CN104605841A - Wearable electrocardiosignal monitoring device and method - Google Patents

Abstract

The invention provides a wearable electrocardiosignal monitoring and processing method and device. Flexible fabric electrocardioelectrodes and a central control box are embedded into an elastic chest belt. The central controller comprises an electrocardiosignal collecting and conditioning unit, a signal processing unit, a data storage unit, a power unit and a wireless communication unit. The signal processing unit carries out filtering, feature extraction and analysis of the heart abnormal change state on the electrocardiosignal collected, amplified and filtered preliminarily by the collecting and conditioning unit. The signal waveform and the analysis result are sent to personal digital management equipment like a mobile phone of a user through the wireless communication unit and can be sent to a remote doctor through a network, and the doctor analyzes the illness state and feeds back medical suggestions and matters needing attentions to the user. The wearable electrocardiosignal monitoring device has the advantages of being comfortable to wear, light, small in size, low in cost and the like, long-term monitoring is facilitated under the conditions of daily life, study, activities and the like, and the intelligent diagnosis and feedback treatment of the heart state are achieved.

Description

Wearable ECG signal monitoring device and method

technical field

The invention relates to the technical field of wearable health care, in particular to a wearable ECG signal monitoring and processing method and device.

Background technique

In recent years, as the prevalence of chronic diseases headed by cardiovascular diseases among the elderly has increased, and they have the characteristics of complex etiology and illness, long duration of illness, high medical costs, and the need for long-term management, etc., more and more Many patients expect to change the health care model from hospital-centered to family-centered, and to curb the rapid rise of chronic diseases in the elderly through personal health management, thereby reducing the medical economic burden. The very important ECG signal is of great significance in the monitoring and treatment of heart disease and other chronic diseases. It is one of the earliest bioelectrical signals studied and applied in clinical medicine. The dynamic electrocardiogram is an important objective basis for clinical analysis of the disease and establishment of a diagnosis, including ECG data under different conditions such as rest, activity, work, study, and sleep. It can more completely record the heart condition and provide a more accurate basis for diagnosis. The large ECG monitoring equipment in the hospital is bulky and requires at least 12 connecting wires for measurement, which brings discomfort to the patient and increases the cost of medical treatment; on the other hand, the ECG sensor directly affects the monitoring signal quality and wearing comfort Especially for long-term real-time monitoring, the material and structure of the electrode are very important. At present, the clinical ECG electrode is a wet electrode with electrolytic gel, which will cause skin allergies and signal attenuation due to skin dehydration, so it is not suitable for use for long-term supervision.

In view of the above problems, a variety of ECG monitoring devices based on wearable fabric electrodes have been proposed in China. Chinese Patent No. 200920089699.8 discloses a wearable ECG electrode vest with a data recording device. The electrodes are woven from silver and composite fibers. The collected ECG signals can be collected and sent; Chinese Patent No. 201120506497.6 discloses a wearable ECG signal measuring device, the ECG electrodes can collect signals, process and send them, and can obtain ECG signals and heart rate values; Patent No. 201120303457.1 discloses a wearable clothing that detects human ECG signals. The ECG electrodes are sewn on the clothes by conductive cloth, and the ECG signal processing circuit and transmitter are placed in a housing. Doctor provides medical reference.

Although the above patents are based on wearable technology using flexible fabric electrodes, which can successfully achieve the purpose of long-term monitoring of ECG signals, fabric electrodes are relatively complicated to manufacture, and even require manufacturing equipment to complete. Detection of ECG in the case of human activity will bring a lot of motion artifact interference. Unlike other interferences that have a specific frequency range, it has a dynamic frequency range and a large amplitude, which can easily damage or overwhelm biological signals, which will lead to unreasonable processing and wrong diagnosis, and bring great challenges to monitoring. Motion artifacts can largely interfere with the validity of the ECG signal, possibly leading to wrong evaluation of ECG signal parameters and triggering of false alarms. Therefore, how to effectively suppress motion artifacts in dynamic ECG signals is a key issue that must be solved in wearable health monitoring.

Contents of the invention

In order to solve the above problems, the present invention provides a wearable ECG signal monitoring and processing method and device, which can monitor the heart activity status of the human body anytime and anywhere without hindering daily activities.

The wearable ECG signal monitoring device is characterized in that: the device is composed of an elastic chest strap, a flexible fabric ECG electrode, a central control box and an internal connection line, and the flexible fabric ECG electrode and the central control box are embedded in the elastic chest strap; in

The flexible fabric ECG electrodes are arranged at appropriate positions inside the chest belt, and the electrodes are protruding to ensure full contact with the skin; preferably, the flexible fabric ECG electrodes are made by printing or directly coating conductive liquid; For the convenience of setting the shapes of different chest straps, the shape of the flexible fabric ECG electrodes can be set to shapes such as circles, ovals or polygons, and can also be set in ways such as beautified patterns.

Preferably, a single-channel lead detection method is adopted, and at least three flexible fabric ECG electrodes are configured on the chest strap, two of which are corresponding to the left chest and the right chest.

The central control box at least includes an electrocardiographic signal acquisition and conditioning unit, an acceleration acquisition unit, a signal processing unit, and a wireless communication unit; in addition, the central control box can also include a data storage unit and a power supply unit, and the storage unit is stored in the ECG signal monitoring unit. The data generated during the process, the characteristics and the analysis data of the abnormal state of the heart, and the power supply unit provides power for the entire device.

The electrocardiographic signal acquisition and conditioning unit is used for collecting electrocardiographic signals, performing power frequency interference, baseline drift and noise preprocessing on electromyographic interference on the collected electrocardiographic signals, and performing gain processing on the processed electrocardiographic signals zoom in;

The acceleration acquisition unit uses an accelerometer to acquire an acceleration signal; wherein, the acceleration signal is used for suppression of motion artifact interference.

The signal processing unit is based on the processed ECG signal obtained by the ECG signal acquisition and conditioning unit and the acceleration signal collected by the acceleration acquisition unit, and performs filtering, feature extraction, and analysis of cardiac abnormalities;

Preferably, the signal processing unit includes:

The ECG signal filtering unit is connected with the ECG signal acquisition and conditioning unit and the acceleration acquisition unit, and adopts the adaptive filtering method to suppress the motion artifact interference noise generated by the human body under the condition of motion, wherein the acceleration signal is used as the adaptive filter reference signal. The signal feature extraction unit is connected with the ECG signal filtering unit, and adopts a signal feature detection algorithm to extract important signal features from the filtered ECG signal. The signal features include at least characteristic indicators such as R wave and heart rate with the largest amplitude. The indicators can be modified according to user needs, and will not be limited and described here. Cardiac abnormality diagnosis unit is used to detect the parameter value of each feature of the ECG signal, analyze these signal features in the time domain and frequency domain, obtain the statistical indicators of the ECG signal, and use the machine learning classification algorithm to analyze the user's heart rate. Status is classified and identified.

The wireless communication unit is used to wirelessly send the processed data to the receiving terminal for observation of human physiology and activity status and analysis and diagnosis of disease; preferably, the receiving terminal can be a mobile terminal, a fixed PC computers, personal digital assistants, laptops, tablets, and more.

The interconnection wire connects the flexible fabric ECG electrode and the central control box; preferably, the interconnection wire is made of a wire or conductive fiber, and is covered with common yarn sewing on the interconnection wire to form an interconnection track.

In addition, the present invention also provides an ECG signal monitoring method based on a wearable ECG signal monitoring device, the device at least includes a flexible fabric ECG electrode, an ECG signal acquisition and conditioning unit, an acceleration acquisition unit, a signal processing unit, The wireless communication unit and the internal connection are characterized in that:

1), signal acquisition and conditioning

Collect the ECG signal through the ECG signal acquisition and conditioning unit, and preliminarily eliminate the interference noise such as myoelectric interference, power frequency interference and baseline drift for the collected ECG signal; and collect the acceleration signal through the acceleration acquisition unit;

2), motion artifact suppression

Through the adaptive filtering algorithm, for the ECG signal obtained in step 1), the adaptive filter is used to suppress the motion artifact interference noise generated by the human body in motion. In this scheme, a conventional adaptive filter can be used In this method, the acceleration signal is used as a reference signal, and the self-adaptive filtering algorithm is used to automatically adjust its own weight coefficients to achieve the best filtering effect.

Preferably, this solution can be implemented using the following adaptive filter, but the application is not limited to the following method: the adaptive filtering algorithm can be used to automatically adjust its own weight coefficient W to achieve the best filtering effect. The adaptive filter has two input signals, one is the ECG signal d(k) with motion artifact interference, and the other is the reference signal x(k); among them, a normalized variable step size minimum mean square error ( Least Mean Squares, LMS) algorithm is used as an adaptive filtering algorithm, and the acceleration signal is used as the reference signal of the adaptive filter;

According to the structure of the adaptive filter, the output of the adaptive filter is the inner product of the reference signal and the weight coefficient, that is, y=x ^T W, then the output error of the entire adaptive filter is the difference between the input signal and the output signal , ie e(k)=d(k)-x ^T W. The LMS algorithm is to minimize the mean square value of the output error of the above formula, so as to suppress the noise signal. According to the LMS algorithm, it can be known that the update formula of the weight coefficient is:

W(k+1)=W(k)+μe(k)x(k)

Among them, μ is the step size factor. The present invention uses a kind of normalized LMS algorithm of variable step size, has faster convergence speed and smaller steady-state error compared with traditional LMS algorithm, and the step size factor of this algorithm is expressed as:

$\mathrm{<span\; class="notranslate">\; \&mu;</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span>\frac{\mathrm{<span\; class="notranslate">\; \&mu;</span>}}{{\mathrm{<span\; class="notranslate">\; x</span>}}^{\mathrm{<span\; class="notranslate">\; T</span>}}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; )</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; )</span>}$

Therefore, the update formula of the weight coefficient becomes:

$\mathrm{<span\; class="notranslate">\; W</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; +</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; W</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; +</span>\frac{\mathrm{<span\; class="notranslate">\; \&mu;</span>}}{{\mathrm{<span\; class="notranslate">\; x</span>}}^{\mathrm{<span\; class="notranslate">\; T</span>}}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; )</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; )</span>}\mathrm{<span\; class="notranslate">\; e</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; )</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; )</span>$

Utilize the acceleration signal of three direction axes (x, y, z) of accelerometer in the present invention as reference signal, i.e. x(k)=[Acc _x (k), Acc _y (k), Acc _z (k)] , the filter weight coefficient vector W=[w ₁ , w ₂ , w ₃ ].

Thus, the steps of the whole adaptive filtering are:

Step 1: Initialize the weight coefficient vector W(0)=[0,0,0];

The second step: estimate the filter error e(k)=d(k)-x ^T W at the current moment;

Step 3: Update the filter weight coefficient vector:

$\mathrm{<span\; class="notranslate">\; W</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; +</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; W</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; +</span>\frac{\mathrm{<span\; class="notranslate">\; \&mu;</span>}}{{\mathrm{<span\; class="notranslate">\; x</span>}}^{\mathrm{<span\; class="notranslate">\; T</span>}}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; )</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; )</span>}\mathrm{<span\; class="notranslate">\; e</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; )</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; )</span>$

Step 4: Increase the time parameter k until the next time, repeat the above steps until the number of iterations is reached.

3), signal feature extraction

With the ECG signal obtained in step 2) after motion artifact suppression, important signal features are extracted through a signal feature detection algorithm. The signal features at least include characteristic indicators such as R waves and heart rate with the largest amplitude. The user needs to be modified, and will not be limited and described here; these signal features are formed into a local feature database for the analysis and evaluation of heart rate variability;

4) Analysis and evaluation of heart disease

The parameter value of each feature of the ECG signal obtained in step 3) forms a feature matrix, and these signal features are analyzed in the time domain and frequency domain to obtain the statistical index of the ECG signal, and according to the statistical index, the user's heart Status is classified and identified. Preferably, in this step, a machine learning classification algorithm can be used to classify and identify the user's heart state according to the statistical indicators of the ECG signal

In addition, the method may further include:

5) Send the ECG signal after signal processing and the classification and recognition results of the user's heart state to the doctor through wireless, and the doctor will feedback the diagnosis result and the suggestion of disease prevention.

The wearable ECG signal monitoring device of the present invention can monitor the heart activity of the wearer for a long time under the conditions of daily life, study and exercise. Appropriate signal processing methods are adopted for the collected ECG signals, which can get rid of the limitation that the ECG signals can only be collected under static conditions, and feature extraction algorithms are used to extract signal features from the processed ECG signals to form a feature database for subsequent intelligent diagnosis and treatment. provide data. The beneficial effects of the above-mentioned technical solution are as follows: (1) Electrocardiographic electrodes are made by using conductive materials that are electrically stimulated to change electrical characteristics. , The central node embedded in the chest strap is light in weight and low in power consumption; (3) It can detect the ECG signal of the user in the exercise state in real time, the signal transmission is stable, the ECG waveform is clear, and the noise is less, which is convenient for signal feature detection; (4) The processed signal can be intelligently analyzed to generate diagnostic results, and wirelessly transmitted to mobile phones, tablet computers or personal management devices in real time, so that users can understand their heart conditions in real time; (5), the The signal and analysis results are transmitted to the diagnosis and treatment doctors through the network, and the doctors evaluate the diagnosis results and provide scientific treatment suggestions to the patients.

Description of drawings

Figure 1 is a schematic structural diagram of a wearable ECG monitoring chest strap;

Fig. 2 is a schematic diagram of the structure of a wearable ECG electrode;

Fig. 3 is a functional block diagram of a wearable ECG monitoring device;

Fig. 4 is a block diagram of a signal processing module of a wearable ECG monitoring device;

Fig. 5 is a wearable ECG motion artifact interference suppression block diagram;

Fig. 6 is a flowchart of a wearable ECG monitoring method.

Among them, 1-elastic chest belt; 2-conductive fabric electrode; 3-central control box; 4-fabric; 5-fabric electrode layer; 6-ECG signal acquisition and conditioning unit; 7-acceleration acquisition unit; 8-signal processing unit 9-wireless communication unit; 10-data storage unit; 11-power supply unit; 12-ECG signal filtering unit; 13-signal feature extraction unit; 14-cardiac abnormality diagnosis unit.

Detailed ways

In order to make the technical problems, technical solutions and advantages to be solved by the present invention clearer, the following will describe in detail with reference to the drawings and specific embodiments.

The present invention aims at the problems of the existing wearable ECG acquisition and measurement devices, such as the complex production process of the ECG electrode, the need for large-scale manufacturing equipment, the high cost, the intelligent detection of the ECG signal in the static state of the user, and the lack of intelligent diagnosis and treatment functions, etc., and provides a solution. A wearable electrocardiographic monitoring diagnosis and treatment method and device that can be used for a long time in the daily life, study, activity and sleep of the user with simple manufacturing process and low cost.

As shown in FIG. 1 , the wearable ECG monitoring diagnosis and treatment device of the present invention includes an elastic chest strap 1 , conductive fabric electrodes 2 , a central control box 3 and interconnecting wires.

The chest strap is made of elastic material, and the connecting buckle of the chest strap is designed into several different positions, so that the tightness of the strap can be adjusted, which is suitable for different fat and thin users, and is easy to wear. Then fasten it. Since the main purpose of the device is to identify the cardiac cycle and analyze the heart rate, a single-channel lead detection method is adopted, and at least three fabric electrodes are arranged on the chest strap, and the shape of the fabric electrodes can be circular, oval or polygonal. Place two of the fabric electrodes on the left and right chest while wearing the chest strap. The central control box is placed in the small pocket of the chest strap and wraps around the user's back along with the chest strap. The central control box is connected to the fabric electrodes through the internal connection line. The internal connection line is made of wire or conductive fiber with better elasticity and thinner size. It is covered by sewing on the inner connecting line to form a regular inner connecting track.

As shown in Figure 2, the textile electrode of the present invention adopts the method of printing or directly coating the conductive liquid. Compared with the woven electrode, the manufacturing process is simple and does not require large-scale manufacturing equipment, so the cost is lower. Utilize the elastic force of the better elastic fabrics such as sponge to ensure that the conductive part can fully contact with the human skin, and the fabric 4 (such as sponge) with a good shape is used to stick or sew on the chest strap 1 with an adhesive. The non-irritating conductive liquid to the skin is printed or directly coated on the fabric 4 to form the fabric electrode layer 5 .

As shown in FIG. 3 , the wearable ECG monitoring device includes an ECG signal acquisition and conditioning unit 6 , an acceleration acquisition unit 7 , a signal processing unit 8 , a wireless communication unit 9 , a data storage unit 10 and a power supply unit 11 . Wherein the electrocardiographic signal acquisition conditioning unit 6 amplifies and filters the electrocardiographic signal, selects a suitable gain to reach the voltage of AD conversion, and preliminarily eliminates interference noises such as myoelectric interference, power frequency interference and baseline drift through the filter circuit; the acceleration acquisition unit 7 Acquisition of the acceleration signal of the human body can provide a reference signal for the filtering of motion artifacts; the signal processing unit 8 has a built-in AD conversion module, which converts the analog signal into a digital signal and processes it through the signal processing unit, mainly for noise filtering and signal feature extraction and calculation and analysis of heart rate variability, the data storage unit 10 stores the collected data in a portable storage medium for the doctor to analyze the disease; the processed data is sent to the user in a wireless manner through the wireless communication unit 9 Mobile phones, tablet computers or personal management devices for users to observe their own heart conditions in real time; power supply unit 11 provides electrical energy for ECG signal acquisition and conditioning unit 6, acceleration acquisition unit 7, calculation processing unit 8 and wireless communication unit 9.

As shown in FIG. 4 , the signal processing unit of the wearable ECG monitoring device includes an ECG signal filtering unit 12 , a signal feature extraction unit 13 and a cardiac abnormality diagnosis unit 14 .

The electrocardiographic signal filtering unit 12 filters the noise that cannot be completely filtered out in the hardware filtering method, wherein, since the motion artifact interference belongs to a non-stationary random signal, has a dynamic frequency range, and has a large amplitude, use hardware filtering or common The software filtering method obviously can't reach the effect of filtering, and needs a filter with good performance and filtering algorithm to complete. In the present invention, an adaptive filter is used to filter, and an acceleration signal having correlation with the motion artifact signal is used as an adaptive filter. The reference signal of the filter, and adopts a reasonable adaptive filtering algorithm to realize real-time ECG monitoring when the user is in motion;

Preferably, this solution can be implemented using the following adaptive filter, but the application is not limited to the following method: an adaptive filtering algorithm can be used to automatically adjust its own weight coefficient W to achieve the best filtering effect. The adaptive filter has two input signals, one is the ECG signal d(k) with motion artifact interference, and the other is the reference signal x(k), where k is the time parameter; among them, a normalized variable step is adopted The Least Mean Squares (LMS) algorithm is used as an adaptive filtering algorithm, and the acceleration signal is used as the reference signal of the adaptive filter;

According to the structure of the adaptive filter, the output of the adaptive filter is the inner product of the reference signal and the weight coefficient, that is, y=x ^T W, then the output error of the entire adaptive filter is the difference between the input signal and the output signal , ie e(k)=d(k)-x ^T W. The LMS algorithm is to minimize the mean square value of the output error of the above formula, so as to suppress the noise signal. According to the LMS algorithm, it can be known that the update formula of the weight coefficient is:

W(k+1)=W(k)+μe(k)x(k)

Among them, μ is the step size factor. The present invention uses a kind of normalized LMS algorithm of variable step size, has faster convergence speed and smaller steady-state error compared with traditional LMS algorithm, and the step size factor of this algorithm is expressed as:

$\mathrm{<span\; class="notranslate">\; \&mu;</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span>\frac{\mathrm{<span\; class="notranslate">\; \&mu;</span>}}{{\mathrm{<span\; class="notranslate">\; x</span>}}^{\mathrm{<span\; class="notranslate">\; T</span>}}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; )</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; )</span>}$

Therefore, the update formula of the weight coefficient becomes:

$\mathrm{<span\; class="notranslate">\; W</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; +</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; W</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; +</span>\frac{\mathrm{<span\; class="notranslate">\; \&mu;</span>}}{{\mathrm{<span\; class="notranslate">\; x</span>}}^{\mathrm{<span\; class="notranslate">\; T</span>}}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; )</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; )</span>}\mathrm{<span\; class="notranslate">\; e</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; )</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; )</span>$

Utilize the acceleration signal of three direction axes (x, y, z) of accelerometer in the present invention as reference signal, i.e. x(k)=[Acc _x (k), Acc _y (k), Acc _z (k)] , the filter weight coefficient vector W=[w ₁ , w ₂ , w ₃ ].

Thus, the steps of the whole adaptive filtering are:

Step 1: Initialize the weight coefficient vector W(0)=[0,0,0];

The second step: estimate the filter error e(k)=d(k)-x ^T W at the current moment;

Step 3: Update the filter weight coefficient vector:

$\mathrm{<span\; class="notranslate">\; W</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; +</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; W</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; +</span>\frac{\mathrm{<span\; class="notranslate">\; \&mu;</span>}}{{\mathrm{<span\; class="notranslate">\; x</span>}}^{\mathrm{<span\; class="notranslate">\; T</span>}}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; )</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; )</span>}\mathrm{<span\; class="notranslate">\; e</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; )</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; )</span>$

Step 4: Increase the time parameter k until the next time, repeat the above steps until the number of iterations is reached.

The signal feature extraction unit 13 is to adopt the signal feature detection algorithm to extract important signal features from the filtered ECG signal, such as the R wave with the largest amplitude, heart rate, etc.; The statistical indicators of the ECG signal are obtained through the analysis of the domain, and the machine learning classification algorithm is used to classify and identify the user's heart state, and it can also be sent to the doctor through the wireless communication module 9, and the doctor will feedback the diagnosis result and the suggestion of disease prevention.

As shown in Figure 5, the human body dynamic electrocardiogram signal that utilizes fabric electrode to collect is as the input signal source 14 of filter, and it is the combined signal of ideal electrocardiogram signal and motion artifact noise, will be collected synchronously with dynamic electrocardiogram signal The three-axis acceleration signal 15 is used as the reference signal of the adaptive filter, and then the adaptive filter 16 is used to suppress motion artifacts to obtain relatively pure ECG signals.

As shown in Figure 6, the ECG signal monitoring method based on the wearable ECG signal monitoring device in the present invention is mainly realized through the following steps:

1), signal acquisition and conditioning

Collect the ECG signal through the ECG signal acquisition and conditioning unit, and preliminarily eliminate the interference noise such as myoelectric interference, power frequency interference and baseline drift for the collected ECG signal; and collect the acceleration signal through the acceleration acquisition unit;

2), motion artifact suppression

Through the adaptive filtering algorithm, for the electrocardiographic signal obtained in step 1), the motion artifact interference noise generated by the human body under the motion condition is suppressed; wherein, the acceleration signal is used as the reference signal of the adaptive filter;

3), signal feature extraction

With the electrocardiographic signal obtained in step 2) after motion artifact suppression, extract important signal features by a signal feature detection algorithm, this signal feature includes at least the R wave and heart rate with the largest amplitude; these signal features are formed into a local A characteristic database for the analysis and evaluation of heart rate variability;

4) Analysis and evaluation of heart disease

The parameter value of each feature of the electrocardiographic signal obtained in step 3) forms a feature matrix, and these signal features are analyzed in the time domain and frequency domain to obtain the statistical index of the electrocardiographic signal, and according to the statistical index to the user's Classification and identification of heart status.

In step 4), a machine learning classification algorithm, such as a neural network, is used to classify and identify the user's heart state according to the statistical indicators of the ECG signal.

5) Send the ECG signal after signal processing and the classification and recognition results of the user's heart state to the doctor through wireless, and the doctor will feedback the diagnosis result and the suggestion of disease prevention.

Preferably, the adaptive filtering algorithm in step 2) automatically adjusts its own weight coefficient W to achieve the best filtering effect. The adaptive filter uses two input signals, one of which is the ECG signal d with motion artifact interference. (k), one path is a reference signal x(k), where k is a time parameter.

The above adaptive filtering algorithm can use the normalized variable step size minimum mean square error algorithm,

The output of the adaptive filter of this adaptive filtering algorithm is the inner product of the reference signal x(k) and the weight coefficient W, that is, y=x ^T W, and the output error of the adaptive filter is e(k)=d( k)-x ^T W; the update formula of the weight coefficient is:

$\mathrm{<span\; class="notranslate">\; W</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; +</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; W</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; +</span>\frac{\mathrm{<span\; class="notranslate">\; \&mu;</span>}}{{\mathrm{<span\; class="notranslate">\; x</span>}}^{\mathrm{<span\; class="notranslate">\; T</span>}}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; )</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; )</span>}\mathrm{<span\; class="notranslate">\; e</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; )</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; .</span>$

The steps of the above adaptive filtering algorithm are as follows:

Step 1: Initialize the weight coefficient vector W(0)=[0,0,0];

The second step: estimate the filter error e(k)=d(k)-x ^T W at the current moment;

Step 3: Update the filter weight coefficient vector:

$\mathrm{<span\; class="notranslate">\; W</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; +</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; W</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; +</span>\frac{\mathrm{<span\; class="notranslate">\; \&mu;</span>}}{{\mathrm{<span\; class="notranslate">\; x</span>}}^{\mathrm{<span\; class="notranslate">\; T</span>}}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; )</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; )</span>}\mathrm{<span\; class="notranslate">\; e</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; )</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; )</span>$

Step 4: Increase the time parameter k until the next time, repeat the above steps until the number of iterations is reached.

Preferably, the acceleration signals of the three direction axes (x, y, z) of the accelerometer can be used as reference signals, that is, x(k)=[Acc _x (k), Acc _y (k), Acc _z (k) ], filter weight coefficient vector W=[w ₁ , w ₂ , w ₃ ].

The above description is a preferred embodiment of the present invention, it should be pointed out that for those of ordinary skill in the art, without departing from the principle of the present invention, some improvements and modifications can also be made, these improvements and modifications It should also be regarded as the protection scope of the present invention.

Claims (16)

1. A wearable ECG signal monitoring device, characterized in that: said device is made up of elastic chest strap, flexible fabric ECG electrode, central control box and interconnection wire, wherein, The flexible fabric ECG electrode is arranged at an appropriate position inside the chest belt, and the electrode is provided as a protrusion, thereby ensuring complete contact between the electrode and the skin; The central control box at least includes an ECG signal acquisition and conditioning unit, an acceleration acquisition unit, a signal processing unit, and a wireless communication unit; The electrocardiographic signal acquisition and conditioning unit is used for collecting electrocardiographic signals, performing power frequency interference, baseline drift and noise preprocessing on electromyographic interference on the collected electrocardiographic signals, and performing gain processing on the processed electrocardiographic signals zoom in; The acceleration acquisition unit is used to acquire acceleration signals; The signal processing unit is based on the processed ECG signal obtained by the ECG signal acquisition and conditioning unit and the acceleration signal collected by the acceleration acquisition unit, and performs filtering, feature extraction, and analysis of cardiac abnormalities; The wireless communication unit is used to wirelessly send the processed data to the receiving terminal; The internal wiring connects the flexible fabric ECG electrodes and the central control box. 2. The device according to claim 1, characterized in that: The monitoring device adopts a single-channel lead detection method, and at least three flexible fabric ECG electrodes are arranged on the chest belt, two of which correspond to the left chest and the right chest. 3. The device according to claim 1, characterized in that: The acceleration signal collected by the acceleration collection unit is used for suppression of motion artifact interference. 4. The device according to claim 1, characterized in that: The interconnecting wires use wires or conductive fibers and sew ordinary yarns on the interconnecting wires to cover them to form interconnecting tracks. 5. The device according to claim 1, characterized in that: The flexible fabric ECG electrode is made by printing or directly coating conductive liquid. 6. The device according to claim 1, characterized in that: The central control box also includes a data storage unit and a power supply unit; the storage unit stores the data, characteristics and analysis data of cardiac abnormalities generated during the ECG signal monitoring process. 7. The device according to claim 1, characterized in that: The receiving terminal is a mobile terminal, a fixed PC, a personal digital assistant, a notebook computer, or a tablet computer. 8. The device according to claim 1, characterized in that: The shape of the flexible fabric ECG electrode is circular, oval or polygonal. 9. The device according to claim 1, characterized in that: The signal processing unit includes: The electrocardiographic signal filtering unit adopts an adaptive filtering method to suppress the motion artifact interference noise generated by the human body in motion, wherein the acceleration signal is used as the reference signal of the adaptive filter; The signal feature extraction unit adopts a signal feature detection algorithm to extract important signal features from the filtered ECG signal, and the signal features include at least the R wave and heart rate with the largest amplitude; The cardiac abnormality diagnosis unit is used to detect the parameter value of each feature of the ECG signal, analyze these signal features in the time domain and frequency domain, obtain the statistical indicators of the ECG signal, and use the machine learning classification algorithm to analyze the user's heart state. Perform classification recognition. 10. An ECG signal monitoring method based on a wearable ECG signal monitoring device, the device at least includes a flexible fabric ECG electrode, an ECG signal acquisition and conditioning unit, an acceleration acquisition unit, a signal processing unit, a wireless communication unit and an internal Wired, characterized by: 1), signal acquisition and conditioning Collect the ECG signal through the ECG signal acquisition and conditioning unit, and preliminarily eliminate the interference noise such as myoelectric interference, power frequency interference and baseline drift for the collected ECG signal; and collect the acceleration signal through the acceleration acquisition unit; 2), motion artifact suppression Through the adaptive filtering algorithm, for the electrocardiographic signal obtained in step 1), the motion artifact interference noise generated by the human body under the motion condition is suppressed; wherein, the acceleration signal is used as the reference signal of the adaptive filtering algorithm; 3), signal feature extraction With the electrocardiographic signal obtained in step 2) after motion artifact suppression, extract important signal features by a signal feature detection algorithm, this signal feature includes at least the R wave and heart rate with the largest amplitude; these signal features are formed into a local A characteristic database for the analysis and evaluation of heart rate variability; 4) Analysis and evaluation of heart disease The parameter value of each feature of the ECG signal obtained in step 3) forms a feature matrix, and these signal features are analyzed in the time domain and frequency domain to obtain the statistical index of the ECG signal, and according to the statistical index, the user's heart Status is classified and identified. 11. The method of claim 10, wherein: In the step 4), a machine learning classification algorithm is used to classify and identify the user's heart state according to the statistical indicators of the ECG signal. 12. The method of claim 10, wherein: It also includes step 5), sending the processed ECG signal and the classification and identification results of the user's heart state to the doctor through wireless, and the doctor will feedback the diagnosis result and the suggestion of disease prevention. 13. The method according to claim 10, characterized in that: the self-adaptive filtering algorithm of said step 2) automatically adjusts the weight coefficient W to achieve the best filtering effect, and the adaptive filtering adopts two-way input Signals, one is an ECG signal d(k) with motion artifact interference, and the other is a reference signal x(k), where k is a time parameter. 14. The method according to claim 13, characterized in that: said adaptive filtering algorithm adopts a normalized variable step size minimum mean square error algorithm, The output of the adaptive filter of this adaptive filtering algorithm is the inner product of the reference signal x(k) and the weight coefficient W, that is, y=x ^T W, and the output error of the adaptive filter is e(k)=d( k)-x ^T W; the update formula of the weight coefficient is: $\mathrm{<span\; class="notranslate">\; W</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; +</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; W</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; +</span>\frac{\mathrm{<span\; class="notranslate">\; \&mu;</span>}}{{\mathrm{<span\; class="notranslate">\; x</span>}}^{\mathrm{<span\; class="notranslate">\; T</span>}}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; )</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; )</span>}\mathrm{<span\; class="notranslate">\; e</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; )</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; .</span>$ 15. The method according to claim 14, characterized in that: the step of the adaptive filtering algorithm is, Step 1: Initialize the weight coefficient vector W(0)=[0,0,0]; The second step: estimate the filter error e(k)=d(k)-x ^T W at the current moment; Step 3: Update the filter weight coefficient vector: $\mathrm{<span\; class="notranslate">\; W</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; +</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; W</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; +</span>\frac{\mathrm{<span\; class="notranslate">\; \&mu;</span>}}{{\mathrm{<span\; class="notranslate">\; x</span>}}^{\mathrm{<span\; class="notranslate">\; T</span>}}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; )</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; )</span>}\mathrm{<span\; class="notranslate">\; e</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; )</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; )</span>$ Step 4: Increase the time parameter k until the next time, repeat the above steps until the number of iterations is reached. 16. method as claimed in claim 13 is characterized in that: utilize the acceleration signal of three direction axes (x, y, z) of accelerometer as reference signal, i.e. x (k)=[Acc _x (k), Acc _y (k), Acc _z (k)], filter weight coefficient vector W=[w ₁ , w ₂ , w ₃ ].