filter effect by calculating.Here the FIR low pass filter transition band width giving Hamming window function method for designing N=36 rank is 10Hz, side lobe attenuation is-52db at least, close-the 60db of 50Hz place amplitude fading simultaneously, namely Hz noise is reduced to original 1 ‰ by the amplitude that wave filter is later, can meet design requirement, the parameter of this wave filter is as follows:

h(0)=-0.001345469739806

h(1)=-0.001549062862164

h(2)=-0.001259886498971

h(3)=0.000000000000000

h(4)=0.002507372631740

h(5)=0.005689420470617

h(6)=0.007820542862620

h(7)=0.006508806744404

h(8)=-0.000000000000000

h(9)=-0.011225871405464

h(10)=-0.023457018064773

h(11)=-0.030157567263946

h(12)=-0.024010891617242

h(13)=0.000000000000000

h(14)=0.041740617113497

h(15)=0.094691303089101

h(16)=0.147166259967970

h(17)=0.185790335806383

h(18)=0.200000000000000

h(19)=0.185790335806383

h(20)=0.147166259967970

h(21)=0.094691303089101

h(22)=0.041740617113497

h(23)=0.000000000000000

h(24)=-0.024010891617242

h(25)=-0.030157567263946

h(26)=-0.023457018064773

h(27)=-0.011225871405464

h(28)=-0.000000000000000

h(29)=0.006508806744404

h(30)=0.007820542862620

h(31)=0.005689420470617

h(32)=0.002507372631740

h(33)=0.000000000000000

h(34)=-0.001259886498971

h(35)=-0.001549062862164

h(36)=-0.001345469739806

The system function of this wave filter is:

Amplitude-frequency response characteristic as shown in Figure 6.

5, by above-mentioned IIR high pass filter with IIR low pass filter series connection or by IIR high pass filter with FIR low pass filter the digital band-pass filter be composed in series, its high pass cut off frequency is 0.2Hz, low-pass cut-off frequencies 10Hz, can realize the separation and extraction to pulse wave signal.

6,3 data buffer zones are distributed.Title and the purposes of these 3 relief areas are as follows:

Buffer1---receive pressure transducer and to sample the data obtained through A/D change-over circuit, the input data of high pass filter take from this relief area;

Buffer2---store the output data of high pass filter, the input data of low pass filter take from this relief area;

Buffer3---store the output data of low pass filter.

7, the flow process of date processing as shown in Figure 4.After completing computing, the data be stored in inside buffer3 are exactly separation and extraction AC signal out (i.e. pulse signal) from pressure transducer output signal.In the application of oscillographic method electronic blood pressure measurement, can the envelope of this matching pulse wave, and according to the feature determination systolic pressure of envelope and the occurrence of diastolic pressure, and in Oximetry application, then according to peak value identification pulse signal, the heart rate of tested individuality can be calculated.

  

List of references:

1, " EMD is eliminating the application of motion artifacts in photoplethysmographic ", Wang Qian/Ding Hongxia/first Room, advanced technology research is circulated a notice of, Vol.4 No.6/Jun.2010;

2, " five notes of traditional Chinese music sequence and the research of human pulse ripple relation ", Liu Wenquan/Wang Zhen/Yu Wenlong etc., advanced technology research is circulated a notice of, Vol.5 No.3/Mar.2011;

Accompanying drawing explanation

Fig. 1 is basic module and the system composition schematic diagram of electronic blood pressure measurement device;

Fig. 2 is that in static pressure signal, direct current is separated schematic diagram with interchange;

Fig. 3 is module and the system composition schematic diagram of the blood pressure measuring device adopting digital filtering;

Fig. 4 is cut-off frequency f1the logarithm amplitude-frequency response characteristic of the 2 rank IIR high pass filters of=0.2Hz ;

Fig. 5 is cut-off frequency f2the amplitude-frequency response characteristic of the 6 rank IIR low pass filters of=10Hz ;

Fig. 6 is cut-off frequency f2the logarithm amplitude-frequency response characteristic of the 36 rank FIR low pass filters of=10Hz ;

Fig. 7 is the schematic flow sheet of digital filter process;

Main devices wherein and module declaration as follows:

MCU: use the STM32F103C8T6 of ST company, 32 ARM of this to be the kernel that a industrial circle is commonly used be Cortex-M3, have the disposal ability of the highest 72MIPS.And containing abundant peripheral circuit, as the analog to digital converter (ADC) of multichannel 12-bit precision.It is with the SRAM of 20K byte.Band flash memorizer, for preserving object code.Running voltage 1.5 ~ 3.3V.

Pressure transducer: use Freescale(Freescale) model of company is the air gauge chip of MPVX5050GP, running voltage 5.0V, measurement range is 0 ~ 375mmHg.

Display module: the LCD display device being customization, for the display of the parameter in display measurement process and measurement result.

Detailed description of the invention

Digital filter runs in MCU with the form of software code, below uses C language to realize, and by embedded-development environment, compiling connects generation object code, is downloaded to MCU by jtag interface.To each digital filter, list entries is x (n), and output sequence is y (n).

// definition data buffer zone data length

#define DATA_LENGTH 200

// definition IIR high pass filter coefficient

double float a[2]={1,-2,1};

double float b[2]={1,-1.982228929792528,0.982385450614125};

// definition IIR low-pass filter coefficients

double float a[2][2]={1,2,1,1,2,1,1,2,1};

double float a[2][2]=

{

1,-1.404384890471582,0.735915191196472,

1,-1.142980502539901,0.412801598096189,

1,-1.032069405319710,0.275707942472944};

// definition FIR low-pass filter coefficients

double float h[36]=

{

0.001345469739806,

-0.001549062862164,

-0.001259886498971,

0.000000000000000,

0.002507372631740,

0.005689420470617,

0.007820542862620,

0.006508806744404,

-0.000000000000000,

-0.011225871405464,

-0.023457018064773,

-0.030157567263946,

-0.024010891617242,

0.000000000000000,

0.041740617113497,

0.094691303089101,

0.147166259967970,

0.185790335806383,

0.200000000000000,

0.185790335806383,

0.147166259967970,

0.094691303089101,

0.041740617113497,

0.000000000000000,

-0.024010891617242,

-0.030157567263946,

-0.023457018064773,

-0.011225871405464,

-0.000000000000000,

0.006508806744404,

0.007820542862620,

0.005689420470617,

0.002507372631740,

0.000000000000000,

-0.001259886498971,

-0.001549062862164,

-0.001345469739806};

// definition data buffer zone

double float buffer1[DATA_LENGTH];

double float buffer2[DATA_LENGTH];

double float buffer3[DATA_LENGTH];

double float tmp_buff1[DATA_LENGTH];

//IIR high pass filter, calls form: High_Pass_Filter (buffer1, buffer2)

void High_Pass_Filter(double float *x, double float *y)

{

unsigned int n=0;

for(n=2,n<DATA_LENGTH;n++)

{

y[n]=(a[0]*x[n]+a[1]*x[n-1]+a[2]*x[n-2]–b[1]*y[n-1]–b[2]*y[n-2])/b[0];

}

}

//IIR low pass filter, calls form: IIR_Low_Pass_Filter (buffer2, buffer3)

void IIR_Low_Pass_Filter(double float *x, double float *y)

{

unsigned int n=0;

for(n=2,n<DATA_LENGTH;n++)

{

tmp_buff[n]=(a[0][0]*x[n]+a[0][1]*x[n-1]+a[0][2]*x[n-2]–b[0][1]*tmp_buff[n-1]–b[0][2]*tmp_buff[n-2])/b[0][0];

}

for(n=2,n<DATA_LENGTH;n++)

{

buffer2[n]=(a[1][0]*tmp_buff[n]+a[1][1]*tmp_buff[n-1]+a[1][2]*tmp_buff[n-2]–b[1][1]*buffer2[n-1]– b[1][2]*buffer2[n-2])/b[1][0];

}

for(n=2,n<DATA_LENGTH;n++)

{

buffer3[n]=(a[2][0]*buffer2[n]+a[2][1]*buffer2[n-1]+a[2][2]*buffer2[n-2]–b[2][1]*buffer3[n-1]–b[2][2]*buffer3[n-2])/b[2][0];

}

}

//FIR low pass filter, calls form: FIR_Low_Pass_Filter (buffer2, buffer3)

void FIR_Low_Pass_Filter(double float *x, double float *y)

{

unsigned int n=0,m=0;

for(n=36;n<DATA_LENGTH;n++)

{

For(m=0;m<36;m++)

{

y[n] += h[m]*x[n-m];

}

}

}

Claims (9)

1. one kind in digital medical equipment from the alternating current-direct current mixed signal that sensor exports the digital filtering method of Extraction and separation human pulse ripple and digital filter, it is characterized in that: by an IIR high pass filter and an IIR low pass filter (or a FIR low pass filter) bandpass digital filter formed of connecting, its passband frequency range is consistent with the scope that pulse wave power spectral energies is concentrated, and high pass filter exists (nearly 0 frequency) place's amplitude fading at least reaches " 0 subharmonic " that-40db causes to suppress and to eliminate many factors, make the pulse wave frequency spectrum of acquisition cleaner, and low pass filter need suppress and eliminate Hz noise, judge that the standard of filter filtering effect is the side lobe attenuation <=-40db of its log magnitude-frequency characteristics.

2. filtering method according to claim 1 and wave filter, is characterized in that: the digital high-pass filter selected within the scope of 0.2Hz ~ 0.5Hz for Extraction and separation pulse wave, cut-off frequency, and its form should be iir filter, instead of FIR filter.

3. filtering method according to claim 1 and wave filter, is characterized in that: comprise one " the nearly 0 IIR high pass filter frequently suppressed ", Yi Ji the judgment basis of the amplitude-frequency decay at (nearly 0 frequency) place, judgment basis is: normalization logarithm amplitude-frequency meets <=-40db.

4. filtering method according to claim 1 and wave filter, it is characterized in that: the low pass filter comprising IIR or FIR, and set intermediate zone width and sidelobe level decay judgment basis, intermediate zone width <=10Hz, side lobe attenuation judgment basis is: sidelobe level decay <=-40db in normalization logarithm amplitude-frequency.

5. filtering method according to claim 1 and wave filter, it is characterized in that: the digital filter of use is made up of two kinds of structures, structure of can choosing any one kind of them uses, one structure is, is composed in series by N rank " the nearly 0 IIR high pass filter frequently suppressed " (N>=2) and N rank IIR low pass filter (N>=4).

6. filtering method according to claim 1 and wave filter, it is characterized in that: the digital filter of use is made up of two kinds of structures, structure of can choosing any one kind of them uses, its two structure is, is composed in series by N rank " the nearly 0 IIR high pass filter frequently suppressed " (N>=2) and N rank FIR low pass filter (N>=30).

7. filtering method according to claim 1 and wave filter, it is characterized in that: the alternating current-direct current mixed signal exported from sensor, after A/D conversion, first be stored in buffer1, then the input of the data of buffer1 as N rank " the nearly 0 IIR high pass filter frequently suppressed " (N>=2) processed by this wave filter, the result be disposed is stored in buffer2 and treats subsequent treatment by this wave filter.

8. filtering method according to claim 1 and wave filter, it is characterized in that: using the input as N rank IIR low pass filter (N>=4) or N rank FIR low pass filter (N>=30) of the data that store in buffer2, by this filter process, the result be disposed is stored in buffer3 by this wave filter.

9. filtering method according to claim 1 and wave filter, is characterized in that: the data exported by sensor, does not process by analog filter circuit in filter process, but has been run by MCU and corresponding digital filter code completely.