CN110215202A - The pre- measuring/correlation method in Cardiac RR interval based on gait nonlinear characteristic - Google Patents
The pre- measuring/correlation method in Cardiac RR interval based on gait nonlinear characteristic Download PDFInfo
- Publication number
- CN110215202A CN110215202A CN201910398150.5A CN201910398150A CN110215202A CN 110215202 A CN110215202 A CN 110215202A CN 201910398150 A CN201910398150 A CN 201910398150A CN 110215202 A CN110215202 A CN 110215202A
- Authority
- CN
- China
- Prior art keywords
- gait
- feature
- follows
- data
- sample
- 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.)
- Withdrawn
Links
Classifications
-
- 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
- A61B5/112—Gait analysis
-
- 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
- A61B5/24—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]
- A61B5/346—Analysis of electrocardiograms
- A61B5/349—Detecting specific parameters of the electrocardiograph cycle
- A61B5/352—Detecting R peaks, e.g. for synchronising diagnostic apparatus; Estimating R-R interval
-
- 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
- A61B5/389—Electromyography [EMG]
-
- 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
- A61B5/725—Details of waveform analysis using specific filters therefor, e.g. Kalman or adaptive filters
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
- A61B5/00—Measuring for diagnostic purposes; Identification of persons
- A61B5/72—Signal processing specially adapted for physiological signals or for diagnostic purposes
- A61B5/7235—Details of waveform analysis
- A61B5/7264—Classification of physiological signals or data, e.g. using neural networks, statistical classifiers, expert systems or fuzzy systems
- A61B5/7267—Classification of physiological signals or data, e.g. using neural networks, statistical classifiers, expert systems or fuzzy systems involving training the classification device
Landscapes
- Health & Medical Sciences (AREA)
- Life Sciences & Earth Sciences (AREA)
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Public Health (AREA)
- Surgery (AREA)
- Veterinary Medicine (AREA)
- General Health & Medical Sciences (AREA)
- Animal Behavior & Ethology (AREA)
- Biophysics (AREA)
- Pathology (AREA)
- Biomedical Technology (AREA)
- Heart & Thoracic Surgery (AREA)
- Medical Informatics (AREA)
- Molecular Biology (AREA)
- Artificial Intelligence (AREA)
- Cardiology (AREA)
- Physiology (AREA)
- Psychiatry (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Signal Processing (AREA)
- Evolutionary Computation (AREA)
- Fuzzy Systems (AREA)
- Mathematical Physics (AREA)
- Dentistry (AREA)
- Oral & Maxillofacial Surgery (AREA)
- Measurement Of The Respiration, Hearing Ability, Form, And Blood Characteristics Of Living Organisms (AREA)
Abstract
The invention discloses the pre- measuring/correlation methods in Cardiac RR interval based on gait nonlinear characteristic.Step of the present invention: the three-dimensional coordinate of 1. acquisition physical feelings, and the corresponding electrocardiogram of synchronous acquisition establish the sample database of different motion mode;2. extracting gait cycle, gait data is subjected to segment processing with every ten gait cycles, the method that quantitative analysis and nonlinear dynamic analysis is then respectively adopted extracts a variety of gait features;3. corresponding ECG signal is pre-processed, the interval RR of ECG signal corresponding with each gait subset is extracted;4. constructing the interaction prediction model of gait feature and Cardiac RR interval;5. trained and tested K ELM neural network, tests feature after single feature and fusion.The present invention is easier to the dynamic nature characteristic that the lively gait for depicting different motion mode contains, and can preferably predict the interval RR of ECG signal.And the advantages such as accuracy rate, rapidity and Generalization Capability are more prominent.
Description
Technical field
The invention belongs to the technical fields such as machine learning and signal processing, medical treatment & health, human motion analysis, are related to one
Cardiac RR interval pre- measuring/correlation method of the kind based on gait nonlinear characteristic.
Background technique
With the development of science and technology the raising of human living standard, people also increasingly pay close attention to the health of itself, therefore right
It is also higher and higher in the requirement of medical service quality.Arrhythmia cordis is a kind of extremely common electrical activity abnormality, when serious
Even it can cause to die suddenly.Currently, cardiovascular and cerebrovascular disease is still to endanger one of principal disease of human life and health.Especially
It is that during the motion, excessive movement causes the consequence of uncomfortable even sudden death etc. especially severes of body.
The gait analysis system of existing instrument and meter mainly includes the movement of the movement measurement system of video, pressure information
Monitoring system, surface electromyogram signal monitor system.But that there are data volumes is big for these methods, is not easy to store;It handles relatively multiple
Miscellaneous, real-time is poor;Information lacks, vulnerable to interference the shortcomings that.Development meaning weight of the research of ECG signal to medical domain
It greatly, is myocardial infarction, the treatment of the patients such as sinus arrhythmia and sudden cardiac death provides very big value.QRS complex
The important component of ECG signal, main method has a neural network and wavelet transformation, but its there are the training of initial stage
The shortcomings that time is longer, and calculation amount is larger, is not suitable for real-time detection.By the two binding, there is important value and meaning
Justice.In medical field, heart disease prediction and monitoring can be carried out, patient is by wearing some portable devices, for detecting
The body parameter of patient is recorded, electrocardiosignal extraction, and synchro measure and record human body exercise intensity is carried out, patient is moved
Make carry out intelligent monitor.In Sports Field, by combining the knowledge such as kinematics that sportsman can be helped not influence physical safety
Situation under more acurrate, targetedly exercise motion, to optimize and be promoted sports performance.Body gait signal essence
On be considered as by complex dynamical systems generate complicated non-stationary signal.In recent years, with the development of technology, Yi Xieshi
Analysis method for nonlinear and non local boundary value problem is suggested, such as Lyapunov index, entropy and complexity etc..Only with
The dynamic nature characteristic that the gait that traditional gait feature parameter is obviously not enough to portray different motion mode contains.
Summary of the invention
In view of the deficiencies of the prior art, it is an object of the present invention to provide a kind of Cardiac RRs based on gait nonlinear characteristic
It is spaced pre- measuring/correlation method.
It is as follows that the present invention solves technical solution used by its technical problem:
The acquisition of three-dimensional gait data and corresponding electrocardiosignal when step 1. human motion;Pass through three-dimensional real time kinematics
The three-dimensional gait data of system acquisition physical feeling, and the corresponding electrocardiogram of synchronous acquisition (ECG) activity are captured, is established different
The sample database of motor pattern;
The pretreatment and feature extraction of step 2. gait data;Gait data is segmented with every ten gait cycles
Processing, the method that quantitative analysis and nonlinear dynamic analysis is then respectively adopted extract a variety of gait features;
Step 3. pre-processes electrocardiosignal (ECG signal), extracts ECG letter corresponding with each gait subset
Number RR spaced features;
Step 4. merges body gait feature, and building gait fusion feature is associated with recurrence with RR interphase feature
Model;
Step 5. training and test interaction prediction model, test the prediction result after single feature and Fusion Features respectively.
Step 1 is implemented as follows:
Experiment is walked and run to operating speed range in the Cybex 770T treadmill of 8~20km/h to acquire five kinds of fortune
The data of dynamic scene;Wherein, the linear acceleration of head, lower back portion and foot and angular velocity data are using Inertial Measurement Unit
Acquisition;It is tracked by the infrared three-dimensional real time motion capture system of Codamotion, is arranged using gait analysis software
Body kinematics position under the sample frequency of 100Hz;Pass through 4 Coda CX1 units being placed in laboratory work space
Detection is to cover range of operation;24 markers are placed with measured, wherein 4 markers are located at head: preceding forehead, a left side,
Right, rear frontal lobe;4 are located at upper limb: left acromion, right acromion, seventh cervical spine, the 5th pelvis rear;In left and right leg outer side shin bone
It is horizontally mounted 24 label clusters;Other 8 are located on foot: left and right with elbow lever, left and right followed by low-level, left elbow and the right side
Elbow, the 5th left toe, right toe;It is walked by acquisition walking and the three-dimensional for running 24 markers of different motion mode lower part of the body body region
State data, at the same record electromyogram (EMG) and electrocardiogram (ECG) activity, acquire 10 leg muscles surface electromyogram signal and
1 channel electrocardiogram, and it is synchronous with the physical feeling kinematics coordinate data of 100Hz.
Step 2 is implemented as follows;
2.1 extract gait cycle;
By the resulting three-dimensional gait data de-noising of step 1, the three-dimensional gait of the mark point acquisition at human body lower limbs position is selected
Data find all wave crests, are a gait cycle by each adjacent two wave crest;
Gait data is divided into multistage using every ten gait cycles as a subset by 2.2;
2.3 are respectively adopted quantitative analysis, nonlinear dynamic analysis method extracts a variety of gait features, specifically such as
Under:
(1) using quantitative analysis method calculate gait space-time and angle character parameter, including gait cycle, step-length,
Stride, step height, step width, leg speed, leg speed acceleration and sufficient drift angle, and its mean value is taken by each section;
Leg speed SV formula is as follows:
SV=S/GC=S/ (n/fs)
Wherein, the sampling number that n includes by a gait cycle, fsFor the sampling frequency of acquisition kinematic data setting
Rate, i.e. fs=100Hz;S is distance length, and GC is gait cycle;
Leg speed acceleration SACC calculation formula is as follows:
SACC=SV/GC=S/GC2
Sufficient drift angle is angle formed by the line and direction of advance of heel and toe;Left and right toe foot drift angle is extracted every
Maximum value in one gait cycle is denoted as MaxFA, and minimum value remembers MinFA, then takes its mean value for each section;
(2) fuzzy entropy, approximate entropy, sample of every ten gait cycles are extracted using the method for nonlinear dynamic analysis
Entropy, LZ complexity, C0 complexity characteristics.
Step 3 is implemented as follows;
3-1. handle to ECG signal by the bandpass filter that cascade low pass and high-pass filter form;
The difference equation of the low-pass filter of second order in bandpass filter are as follows:
Y (nT)=2y (nT-T)-y (nT-2T)+x (nT) -2x (nT-6T)+x (nT-12T)
Wherein, T is the sampling period, and x (nT) indicates the input of discrete-time system electrocardiosignal time series, and y (nT) is
Electrocardiosignal after low-pass filtering, the cutoff frequency of low-pass filter are set as 12Hz, gain 2, and processing delay is about adopted for 6
The sample period;
The cutoff frequency of high-pass filter is about 5Hz, the delay in 32,16 sampling periods of gain;Its difference equation is such as
Under:
Y (mT)=32x (mT-16T)-[y (mT-T)+x (mT)-x (mT-32T)]
3-2. uses derivative filter to provide QRS negative slope information by bandpass filter treated ECG signal;It leads
The difference equation of wavenumber filter is as follows:
Y (mT)=(1/8T) [- x (mT-2T) -2x (mT-T)+2x (mT+T)+x (mT+2T)]
The QRS negative slope information that derivative filter provides is put into non-linear squares function by 3-3., and signal is put down point by point
Side emphasizes that the derivative output of high frequency carries out nonlinear amplification;
Y (mT)=[x (mT)]2
3-4. obtains wave character information by importing moving window integrator;Import the difference of moving window integrator
Equation functions are as follows:
Wherein, N is the sample size that Moving Window includes;The sample frequency of electrocardiosignal is 1000HZ, samples this quantity
About 150, the width period of corresponding window is 0.15 second;
ECG signal after the above-mentioned Integral Processing of 3-5. training obtains Initial Hurdle, constantly adjustment threshold values, to distinguish R
The position of wave or QRS complex extracts the RR spaced features of ecg information;
If the ECG signal after Integral Processing is Y (n), ECG signal is trained by following formula, is obtained just
Beginning threshold values, it may be assumed that
Wherein, using 2s as the training period of initialization threshold values, then n=2fs, fs are the sample frequency of ECG signal, THRsig
Signal amplitude Initial Hurdle, THRnoiseTo interfere Initial Hurdle.In order to enable initial threshold has more reasonability, the number of winning the confidence X (n)
Preceding 20s sample data, and be divided into 10 groups of carry out initial threshold value operations, 10 threshold values calculated according to above-mentioned formula, then
The maximum value and minimum value in 10 groups are removed, in case situations such as causing the spike noise being likely to occur to interfere causes threshold value excessive
Or too small caused error.Finally again remaining 8 values are averaged to obtain THRsigAnd THRnoiseRespectively as signal width
It is worth initial threshold and interference initial threshold.If the signal peak detected is PEAK, if PEAK > THRsig, then peak value institute is right
QRS wave of the position answered as pre-selection, and obtain an estimation signal level value LEVsig。
The estimation signal level LEV of updatesigElectricity LEV is interfered with estimationnoise, threshold values is adjusted, formula is as follows:
THRsig=α LEVsig+β·LEVnoise
Wherein, α, β are that the weighted factor of threshold values adjustment contribution takes α=0.25, β=0.75 by many experiments.?
R wave is detected in QRS complex, and the time for calculating current two neighboring R peak is the interval RR, is averaged.
Building body gait feature described in step 4 is associated with regression model with RR interphase feature, is implemented as follows:
The gait feature that step 2 is obtained splices, and the gait feature vector of 13 dimensions is formed after splicing fusion, will be sub
The gait feature vector of collection is as input sample collection, and the RR interphase feature of corresponding ecg information is as desired output sample
Collection establishes the interaction prediction model of the two, specifically such as:
The learning machine that transfinites based on core introduces regularization coefficient γ, γ a > 0, then into the learning machine that transfinites first
Core transfinites the nuclear matrix Ω in learning machineELMIt indicates are as follows: ΩELM=HHT, it may be assumed that
ΩELMi,j=h (xi)h(xj)=K (xi,xj)
Then output function f (x) can be indicated are as follows:
Establishing core transfinites learning machine model, then the Feature Mapping h (x) of hidden layer does not need to be known, instead
By calculating its corresponding core K (u, v);And core transfinites the kernel function in learning machine using radial base Gaussian kernel;
Gait feature vector is normalized;It is normalized to [- 1,1], normalized function expression formula are as follows:
Wherein, i=1,2 ..., N, N indicate that sample size, j=1,2 ..., p, p indicate gait feature number.
Step 5 is implemented as follows:
Training sample set and test sample collection accounting are 4:1, respectively by the gait feature vector of training sample set and electrocardio
It is trained in the KELM network model of both RR interphase feature inputs of information, and determines that core transfinites and learn machine neural network
Then parameter needed for model is associated forecast analysis using remaining feature as test set;
Come the accuracy and validity of valuation prediction models by using root-mean-square error, RMSE is smaller, prediction result
Better;Expression formula is as follows:
Wherein, n indicates the number of prediction data, yiIt is actual prediction data value, tiRepresent desired output.
The present invention has the beneficial effect that:
1, the present invention is proposed a kind of non-based on gait by three-dimensional motion tracking system synchronous acquisition gait and ECG signal
The pre- measuring/correlation method in Cardiac RR interval of linear character is carried out the prediction association analysis of the two by intelligent algorithm, is easier to it
Realization, prediction are more accurate.Find the fuzzy entropy feature of Nonlinear Dynamics extraction to the interval RR by experimental result
Prediction has preferable effect, the feature newly extracted is merged with traditional characteristic, the effect of prediction is more preferable.
2, compared with traditional gait feature abstracting method, it is dynamic that the method for the present invention is directed to the three-dimensional under different motion mode
Make data and carry out nonlinear dynamic analysis, is easier to find the gait of different motion mode difference in nonlinear kinetics, with
And the similitude between same movement mode, the dynamic nature that the more lively gait for depicting different motion mode contains are special
Property.
3, machine learning algorithm is applied to body gait prediction Cardiac RR interval by the present invention, establishes human body different motion
The core of mode transfinites learning machine prediction model, and discovery core learning machine (KELM) algorithm that transfinites exists compared to traditional neural network
The advantages such as its accuracy rate, rapidity and Generalization Capability are more prominent, and the effect of prediction is preferable.
Detailed description of the invention
Fig. 1 is that the body gait of the embodiment of the present invention predicts the overall flow figure of ecg information;
Fig. 2 is the ECG signal processing flow chart of the embodiment of the present invention;
Fig. 3 is the Nonlinear Dynamical Characteristics distribution map of the embodiment of the present invention;
Fig. 4 is the fuzzy entropy feature box traction substation of five kinds of motor patterns of the embodiment of the present invention;
The every kind of feature of nonlinear kinetics and fused feature prediction result statistical chart of Fig. 5 for the embodiment of the present invention;
Fig. 6 is the prediction result after five kinds of motor pattern Fusion Features.
Specific embodiment
In order to be more clear technical solution of the present invention, with reference to the accompanying drawings and examples, the present invention is made further
Detailed description.
Above-mentioned background and there are aiming at the problem that, the present invention by used in laboratory velocity interval 8~
The enterprising walking of the Cybex 770T treadmill of 20km/h and race related experiment, are acquired the data of five kinds of moving scenes, such as
Shown in Fig. 4.Wherein, head, lower back portion (i.e. the 4th vertebra to the position of the 5th vertebra) and foot linear acceleration and angular speed
Data are acquired using Inertial Measurement Unit (IMUs).Including three axis accelerometer, gyroscope and magnetometer, and using fixed
The adhesive tape of system is fixed, and is connect by WiFi with computer.Then pass through the infrared three-dimensional real time motion capture system of Codamotion
It is tracked, uses the body kinematics position under the sample frequency of gait analysis software setting 100Hz.By being placed on experiment
4 Coda CX1 units in the working space of room are detected to cover range of operation.We place 24 with measured
Marker label, wherein 4 markers are located at head (preceding forehead, left and right, rear frontal lobe), 4 positioned at upper limb (left and right acromion,
Seventh cervical spine, the 5th pelvis rear), 24 label clusters are horizontally mounted in left and right leg outer side shin bone, in addition 8 are located on foot
(left and right with elbow lever, left and right followed by low-level, left elbow and right elbow, the 5th left and right toe).It acquires walking and runs different fortune
The three-dimensional motion data of the dynamic model formula lower part of the body 24 markers of body region, for portraying the signal of body gait.TrignoTMWirelessly
System records electromyogram (EMG) and electrocardiogram (ECG) activity, acquire 10 leg muscles surface electromyogram signal and 1 it is logical
Road electrocardiogram, and it is synchronous with the physical feeling kinematics coordinate data of 100Hz.
The present invention carries out the association analysis of the prediction of Cardiac RR interval and the two using gait signal, by proposing
A kind of pre- measuring/correlation method in Cardiac RR interval based on gait nonlinear characteristic, is analyzed and processed data collected, mentions
Take a variety of gait features.And merge feature, it is pre- by establishing fused gait feature and being associated with for Cardiac RR interval
Analysis model is surveyed, we are preferable by the resulting prediction effect of this method.Gait cannot be portrayed by overcoming traditional analysis
The shortcomings that dynamic nature characteristic contained, and there is good prediction effect.Prevent arrhythmia cordis and trouble in human motion
Person, which carries out athletic rehabilitation treatment etc., has good monitoring effect, can remind people at highest risk and its danger to a certain extent
Dangerous degree.
As depicted in figs. 1 and 2, the overall flow figure at Cardiac RR interval, packet are predicted for the body gait of the embodiment of the present invention
Include following steps:
The acquisition of the three-dimensional motion data of step 1. human motion and corresponding electrocardiosignal;It is caught by three-dimensional real time kinematics
The three-dimensional coordinate of system acquisition physical feeling, and the corresponding electrocardiogram of synchronous acquisition (ECG) activity are caught, different motion mould is established
The sample database of formula.
By using a velocity interval in the enterprising walking of Cybex 770T treadmill of 8~20km/h in laboratory
With race related experiment, the data of five kinds of moving scenes are acquired.Pass through the infrared three-dimensional real time motion capture system of Codamotion
System is tracked, and the movement three-dimensional coordinate number of body position is captured under the sample frequency using gait analysis software setting 100Hz
Accordingly and TrignoTMWireless system records electromyogram (EEG) and electrocardiogram (ECG) activity, acquires the table of 10 leg muscles
Facial muscle electric signal and 1 channel electrocardiogram, and it is synchronous with the physical feeling kinematics coordinate data of 100Hz.Wherein, electrocardio is believed
Number sample frequency be 1000Hz.
Wherein, 17 healthy volunteer volunteers participate in data record, and all participants are not diagnosed any disease
Disease, and be normal type (170 ± 7.6cm, 64 ± 11.5kg), and be all keen on sports.Their body-mass index
(22.3±2kg/m2) meet " body mass index (BMI) classification " in " normal " range and global body mass index database (WHO,
2006).All subjects use same equipment, identical label sets and same experiment condition.Experiment scene setting
Include: that (1) comfortably walks: participant improves treadmill speed, is denoted as V1 until reaching the comfortable speed of travel, and with this speed
Spend walking 2min;(2) walking: participant increases or decreases treadmill speed, until 4km/h, and with the walking of this speed
2min remembers that the speed is V2;(3) maximum speed is walked: participant improves speed and is denoted as until its maximum walking speed
V3, and with this speed walking 1min, then participant rests;(4) after resting, the speed of treadmill is increased to by participant
V3, walking 30s, then runs 1min at a same speed;(5) run: participant improves speed, until he run the limit,
The speed of service is denoted as V4.Data record continues two minutes.
The pretreatment and feature extraction of step 2. gait data;Gait data is segmented with every ten gait cycles
Processing, the method that quantitative analysis and nonlinear dynamic analysis is then respectively adopted extracts the various features of gait, such as Fig. 3 institute
Show;
2.1 extract gait cycle;
The three-dimensional motion data that noise is removed as obtained by step 1, select the mark point at human body lower limbs position to be divided
Analysis.All wave crests are found, are a gait cycle by each adjacent two wave crest.
Data are divided into multistage using every ten gait cycles as a subset by 2.2;
2.3 are respectively adopted quantitative analysis, the method for nonlinear dynamic analysis extracts the more characteristic parameters of gait.Specifically
It is as follows:
(2) using quantitative analysis method calculate gait space-time and angle character parameter, including gait cycle, step-length,
Stride, step height, step width, leg speed, leg speed acceleration and sufficient drift angle, and its mean value is taken by each section.
Gait cycle (GC) refers to same foot (from right crus of diaphragm to right crus of diaphragm, or from left foot to left foot) from initial foot over the ground
Touch the time interval between two that foot next time contacts over the ground continuous moments.Step-length (SL) is defined as left and right heel
Continuously contact with the vertical linear distance between ground;Stride (STL) refers to the left step-length and the sum of right step-length of a gait cycle;
Walk the maximum height of the heel lift during high (SH) is gait;It is laterally straight between two heels when step width (WB) refers to people's walking
Linear distance;Leg speed (SV) refers to the average speed of human locomotion;It is calculated according to stride (STL) and gait cycle (GC), leg speed
Formula is as follows:
SV=S/GC=S/ (n/fs)
Wherein, the sampling number that n includes by a gait cycle, fsFor the sampling frequency of acquisition kinematic data setting
Rate, i.e. fs=100Hz.S is distance length, and GC is gait cycle.
Step acceleration (SACC) is the mean change amount of heel forward speed in the unit time, it is used to describe human body row
The physical quantity of velocity variations speed is walked, calculation formula is as follows:
ACC=SV/GC=S/GC2
Sufficient drift angle (FA) is angle formed by the line and direction of advance of heel and toe.The present invention extracts left and right respectively
Maximum value of the toe foot drift angle in each gait cycle is denoted as MaxFA, and minimum value remembers MinFA, then takes it for each section
Value.
(3) fuzzy entropy, approximate entropy, sample of every ten gait cycles are extracted using the method for nonlinear dynamic analysis
Entropy, LZ complexity, C0 complexity characteristics;It specifically includes following:
What fuzzy entropy (FuzzyEn) was measured is the probability size that new model generates, and measure value is bigger, what new model generated
Probability is bigger, i.e., sequence complexity is higher, indicates that time series has more randomness.Specific algorithm is described as follows:
A. for given N-dimensional time series [u (1), u (2) ..., u (N)];
B. phase space dimension m (m≤N-2) and similar content r are defined, phase space reconstruction:
X (i)=[u (i), u (i+1) ..., u (i+m-1)]-u0(i), i=1,2 ..., N-m+1
Wherein, parameter N=1,2,3,4 ..., N,
C. fuzzy membership functions is introduced:
Wherein, m=2, r=0.2*SD (standard deviation that SD is original series).For i=1,2 ..., N-m+1, calculate:
Wherein, For window
Maximum absolute distance between vector X (i) and X (j).
D. it is directed to each i, is averaged, obtains:
The then fuzzy entropy (FuzzyEn) of former gait time sequence are as follows:
Wherein,
E. the fuzzy entropy (FuzzyEn) of former time series are as follows:
FuzzyEn (m, r)=lim [ln Φm(r)-lnΦm+1(r)]
For finite data collection, fuzzy entropy estimation are as follows:
FuzzyEn (m, r, N)=ln Φm(r)-lnΦm+1(r)
Approximate entropy (ApEn) is the Nonlinear Dynamic of a kind of regularity for the fluctuation of quantization time sequence and unpredictability
Mechanics parameter reflects a possibility that new information occurs in time series, and the bigger entropy the more complicated.Specific algorithm is described as follows:
A. for given N-dimensional time series [u (1), u (2) ..., u (N)];
B. algorithm relevant parameter m=2, r=0.2*SD are defined;
C. m dimensional vector X (1) is reconstructed, X (2) ..., X (N-m+1), wherein
X (i)=[u (i), u (i+1) ..., u (i+m-1)]
D. for 1≤i≤N-m+1, statistics meets the vector number of the following conditions
Wherein,U (a) is the element of vector X, d be vector X (i) and X (j) away from
From being determined by the maximum difference of corresponding element, the value range of j is [1, N-m+1], including j=i.
E. it defines
F. then approximate entropy (ApEn) is defined as
ApEn=Φm(r)-Φm+1(r)
Sample Entropy (SampEn) is a kind of improvement side for measuring period sequence complexity based on approximate entropy (ApEn)
Method is a kind of new time series Complexity Measurement method.The specific algorithm of Sample Entropy is as follows:
Previous section is consistent with Sample Entropy;
C. for 1≤i≤N-m+1, statistics meets the vector number of the following conditions
Wherein, i ≠ j, d [X, X*] is defined as:
D. it asksTo the average value of all values, B is rememberedm(r) it is, i.e.,
E. k=m+1 is enabled, step c is repeated, obtains
Wherein,
F. then Sample Entropy (SampEn) defines: SampEn=-ln [Ak(r)/Bm(r)]
Lempel-Ziv complexity (LZ) reflects a time series and the speed of new model occurs with the growth of its length
Degree.Complexity value is bigger, and the new change for illustrating that data occur at any time within length of window period is more, and new change occurs
Rate is faster, shows that the data variation in this period is unordered and complicated;Conversely, complexity is smaller, then illustrate to occur newly to become
The rate of change is slower, data variation be it is regular, periodically it is stronger.Specific algorithm is as follows:
A. using binarization method to sequence X={ X1,X2,…,XnCoarse processing is carried out, form " 0-1 " sequence P=
{S1,S2,…,Sn}.
B. to obtained " 0-1 " sequence above, new character strings therein are successively retrieved.New character strings need to meet uniquely
Property and continuity, and will be separated with one " ", wherein the retrieving of new character strings is as follows in detail:
The partial character string of sequence P, i.e. S={ S are indicated with S1,S2,…,Sr}(r≤n);The substring of S, i.e. Q are indicated with Q
=| Q1,Q2,…,Qm|;Indicate that the tandem compound of S, Q, i.e. SQ={ S, Q } are indicated to leave out the last character of SQ with SQ π with SQ
Accord with resulting character string;Substring set obtained in SQ π is indicated with V (SQ π);The number of different substrings is indicated with d (n).For
Given sequence, P={ S1,S2,…,Sn, when beginning, take S=S1, Q=S2, SQ π=S1, d (n)=1.Under normal circumstances, S is enabled
={ S1,S2,…,Sr(r=2,3 ..., n-1);So SQ π={ S1,S2,…,Sr(r=2,3 ..., n-1), Q=Sr+1, sentence
Disconnected Q whether a substring for being S, if Q belongs to V (SQ π), Q is a substring of SQ π rather than new substring, S are remained unchanged,
Enable Q=Sr+1Sr+2, only until Q is not belonging to V (SQ π), that is, have Q={ Sr+1,Sr+2,…,Sr+iIt is not belonging to SQ π={ S1,
S2,…,Sr+i-1, d (n) plus 1, once repeats the above process a to the last character at this time.One is calculated with " "
It is divided into the number of the character string of section, defines complexity d (n).
C. complexity is normalized;In order to obtain the complexity independent of sequence length, need d (n) normalizing
Change.It was found that all symbol sebolic addressing complexity d (n) tend to a determining value, it may be assumed that
Wherein n is sequence length, and a is the number (" 0-1 " sequence a=2) of kinds of characters in character string, by b (n) to d
(n) it is normalized, show that normalized LZ complexity is as follows:
C0 product complexity theory thought is mainly that the gait time argument sequence that will be analyzed is decomposed, and is divided into stochastic ordering
The part and sequence of rules part of column, when being defined as doing the gait of random motion part to the complexity prediction of sequence
Between argument sequence length and area ratio that all original gait time argument sequence length surrounds between time shaft respectively.
This method is specific as follows:
A. gait time argument sequence that a given length is N { x (t), t=0,1,2 ..., N-1 } is carried out
Discrete Fourier transform:
I is imaginary unit in formula,
B. the mean-square value of f (k) is found out:
C. a normal number δ greater than 1 is defined, the frequency spectrum in f (k) sequence being more than δ times of mean-square value is retained, instead
Then its zero setting is handled, the sequence after being converted:
D. carrying out inverse Fourier transform to above formula can obtain:
WhereinIt is defined according to C0 product complexity theory, C0 complicated dynamic behaviour value can be obtained:
The pretreatment of step 3.ECG signal extracts the interval RR of ECG signal corresponding with each gait subset, specifically
Steps are as follows:
3-1. inhibition baseline drift and T wave interfere, and pass through the band logical of cascade low pass and high-pass filter composition
Filter handle to ECG signal;The difference equation of the low-pass filter of second order in bandpass filter are as follows:
Y (nT)=2y (nT-T)-y (nT-2T)+x (nT) -2x (nT-6T)+x (nT-12T)
Wherein, T is the sampling period, and x (nT) indicates the input of discrete-time system electrocardiosignal time series, and y (nT) is
Electrocardiosignal after low-pass filtering, the cutoff frequency of low-pass filter are set as 12Hz, gain 2, and processing delay is about adopted for 6
The sample period.The cutoff frequency of high-pass filter is about 5Hz, the delay in 32,16 sampling periods of gain.Its difference equation is such as
Under:
Y (mT)=32x (mT-16T)-[y (mT-T)+x (mT)-x (mT-32T)]
3-2. uses derivative filter to provide QRS negative slope information by bandpass filter treated ECG signal;It leads
The difference equation of wavenumber filter is as follows:
Y (mT)=(1/8T) [- x (mT-2T) -2x (mT-T)+2x (mT+T)+x (mT+2T)]
The QRS negative slope information that derivative filter provides is put into non-linear squares function by 3-3., and signal is put down point by point
Side emphasizes that the derivative output of high frequency (being mainly ECG frequency) carries out nonlinear amplification;
Y (mT)=[x (mT)]2
3-4. obtains wave character information by importing moving window integrator;Import the difference of moving window integrator
Equation functions are as follows:
Wherein, N is the sample size that Moving Window includes.The sample frequency of electrocardiosignal is 1000HZ, samples this quantity
About 150, the width period of corresponding window is 0.15 second.
ECG signal after the above-mentioned Integral Processing of 3-5. training obtains Initial Hurdle, constantly adjustment threshold values, to distinguish R
The position of wave or QRS complex extracts the RR spaced features of ecg information;
If the ECG signal after Integral Processing is Y (n), ECG signal is trained by following formula, is obtained just
Beginning threshold values, it may be assumed that
Wherein, using 2s as the training period of initialization threshold values, then n=2fs, fs are the sample frequency of ECG signal, THRsig
Signal amplitude Initial Hurdle, THRnoiseTo interfere Initial Hurdle.In order to enable initial threshold has more reasonability, the number of winning the confidence X (n)
Preceding 20s sample data, and be divided into 10 groups of carry out initial threshold value operations, 10 threshold values calculated according to above-mentioned formula, then
The maximum value and minimum value in 10 groups are removed, in case situations such as causing the spike noise being likely to occur to interfere causes threshold value excessive
Or too small caused error.Finally again remaining 8 values are averaged to obtain THRsigAnd THRnoiseRespectively as signal width
It is worth initial threshold and interference initial threshold.If the signal peak detected is PEAK, if PEAK > THRsig, then peak value institute is right
QRS wave of the position answered as pre-selection, and obtain an estimation signal level value LEVsig。
The estimation signal level LEV of updatesigElectricity LEV is interfered with estimationnoise, threshold values is adjusted, formula is as follows:
THRsig=α LEVsig+β·LEVnoise
Wherein, α, β are that the weighted factor of threshold values adjustment contribution takes α=0.25, β=0.75 by many experiments.?
R wave is detected in QRS complex, and the time for calculating current two neighboring R peak is the interval RR, is averaged.
Step 4. building body gait feature is associated with regression model with RR interphase feature;
Spliced by the gait feature that step 2 obtains, the feature vector of 13 dimensions is formed after splicing fusion, by subset
Feature vector as input sample collection, the RR interphase feature of corresponding ecg information is as desired output sample set.It establishes
The interaction prediction model of the two.
The learning machine that transfinites based on core introduces a regularization coefficient γ (γ > 0) into the learning machine that transfinites first,
Core transfinites the nuclear matrix Ω in learning machineELMIt indicates are as follows: ΩELM=HHT, i.e.,
Then output function f (x) can be indicated are as follows:
Establishing core transfinites learning machine model, then the Feature Mapping h (x) of hidden layer does not need to be known, instead
By calculating its corresponding core K (u, v).Core transfinites there are many kernel functions in learning machine, such as Polynomial kernel function, linear kernel
Function, Wavelet Kernel Function, radial base Gaussian kernel etc.;The present invention is using radial base Gaussian kernel (RBF kernel).
In order to avoid in training process as attribute value is small and caused by the too small influence experimental result of contribution, by step 3 institute
The gait feature parameter obtained is normalized;It is normalized to [- 1,1], normalized function expression formula are as follows:
Wherein, i=1,2 ..., N, N indicate that sample size, j=1,2 ..., p, p indicate gait feature number.
Step 5. training and test interaction prediction model, test the prediction result after single feature and Fusion Features respectively;
Training sample set and test sample collection accounting are 4:1, respectively will be between the gait feature of training sample set and Cardiac RR
It is trained in the KELM network model of both phase feature vector inputs, and determines that core transfinites learning machine neural network model institute
The parameter needed, is then associated forecast analysis using remaining feature as test set.
Come the accuracy and validity of valuation prediction models by using root-mean-square error (RMSE), RMSE is smaller, in advance
It is better to survey result;Expression formula is as follows:
Wherein, n indicates the number of prediction data, yiIt is actual prediction data value, tiRepresent desired output.
As shown in figure 5, for single feature in KELM model measurement as a result, as seen from the figure, in addition to the gait of quantitative analysis
Period and leg speed acceleration have preferable effect, and the fuzzy entropy feature in nonlinear kinetics also has preferable prediction
Effect, the prediction effect after Fusion Features is more preferable compared to the prediction effect of single feature, embodies better prediction advantage.
As shown in fig. 6, for the prediction result after five kinds of motor pattern Fusion Features, as seen from the figure 3- quickly walk, 5- most
It is more preferable that big speed runs prediction effect, shows as that speed is faster, and prediction effect is better, and the result of prediction is more accurate.
Claims (6)
1. the pre- measuring/correlation method in Cardiac RR interval based on gait nonlinear characteristic, it is characterised in that include the following steps:
The acquisition of three-dimensional gait data and corresponding electrocardiosignal when step 1. human motion;Pass through three-dimensional real time motion capture system
The three-dimensional gait data of system acquisition physical feeling, and the corresponding electrocardiogram of synchronous acquisition (ECG) activity, establish different motion mode
Sample database;
The pretreatment and feature extraction of step 2. gait data;Gait data is subjected to segment processing with every ten gait cycles,
Then the method that quantitative analysis and nonlinear dynamic analysis is respectively adopted extracts a variety of gait features;
Step 3. pre-processes electrocardiosignal (ECG signal), extracts the RR of ECG signal corresponding with each gait subset
Spaced features;
Step 4. merges body gait feature, and building gait fusion feature is associated with regression model with RR interphase feature;
Step 5. training and test interaction prediction model, test the prediction result after single feature and Fusion Features respectively.
2. the Cardiac RR interval pre- measuring/correlation method according to claim 1 based on gait nonlinear characteristic, feature exist
It is implemented as follows in step 1:
Experiment is walked and run to operating speed range in the Cybex 770T treadmill of 8~20km/h to acquire five kinds of moving scenes
Data;Wherein, the linear acceleration of head, lower back portion and foot and angular velocity data are acquired using Inertial Measurement Unit;It is logical
It crosses the infrared three-dimensional real time motion capture system of Codamotion to be tracked, uses the sampling of gait analysis software setting 100Hz
Body kinematics position under frequency;It is transported by the 4 Coda CX1 units detection being placed in laboratory work space with covering
Line range;24 markers are placed with measured, wherein 4 markers are located at head: preceding forehead, left and right, rear frontal lobe;4
Positioned at upper limb: left acromion, right acromion, seventh cervical spine, the 5th pelvis rear;24 marks are horizontally mounted in left and right leg outer side shin bone
Remember cluster;Other 8 are located on foot: left and right with elbow lever, left and right followed by low-level, left elbow and right elbow, the 5th left toe, the right side
Toe;By the three-dimensional gait data of acquisition walking and race 24 markers of different motion mode lower part of the body body region, while recording flesh
Electrograph (EMG) and electrocardiogram (ECG) activity, acquire the surface electromyogram signal and 1 channel electrocardiogram of 10 leg muscles, and with
The physical feeling kinematics coordinate data of 100Hz is synchronous.
3. the Cardiac RR interval pre- measuring/correlation method according to claim 1 or 2 based on gait nonlinear characteristic, feature
It is that step 2 is implemented as follows;
2.1 extract gait cycle;
By the resulting three-dimensional gait data de-noising of step 1, the three-dimensional gait data of the mark point acquisition at human body lower limbs position are selected,
All wave crests are found, are a gait cycle by each adjacent two wave crest;
Gait data is divided into multistage using every ten gait cycles as a subset by 2.2;
2.3 are respectively adopted quantitative analysis, nonlinear dynamic analysis method extracts a variety of gait features, specific as follows:
(1) using quantitative analysis method calculate gait space-time and angle character parameter, including gait cycle, step-length, stride,
Height, step width, leg speed, leg speed acceleration and sufficient drift angle are walked, and takes its mean value for each section;
Leg speed SV formula is as follows:
SV=S/GC=S/ (n/fs)
Wherein, the sampling number that n includes by a gait cycle, fsFor the sample frequency of acquisition kinematic data setting, i.e. fs
=100Hz;S is distance length, and GC is gait cycle;
Leg speed acceleration SACC calculation formula is as follows:
SACC=SV/GC=S/GC2
Sufficient drift angle is angle formed by the line and direction of advance of heel and toe;Left and right toe foot drift angle is extracted in each step
Maximum value in the state period is denoted as MaxFA, and minimum value remembers MinFA, then takes its mean value for each section;
(2) fuzzy entropy, approximate entropy, Sample Entropy, the LZ of every ten gait cycles are extracted using the method for nonlinear dynamic analysis
Complexity, C0 complexity characteristics.
4. the Cardiac RR interval pre- measuring/correlation method according to claim 3 based on gait nonlinear characteristic, feature exist
It is implemented as follows in step 3;
3-1. handle to ECG signal by the bandpass filter that cascade low pass and high-pass filter form;Band logical
The difference equation of the low-pass filter of second order in filter are as follows: y (nT)=2y (nT-T)-y (nT-2T)+x (nT) -2x (nT-6T)
+x(nT-12T)
Wherein, T is the sampling period, and x (nT) indicates the input of discrete-time system electrocardiosignal time series, and y (nT) is low pass
Filtered electrocardiosignal, the cutoff frequency of low-pass filter are set as 12Hz, gain 2, processing delay about 6 sampling weeks
Phase;
The cutoff frequency of high-pass filter is about 5Hz, the delay in 32,16 sampling periods of gain;Its difference equation is as follows:
Y (mT)=32x (mT-16T)-[y (mT-T)+x (mT)-x (mT-32T)]
3-2. uses derivative filter to provide QRS negative slope information by bandpass filter treated ECG signal;Derivative filter
The difference equation of wave device is as follows:
Y (mT)=(1/8T) [- x (mT-2T) -2x (mT-T)+2x (mT+T)+x (mT+2T)]
The QRS negative slope information that derivative filter provides is put into non-linear squares function by 3-3., point-by-point square of signal, is emphasized
The derivative output of high frequency carries out nonlinear amplification;
Y (mT)=[x (mT)]2
3-4. obtains wave character information by importing moving window integrator;Import the difference equation letter of moving window integrator
Number are as follows:
Wherein, N is the sample size that Moving Window includes;The sample frequency of electrocardiosignal is 1000HZ, and the quantity for sampling this is about
150, the width period of corresponding window is 0.15 second;
ECG signal after the above-mentioned Integral Processing of 3-5. training obtains Initial Hurdle, constantly adjustment threshold values, thus distinguish R wave or
The position of QRS complex extracts the RR spaced features of ecg information;
If the ECG signal after Integral Processing is Y (n), ECG signal is trained by following formula, obtains initial valve
Value, it may be assumed that
Wherein, using 2s as the training period of initialization threshold values, then n=2fs, fs are the sample frequency of ECG signal, THRsigSignal
Amplitude Initial Hurdle, THRnoiseTo interfere Initial Hurdle;In order to enable initial threshold has more reasonability, before the number of winning the confidence X (n)
The sample data of 20s, and it is divided into 10 groups of carry out initial threshold value operations, 10 threshold values are calculated according to above-mentioned formula, then remove
Maximum value and minimum value in 10 groups, in case situations such as causing the spike noise being likely to occur to interfere causes threshold value excessive or too small
Caused error;Finally again remaining 8 values are averaged to obtain THRsigAnd THRnoiseRespectively as the initial threshold of signal amplitude
Value and interference initial threshold;If the signal peak detected is PEAK, if PEAK > THRsig, then position corresponding to the peak value
As the QRS wave of pre-selection, and obtain an estimation signal level value LEVsig;
The estimation signal level LEV of updatesigElectricity LEV is interfered with estimationnoise, threshold values is adjusted, formula is as follows:
THRsig=α LEVsig+β·LEVnoise
Wherein, α, β are that the weighted factor of threshold values adjustment contribution takes α=0.25, β=0.75 by many experiments;In QRS complex
In detect R wave, and the time for calculating current two neighboring R peak is the interval RR, is averaged.
5. the Cardiac RR interval pre- measuring/correlation method according to claim 4 based on gait nonlinear characteristic, feature exist
Building body gait feature described in step 4 is associated with regression model with RR interphase feature, is implemented as follows:
The gait feature that step 2 is obtained splices, and the gait feature vector of 13 dimensions is formed after splicing fusion, by the step of subset
State feature vector is established as input sample collection, the RR interphase feature of corresponding ecg information as desired output sample set
The interaction prediction model of the two, specifically such as:
The learning machine that transfinites based on core introduces regularization coefficient γ, γ a > 0, then into the learning machine that transfinites first
Core transfinites the nuclear matrix Ω in learning machineELMIt indicates are as follows: ΩELM=HHT, it may be assumed that
Then output function f (x) can be indicated are as follows:
Establishing core transfinites learning machine model, then the Feature Mapping h (x) of hidden layer does not need to be known, instead passes through
Calculate its corresponding core K (u, v);And core transfinites the kernel function in learning machine using radial base Gaussian kernel;
Gait feature vector is normalized;It is normalized to [- 1,1], normalized function expression formula are as follows:
Wherein, i=1,2 ..., N, N indicate that sample size, j=1,2 ..., p, p indicate gait feature number.
6. the Cardiac RR interval pre- measuring/correlation method according to claim 5 based on gait nonlinear characteristic, feature exist
It is implemented as follows in step 5:
Training sample set and test sample collection accounting are 4:1, respectively by the gait feature vector of training sample set and ecg information
Both RR interphase feature inputs KELM network model in be trained, and determine that core transfinites learning machine neural network model institute
The parameter needed, is then associated forecast analysis using remaining feature as test set;
Come the accuracy and validity of valuation prediction models by using root-mean-square error, RMSE is smaller, and prediction result is better;
Expression formula is as follows:
Wherein, n indicates the number of prediction data, yiIt is actual prediction data value, tiRepresent desired output.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910398150.5A CN110215202A (en) | 2019-05-14 | 2019-05-14 | The pre- measuring/correlation method in Cardiac RR interval based on gait nonlinear characteristic |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910398150.5A CN110215202A (en) | 2019-05-14 | 2019-05-14 | The pre- measuring/correlation method in Cardiac RR interval based on gait nonlinear characteristic |
Publications (1)
Publication Number | Publication Date |
---|---|
CN110215202A true CN110215202A (en) | 2019-09-10 |
Family
ID=67821043
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201910398150.5A Withdrawn CN110215202A (en) | 2019-05-14 | 2019-05-14 | The pre- measuring/correlation method in Cardiac RR interval based on gait nonlinear characteristic |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN110215202A (en) |
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN113812964A (en) * | 2021-08-02 | 2021-12-21 | 杭州航弈生物科技有限责任公司 | Method and device for proxy measurement of electroencephalogram characteristics and pseudo-multi-mode frozen gait detection |
TWI823684B (en) * | 2021-11-30 | 2023-11-21 | 日商三菱電機股份有限公司 | Learning device, abnormal sign detection device, abnormal sign detection system, learning method and recording medium |
Citations (16)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
KR101205908B1 (en) * | 2011-05-26 | 2012-11-28 | 건국대학교 산학협력단 | Mild cognitive impairment and Alzheimer's disease evaluation system using Gait dynamics and HRV monitoring and the method thereof |
CN102824177A (en) * | 2012-07-25 | 2012-12-19 | 王哲龙 | Three-dimensional human body gait quantitative analysis system and method |
CN103549950A (en) * | 2013-11-19 | 2014-02-05 | 上海理工大学 | Improved difference threshold detection algorithm for mobile ECG (electrocardiogram) monitoring |
CN104134077A (en) * | 2014-07-10 | 2014-11-05 | 华南理工大学 | Deterministic learning theory based gait recognition method irrelevant to visual angle |
CN104720794A (en) * | 2015-04-13 | 2015-06-24 | 康泰医学系统(秦皇岛)股份有限公司 | Method for detecting electrocardiogram wave form R points through superposing |
CN105512609A (en) * | 2015-11-25 | 2016-04-20 | 北京工业大学 | Multi-mode fusion video emotion identification method based on kernel-based over-limit learning machine |
CN105740619A (en) * | 2016-01-28 | 2016-07-06 | 华南理工大学 | On-line fault diagnosis method of weighted extreme learning machine sewage treatment on the basis of kernel function |
CN105919590A (en) * | 2016-06-02 | 2016-09-07 | 浙江铭众科技有限公司 | QRS automatic delineation method for multichannel electrocardiogram |
CN106845458A (en) * | 2017-03-05 | 2017-06-13 | 北京工业大学 | A kind of rapid transit label detection method of the learning machine that transfinited based on core |
CN107041743A (en) * | 2017-04-05 | 2017-08-15 | 南京大学 | A kind of real-time R wave detecting methods of electrocardiosignal |
CN107174255A (en) * | 2017-06-15 | 2017-09-19 | 西安交通大学 | Three-dimensional gait information gathering and analysis method based on Kinect somatosensory technology |
CN107832785A (en) * | 2017-10-30 | 2018-03-23 | 天津理工大学 | A kind of non-linear limit learning machine algorithm |
CN108564043A (en) * | 2018-04-17 | 2018-09-21 | 中国民航大学 | A kind of Human bodys' response method based on time-space distribution graph |
CN109009143A (en) * | 2018-07-12 | 2018-12-18 | 杭州电子科技大学 | A method of ecg information is predicted by body gait |
CN109446884A (en) * | 2018-09-06 | 2019-03-08 | 杭州电子科技大学 | A kind of gait recognition method that the speed based on nonlinear kinetics index is unrelated |
CN109717833A (en) * | 2018-11-26 | 2019-05-07 | 中国科学院软件研究所 | A kind of neurological disease assistant diagnosis system based on human motion posture |
-
2019
- 2019-05-14 CN CN201910398150.5A patent/CN110215202A/en not_active Withdrawn
Patent Citations (16)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
KR101205908B1 (en) * | 2011-05-26 | 2012-11-28 | 건국대학교 산학협력단 | Mild cognitive impairment and Alzheimer's disease evaluation system using Gait dynamics and HRV monitoring and the method thereof |
CN102824177A (en) * | 2012-07-25 | 2012-12-19 | 王哲龙 | Three-dimensional human body gait quantitative analysis system and method |
CN103549950A (en) * | 2013-11-19 | 2014-02-05 | 上海理工大学 | Improved difference threshold detection algorithm for mobile ECG (electrocardiogram) monitoring |
CN104134077A (en) * | 2014-07-10 | 2014-11-05 | 华南理工大学 | Deterministic learning theory based gait recognition method irrelevant to visual angle |
CN104720794A (en) * | 2015-04-13 | 2015-06-24 | 康泰医学系统(秦皇岛)股份有限公司 | Method for detecting electrocardiogram wave form R points through superposing |
CN105512609A (en) * | 2015-11-25 | 2016-04-20 | 北京工业大学 | Multi-mode fusion video emotion identification method based on kernel-based over-limit learning machine |
CN105740619A (en) * | 2016-01-28 | 2016-07-06 | 华南理工大学 | On-line fault diagnosis method of weighted extreme learning machine sewage treatment on the basis of kernel function |
CN105919590A (en) * | 2016-06-02 | 2016-09-07 | 浙江铭众科技有限公司 | QRS automatic delineation method for multichannel electrocardiogram |
CN106845458A (en) * | 2017-03-05 | 2017-06-13 | 北京工业大学 | A kind of rapid transit label detection method of the learning machine that transfinited based on core |
CN107041743A (en) * | 2017-04-05 | 2017-08-15 | 南京大学 | A kind of real-time R wave detecting methods of electrocardiosignal |
CN107174255A (en) * | 2017-06-15 | 2017-09-19 | 西安交通大学 | Three-dimensional gait information gathering and analysis method based on Kinect somatosensory technology |
CN107832785A (en) * | 2017-10-30 | 2018-03-23 | 天津理工大学 | A kind of non-linear limit learning machine algorithm |
CN108564043A (en) * | 2018-04-17 | 2018-09-21 | 中国民航大学 | A kind of Human bodys' response method based on time-space distribution graph |
CN109009143A (en) * | 2018-07-12 | 2018-12-18 | 杭州电子科技大学 | A method of ecg information is predicted by body gait |
CN109446884A (en) * | 2018-09-06 | 2019-03-08 | 杭州电子科技大学 | A kind of gait recognition method that the speed based on nonlinear kinetics index is unrelated |
CN109717833A (en) * | 2018-11-26 | 2019-05-07 | 中国科学院软件研究所 | A kind of neurological disease assistant diagnosis system based on human motion posture |
Cited By (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN113812964A (en) * | 2021-08-02 | 2021-12-21 | 杭州航弈生物科技有限责任公司 | Method and device for proxy measurement of electroencephalogram characteristics and pseudo-multi-mode frozen gait detection |
CN113812964B (en) * | 2021-08-02 | 2023-08-04 | 杭州航弈生物科技有限责任公司 | Proxy measurement and pseudo-multimode frozen gait detection method and device for electroencephalogram characteristics |
TWI823684B (en) * | 2021-11-30 | 2023-11-21 | 日商三菱電機股份有限公司 | Learning device, abnormal sign detection device, abnormal sign detection system, learning method and recording medium |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN106971059B (en) | Wearable equipment based on neural network self-adaptation health monitoring | |
Janssen et al. | Recognition of emotions in gait patterns by means of artificial neural nets | |
Burns et al. | Upper limb movement classification via electromyographic signals and an enhanced probabilistic network | |
CN110236523A (en) | Gait-Cardiac RR interval the correlating method returned based on Gauss | |
CN111027417B (en) | Gait recognition method and gait evaluation system based on human body key point detection algorithm | |
CN103519819A (en) | Gait analysis method and gait analysis system | |
CN110110633B (en) | Method for automatically identifying and analyzing hemiplegic gait based on machine learning | |
CN109446972B (en) | Gait recognition model establishing method, recognition method and device based on electromyographic signals | |
Lu et al. | Evaluation of classification performance in human lower limb jump phases of signal correlation information and LSTM models | |
CN109567812B (en) | Gait analysis system based on intelligent insole | |
CN109805935A (en) | A kind of intelligent waistband based on artificial intelligence hierarchical layered motion recognition method | |
CN106725376A (en) | Sign detection method and device | |
CN112263244A (en) | Gait-based fatigue degree evaluation system and method | |
CN110215202A (en) | The pre- measuring/correlation method in Cardiac RR interval based on gait nonlinear characteristic | |
CN113066260A (en) | Parkinson's disease early warning system based on daily behavior analysis | |
Chuang et al. | A wearable activity sensor system and its physical activity classification scheme | |
Aubol et al. | Tibial acceleration reliability and minimal detectable difference during overground and treadmill running | |
CN113456060B (en) | Extraction device for motion function characteristic parameters | |
Alcaraz et al. | Monitoring the rehabilitation progress using a DCNN and kinematic data for digital healthcare | |
CN113768471A (en) | Parkinson disease auxiliary diagnosis system based on gait analysis | |
Perez et al. | A smartphone-based system for clinical gait assessment | |
CN108325201A (en) | A kind of cadence detection method and equipment based on Multi-source Information Fusion | |
Chee et al. | Gait acceleration-based diabetes detection using hybrid deep learning | |
CN111178338A (en) | Method for establishing database and standardized model in gait analysis system | |
CN111436940A (en) | Gait health assessment method and device |
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 | ||
WW01 | Invention patent application withdrawn after publication | ||
WW01 | Invention patent application withdrawn after publication |
Application publication date: 20190910 |