CN111657900B - Korotkoff sound time phase classification and identification method and system - Google Patents

Korotkoff sound time phase classification and identification method and system Download PDF

Info

Publication number
CN111657900B
CN111657900B CN202010350121.4A CN202010350121A CN111657900B CN 111657900 B CN111657900 B CN 111657900B CN 202010350121 A CN202010350121 A CN 202010350121A CN 111657900 B CN111657900 B CN 111657900B
Authority
CN
China
Prior art keywords
data
audio data
stethoscope
sequence
korotkoff sound
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Active
Application number
CN202010350121.4A
Other languages
Chinese (zh)
Other versions
CN111657900A (en
Inventor
潘帆
何培宇
郑定昌
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Sichuan University
Original Assignee
Sichuan University
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Sichuan University filed Critical Sichuan University
Priority to CN202010350121.4A priority Critical patent/CN111657900B/en
Publication of CN111657900A publication Critical patent/CN111657900A/en
Application granted granted Critical
Publication of CN111657900B publication Critical patent/CN111657900B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B5/00—Measuring for diagnostic purposes; Identification of persons
    • A61B5/02—Detecting, measuring or recording pulse, heart rate, blood pressure or blood flow; Combined pulse/heart-rate/blood pressure determination; Evaluating a cardiovascular condition not otherwise provided for, e.g. using combinations of techniques provided for in this group with electrocardiography or electroauscultation; Heart catheters for measuring blood pressure
    • A61B5/021—Measuring pressure in heart or blood vessels
    • A61B5/022—Measuring pressure in heart or blood vessels by applying pressure to close blood vessels, e.g. against the skin; Ophthalmodynamometers
    • A61B5/02208—Measuring pressure in heart or blood vessels by applying pressure to close blood vessels, e.g. against the skin; Ophthalmodynamometers using the Korotkoff method
    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B5/00—Measuring for diagnostic purposes; Identification of persons
    • A61B5/72—Signal processing specially adapted for physiological signals or for diagnostic purposes
    • A61B5/7235—Details of waveform analysis
    • A61B5/7264—Classification of physiological signals or data, e.g. using neural networks, statistical classifiers, expert systems or fuzzy systems
    • A61B5/7267—Classification of physiological signals or data, e.g. using neural networks, statistical classifiers, expert systems or fuzzy systems involving training the classification device

Landscapes

  • Health & Medical Sciences (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Artificial Intelligence (AREA)
  • Medical Informatics (AREA)
  • General Health & Medical Sciences (AREA)
  • Biophysics (AREA)
  • Pathology (AREA)
  • Cardiology (AREA)
  • Biomedical Technology (AREA)
  • Heart & Thoracic Surgery (AREA)
  • Vascular Medicine (AREA)
  • Molecular Biology (AREA)
  • Surgery (AREA)
  • Animal Behavior & Ethology (AREA)
  • Physiology (AREA)
  • Public Health (AREA)
  • Veterinary Medicine (AREA)
  • Ophthalmology & Optometry (AREA)
  • Evolutionary Computation (AREA)
  • Fuzzy Systems (AREA)
  • Mathematical Physics (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Psychiatry (AREA)
  • Signal Processing (AREA)
  • Measuring Pulse, Heart Rate, Blood Pressure Or Blood Flow (AREA)

Abstract

The invention discloses a Korotkoff sound time phase classification and identification method and a Korotkoff sound time phase classification and identification system, aiming at the properties that different time phases (I, II, III, IV and V time phases) of Korotkoff sounds have different acoustics and time-frequency characteristics, preprocessing collected data, and automatically classifying and identifying Korotkoff sound signals collected from a stethoscope based on a deep neural network and a related processing algorithm, thereby being capable of distinguishing the different time phases of the Korotkoff sounds and outputting the identification results of the time phases. The method and the system provided by the invention can realize accurate and efficient identification of different time phases of the Korotkoff sounds.

Description

Korotkoff sound time phase classification and identification method and system
Technical Field
The invention relates to the technical field of blood pressure measurement, in particular to a Korotkoff's sound time phase classification and identification method and system.
Background
When the artificial auscultation method is used for measuring blood pressure, the cuff wound on the upper arm needs to be inflated and pressurized to block the blood circulation of brachial artery, and then the cuff is deflated at a constant speed to reduce pressure. As the pressure drops, the blood flow re-washes away the blood vessels, producing the same rhythmic tone as the beat of the cardiac cycle, called the korotkoff tone. The characteristics of korotkoff sounds change with the decrease of the cuff pressure, and the process is divided into five phases: I. the phase II, III, IV and V are also called the sound, murmur, clap, cover and silence, respectively.
For most adults, the cuff pressures corresponding to the first and last heard korotkoff sounds are recorded as systolic and diastolic pressures, respectively. There are several papers and patents that propose methods that can automatically recognize these two korotkoff sounds. However, for some special populations (e.g. children under 12 years old, pregnant women, elderly, severe anemia, hyperthyroidism, aortic insufficiency, etc.), reading of the diastolic pressure at the IV phase of korotkoff sounds is required. Therefore, the correct identification of the phase IV of korotkoff sounds is the key to accurately measure the blood pressure of this particular group of people. However, most of the existing automatic blood pressure measuring methods represented by the oscillometric method estimate the blood pressure value based on the pressure oscillation wave in the cuff, and cannot judge the korotkoff sound signal at all, let alone distinguish the time phase of the korotkoff sound signal. Although the blood pressure estimation empirical coefficient can also be calibrated by a statistical rule, the accuracy is still low. Therefore, the oscillometric method cannot perform accurate measurement for a particular population. Although the manual auscultation method can realize the measurement of the blood pressure of special people, the correlation between the measurement result and the experience and the operation accuracy of a measuring doctor is very large, the reliability of the measurement result is difficult to guarantee, and the manual auscultation method is not convenient for people to monitor the blood pressure at home. In addition, some automatic auscultation methods have been proposed in the prior art, which can identify and judge the appearance and disappearance of korotkoff sounds, but none of these methods and techniques involve identifying and judging the phase of korotkoff sounds. Therefore, the existing non-invasive blood pressure measuring method cannot meet the blood pressure measuring requirement of special people.
Disclosure of Invention
The present invention is directed to solve the above problems and to provide a method and system for accurately and effectively classifying and recognizing the time phase of korotkoff sounds.
The invention discloses a Korotkoff sound time phase classification and identification method, which comprises the following steps of:
s1, acquiring cuff pressure data P and stethoscope audio data K;
s2, extracting pressure pulse wave data P from the obtained cuff pressure dataODetecting the vertex of the pressure pulse wave data to obtain a vertex position sequence Pi；
S3, pre-emphasis processing is carried out on the stethoscope audio data K to obtain the stethoscope audio data K1；
S4, using vertex position sequence PiAs a reference point, for the stethoscope audio data K1Performing segmentation to form a stethoscope audio data frame sequence Ks；
S5, combining the stethoscope audio data frame sequence KsConverting into spectrogram, and forming stethoscope audio data spectrogram sequence data K according to sequenceis；
S6, the peak point position sequence K of each Korotkoff sound in the stethoscope audio data KpiThe corresponding pressure pulse wave vertex position sequence PiSubtracting, and dividing by the sampling rate fs of the audio data to obtain the peak-to-peak time delay sequence PKis；
S7, spectrogram sequence data KisHefeng-peak time delay sequence PKisAnd sending the data into a deep neural network for identification, and outputting a corresponding Korotkoff sound time phase classification identification result by the deep neural network for each pair of spectrogram sequence data and peak-peak time delay data.
Optionally, the acoustic transducer converts acoustic signals collected in the stethoscope into electrical signals, and the electrical signals are sequentially amplified and subjected to AD conversion and then converted into audio digital signals, so as to obtain stethoscope audio data K; the pressure signal in the cuff is converted into an electric signal through the pressure sensor, and the electric signal is converted into a pressure digital signal after being sequentially amplified and subjected to AD conversion, so that cuff pressure data P is obtained.
Optionally, band-pass filtering is performed on the cuff pressure data P, the band-pass frequency of the band-pass filter is 0.05Hz to 20Hz, and the filtered cuff pressure data P is obtained1Removing P by polynomial fitting1Obtaining the pressure pulse wave data PO。
Optionally, in step S3, the filter transfer function of the pre-emphasis process is h (z) ═ 1-az-1Wherein a is 0.928.
Optionally, based on enhanced spectral subtraction, for stethoscope audio data K1Noise suppression processing is carried out to obtain the stethoscope audio data K after noise reduction2。
Alternatively, in step S4, when the division is performed, P is usedisAs starting point of slicing data frame, PieAs an end point of slicing a data frame, Pis＝Pi–(Or*0.3)，Pie＝Pi+(Or0.5), wherein, OrIs the average pulse wave period of the pulse wave,
Figure GDA0003097741950000031
n is the number of vertex sequences.
Alternatively, in step S5, the sequence K of stethoscope audio data frames is transformed by short-time fourier transform and energy calculationsAnd converting into a spectrogram.
The invention also provides a korotkoff sound time phase classification and identification system, which comprises:
the cuff pressure data acquisition unit is used for acquiring cuff pressure data P;
the stethoscope audio data acquisition unit is used for acquiring stethoscope audio data K;
the main control and arithmetic processing unit executes the korotkoff sound time phase classification and identification method according to claim 1 based on the cuff pressure data P and the stethoscope audio data K.
Optionally, the stethoscope audio data acquisition unit includes a stethoscope, an acoustic transducer, a first signal amplifier module and a first AD converter, the acoustic transducer is used for converting acoustic signals acquired in the stethoscope into electrical signals, the electrical signals are amplified by the first signal amplifier module and then transmitted to the first AD converter, and digital signals converted by the first AD converter are input to the main control and operation processing unit.
Optionally, the cuff pressure data acquisition unit includes an inflatable cuff, a pressure sensor, a second signal amplifier module and a second AD converter, the pressure sensor is configured to convert a pressure signal in the inflatable cuff into an electrical signal, the electrical signal is amplified by the second amplifier module and then transmitted to the second AD converter, and a digital signal converted by the second AD converter is input to the main control and operation processing unit.
Compared with the prior art, the remarkable progress of the invention is at least reflected in that:
aiming at the property that different phases (I, II, III, IV and V phases) of the Korotkoff sounds have different acoustics and time-frequency characteristics, the Korotkoff sound signals collected from the stethoscope are automatically classified and identified based on a deep neural network and a related preprocessing algorithm, so that the different phases of the Korotkoff sounds can be distinguished, and the identification result of the phases is output (namely, the input identified Korotkoff sounds are I, II, III, IV or V phases).
Drawings
FIG. 1 is a schematic diagram of a Korotkoff's sound time-phase classification and identification method according to an embodiment of the present invention;
FIG. 2 is a schematic diagram of peak-to-peak delay in accordance with an embodiment of the present invention;
FIG. 3 is a schematic diagram of a deep neural network according to an embodiment of the present invention;
FIG. 4 is a schematic diagram illustrating a Korotkoff's sound time phase classification and identification system according to an embodiment of the present invention;
fig. 5 is a schematic diagram illustrating an execution flow of the main control and arithmetic processing unit according to an embodiment of the present invention.
Detailed Description
The present invention is described in detail below with reference to the accompanying drawings and specific embodiments, it should be noted that the embodiments of the present invention are not limited to the specific embodiments listed below, and those skilled in the art can make equivalent substitutions for the embodiments of the present invention without departing from the spirit of the present invention.
Referring to fig. 1, the korotkoff sound time phase classification and identification method of the present invention includes the following steps:
s1, acquiring cuff pressure data P and stethoscope audio data K;
s2, extracting pressure pulse wave data P from the obtained cuff pressure dataODetecting the vertex of the pressure pulse wave data to obtain a vertex position sequence Pi；
S3, pre-emphasis processing is carried out on the stethoscope audio data K to obtain the stethoscope audio data K1；
S4, using vertex position sequence PiAs a reference point, for the stethoscope audio data K1Performing segmentation to form a stethoscope audio data frame sequence Ks；
S5, combining the stethoscope audio data frame sequence KsConverting into spectrogram, and forming stethoscope audio data spectrogram sequence data K according to sequenceis；
S6, the peak point position sequence K of each Korotkoff sound in the stethoscope audio data KpiThe corresponding pressure pulse wave vertex position sequence PiSubtracting, and dividing by the sampling rate fs of the audio data to obtain the peak-to-peak time delay sequence PKis；
S7, spectrogram sequence data KisHefeng-peak time delay sequence PKisAnd sending the data into a deep neural network for identification, and outputting a corresponding Korotkoff sound time phase classification identification result by the deep neural network for each pair of spectrogram sequence data and peak-peak time delay data.
Optionally, in step S1, the acoustic transducer may convert the acoustic signal collected in the stethoscope into an electrical signal, and the electrical signal is sequentially amplified and subjected to AD conversion (analog-to-digital conversion) and then converted into an audio digital signal, so as to obtain stethoscope audio data K; the pressure sensor can also be used for converting the pressure signal in the cuff into an electric signal, and the electric signal is converted into a pressure digital signal after being sequentially amplified and subjected to AD conversion, so that cuff pressure data P can be obtained.
Optionally, in step S2, the method for extracting pressure pulse wave data from the acquired cuff pressure data specifically includes: firstly, band-pass filtering is carried out on the cuff pressure data P, the pass band frequency of a band-pass filter is preferably 0.05Hz to 20Hz, and the filtered cuff pressure data P is obtained1(ii) a Further, considering that the acquired digital pressure signal may contain a baseline interference signal (low-frequency noise) which may adversely affect the signal analysis, the cuff pressure data P may be removed by a polynomial fitting method1To obtain the pressure pulse wave data PO. Still further, the sequence of vertex positions P may be obtained by detecting each vertex in the pressure pulse wave datai. It should be noted that the method of removing the baseline by polynomial fitting listed above is only one preferred embodiment, and those skilled in the art can also remove the baseline by other methods.
It should be noted that in step S3, the stethoscope audio data is pre-emphasized, which is aimed at enhancing the high-frequency signals in the features of the korotkoff sounds in the stethoscope audio signal, so as to facilitate the identification of the phases of korotkoff sounds II and III. As a preferred example, the filter transfer function of the pre-emphasis process is h (z) ═ 1-az-1Based on a summary of a plurality of experiments, in the present embodiment, a preferred value of the pre-emphasis transfer function coefficient a is 0.928.
Furthermore, to avoid noise signal interference in the audio data of the stethoscope, the audio data K of the stethoscope can be processed1Noise suppression processing is performed. Preferably, the stethoscope audio data K can be subjected to spectral subtraction1Noise suppression processing is carried out to obtain the stethoscope audio data K after noise reduction2. It will be appreciated that the spectral subtraction is a subtraction of the noise signal from the frequency spectrum of the noisy signalMethod for processing the spectrum of a number. Those skilled in the art can implement the modified/enhanced spectral subtraction of the prior art for the stethoscope audio data K1Noise suppression processing is performed.
Alternatively, in step S4, the vertex position sequence P is usediAs a reference point, for the stethoscope audio data K1When making the slicing, with PisAs starting point of slicing data frame, PieAs an end point of slicing a data frame, Pis＝Pi–(Or*0.3)，Pie＝Pi+(Or0.5), wherein, OrThe calculation formula is the average pulse wave period:
Figure GDA0003097741950000061
n is the number of vertex sequences in the pressure pulse wave data.
Alternatively, in step S5, the sequence K of stethoscope audio data frames may be transformed by short-time fourier transform and energy calculationsAnd converting into a spectrogram.
It should be noted that, in step S6, the peak point position sequence K may be obtained by detecting the peak points of the kotkoff sounds in the stethoscope audio data KpiThe peak point position sequence K of each Korotkoff sound in the audio data K of the stethoscope is calculatedpiThe corresponding pressure pulse wave vertex position sequence PiSubtracting, and then subtracting the difference Pi-KpiDividing the sampling rate fs of the audio data of the stethoscope to obtain a peak-to-peak time delay sequence PKis. Referring to fig. 2, a schematic diagram of peak-to-peak time delay is shown. In this embodiment, the peak-to-peak time delay PK between the peak point of Korotkoff's sound and the peak of the pressure pulse wave is introducedisBased on applicants' discovery, PKisThe cuff pressure is gradually prolonged along with the reduction of the cuff pressure, the time delay duration of different time phases of the Korotkoff sounds is different, and the time delay duration has stronger correlation with the time phases of the Korotkoff sounds, so that the accurate identification of the Korotkoff sounds is facilitated.
Further, in step S7, the spectrogram sequence data KisHefeng-peak time delay sequence PKisSending the data into a deep neural network for identification,two sequence data correspond one to one, i.e. K1sThe corresponding peak-to-peak time delay is PK1s，K2sThe corresponding peak-to-peak time delay is PK2sIf the peak-to-peak time delay cannot be calculated, the value is not 0; and outputting a corresponding korotkoff sound time phase classification identification result by the deep neural network for each pair of spectrogram data sequence and peak-peak time delay data. Specifically, if phases I, II, III, IV and V are identified, the identification results 1, 2, 3, 4 and 0 are output respectively. Referring to fig. 3, a structural diagram of a deep neural network according to an embodiment is shown, where FC denotes a fully-connected layer, Concatenate denotes a connected layer, CNN denotes a convolutional neural network, and LSTM denotes a long-term memory cyclic network. It will be appreciated that the network needs to be trained with manually annotated phase-back stethoscope audio data before recognition.
In the scheme, a multi-input parameter deep neural network model is adopted, and spectrogram sequence data K is obtainedisHefeng-peak time delay sequence PKisAs the input of the network, the output of the Korotkoff sound and the output of the Korotkoff sound are combined through a connecting layer (Concatenate), the difference of different time phases of the Korotkoff sound on time-frequency-energy characteristics is utilized, the discovery that different time phases of the Korotkoff sound have different peak-peak time delays is also utilized, and the accuracy of identifying the Korotkoff sound time phases is improved. Further, the deep neural network structure is specifically designed and identified in a sequence rather than a single audio frame. The deep neural network combines a Convolutional Neural Network (CNN) and a long-time and short-time memory network (LSTM), the CNN is used for analyzing the characteristics of time, frequency, energy and the like of a single-frame Korotkoff sound signal, and the context relationship of the Korotkoff sound signal in an LSTM learning sequence and the relationship of the change of the peak-peak time delay before and after are used for improving the identification accuracy; in the deep neural network, the weight of the CNN network corresponding to each frame of the spectrogram sequence is shared. The classifier of the deep neural network designed by the invention can output 5 types of Korotkoff sound identification results (corresponding to the phases I, III, IV and V of Korotkoff sounds respectively).
Referring to fig. 4, the present invention further provides a korotkoff sound time phase classification and identification system, including:
the cuff pressure data acquisition unit is used for acquiring cuff pressure data P;
the stethoscope audio data acquisition unit is used for acquiring stethoscope audio data K;
the main control and arithmetic processing unit performs korotkoff sound time phase classification and identification processing based on the cuff pressure data P and the stethoscope audio data K, and a processing flow diagram is shown in fig. 5.
It should be noted that, in the processing flow, the cuff pressure data P collected by the cuff pressure data collecting unit and the stethoscope audio data K collected by the stethoscope audio data collecting unit may be used as input data of the main control and arithmetic processing unit. As another alternative embodiment, the cuff pressure data P and the stethoscope audio data K may be pre-stored in a memory, and the main control and arithmetic processing unit may directly call the data stored in the memory to perform korotkoff sound time phase classification and identification processing;
further, pressure pulse wave data P is extracted from the acquired cuff pressure dataOAnd performing vertex detection, i.e. detecting the vertexes of the pressure pulse wave data to obtain a vertex position sequence Pi；
Further, pre-emphasis processing is carried out on the stethoscope audio data K to obtain the stethoscope audio data K1The filter transfer function of the pre-emphasis process is H (Z) 1-az-1Wherein, the value of the pre-emphasis transfer function coefficient a is that a is 0.928;
further, a segmentation process is carried out, and a vertex position sequence P is used during the segmentation processiAs a reference point, for the stethoscope audio data K1Performing segmentation to form a stethoscope audio data frame sequence KsPreferably, with PisAs starting point of slicing data frame, PieAs an end point of slicing a data frame, Pis＝Pi–(Or*0.3)，Pie＝Pi+(Or0.5), wherein, OrTo average the pulse wave period, the formula is
Figure GDA0003097741950000081
n is pressure pulse waveThe number of vertex sequences in the data;
further, a stethoscope audio data frame sequence KsConverting into spectrogram, and forming stethoscope audio data spectrogram sequence data K according to sequenceis；
The position sequence K of the peak point of each Korotkoff sound in the audio data K of the stethoscope is determinedpiThe corresponding pressure pulse wave vertex position sequence PiSubtracting, and dividing the difference by sampling rate fs of audio data to obtain peak-to-peak time delay sequence PKis；
Still further, the spectrogram sequence data KisHefeng-peak time delay sequence PKisAnd sending the data into a deep neural network for identification, and outputting a corresponding Korotkoff sound time phase classification identification result by the deep neural network for each pair of spectrogram sequence data and peak-peak time delay data.
Optionally, the stethoscope audio data acquisition unit includes a stethoscope, an acoustic transducer, a first signal amplifier module and a first AD converter, the acoustic transducer is used for converting acoustic signals acquired in the stethoscope into electrical signals, the electrical signals are amplified by the first signal amplifier module and then transmitted to the first AD converter, and digital signals converted by the first AD converter are input to the main control and operation processing unit. It should be noted that the acoustic transducer may be embedded in a stethoscope, and the acoustic transducer may be a condenser microphone, a micro-electro-mechanical system (MEMS) microphone, or a piezoelectric film microphone, so as to convert an acoustic signal into an electrical signal.
Optionally, the cuff pressure data acquisition unit includes an inflatable cuff, a pressure sensor, a second signal amplifier module and a second AD converter, the pressure sensor is configured to convert a pressure signal in the inflatable cuff into an electrical signal, the electrical signal is amplified by the second amplifier module and then transmitted to the second AD converter, and a digital signal converted by the second AD converter is input to the main control and operation processing unit. As a further preferred, the inflatable cuff comprises a cuff body, and the cuff body is connected with the pressure sensor, the air pump and the air valve respectively through the conduits. Therefore, the pressure sensor can be used for collecting pressure signals in the cuff, and the operations of inflating the cuff, deflating the cuff and the like can be realized through the air pump and the air valve. Still further, the air pump and the air valve are respectively connected with the main control and operation processing unit through a driver, and the air pump is matched with the driver to inflate the cuff, generally to 200 mmHg. The air valve is matched with the driver to deflate the cuff, and the deflation speed is controlled to be 2-3 mmHg/s. Optionally, the system further comprises a display module connected with the main control and arithmetic processing unit for displaying the korotkoff sound classification and identification result. Therefore, the main control and operation processing unit has at least two functions, one of which is to drive the air valve and the air pump to inflate and deflate the cuff, acquire the data and control the power-on time sequence logic. And secondly, classifying and identifying the Korotkoff sound time phase according to the acquired stethoscope audio data and cuff pressure data, and outputting and displaying an identification result.
In order to better embody the implementation effect of the korotkoff sound time phase classification and identification method and system provided by the invention, the applicant performs relevant clinical verification, the manually labeled korotkoff sound time phase is used as a reference value, the deep neural network and the network without peak-to-peak time delay in the invention are compared, the data are collected from 40 volunteers, and each volunteer collects 18 pieces of data in total, so that 720 pieces of korotkoff sound data and corresponding cuff pressure data are totally collected. The verification was performed by 10-fold cross-validation, and 40 volunteers were divided into 10 groups, and the experimental results are shown in table 1.
TABLE 1
Figure GDA0003097741950000101
As can be seen from Table 1, the network of the present invention has an average recognition accuracy of about 2.2% higher than that of the unused peak-to-peak delay network.
Because the phase II, III of the korotkoff sounds have a significant high frequency component relative to the phase I, IV. In addition, the method performs pre-emphasis on the Korotkoff sound data when the Korotkoff sound data are processed. The high-frequency component of the Korotkoff sound can be emphasized by pre-emphasis, so that the high-frequency resolution of the Korotkoff sound is increased, and the subsequent deep neural network can be favorable for distinguishing the Korotkoff sound time phase. The applicant also performed experiments, and the experimental data and methods are the same as above, and the networks are the deep neural networks of the present invention, and the only difference is whether pre-emphasis is performed, and the experimental results are shown in table 2.
TABLE 2
Figure GDA0003097741950000111
As can be seen from table 2, the accuracy of classification and identification is improved by 0.4% after pre-emphasis is introduced.
It is to be understood that in the description of the embodiments of the present invention, the terms "first", "second", "third", "fourth" are used for descriptive purposes only and are not to be construed as indicating or implying relative importance or implying any number of indicated technical features. Thus, features defined as "first", "second", "third", "fourth" may explicitly or implicitly include one or more of the features. In the description of the present invention, "a plurality" means two or more unless otherwise specified.
In the description of the embodiments of the present invention, it should be noted that, unless otherwise explicitly specified or limited, the terms "mounted," "connected," and "assembled" are to be construed broadly and may, for example, be fixedly connected, detachably connected, or integrally connected; they may be connected directly or indirectly through intervening media, or they may be interconnected between two elements. The specific meanings of the above terms in the present invention can be understood in specific cases to those skilled in the art.
In the description of the embodiments of the invention, the particular features, structures, materials, or characteristics may be combined in any suitable manner in any one or more embodiments or examples.
In the description of the embodiments of the present invention, it should be understood that "-" and "-" indicate the same range of two numerical values, and the range includes the endpoints. For example, "A-B" means a range greater than or equal to A and less than or equal to B. "A to B" means a range of not less than A and not more than B.
In the description of the embodiments of the present invention, the term "and/or" herein is only one kind of association relationship describing an associated object, and means that there may be three kinds of relationships, for example, a and/or B, and may mean: a exists alone, A and B exist simultaneously, and B exists alone. In addition, the character "/" herein generally indicates that the former and latter related objects are in an "or" relationship.
Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that changes, modifications, substitutions and alterations can be made in these embodiments without departing from the principles and spirit of the invention, the scope of which is defined in the appended claims and their equivalents.

Claims (10)

1. The Korotkoff sound time phase classification and identification method is characterized by comprising the following steps of:
s1, acquiring cuff pressure data P and stethoscope audio data K;
s2, extracting pressure pulse wave data P from the obtained cuff pressure dataODetecting the vertex of the pressure pulse wave data to obtain a vertex position sequence Pi；
S3, pre-emphasis processing is carried out on the stethoscope audio data K to obtain the stethoscope audio data K1；
S4, using vertex position sequence PiAs a reference point, for the stethoscope audio data K1Performing segmentation to form a stethoscope audio data frame sequence Ks；
S5, combining the stethoscope audio data frame sequence KsConverting into spectrogram, and forming stethoscope audio data spectrogram sequence data K according to sequenceis；
S6, the peak point position sequence K of each Korotkoff sound in the stethoscope audio data KpiThe corresponding pressure pulse wave vertex position sequence PiSubtracting, and dividing the difference by the sampling rate fs of the audio data to obtain the peak-to-peak time delay sequence PKis；
S7, spectrogram sequence data KisHefeng-peak time delay sequence PKisAnd sending the data into a deep neural network for identification, and outputting a corresponding Korotkoff sound time phase classification identification result by the deep neural network for each pair of spectrogram sequence data and peak-peak time delay data.
2. The korotkoff sound time phase classification and identification method according to claim 1, wherein the acoustic transducer converts the acoustic signal collected in the stethoscope into an electrical signal, and the electrical signal is amplified and AD-converted into an audio digital signal to obtain stethoscope audio data K; the pressure signal in the cuff is converted into an electric signal through the pressure sensor, and the electric signal is converted into a pressure digital signal after being sequentially amplified and subjected to AD conversion, so that cuff pressure data P is obtained.
3. The korotkoff sound time-phase classification and identification method according to claim 1, wherein the cuff pressure data P is band-pass filtered to obtain filtered cuff pressure data P1Removing P by polynomial fitting1Obtaining the pressure pulse wave data PO。
4. The Korotkoff's sound phase classification and identification method according to claim 3, wherein, when performing the band-pass filtering, the frequency of the filter pass band is selected to be 0.05Hz to 20 Hz.
5. The korotkoff sound phase classification recognition method according to claim 3, wherein in step S3, the filter transfer function of the pre-emphasis process is h (z) -1-az-1Wherein a is 0.928.
6. The Korotkoff sound phase classification of claim 5Identification method, characterized in that, based on enhanced spectral subtraction, the stethoscope audio data K is subjected to1Noise suppression processing is carried out to obtain the stethoscope audio data K after noise reduction2。
7. The korotkoff sound phase classification/recognition method according to claim 1, wherein the segmentation is performed in step S4 using PisAs starting point of slicing data frame, PieAs an end point of slicing a data frame, Pis＝Pi–(Or*0.3)，Pie＝Pi+(Or0.5), wherein, OrIs the average pulse wave period of the pulse wave,
Figure FDA0003097741940000021
n is the number of vertex sequences.
8. A korotkoff sound phase classification recognition system, comprising:
the cuff pressure data acquisition unit is used for acquiring cuff pressure data P;
the stethoscope audio data acquisition unit is used for acquiring stethoscope audio data K;
the main control and arithmetic processing unit executes the korotkoff sound time phase classification and identification method according to claim 1 based on the cuff pressure data P and the stethoscope audio data K.
9. The korotkoff sound phase classification and identification system according to claim 8, wherein the stethoscope audio data collecting unit includes a stethoscope, an acoustic transducer, a first signal amplifier module and a first AD converter, the acoustic transducer is configured to convert an acoustic signal collected in the stethoscope into an electrical signal, the electrical signal is amplified by the first signal amplifier module and transmitted to the first AD converter, and a digital signal converted by the first AD converter is input to the main control and operation processing unit.
10. The korotkoff sound time-phase classification and identification system according to claim 8 or 9, wherein the cuff pressure data acquisition unit comprises an inflatable cuff, a pressure sensor, a second signal amplifier module and a second AD converter, the pressure sensor is configured to convert a pressure signal in the inflatable cuff into an electrical signal, the electrical signal is amplified by the second amplifier module and transmitted to the second AD converter, and a digital signal converted by the second AD converter is input to the main control and operation processing unit.
CN202010350121.4A 2020-04-28 2020-04-28 Korotkoff sound time phase classification and identification method and system Active CN111657900B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202010350121.4A CN111657900B (en) 2020-04-28 2020-04-28 Korotkoff sound time phase classification and identification method and system

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202010350121.4A CN111657900B (en) 2020-04-28 2020-04-28 Korotkoff sound time phase classification and identification method and system

Publications (2)

Publication Number Publication Date
CN111657900A CN111657900A (en) 2020-09-15
CN111657900B true CN111657900B (en) 2021-08-17

Family

ID=72382925

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202010350121.4A Active CN111657900B (en) 2020-04-28 2020-04-28 Korotkoff sound time phase classification and identification method and system

Country Status (1)

Country Link
CN (1) CN111657900B (en)

Families Citing this family (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN112603262A (en) * 2020-12-02 2021-04-06 珠海中科先进技术研究院有限公司 Human body state identification method, system and medium

Family Cites Families (12)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CA1163327A (en) * 1979-11-14 1984-03-06 Ethicon, Inc. Automated blood pressure measurement during physical exercise
DE3424536A1 (en) * 1984-07-04 1986-01-09 Robert Bosch Gmbh, 7000 Stuttgart Method and device for the non-invasive blood pressure measurement
JP2008049011A (en) * 2006-08-28 2008-03-06 Parama Tec:Kk Korotkoff sound measuring apparatus
US9974449B2 (en) * 2007-07-11 2018-05-22 Meng-Sun YU Method and apparatus for arterial blood pressure measurement and individualized rectifying technology
CN101664307A (en) * 2009-09-24 2010-03-10 北京航空航天大学 Acquisition and processing method for measuring kororkoff sound information of blood pressure with auscultatory method and device
JP2011206322A (en) * 2010-03-30 2011-10-20 Omron Healthcare Co Ltd Blood pressure measuring device, and program for blood pressure measurement
CN102085094A (en) * 2011-01-24 2011-06-08 罗万前 Method for determining sound intensity peak slopes of originating point and vanishing point of Korotkoff sounds
CN106037698A (en) * 2016-05-12 2016-10-26 京东方科技集团股份有限公司 Blood pressure gauge, and Korotkoff sound recognition method and device
CN105942980B (en) * 2016-05-30 2019-11-26 珠海脉动时代健康科技有限公司 A kind of method and system of stethoscopy sphygmomanometer feature extraction and classification
CN106725401B (en) * 2017-01-12 2020-01-17 成都信息工程大学 Stethoscope audio data processing method based on deep learning
CN108309274A (en) * 2018-03-09 2018-07-24 上海由泰医疗器械科技有限公司 A kind of method that automatic detection obtains blood pressure Korotkoff's Sound
CN109044312A (en) * 2018-08-30 2018-12-21 东南大学 A kind of electronic sphygmomanometer and its blood pressure measuring method based on Korotkoff's Sound

Also Published As

Publication number Publication date
CN111657900A (en) 2020-09-15

Similar Documents

Publication Publication Date Title
CN103313662B (en) System, the stethoscope of the risk of instruction coronary artery disease
JP4409878B2 (en) Method and apparatus for determining blood pressure using a pressure pulse duty cycle
JP2012513858A (en) Method and system for processing heart sound signals
EP1389957A1 (en) Heart diagnosis system
WO2005099562A1 (en) Non-invasive measurement of second heart sound components
US6520918B1 (en) Method and device for measuring systolic and diastolic blood pressure and heart rate in an environment with extreme levels of noise and vibrations
WO2012043013A1 (en) Device for measuring blood pressure information and method for measuring blood pressure information
Shukla et al. Noninvasive cuffless blood pressure measurement by vascular transit time
WO2023142336A1 (en) Blood pressure measurement method, device and apparatus and readable storage medium
CN111657900B (en) Korotkoff sound time phase classification and identification method and system
KR102130340B1 (en) blood pressure monitor with a microphone for detecting a sound outside the cuff
CN109497981A (en) A kind of Korotkoff's Sound blood pressure detector and its detection method with pulse wave detection
CN106725401B (en) Stethoscope audio data processing method based on deep learning
JP2004000422A (en) Sphygmomanometer having waveform analyzing function
US5680868A (en) Method and apparatus for detecting blood pressure by blood pressure sounds in the presence of significant noise
JPS59501895A (en) electronic blood pressure monitor
US20200260968A1 (en) Blood pressure measuring apparatus capable of estimating arteriosclerosis
CN114587314A (en) Internet blood pressure measuring device and control method thereof
Park et al. Novel method of automatic auscultation for blood pressure measurement using pulses in cuff pressure and korotkoff sound
Li et al. A study for the development of K-sound based automatic blood pressure device using PVDF film
JPH0739529A (en) Electronic sphygmomanometer
Yu et al. A novel method for Korotkoff vibration blood pressure measurement based on oscillometric
Naufal et al. Blood Pressure Measuring Device Based on Korotkoff Sound's Tapping Period and Frequency Detection
JP2001309894A (en) Equipment for measuring peripheral venous pressure and its method
Naqvi et al. Noninvasive method for determining blood pressure and contours of arterial and volume pulses

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
GR01 Patent grant
GR01 Patent grant