CN116269413A - Continuous electrocardiographic waveform reconstruction system and method using smart wristband motion sensor - Google Patents
Continuous electrocardiographic waveform reconstruction system and method using smart wristband motion sensor Download PDFInfo
- Publication number
- CN116269413A CN116269413A CN202211090977.8A CN202211090977A CN116269413A CN 116269413 A CN116269413 A CN 116269413A CN 202211090977 A CN202211090977 A CN 202211090977A CN 116269413 A CN116269413 A CN 116269413A
- Authority
- CN
- China
- Prior art keywords
- body surface
- electrocardiographic waveform
- local
- heartbeat
- surface vibration
- 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.)
- Pending
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
- A61B5/24—Detecting, measuring or recording bioelectric or biomagnetic signals of the body or parts thereof
- A61B5/316—Modalities, i.e. specific diagnostic methods
- A61B5/318—Heart-related electrical modalities, e.g. electrocardiography [ECG]
-
- 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
- A61B5/0205—Simultaneously evaluating both cardiovascular conditions and different types of body conditions, e.g. heart and respiratory condition
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
- A61B5/00—Measuring for diagnostic purposes; Identification of persons
- A61B5/103—Detecting, measuring or recording devices for testing the shape, pattern, colour, size or movement of the body or parts thereof, for diagnostic purposes
- A61B5/11—Measuring movement of the entire body or parts thereof, e.g. head or hand tremor, mobility of a limb
- A61B5/1101—Detecting tremor
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
- A61B5/00—Measuring for diagnostic purposes; Identification of persons
- A61B5/103—Detecting, measuring or recording devices for testing the shape, pattern, colour, size or movement of the body or parts thereof, for diagnostic purposes
- A61B5/11—Measuring movement of the entire body or parts thereof, e.g. head or hand tremor, mobility of a limb
- A61B5/1126—Measuring movement of the entire body or parts thereof, e.g. head or hand tremor, mobility of a limb using a particular sensing technique
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
- A61B5/00—Measuring for diagnostic purposes; Identification of persons
- A61B5/24—Detecting, measuring or recording bioelectric or biomagnetic signals of the body or parts thereof
- A61B5/25—Bioelectric electrodes therefor
- A61B5/251—Means for maintaining electrode contact with the body
- A61B5/256—Wearable electrodes, e.g. having straps or bands
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
- A61B5/00—Measuring for diagnostic purposes; Identification of persons
- A61B5/24—Detecting, measuring or recording bioelectric or biomagnetic signals of the body or parts thereof
- A61B5/25—Bioelectric electrodes therefor
- A61B5/279—Bioelectric electrodes therefor specially adapted for particular uses
- A61B5/28—Bioelectric electrodes therefor specially adapted for particular uses for electrocardiography [ECG]
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
- A61B5/00—Measuring for diagnostic purposes; Identification of persons
- A61B5/68—Arrangements of detecting, measuring or recording means, e.g. sensors, in relation to patient
- A61B5/6801—Arrangements of detecting, measuring or recording means, e.g. sensors, in relation to patient specially adapted to be attached to or worn on the body surface
- A61B5/6802—Sensor mounted on worn items
- A61B5/681—Wristwatch-type devices
-
- 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
- A61B5/7203—Signal processing specially adapted for physiological signals or for diagnostic purposes for noise prevention, reduction or removal
-
- 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
- A61B5/7253—Details of waveform analysis characterised by using transforms
- A61B5/726—Details of waveform analysis characterised by using transforms using Wavelet transforms
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02A—TECHNOLOGIES FOR ADAPTATION TO CLIMATE CHANGE
- Y02A90/00—Technologies having an indirect contribution to adaptation to climate change
- Y02A90/10—Information and communication technologies [ICT] supporting adaptation to climate change, e.g. for weather forecasting or climate simulation
Landscapes
- Health & Medical Sciences (AREA)
- Life Sciences & Earth Sciences (AREA)
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Biophysics (AREA)
- Biomedical Technology (AREA)
- Molecular Biology (AREA)
- General Health & Medical Sciences (AREA)
- Pathology (AREA)
- Veterinary Medicine (AREA)
- Heart & Thoracic Surgery (AREA)
- Medical Informatics (AREA)
- Surgery (AREA)
- Animal Behavior & Ethology (AREA)
- Public Health (AREA)
- Physiology (AREA)
- Cardiology (AREA)
- Signal Processing (AREA)
- Artificial Intelligence (AREA)
- Dentistry (AREA)
- Oral & Maxillofacial Surgery (AREA)
- Theoretical Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Psychiatry (AREA)
- Computational Linguistics (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Pulmonology (AREA)
- Measurement And Recording Of Electrical Phenomena And Electrical Characteristics Of The Living Body (AREA)
Abstract
The invention provides a continuous electrocardiographic waveform reconstruction system and method by using an intelligent wristband motion sensor, and belongs to the technical field of mobile computing application. The system includes a smart wristband device with a built-in motion sensor and a processing unit. The motion sensor of the smart wristband device is used for acquiring motion sensing signals of the wrist of the user and sending the signals to the processing unit. The processing unit processes the motion sensor and is used for acquiring a human motion sensing signal, extracting a body surface vibration signal related to the heartbeat, dividing the body surface vibration signal into single-heartbeat period fragments, and finally reconstructing an electrocardiographic waveform corresponding to the heartbeat period. The invention can accurately reconstruct the electrocardio waveform based on the body surface vibration signals related to the heartbeat. The system is convenient to use, the electrocardiographic waveform can be continuously reconstructed, the measurement process is transparent to the user, and the user does not need to participate.
Description
Technical Field
The invention relates to a continuous electrocardiographic waveform reconstruction system, in particular to a continuous electrocardiographic waveform reconstruction system and method by using an intelligent wristband motion sensor, and belongs to the technical field of mobile computing application.
Background
The continuous electrocardiographic waveform measurement system has great potential in the aspects of fatigue driving early warning, biological feature recognition, emotion recognition and the like. In a conventional electrocardiographic waveform measurement system, the potential change is measured on the body surface of a human body by using electrodes and electrolyte gel. However, since such a system depends on special equipment and is complicated to operate, it is difficult to widely apply in daily life. To enable convenient home electrocardiographic waveform measurements, some systems integrate electrodes into home appliances and furniture (e.g., toilets and chairs). However, these methods can only be used when the user touches the measurement device, and continuous electrocardiographic waveforms cannot be acquired for a long time. Currently, some systems attempt to stitch fabric electrodes into textiles (e.g., pajamas and waistbands) for electrocardiographic waveform measurement, but the expensive cost has prevented widespread deployment of such systems. In recent years, some new commercial smartwatches have also deployed electrodes for measuring electrocardiographic waveforms, but such devices, when in use, rely on the user to perform certain actions (e.g., covering the electrodes with a finger), and still cannot continuously acquire the electrocardiographic waveforms of the wearer.
In addition, some systems reconstruct an electrocardiographic waveform using signals associated with the heart cycle. Unlike direct measurement of electrocardiogram with body surface potential change, the system needs no electrode contact by the user, and the corresponding electrocardiogram waveform can be accurately reconstructed by utilizing the correlation between the acquired sensing signal and the heartbeat period. For example, chest minute vibrations caused by heart beat are measured using a wireless signal, and an electrocardiographic waveform is predicted. However, during acquisition the user is required to remain stationary, avoiding body movements interfering with the perception of chest vibrations. Still other systems use vibration sensors deployed within the mattress to reconstruct the electrocardiographic waveform of the user while sleeping, however, they can only work when the user is lying on his/her side or after leaving the mattress, and measurements are forced to be interrupted. In addition, there are systems that record pulse signals at the finger and predict electrocardiographic waveforms using a oximeter worn at the fingertip, but wearing a measurement device at the finger for a long time may affect the hand touch function and cause discomfort.
In view of the above, the existing systems suffer from various drawbacks and deficiencies.
Disclosure of Invention
The invention aims to overcome the technical defects that the existing electrocardiographic waveform measurement and electrocardiographic waveform reconstruction system is high in cost and inconvenient to use, particularly the electrocardiographic waveform cannot be continuously acquired, special behaviors are required to be executed by a user, and the like, and creatively provides a continuous electrocardiographic waveform reconstruction system and method by using an intelligent wristband motion sensor.
A continuous electrocardiographic waveform reconstruction system utilizing a smart wristband motion sensor includes a smart wristband device with a built-in motion sensor and a processing unit.
The motion sensor of the intelligent wrist strap device is used for collecting motion sensing signals of the wrist of the user and sending the signals to the processing unit.
The processing unit processes the motion sensor and is used for acquiring a human motion sensing signal, extracting a body surface vibration signal related to the heartbeat, dividing the body surface vibration signal into single-heartbeat period fragments, and finally reconstructing an electrocardiographic waveform corresponding to the heartbeat period.
The implementation method of the system comprises the following steps:
step 1: a motion sensor (e.g., a gyroscope) of the smart wristband device is used to collect motion sensing signals of the target user's wrist.
Specifically, the user wears a smart wristband device with a motion sensor built in, which continuously acquires motion sensing signals of the wrist.
Step 2: the processing unit extracts body surface vibration related to heartbeat in the motion sensing signal. The purpose is to extract the body surface vibration caused by the constant change of the gravity center of the blood flow in the heartbeat period from the disordered motion sensing signals.
Step 2.1: and (3) processing the wrist motion sensing signals acquired in the step (1) by using a band-pass filter to remove irrelevant noise.
Step 2.2: and (2) further removing noise in the disordered motion sensing signal based on the stable wavelet transformation according to the filtered motion sensing signal extracted in the step (2.1), and extracting a body surface vibration signal related to heartbeat.
Step 3: the processing unit cuts off the body surface vibration signals related to the heart beat. The body surface vibration signal is split into single-heart cycle fragments according to the special waveform of the body surface vibration signal related to the heart beat.
Step 3.1: detecting local maximum value and local minimum value points in the body surface vibration signals related to the heartbeat, and constructing a local triangle. From the local triangle feature, a peak associated with ventricular contractions is identified.
Step 3.2: and (3) segmenting the body surface vibration signals related to the heart beat into segments with single heart beat period according to the peak points related to the ventricular contractions, which are identified in the step 3.1.
Step four: and reconstructing a corresponding electrocardio waveform by using the body surface vibration signal segments related to the heartbeat. The purpose is to establish the corresponding relation between the body surface vibration signal and the electrocardiographic waveform. And reconstructing a corresponding electrocardiographic waveform by using the body surface vibration signals.
Step 4.1: and (3) establishing an encoder-decoder network model capable of reconstructing an electrocardio waveform according to the segments of the body surface vibration signals extracted in the step 3.2.
Step 4.2: and (3) training the encoder-decoder network model established in the step 4.2 by utilizing the generation countermeasure network, and accurately reconstructing the corresponding electrocardio waveform according to the body surface vibration signals related to the heart beat.
Advantageous effects
Compared with the prior art, the invention has the following advantages:
1. according to the invention, wrist motion sensing information of a target user is acquired by using the intelligent wrist strap device, and an electrocardiographic waveform is accurately reconstructed based on body surface vibration signals related to heartbeat.
2. The invention has convenient use, can reconstruct the electrocardio waveform continuously, and the measuring process is transparent to the user without the participation of the user.
Drawings
FIG. 1 is a schematic diagram of the present invention.
Fig. 2 is a schematic diagram of three coordinate axis directions of a gyroscope of a smart wristband device and sensing signals and electrocardiographic waveforms acquired in the three coordinate axes according to an embodiment of the present invention.
Fig. 3 is a schematic diagram of motion signals of body surface vibration related to heartbeat extracted by using stationary wavelet transform according to an embodiment of the present invention.
FIG. 4 is a schematic diagram of constructing a partial triangle and partial triangle features according to an embodiment of the present invention.
FIG. 5 is a schematic diagram of an encoder-decoder network model developed in accordance with an embodiment of the present invention.
Fig. 6 is a schematic diagram of a generating countermeasure network constructed according to an embodiment of the present invention.
FIG. 7 is a prototype diagram of an embodiment of the present invention.
Fig. 8 is a waveform reconstruction error for reconstructing an electrocardiographic waveform according to an embodiment of the present invention.
Fig. 9 is a correlation coefficient of a reconstructed electrocardiographic waveform according to an embodiment of the present invention.
FIG. 10 illustrates waveform reconstruction errors and correlation coefficients at different sampling frequencies according to an embodiment of the present invention.
FIG. 11 is a graph showing waveform reconstruction errors and correlation coefficients for an embodiment of the present invention with wristband devices worn at different wrist positions.
Detailed Description
The principles and features of the present invention are described in further detail below with reference to the examples and the attached drawings. The examples are given solely for the purpose of illustration and are not intended to limit the scope of the invention.
A continuous electrocardiographic waveform reconstruction system utilizing a smart wristband motion sensor includes a smart wristband device with a built-in motion sensor and a processing unit.
The motion sensor of the intelligent wrist strap device is used for collecting motion sensing signals of the wrist of the user and sending the signals to the processing unit.
The processing unit processes the motion sensor and is used for acquiring a human motion sensing signal, extracting a body surface vibration signal related to the heartbeat, dividing the body surface vibration signal into single-heartbeat period fragments, and finally reconstructing an electrocardiographic waveform corresponding to the heartbeat period.
Fig. 1 shows a schematic diagram of an embodiment of the invention. The center of gravity of blood changes periodically as blood flows periodically in the blood vessel with the heartbeat. The user's body is affected by the change in the center of gravity of the blood, generating interactive opposite impulses, in particular rotational energy, causing minor vibrations of the body. Thus, body surface vibrations associated with heart beat are acquired with motion sensors, in particular gyroscopes. Continuous electrocardiographic measurements are achieved by analyzing the body surface vibration signals associated with the heart beat to estimate corresponding electrocardiographic signals.
The implementation method of the system comprises the following steps:
step 1: a gyroscope using a smart wristband device collects motion sensing signals of a target user's wrist.
Step 1.1: the user wears a smart wristband device with a motion sensor built in. The motion sensor continuously acquires motion sensing signals of the wrist.
The method comprises the following steps:
motion sensors include three-axis accelerometers, three-axis gyroscopes, and three-axis magnetometers, which are sensitive to minute movements. Analyzing the triaxial gyroscope data, three coordinate axes of a triaxial gyroscope X, Y, Z built in the smart wristband device are shown in fig. 2 (1), and sensing signals of the triaxial gyroscope X, Y, Z axes are respectively defined as G X 、G Y And G Z . As shown in FIG. 2 (2), an electrocardiogram and a sensing signal G X 、G Y And G Z Has obvious correlation, wherein, the key inflection point of the electrocardiogram is in contact with G X The key inflection points have a corresponding relationship, so that an X-axis sensing signal G of a triaxial gyroscope of the smart wristband device is used X As a motion sensing signal for the wrist.
Step 2: the processing unit extracts body surface vibration related to heartbeat in the motion sensing signal.
Step 2.1: and (2) processing the wrist motion sensing signal acquired in the step (1.1) by using a band-pass filter, and primarily removing irrelevant noise.
Because body surface vibration related to heart beat, which is collected by the intelligent wrist strap device, is inevitably affected by body movement of a user, a band-pass filter is used for removing signals of other frequency bands, and preliminary denoising is carried out on the movement sensing signals.
Since body surface vibrations associated with the heart beat of the human body are mainly distributed between 5Hz and 30Hz, the present invention preferably retains the portion of the motion-sensing signal having a frequency in the range of 5Hz to 30 Hz. However, other settings belonging to [0Hz,50Hz ] are within the scope of the present invention.
Step 2.2: and (2) further removing noise in the disordered motion sensing signal based on the stable wavelet transformation according to the filtered motion sensing signal extracted in the step (2.1), and extracting a body surface vibration signal related to heartbeat.
The method comprises the following steps:
first, a stationary wavelet transform is applied to the motion-sensing signal extracted in step 2.1. Let the mother wave be db4, decompose it into J approximate components a containing low-frequency information according to the frequency distribution range 1 ,a 2 ,…,a J And J detail components d containing high frequency information 1 ,d 2 ,…,d J J is the order set during the stationary wavelet transform. Preferably, the invention chooses a J value of 6, but others are of the order [2,10 ]]Is also within the scope of the present invention.
As shown in fig. 3 (1), the body surface vibration significantly associated with the heartbeat cannot be observed due to the influence of the body movement noise in the movement sensing signal. As shown in fig. 3 (2), after the stationary wavelet transform is applied, periodic vibration can be observed in each detail component of the motion sensing signal.
Next, each detail component d is calculated j Is a short-term energy e of (2) j And detecting short-time energy e j All local maximum amplitudes P j (k) K=1, 2, …, K, marking the time at which the local maximum amplitude occurs. Defining a threshold
And->
Is the detail component d j Is a short-term energy e of (2) j All local maximum amplitude P j (k) Mean and variance of (c). This is because most body movement noise-induced movement sensing signals have a greater amplitude than body surface vibration signals associated with heart beats. Comparing each P separately j (k) Amplitude of (2) is>
If->
Then classify P j (k) Is the body surface vibration associated with the heartbeat.
In addition, there is a small portion of body movement noise that causes movement sensing signals that have similar magnitudes as body surface vibration signals associated with heart beats. According to the characteristic that the motion sensing signal caused by the body motion noise does not have periodicity and the body surface vibration signal related to heart beat has periodicity, other unclassified P is further processed by utilizing the time interval of occurrence of the adjacent local maximum amplitude j (k)。
Defining two adjacent local maximum amplitudes P j (k-1) and P j (k) The time interval of occurrence is I j (k) A. The invention relates to a method for producing a fibre-reinforced plastic composite Setting a threshold value
E is j All adjacent local maxima occurrence time interval I j (k) Average value of (2). If->
And is also provided with
Then classify P j (k) Is the body surface vibration associated with the heartbeat.
Finally, using local maximum amplitude P of body surface vibrations classified as heart beat-related j (k) Separation d j Body surface vibrations and noise associated with the heart beat.
Specifically, the maximum amplitude P of each of the body surface vibrations associated with the heart beat is checked j (k) Searching forward and backward for local minimum occurrence times, determining start and end times of body surface vibrations associated with heart beat, and retaining d j Corresponding data in (a), determining other data as noise, and d j The corresponding data in (c) is replaced with 0. And performing inverse stationary wavelet transformation to obtain a body surface vibration signal related to the heartbeat.
As shown in fig. 3 (3), after inverse stationary wavelet transform is applied, the influence of body movement noise is eliminated in the movement sensing signal, and a body surface vibration signal related to the heartbeat is extracted.
Step 3: the processing unit cuts off the body surface vibration signals related to the heart beat.
Step 3.1: local maximum value and local minimum value points in the body surface vibration signals related to the heart beat are detected, a local triangle is constructed, and peak points related to ventricular contraction are identified according to the characteristics of the local triangle.
Since the body surface vibration signal associated with the heart beat has dynamic characteristics (e.g., the intervals between successive heart beats differ, the waveforms in each individual heart beat cycle differ), there is a lack of clear features for determining the boundaries of the individual heart beat cycles. By observation, the body surface vibration signal associated with the heart beat always shows a significant spike point when the heart chamber contracts. Body surface vibration signals associated with heart beat are segmented by picking peak points associated with ventricular contractions.
Specifically, all local maxima and local minima points are searched in the body surface vibration signals related to the heartbeat, each local maxima point and two adjacent left and right local minima points are utilized to construct a local triangle, and features are extracted from the local triangle. As shown in fig. 4, a local maximum point in the local triangle is defined as a vertex, an adjacent left local minimum point is a left bottom point, an adjacent right local minimum point is a right bottom point, a sum a+b of distances from the vertex to the left and right bottom stores of the local triangle is selected, a difference d between the vertical coordinates of the vertex and the left bottom point, a difference e between the vertical coordinates of the vertex and the right bottom point, and an angle f of an angle where the vertex in the local triangle is located are characterized.
In addition, the difference c between the left and right base points in the abscissa can be selected, the angle g of the left base point in the partial triangle and the angle h of the right base point in the partial triangle can be used as supplementary features to assist in identifying the peak point.
A random forest classifier is trained to identify whether the vertices in the local triangle (i.e., local maxima points) are peak points associated with ventricular contractions. During recognition, the selected characteristics are input into a random forest classifier, the probability that the vertex belongs to a peak point is predicted, and if the probability is larger than a set value (such as 0.5), the vertex is the peak point related to ventricular contraction.
Step 3.2: and (3) segmenting the body surface vibration signals related to the heart beat into segments corresponding to a single heart beat period according to the peak points related to the ventricular contractions, which are identified in the step 3.1.
Further, in order to obtain a heart beat-related body surface vibration signal containing a single heart beat cycle, 250 milliseconds before a peak point may be determined as the start point of one heart beat cycle, and 250 milliseconds before the next peak point may be determined as the end point of this heart beat cycle. However, other settings belonging to [100 ms, 400 ms ] are within the scope of the invention.
Step 4: and reconstructing a corresponding electrocardio waveform by using the body surface vibration signal segments related to the heartbeat.
Step 4.1: and (3) establishing an encoder-decoder network model capable of reconstructing an electrocardio waveform according to the segments of the body surface vibration signals extracted in the step 3.2.
Specifically, an encoder-decoder network model is developed based on a long-short-term memory neural network (LSTM), and as shown in fig. 5, a segment of the body surface vibration signal is input into the encoder-decoder network model. Due to the inconsistent length of the input segments, about 0.6 seconds to about 1.1 seconds. The signal length was stretched to 1.2 seconds (corresponding to 120 sample data at a sampling rate of 100 Hz) using a linear interpolation method. The encoder uses a bi-directional long and short time memory neural network (BLSTM) to extract hidden information related to the heartbeat and integrate it into a contraction feature consisting of N samples (e.g., 30). N samples in the extracted contraction feature of the encoder are respectively assigned weights by using an attention mechanism (applying a softmax function). The decoder is composed of two unidirectional long and short time memory neural network layers, and reconstructs electrocardio waveforms according to contraction characteristics and weights corresponding to the contraction characteristics. Finally, the reconstructed electrocardiographic waveforms are 1.2 seconds in length, and the length is adjusted to be the original segment length by applying inverse interpolation.
Step 4.2: and (3) training the encoder-decoder network model established in the step 4.2 by utilizing the generation countermeasure network, and accurately reconstructing a corresponding electrocardio waveform according to the body surface vibration signal related to the heartbeat.
Specifically, in order to realize accurate electrocardio waveform reconstruction at any time and any place, an encoder-decoder network model training method is established. To this end, a deep learning model based on generating an countermeasure network is built to assist in the training of the encoder-decoder network model in an off-line manner.
As shown in fig. 6, the established generation countermeasure network model is composed of a generator of the reconstructed electrocardiographic waveform and a discriminator that distinguishes the reconstructed electrocardiographic waveform from the true electrocardiographic waveform. Wherein, the encoder-decoder network model established in the step 4.1 is used as a generator to learn the complex mapping from the body surface vibration signal related to the heartbeat to the electrocardio waveform. The discriminator takes the reconstructed electrocardio waveform and the corresponding real electrocardio waveform output by the generator as input respectively, and stretches the input data length to be 1.2 seconds by utilizing linear interpolation. Then, using two bidirectional long and short-term memory neural network layers, a full connection layer and a decision layer (using softmax function) to identify the currently input signal as a true electrocardiographic waveform or a reconstructed electrocardiographic waveform. The parameters of the generator and the discriminator are trained in turn, so that the difference between the generated data and the real data is reduced, and the generator model accurately reconstructs the electrocardio waveform.
The reconstructed electrocardiographic waveform is denoted as e= { E 1 ,E 2 ,…,E i ,…,E L The true electrocardiographic waveform corresponding to E is denoted as a= { a } 1 ,A 2 ,…,A i ,…,A L }，E i For the amplitude of the electrocardiographic waveform signal, A i The amplitude of the real electrocardiosignal is L, and the length of the two waveforms is L. Waveform reconstruction error L for reconstructing electrocardiographic waveform by generator of Loss function Loss e And a discriminator for discriminating a discrimination error L of a real electrocardiographic waveform from a reconstructed electrocardiographic waveform a Common composition, loss=l e +L a . Setting:
L a ＝log[1-P EA ] (2)
wherein P is EA The reconstructed electrocardiographic waveform is incorrectly identified to the discriminator as a proportion of the actual electrocardiographic waveform. In the training process, the parameters in the countermeasure network are generated continuously and iteratively until the Loss converges.
Test verification
To verify the benefits of the present invention, the present invention was developed into a wristband prototype system for testing, as shown in fig. 7. The prototype consists of an integrated motion sensor (which can acquire three-axis gyroscope sensor signals) and an adjustable wrist strap.
A total of 20 healthy volunteers (10 men and 10 women, ages 20-33) were enrolled in data collection. During the data acquisition process, each volunteer was wearing a wristband prototype device and a medical electrocardiograph measuring device to record about 30 minutes of data for analysis and training, with no body movement and with body movement (sloshing forearm, sloshing forearm and walking), respectively.
The waveform reconstruction error and correlation coefficient are used for system performance evaluation. Wherein the waveform reconstruction error is defined as: and (3) an average value of the ratio of the absolute value of the amplitude difference between the reconstructed electrocardiographic waveform and the corresponding real electrocardiographic waveform to the amplitude of the real electrocardiographic waveform, namely the formula (1). The waveform reconstruction error approaching 0 indicates that the system can accurately reconstruct the electrocardiographic waveform. The correlation coefficient is defined as:
wherein the method comprises the steps of
Average value of the amplitude of the reconstructed electrocardiographic waveform E, < >>
For the average value of the amplitude value of the real electrocardio waveform A corresponding to E, the correlation coefficient is close to 1, which indicates that the system can accurately reconstruct the electrocardio waveform.
The overall performance of the invention was first tested. The data training of 20 volunteers is used for generating an countermeasure network, so that the generator can accurately reconstruct body surface vibration signals related to heartbeat into corresponding electrocardio waveforms. Fig. 8 shows a waveform reconstruction error box plot for performing a four-fold cross-validation (75% of all experimental data randomly extracted for model training and the remaining 25% for model testing), with boxes plotted from bottom quartile to top quartile, with middle horizontal line representing median. The average value of the waveform reconstruction errors of all 20 volunteers was 5.989% and the standard deviation was 2.496%. Fig. 9 shows the correlation coefficients for performing four-fold cross-validation, with an average of 0.926 for all 20 volunteers and a standard deviation of 0.030. The invention can accurately reconstruct the electrocardio waveform.
Then testing the performance of the invention when different sampling rates are applied, the invention proves that the invention can achieve lower waveform reconstruction error and higher correlation coefficient under various sampling rates. All volunteers acquired motion sensing signals of the wrist at 60Hz,100Hz,150Hz and 200Hz, respectively. Fig. 10 shows waveform reconstruction errors and correlation coefficients for reconstructing an electrocardiographic waveform using acquired data for four sample rate cases. As the sampling rate increases, the waveform reconstruction error decreases, the correlation coefficient increases, and in all cases, the waveform reconstruction error is less than 10% and the correlation coefficient is greater than 0.8. Experiments prove that the invention can accurately reconstruct the electrocardio waveform under the condition of various sampling rates. Since the motion sensor built in the commercial smart wristband device often supports a 60-200Hz sampling rate, the present invention can be applied to commercial smart wristband devices with different sampling rates.
Finally, the performance of the intelligent wrist strap equipment worn at different positions of the wrist is tested, and the intelligent wrist strap equipment can achieve higher precision at different wearing positions. All volunteers acquired wrist motion sensor signals 1 cm, 2 cm, and 3 cm above the ulnar styloid process (toward the elbow), respectively. Fig. 11 shows the waveform reconstruction error and correlation coefficient at different acquisition positions. The waveform reconstruction error is less than 10% under all three conditions, and the correlation coefficient is greater than 0.8. The invention can accurately reconstruct the electrocardio waveform under the condition of various wearing positions.
The foregoing embodiments are further illustrative of the present invention and are not to be construed as limiting the scope of the invention, which is defined by the appended claims.
Claims (10)
1. The continuous electrocardiographic waveform reconstruction system utilizing the intelligent wrist strap motion sensor is characterized by comprising intelligent wrist strap equipment with a built-in motion sensor and a processing unit;
the motion sensor of the intelligent wrist strap device is used for collecting motion sensing signals of the wrist of the user and sending the signals to the processing unit;
the processing unit processes the motion sensor and is used for acquiring a human motion sensing signal, extracting a body surface vibration signal related to the heartbeat, dividing the body surface vibration signal into single-heartbeat period fragments, and finally reconstructing an electrocardiographic waveform corresponding to the heartbeat period.
2. The continuous electrocardiographic waveform reconstruction method by using the intelligent wristband motion sensor is characterized by comprising the following steps of:
step 1: collecting a motion sensing signal of a wrist of a target user by using a motion sensor of the intelligent wrist strap device;
step 2: the processing unit extracts body surface vibration related to heartbeat in the motion sensing signal;
step 2.1: processing the wrist motion sensing signal acquired in the step 1 by using a band-pass filter, removing irrelevant noise, and reserving a part of the motion sensing signal with the frequency range of 0Hz-50 Hz;
step 2.2: according to the filtered motion sensing signals extracted in the step 2.1, based on stable wavelet transformation, further removing noise in the disordered motion sensing signals, and extracting body surface vibration signals related to heartbeat;
step 3: the processing unit cuts the body surface vibration signals related to the heart beat;
step 3.1: detecting local maximum value and local minimum value points in body surface vibration signals related to heartbeat, and constructing a local triangle; identifying a peak associated with ventricular contractions based on the local triangle feature;
step 3.2: dividing the body surface vibration signal related to the heartbeat into segments with a single heartbeat period according to the peak point related to the ventricular contraction identified in the step 3.1;
wherein, X milliseconds before the peak point is determined as the starting point of a heart cycle, X milliseconds before the next peak point is the ending point of the heart cycle, and the value range of X is 100 milliseconds to 400 milliseconds;
step 4: reconstructing a corresponding electrocardiographic waveform by using body surface vibration signal fragments related to the heartbeat;
step 4.1: according to the segments of the body surface vibration signals extracted in the step 3.2, an encoder-decoder network model capable of reconstructing an electrocardio waveform is established;
step 4.2: and (3) training the encoder-decoder network model established in the step 4.2 by utilizing the generation countermeasure network, and accurately reconstructing the corresponding electrocardio waveform according to the body surface vibration signals related to the heart beat.
3. The continuous electrocardiographic waveform reconstruction system utilizing a smart wristband motion sensor according to claim 1 wherein the motion sensor includes a tri-axis accelerometer, a tri-axis gyroscope and a tri-axis magnetometer, the tri-axis gyroscope data is analyzed, a tri-axis gyroscope X, Y, Z built in the smart wristband device, and sensing signals of the tri-axis gyroscope X, Y, Z axes are respectively defined as G X 、G Y And G Z Electrocardiogram and sensing signal G X 、G Y And G Z Has obvious correlation, wherein, the key inflection point of the electrocardiogram is in contact with G X The key inflection points have corresponding relations, and an X-axis sensing signal G of a triaxial gyroscope of the intelligent wristband equipment is used X As a motion sensing signal for the wrist.
4. The method for continuous electrocardiographic waveform reconstruction with a smart wristband motion sensor according to claim 2 wherein in step 2.1, the portion of the motion sensing signal having a frequency in the range of 5Hz-30Hz is retained.
5. The continuous electrocardiographic waveform reconstruction method using a smart wristband motion sensor according to claim 2, wherein in step 2.2, a stationary wavelet transform is first applied to the motion sensing signal extracted in step 2.1; let the mother wave be db4, decompose it into J approximate components a containing low-frequency information according to the frequency distribution range 1 ,a 2 ,…,a J And J detail components d containing high frequency information 1 ,d 2 ,…,d J J is the order set in the process of stable wavelet transformation, and the value range of the J value is [2,10]；
Next, each detail component d is calculated j Is a short-term energy e of (2) j And detecting short-time energy e j All local maximum amplitudes P j (k) K=1, 2, …, K, marking the time at which the local maximum amplitude occurs; defining a threshold
And
is the detail component d j Is a short-term energy e of (2) j All local maximum amplitude P j (k) Mean and variance of (a); comparing each P separately j (k) Amplitude of (2) is>
If->
Then classify P j (k) Is body surface vibration related to heartbeat;
further processing of otherwise uncategorized P by adjacent local maximum amplitude occurrence time intervals j (k) The method comprises the steps of carrying out a first treatment on the surface of the Defining two adjacent local maximum amplitudes P j (k-1) and P j (k) The time interval of occurrence is I j (k) The method comprises the steps of carrying out a first treatment on the surface of the Setting a threshold value
E is j All adjacent local maxima occurrence time interval I j (k) Average value of (2); if->
And->
Then classify P j (k) Is body surface vibration related to heartbeat;
finally, using local maximum amplitude P of body surface vibrations classified as heart beat-related j (k) Separation d j Middle and heart beatRelated body surface vibrations and noise; checking the maximum amplitude P of each body surface vibration related to the heartbeat j (k) Searching forward and backward for local minimum occurrence times, determining start and end times of body surface vibrations associated with heart beat, and retaining d j Corresponding data in (a), determining other data as noise, and d j The corresponding data in (a) is replaced by 0; and performing inverse stationary wavelet transformation to obtain a body surface vibration signal related to the heartbeat.
6. The continuous electrocardiographic waveform reconstruction method utilizing a smart wristband motion sensor according to claim 5 wherein the J value is selected to be 6.
7. The continuous electrocardiographic waveform reconstruction method using a smart wristband motion sensor according to claim 2, wherein in step 3.1, all local maxima and local minima points are searched in the body surface vibration signal related to the heartbeat, each local maxima point and two local minima points adjacent to the local maxima point are used for constructing a local triangle, and features are extracted from the local triangle;
defining local maximum points in the local triangle as vertexes, wherein adjacent left local minimum points are left bottom points, adjacent right local minimum points are right bottom points, selecting the sum a+b of distances from the vertexes of the local triangle to the left bottom point and the right bottom point, the difference d between the vertexes and the ordinate of the left bottom point, the difference e between the vertexes and the ordinate of the right bottom point, and the angle f of the angle where the vertexes in the local triangle are positioned as characteristics;
in addition, the difference c between the left and right base points in the abscissa can be selected, the angle g of the left base point in the local triangle and the angle h of the right base point in the local triangle can be used as supplementary features to assist in identifying the peak point;
training a random forest classifier to identify whether the vertices in the local triangle, i.e., the local maxima points, are peak points associated with ventricular contractions; during recognition, the selected characteristics are input into a random forest classifier, the probability that the vertex belongs to a peak point is predicted, and if the probability is larger than a set value, the vertex is the peak point related to ventricular contraction.
8. The continuous electrocardiographic waveform reconstruction method according to claim 7, wherein in step 3.1, a difference c between left and right nadir abscissas, an angle g of a corner where the left nadir is located in the partial triangle, and an angle h of a corner where the right nadir is located in the partial triangle are selected as supplementary features to assist in identifying the peak point.
9. The continuous electrocardiographic waveform reconstruction method using a smart wristband motion sensor according to claim 2 wherein in step 3.2, 250 ms before a peak is determined as a start point of a heart cycle and 250 ms before the next peak is determined as an end point of the heart cycle.
10. The continuous electrocardiographic waveform reconstruction method using the smart wristband motion sensor according to claim 2, wherein in step 4.1, an encoder-decoder network model is developed based on a long-short-term memory neural network, and the segment of the body surface vibration signal is input into the encoder-decoder network model;
stretching the signal length to 1.2 seconds by using a linear interpolation method, and extracting hidden information related to heartbeat by using a bidirectional long-short-time memory neural network by using an encoder and integrating the hidden information into a contraction characteristic consisting of X samples; respectively distributing weights to X samples in the contraction characteristics extracted by the encoder by using an attention mechanism; the decoder is composed of two unidirectional long and short-time memory neural network layers, and reconstructs an electrocardiographic waveform according to contraction characteristics and weights corresponding to the contraction characteristics; finally, the length of the reconstructed electrocardiographic waveform is 1.2 seconds, and the length is adjusted to be the original fragment length by applying inverse interpolation;
in step 4.2, establishing a deep learning model based on generating an countermeasure network, and assisting the training of the encoder-decoder network model in an offline mode;
the established generating countermeasure network model consists of a generator for reconstructing an electrocardiographic waveform and a discriminator for distinguishing the reconstructed electrocardiographic waveform from a real electrocardiographic waveform, wherein the encoder-decoder network model established in the step 4.1 is used as the generator for learning the complex mapping from the body surface vibration signals related to the heartbeat to the electrocardiographic waveform; the discriminator takes the reconstructed electrocardio waveform and the corresponding real electrocardio waveform output by the generator as input respectively, and stretches the input data length for 1.2 seconds by utilizing linear interpolation; then, using two-way long-short-term memory neural network layers, a full-connection layer and a decision layer to identify the current input signal as a real electrocardio waveform or a reconstructed electrocardio waveform; the parameters of the generator and the discriminator are trained in turn, so that the difference between the generated data and the real data is reduced, and the generator model accurately reconstructs the electrocardio waveform;
the reconstructed electrocardiographic waveform is denoted as e= { E 1 ,E 2 ,…,E i ,…,E L The true electrocardiographic waveform corresponding to E is denoted as a= { a } 1 ,A 2 ,…,A i ,…,A L }，E i For the amplitude of the electrocardiographic waveform signal, A i The amplitude of the real electrocardiosignal is L, and the length of two waveforms is L; waveform reconstruction error L for reconstructing electrocardiographic waveform by generator of Loss function Loss e And a discriminator for discriminating a discrimination error L of a real electrocardiographic waveform from a reconstructed electrocardiographic waveform a Common composition, loss=l e +L a ；
Setting:
L a ＝log[1-P EA ]
wherein P is EA The reconstructed electrocardiographic waveform is wrongly identified as the proportion of the real electrocardiographic waveform for the discriminator; in the training process, the parameters in the countermeasure network are generated continuously and iteratively until the Loss converges.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211090977.8A CN116269413A (en) | 2022-09-07 | 2022-09-07 | Continuous electrocardiographic waveform reconstruction system and method using smart wristband motion sensor |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211090977.8A CN116269413A (en) | 2022-09-07 | 2022-09-07 | Continuous electrocardiographic waveform reconstruction system and method using smart wristband motion sensor |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN116269413A true CN116269413A (en) | 2023-06-23 |
Family
ID=86782100
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202211090977.8A Pending CN116269413A (en) | 2022-09-07 | 2022-09-07 | Continuous electrocardiographic waveform reconstruction system and method using smart wristband motion sensor |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN116269413A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117122308A (en) * | 2023-07-24 | 2023-11-28 | 苏州大学 | Electrocardiogram measurement method and system based on mobile phone built-in acceleration sensor |
-
2022
- 2022-09-07 CN CN202211090977.8A patent/CN116269413A/en active Pending
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117122308A (en) * | 2023-07-24 | 2023-11-28 | 苏州大学 | Electrocardiogram measurement method and system based on mobile phone built-in acceleration sensor |
| CN117122308B (en) * | 2023-07-24 | 2024-04-12 | 苏州大学 | Electrocardiogram measurement method and system based on mobile phone built-in acceleration sensor |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN106971059B (en) | Wearable equipment based on neural network self-adaptation health monitoring | |
| CN110946556B (en) | Parkinson resting state tremor evaluation method based on wearable somatosensory network | |
| Yoneyama et al. | Accelerometry-based gait analysis and its application to Parkinson's disease assessment—part 1: detection of stride event | |
| Hou et al. | A real-time QRS detection method based on phase portraits and box-scoring calculation | |
| CN109222969A (en) | A kind of wearable human upper limb muscular movement fatigue detecting and training system based on Fusion | |
| WO2019096175A1 (en) | Vital sign signal analysis processing method and vital signal monitoring device | |
| EP3065628A1 (en) | Biomechanical activity monitoring | |
| CN107137071A (en) | It is a kind of to analyze the method that heart impact signal is used for calculating short-term heart beat value | |
| CN107530015B (en) | Vital sign analysis method and system | |
| Zhou et al. | Accurate recognition of lower limb ambulation mode based on surface electromyography and motion data using machine learning | |
| CN110801212B (en) | BCG signal heart rate extraction method based on neural network | |
| Fang et al. | Robust sEMG electrodes configuration for pattern recognition based prosthesis control | |
| CN112294264A (en) | Sleep staging method based on BCG and blood oxygen saturation rate | |
| CN103690169A (en) | Respiration information detection method and system | |
| CN108523868A (en) | Self-calibration system and method for blood pressure measurement | |
| Tepe et al. | Classification of emg finger data acquired with myo armband | |
| CN112274120A (en) | Noninvasive arteriosclerosis detection method and device based on one-way pulse wave | |
| CN116269413A (en) | Continuous electrocardiographic waveform reconstruction system and method using smart wristband motion sensor | |
| Zhang et al. | Gait pattern recognition based on plantar pressure signals and acceleration signals | |
| KR101870630B1 (en) | Method and device for the measurement of energy consumption based on vital/motion signals | |
| Kang et al. | A Precise Muscle activity onset/offset detection via EMG signal | |
| CN105796091B (en) | A kind of intelligent terminal for removing electrocardiosignal vehicle movement noise | |
| Febriana et al. | Sleep monitoring system based on body posture movement using Microsoft Kinect sensor | |
| CN113749644A (en) | Intelligent garment capable of monitoring lumbar movement of human body and automatically correcting posture | |
| TW201838587A (en) | Method of analyzing ballistocardiogram signal to calculate short-term heart rate value capable of fast and accurately obtaining the short term average heart rate under a low calculation amount condition |
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 |