CN111067503A - Sleep staging method based on heart rate variability - Google Patents
Sleep staging method based on heart rate variability Download PDFInfo
- Publication number
- CN111067503A CN111067503A CN201911412135.8A CN201911412135A CN111067503A CN 111067503 A CN111067503 A CN 111067503A CN 201911412135 A CN201911412135 A CN 201911412135A CN 111067503 A CN111067503 A CN 111067503A
- Authority
- CN
- China
- Prior art keywords
- sleep
- stage
- transition
- pure
- staging
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 230000007958 sleep Effects 0.000 title claims abstract description 191
- 238000000034 method Methods 0.000 title claims abstract description 71
- 230000008667 sleep stage Effects 0.000 claims abstract description 77
- 238000000605 extraction Methods 0.000 claims abstract description 13
- 230000007704 transition Effects 0.000 claims description 88
- 206010062519 Poor quality sleep Diseases 0.000 claims description 76
- 238000006243 chemical reaction Methods 0.000 claims description 47
- 230000004424 eye movement Effects 0.000 claims description 39
- 230000008569 process Effects 0.000 claims description 30
- 230000016507 interphase Effects 0.000 claims description 19
- 238000004422 calculation algorithm Methods 0.000 claims description 14
- 238000001514 detection method Methods 0.000 claims description 14
- 239000011159 matrix material Substances 0.000 claims description 11
- 238000012549 training Methods 0.000 claims description 9
- 238000009499 grossing Methods 0.000 claims description 7
- 230000004461 rapid eye movement Effects 0.000 claims description 6
- 230000006870 function Effects 0.000 claims description 5
- 230000010354 integration Effects 0.000 claims description 4
- 238000005259 measurement Methods 0.000 claims description 4
- 230000033764 rhythmic process Effects 0.000 claims description 4
- 238000005314 correlation function Methods 0.000 claims description 3
- 238000010801 machine learning Methods 0.000 claims description 3
- 238000012706 support-vector machine Methods 0.000 claims description 3
- 230000002618 waking effect Effects 0.000 claims description 3
- 230000000694 effects Effects 0.000 abstract description 3
- 238000007781 pre-processing Methods 0.000 abstract description 3
- 230000008901 benefit Effects 0.000 abstract description 2
- 230000008859 change Effects 0.000 description 9
- 230000037007 arousal Effects 0.000 description 8
- 238000011160 research Methods 0.000 description 7
- 230000029058 respiratory gaseous exchange Effects 0.000 description 6
- 230000033228 biological regulation Effects 0.000 description 4
- 238000004364 calculation method Methods 0.000 description 4
- 238000010586 diagram Methods 0.000 description 4
- 210000005036 nerve Anatomy 0.000 description 4
- 230000002889 sympathetic effect Effects 0.000 description 4
- 238000001228 spectrum Methods 0.000 description 3
- QVGXLLKOCUKJST-UHFFFAOYSA-N atomic oxygen Chemical compound [O] QVGXLLKOCUKJST-UHFFFAOYSA-N 0.000 description 2
- 230000006399 behavior Effects 0.000 description 2
- 239000008280 blood Substances 0.000 description 2
- 210000004369 blood Anatomy 0.000 description 2
- 230000036760 body temperature Effects 0.000 description 2
- 238000000354 decomposition reaction Methods 0.000 description 2
- 230000007547 defect Effects 0.000 description 2
- 201000010099 disease Diseases 0.000 description 2
- 208000037265 diseases, disorders, signs and symptoms Diseases 0.000 description 2
- 239000012634 fragment Substances 0.000 description 2
- 230000004927 fusion Effects 0.000 description 2
- 230000036541 health Effects 0.000 description 2
- 230000009063 long-term regulation Effects 0.000 description 2
- 230000007246 mechanism Effects 0.000 description 2
- 238000012806 monitoring device Methods 0.000 description 2
- 229910052760 oxygen Inorganic materials 0.000 description 2
- 239000001301 oxygen Substances 0.000 description 2
- 230000001734 parasympathetic effect Effects 0.000 description 2
- 210000005037 parasympathetic nerve Anatomy 0.000 description 2
- 230000000241 respiratory effect Effects 0.000 description 2
- 230000011218 segmentation Effects 0.000 description 2
- 208000019116 sleep disease Diseases 0.000 description 2
- 230000004617 sleep duration Effects 0.000 description 2
- 230000003860 sleep quality Effects 0.000 description 2
- 230000003595 spectral effect Effects 0.000 description 2
- 210000001186 vagus nerve Anatomy 0.000 description 2
- 239000000654 additive Substances 0.000 description 1
- 230000000996 additive effect Effects 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000000747 cardiac effect Effects 0.000 description 1
- 230000010351 cardiac pulsation Effects 0.000 description 1
- 238000010924 continuous production Methods 0.000 description 1
- 238000012937 correction Methods 0.000 description 1
- 230000003111 delayed effect Effects 0.000 description 1
- 238000003745 diagnosis Methods 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 230000006996 mental state Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000012544 monitoring process Methods 0.000 description 1
- 230000008520 organization Effects 0.000 description 1
- 238000013139 quantization Methods 0.000 description 1
- 208000020685 sleep-wake disease Diseases 0.000 description 1
- 238000007619 statistical method Methods 0.000 description 1
- 238000012360 testing method Methods 0.000 description 1
Images
Classifications
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
- A61B5/00—Measuring for diagnostic purposes; Identification of persons
- A61B5/02—Detecting, measuring or recording pulse, heart rate, blood pressure or blood flow; Combined pulse/heart-rate/blood pressure determination; Evaluating a cardiovascular condition not otherwise provided for, e.g. using combinations of techniques provided for in this group with electrocardiography or electroauscultation; Heart catheters for measuring blood pressure
- A61B5/024—Detecting, measuring or recording pulse rate or heart rate
- A61B5/02405—Determining heart rate variability
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
- A61B5/00—Measuring for diagnostic purposes; Identification of persons
- A61B5/48—Other medical applications
- A61B5/4806—Sleep evaluation
- A61B5/4809—Sleep detection, i.e. determining whether a subject is asleep or not
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
- A61B5/00—Measuring for diagnostic purposes; Identification of persons
- A61B5/48—Other medical applications
- A61B5/4806—Sleep evaluation
- A61B5/4812—Detecting sleep stages or cycles
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
- A61B5/00—Measuring for diagnostic purposes; Identification of persons
- A61B5/48—Other medical applications
- A61B5/4806—Sleep evaluation
- A61B5/4815—Sleep quality
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
- A61B5/00—Measuring for diagnostic purposes; Identification of persons
- A61B5/72—Signal processing specially adapted for physiological signals or for diagnostic purposes
- A61B5/7203—Signal processing specially adapted for physiological signals or for diagnostic purposes for noise prevention, reduction or removal
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
- A61B5/00—Measuring for diagnostic purposes; Identification of persons
- A61B5/72—Signal processing specially adapted for physiological signals or for diagnostic purposes
- A61B5/7235—Details of waveform analysis
- A61B5/7264—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
Abstract
The invention discloses a sleep staging method based on heart rate variability, which comprises the following steps: performing smooth preprocessing on the sleep stage signal; extracting RR intervals from the smoothed ECG signal; performing heart rate variability feature extraction by using RR intervals; the method has the advantages that the proper heart rate variability characteristics are selected to train the sleep staging model, the trained model is utilized to complete sleep staging, the noise signals are smoothed, original information is better kept, the sleep stages are more accurately divided, classification is more refined, and the sleep staging effect is better.
Description
Technical Field
The invention belongs to the field of intelligent health assistance, and particularly relates to a sleep staging method based on heart rate variability.
Background
According to the survey of the world health organization, people of about 1/3 people worldwide suffer from sleep disorder, and the disease proportion of various sleep disorders in China is as high as 38.2%; the sleep quality index of a human body is not only related to the sleep duration but also related to the sleep depth, and when people wake up from different sleep stages, different influences are generated on the mental state of the people, in medical science, a user needs to carry out a complete sleep monitoring, a professional PSG monitoring device needs to be worn in a hospital, so that not only is the cost high and the operation difficult, but also a plurality of patients can change the sleep behavior along with the change of the sleep environment, so that the measurement result in the hospital is different from the usual sleep behavior, the analysis result is inaccurate, the doctor's condition analysis is misled, and the diagnosis and treatment are delayed;
sleep staging is an important basis for evaluating sleep quality, but the existing sleep staging methods mostly use electroencephalogram signals as a main part, the most widely used method is to collect electroencephalogram, electrocardio, myoelectricity and other signals by utilizing a sleep multi-lead map, and a sleep expert completes staging; on one hand, the equipment is complex and needs professionals to be carried out in a specific sleep test place, on the other hand, a plurality of sensors need to be pasted on a subject, so that the equipment causes serious intrusion to natural sleep, and the equipment is expensive, complex to operate, not suitable for family use and incapable of being popularized to the public;
the novel physiological signals, such as respiration characteristics, heart rate variability characteristics, body movement and other easily acquired physiological characteristics are used for sleep staging, the non-contact mode is more easily accepted by the public, relevant research is already carried out in relevant documents, HRV and respiration signals are acquired from ECG signals, the respiration and the heart rate are used for identifying sleep stages, the HRV is obtained by comprehensively using a body movement recorder and ECG data for sleep staging, and the research firstly uses the body movement recorder to analyze whether an experimenter is in an awake state or a sleep state; then, analyzing whether the sleep stage is in a light sleep stage or a deep sleep stage by using the time domain and frequency domain characteristics of the HRV; there is also proposed in the literature a sleep staging method taking into account individual characteristics, using physiological signals including heart rate, respiration rate, body movement and blood oxygen saturation, wherein the respiration rate and body movement are measured using a micromotion sensitive mattress, and the heart rate and blood oxygen saturation are obtained by an oximeter;
the sleep stage of multi-parameter information fusion is characterized in that under the condition of not using electroencephalogram, basic physiological parameters such as cardiac cycle, respiration, body movement and the like which are easily obtained are utilized, rules and information related to the sleep process and changes of the sleep process are extracted, a knowledge rule base is established, multi-parameter sleep information fusion calculation is carried out by adopting an uncertain reasoning evidence theory, and the sleep structure stage is realized.
Disclosure of Invention
In order to overcome the defects of the prior art, the invention aims to provide a sleep staging method based on heart rate variability, which performs smooth preprocessing on a sleep stage signal; extracting RR intervals from the smoothed ECG signal; performing heart rate variability feature extraction by using RR intervals; the sleep staging model is trained by selecting proper heart rate variability characteristics, the trained model is utilized to complete sleep staging, noise signals are smoothly removed, original information is better kept, sleep stages are divided more accurately, classification is more detailed, and the sleep staging effect is better.
In order to achieve the purpose, the technical scheme of the invention is realized as follows:
a sleep staging method based on HRV specifically comprises the following steps:
the method comprises the following steps: smoothing the sleep stage signal to smooth out the stage transition which is impossible to occur during sleep and storing the original sleep stage information;
step two: extracting RR intervals from the ECG signal subjected to the smoothing treatment in the step one;
step three: performing heart rate variability feature extraction by using the RR intervals extracted in the second step;
step four: and (4) training a sleep staging model through the features extracted in the step three to finish sleep staging.
Further, the sleep stages of step one include 4 pure sleep stages, namely, light sleep L, deep sleep D, wakefulness W, and rapid eye movement period R; the sleep stages further include 12 stage transitions, namely, wakefulness transition to light sleep WL, wakefulness transition to deep sleep WD, wakefulness transition to fast eye movement period WR, light sleep transition to wakefulness LW, light sleep transition to deep sleep LD, light sleep transition to fast eye movement period LR, deep sleep transition to wakefulness DW, deep sleep transition to light sleep DL, deep sleep transition to fast eye movement period DR, fast eye movement period transition to wakefulness RW, fast eye movement period transition to light sleep RL, and fast eye movement period transition to deep sleep RD.
Further, the specific process of extracting RR intervals from the ECG signal in step two is as follows:
s1, removing noise in an ECG signal by using wavelet transform;
s2, detecting R wave peak points based on a sliding window to obtain an R wave horizontal coordinate;
s3, subtracting the coordinate of the former peak point from the coordinate of the next continuous peak point of the obtained abscissa of the peak point to obtain an RR interval;
s4, performing error detection and omission detection of RR intervals by using a criterion based on 3 sigma.
Further, the heart rate variability feature extraction on the RR intervals in the step three includes extraction of three-dimensional features of a time domain, a frequency domain and a nonlinear domain.
Further, the specific step of selecting a suitable heart rate variability feature in step four includes:
s1, dividing a sleep process into a sleep transition stage and a pure sleep stage, and classifying the sleep transition stage and the pure sleep stage;
s2, classifying sleep transition stages in sections: the last pure sleep T of the sleep transition stagelatterAnd previous pure sleep TfommerThe characteristic ratio of (a) is used as a new characteristic of the sleep transition stage and the pure sleep stage classification;
s3, carrying out section classification on the pure sleep stages: taking the previous sleep stage and the night-long sleep cycle of the sleep stage as classification characteristics of pure sleep;
and S4, aiming at the time domain, the frequency domain and the nonlinear domain characteristics of the heart rate in the sleep conversion stage and the pure sleep stage, selecting the characteristics by using a maximum correlation minimum redundancy algorithm.
Further, the specific process of using the maximum correlation minimum redundancy algorithm to perform the feature selection in S4 is as follows:
(1) taking the time domain characteristic of the heart rate in the stage as a time dimension, the frequency domain characteristic as an energy dimension and the nonlinear characteristic as complexity measurement;
(2) given two random variables x and y, their probability density functions (corresponding to continuous variables) are p (x), p (y), p (x, y), then the mutual information is:
I(x,y)＝∫∫p(x,y)log[p(x,y)/p(x)p(y)]dxdy
the goal is to find a solution containing m { x }iA feature subset S of the features, if a discrete variable, the maximum correlation is calculated as:
maxD(S,c),D＝(1/|S|)∑xi∈SI(xi,c)
wherein: x is the number ofiIs the ith feature, c is a category variable, and S is a feature subset;
the minimum redundancy is then:
minR(S,c),R＝(1/|S|2)∑xi,xj∈SI(xi；xj)
if it is a continuous variable, the maximum correlation:
maxDF,DF＝(1/|S|)∑xi∈SF(xi,c)
wherein: dFIs a correlation, (x)iAnd c) is the statistic of F;
the minimum redundancy is:
minRc,Rc＝(1/|S|2∑xi,xj∈Sc(xi,xj)
wherein: rcFor redundancy, c (x)i,xj) Is a correlation function;
(3) then, integrating the maximum correlation and the minimum redundancy, and adopting addition integration, then:
maxΦ(D,R),Φ＝D–R
(4) finding approximately optimal characteristics by using an incremental search method; suppose we have a set of features
Sm-1Our task is to derive the remaining features X-Sm-1By selecting features such thatPhi is maximum; the incremental algorithm optimizes the following conditions:
maxxj∈X-Sm-1[I(xj,c)-[∑xi∈Xm-1I(xj,xi)]/(m-1)]
(5) aiming at pure sleep, the first 12 characteristics are selected, and the 12 characteristics of the pure sleep are respectively the minimum value of an RR interphase, the variance of the RR interphase, the Shannon information entropy, the standard deviation of normal sinus rhythm, the root mean square of the difference value of adjacent RR interphase, the mean value of the RR interphase, the root mean square of the difference value of adjacent RR interphase, the variation degree of an RR sequence, the maximum value of the RR interphase, the energy value of an extremely-low frequency band, the energy value of a low frequency band and the energy of an HF high frequency band; for the sleep conversion stage, the selected characteristics are r _ Mean, r _ Min minimum, ratio of characteristic values before and after r _ TP, r _ CV variation coefficient, r _ vLF extremely low frequency band energy value, r _ LF low frequency band energy value, r _ var variance, r _ SDNN standard deviation, r _ Max maximum, r _ Renyi shannon information entropy, ratio of r _ LF/HF low frequency energy value to high frequency energy value, r _ HF high frequency band energy value, r _ RMSSD difference root Mean square, r _ SDSD difference standard deviation, and Sam _ entry sample entropy.
Further, the sleep staging model in the fourth step is specifically divided into two layers of sleep models, wherein the first layer performs second classification on pure sleep and stage transition, the second layer performs segmental classification and fragment classification on the pure sleep, and performs coarse-grained classification and fine-grained classification on the stage transition.
Further, the step four of using the trained model to complete the sleep staging includes the specific steps of:
s1, selecting a support vector machine classification method and an integrated learning classifier classification method in machine learning to carry out sleep stage model training;
s2, constructing a sleep stage identification model based on a hidden Markov decoding process;
and S3, using an SVM classifier to divide stage conversion and pure sleep, further using an HMM classifier to finish pure sleep classification, then using an AdaBoost classifier to finish coarse-grained classification and fine-grained classification of the stage conversion, and finally finishing sleep staging.
Further, the hidden markov decoding process used in step S2 includes a specific process of constructing a sleep stage identification model:
1) set of states S ═ S1，S2，S3，S4A sleep stage, namely { waking, light sleep, deep sleep, rapid eye movement };
2) obtaining initial state probability distribution, and taking value as the proportion occupied by each sleep stage of 456 individual sleep1＝0.2641，Π2＝0.4527,Π3＝0.1322，Π4＝0.1510；
3) Determining an observation sequence O ═ { O ] according to the heart rate feature set G1，O2，O3，O4{ Min, Var, Renyi, SDNN, RMSSD, Mean, SDSD, CV, Max, vLF, LF, HF };
4) obtaining a sequence of states Q ═ Q1，q2，q3，…,qtAnd q isi∈{S1，S2，S3，S4In which Q represents the state sequence, QtFor the last state, qiIs one of t states;
5) performing probability statistics on sleep stage transition by using training set data to obtain a state transition probability matrix, wherein the state transition probability matrix A is { a ═ aij},aij＝P(qt+1＝Sj|qt＝Si) I is more than or equal to 1 and j is less than or equal to 4, wherein aijIs in a state of SjHas a probability and a state of SiThe ratio of the probabilities of (a);
6) performing probability statistics on HRV characteristics corresponding to the sleep stage to obtain an observation probability distribution matrix:
B＝{bj(k)},bj(k)＝P(ot＝vk|qt＝Sj),1≤j≤4,vkfor the t-th feature O of the observation sequence OtActual value of bj(k) Being elements of a probability distribution matrix, i.e. in state SjNext, the probability of occurrence of the tth feature;
7) the model is denoted as λ ═ (pi, a, B);
8) and solving the most possible hidden state sequence, namely the sleep stage sequence according to the observation sequence O and the hidden Markov model lambda (pi, A and B).
The invention has the beneficial effects that: compared with the prior art, the improvement of the invention is that the sleep stage signal is subjected to smooth preprocessing because of the HRV-based sleep stage classification method provided by the invention; extracting RR intervals from the smoothed ECG signal; performing heart rate variability feature extraction by using RR intervals; selecting a proper heart rate variability characteristic to train a sleep staging model, and finishing sleep staging by using the trained model; therefore, by adopting the steps, the noise signals are smoothed, the original information is better kept, and the method has the characteristics of more accurate division of sleep stages, more detailed classification and better sleep staging effect.
Drawings
FIG. 1 is a flow chart of the main algorithm steps of the sleep staging method based on heart rate variability of the present invention.
FIG. 2 is a schematic diagram of the classification of pure sleep segments according to the present invention.
Fig. 3 is a schematic diagram illustrating slice classification for pure sleep according to the present invention.
FIG. 4 is a diagram illustrating fine-grained classification of phase transitions in the present invention.
Detailed Description
In order to make those skilled in the art better understand the technical solution of the present invention, the following further describes the technical solution of the present invention with reference to the drawings and the embodiments.
As shown in fig. 1 to 4, an HRV-based sleep staging method specifically includes the following steps:
the method comprises the following steps: smoothing the sleep stage signal to smooth out the stage transition which is impossible to occur during sleep, and storing the original sleep stage information: specifically, sleep is a continuous process, but after PSG signal acquisition and then preliminary judgment, there may be some other stages that may be caused by some external factors, such as diseases and apparatus; the smoothing rule is an attribute showing a stable sleep duration and sleep continuity; that is, in a stable sleep stage, the stages are unlikely to change suddenly, so we smooth the data, using Local Clustering Algorithm (LCA), keeping the original sleep stage information as much as possible while smoothing out the stage transitions that are unlikely to occur;
the sleeping structure is not stable and unchangeable, and people sleep unstably in one night and go through 3-5 sleeping cycles, wherein each cycle lasts for 90-110 minutes; in each cycle, an arousal W, a light sleep L, a deep sleep D and a rapid eye movement period R occur; by analyzing the sleep data of a whole night, a certain pure sleep occupies a main part in a certain period of time, which indicates that the sleep is not disordered and is not switched randomly, and the sleep is a regular and slow change process, wherein the specific change process is divided into 12 sleep conversion stages, namely, arousal conversion to light sleep WL, arousal conversion to deep sleep WD, arousal conversion to fast eye movement WR, light sleep conversion to arousal LW, light sleep conversion to deep sleep LD, light sleep conversion to fast eye movement LR, deep sleep conversion to arousal DW, deep sleep conversion to light sleep DL, deep sleep conversion to fast eye movement DR, fast eye movement conversion to RW, fast eye movement conversion to light sleep RL and fast eye movement conversion to deep sleep RD;
step two: extracting RR intervals from the smoothed preprocessed ECG signal; the specific process comprises the following steps:
s1, removing noise in an ECG signal by using wavelet transform;
(1) performing wavelet decomposition on the image signal;
(2) carrying out threshold quantization on the high-frequency coefficient subjected to hierarchical decomposition;
(3) reconstructing an image signal using a two-dimensional wavelet;
s2, detecting R wave peak points based on a sliding window to obtain an R wave horizontal coordinate;
identifying all peak points by using a fixed sliding window method; the method for detecting the peak point is selected because all R waves are detected due to the prominent expression of the peak; the tool kit for extracting the R wave peak value in the TOMPKINS 'A Real-Time QRS detection algorithm' is used for carrying out peak value detection to obtain the horizontal coordinate of the R wave;
s3, subtracting the coordinate of the former peak point from the coordinate of the next continuous peak point of the obtained abscissa of the peak point to obtain an RR interval;
s4, performing error detection and missing detection of RR interphase by using a criterion based on 3 sigma;
when the peak point detection is carried out by using a sliding window method, the accuracy rate of the peak point detection has a certain relation with the width of a window, a larger window can cause the missing detection of the peak point, and a smaller window can cause the false detection of the peak point; therefore, the correction algorithm for false detection and missed detection on the premise that most heart rate intervals are correct is also provided;
step three: performing heart rate variability feature extraction by using RR intervals, wherein the heart rate variability feature extraction on the RR intervals comprises extracting three-dimensional features of a time domain, a frequency domain and a nonlinear domain; the specific characteristics are as follows:
(1) the time domain feature extraction of the HRV comprises the following specific processes:
s1, continuously recording normal sinus cardiac pulsation of a person in a sleeping process by adopting an electrocardio monitoring device;
s2, arranging the numerical values of the RR intervals according to time or heart beat sequence, wherein the numerical values are RR interval sequences;
s3, performing time domain feature analysis on the RR interval sequence, and calculating the mean value, the variance and the maximum and minimum value of the RR intervals, the standard deviation of a normal sinus rhythm, the root mean square of the difference value of adjacent RR intervals, the standard deviation of the difference value of adjacent RR intervals, wherein the difference value accounts for all percentages above 50ms and the variation degree of the RR sequence;
(2) the frequency domain feature extraction of the HRV comprises the following specific processes:
mainly analyze RR interval time sequence's power spectral density estimation diagram, it describes HRV frequency spectrum along with the distribution situation of the time sequence, through studying and analyzing the frequency domain value in different frequency ranges (high-low frequency, ultralow frequency) and then realize the assessment to the law of change of heart rate, the invention is with the help of Auto-regressive model (AR), the frequency domain analysis of the power spectrum estimation method heart rate variability;
the AR model is an all-pole model, and the signal to be analyzed, x (n), is represented using the difference equation:
x(n)＝-∑p i＝1ap(i)x(n-1)+u(n)
wherein: u (n) denotes mean 0 and variance σ2P denotes the order of the AR model, ap(i) I is 1,2 …, p represents the corresponding parameters of the AR model at different orders;
the calculation method of the AR model power spectrum estimation is as follows:
P(w)＝σ2/|1+∑p i＝1aie-jwi|2
wherein: e.g. of the type-jwiRepresenting a value trend for the AR model function, aiRepresenting the corresponding parameters of the AR model at the i order.
The power spectral density of the HRV is divided into 5 parts, namely very Low Frequency (Low Frequency, vLF, 0.04HZ), Low Frequency (Low Frequency, LF, 0.15HZ), High Frequency (High Frequency, HF, 0.4HZ), very High Frequency (very High Frequency, vHF, 0.5HZ), Total Frequency power (Total Frequency, TF, 0HZ-0.5HZ), and according to the above division criteria, the following Frequency domain features of the HRV are extracted for model training in the present invention:
1) and vLF: representing the energy value of the extremely low frequency band, and researches show that the frequency band reflects the related long-term regulation mechanisms such as humoral factors, body temperature regulation and the like;
2) LF: represents the energy value of a low-frequency band, and related researches show that the frequency band reflects the joint regulation of sympathetic nerves and parasympathetic nerves;
3) HF: representing high-frequency band energy, and research shows that the band energy reflects parasympathetic activity and reflects respiratory change of a human body;
4) vHF: represents the energy of the extremely high frequency band;
5) TP: represents the sum of the energies of the power of the whole frequency band, and the parameter reflects the variability of the whole signal in the period of time;
6) LF/HF: the energy value ratio of the low-frequency and high-frequency bands is represented and is a quantitative index of the balance between the sympathetic nerve and the vagus nerve;
(3) extracting nonlinear features of HRV: calculating Shannon information entropy as the description of RR interval complexity, including sample entropy;
the specific calculation process of the Shannon information entropy is as follows: assuming a discrete variable X with a probability distribution of X ═ X1, X2, …, xn, the Renyi entropy is defined as follows:
Hα(X)＝log2∑n i＝1pi α/(1-α)
wherein: p is a radical ofiIs event X ═ Xiα represents the order of Renyi entropy, and the value range is positive integer;
herein, mainly using 1-order Renyi entropy, when α is equal to 1, as can be known from the definition of Renyi entropy, the denominator thereof is close to zero, and the limit value is solved by the lobbie method to obtain the entropy value, and the calculation formula is as follows:
H1(X)＝limα→1Hα(X)＝∑n i＝1[pilog2(pi)]
at this time, the Renyi entropy is also called Shannon (Shannon) information entropy, and can be used as RR interval complexity description.
Step four: selecting a proper heart rate variability characteristic to train a sleep staging model, and finishing sleep staging by using the trained model; specifically, the method comprises three parts:
the first part is the process of selecting suitable heart rate variability characteristics;
the second part is layering of a sleep staging model;
and the third part is the process of completing sleep staging by using the trained model.
The process of selecting the appropriate heart rate variability feature in the first part is:
s1, dividing a sleep process into a sleep transition stage and a pure sleep stage, and distinguishing the sleep transition stage from the pure sleep stage;
s2, classifying sleep conversion stages in sections: the last pure sleep T of the sleep transition stagelatterAnd previous pure sleep TfommerThe characteristic ratio of (a) is used as a new characteristic of the sleep transition stage and the pure sleep stage classification;
s3, carrying out section classification on the pure sleep stages: taking the previous sleep stage and the night-long sleep cycle of the sleep stage as classification characteristics of pure sleep;
s4, aiming at time domain (time dimension), frequency domain (energy dimension) and nonlinear features (complexity measurement) of the heart rate in the sleep conversion stage and the pure sleep stage, selecting the features by using a maximum correlation minimum redundancy algorithm;
the specific selection algorithm is as follows: given two random variables x and y, their probability density functions (corresponding to continuous variables) are p (x), p (y), p (x, y), then the mutual information is:
I(x；y)＝∫∫p(x,y)log[p(x,y)/p(x)p(y)]dxdy
the goal is to find a solution containing m { x }iA feature subset S of the features, if a discrete variable, the maximum correlation is calculated as:
maxD(S,c),D＝(1/|S|)∑xi∈SI(xi,c)
xiis the ith feature, c is a category variable, and S is a feature subset;
the minimum redundancy is then:
minR(S,c),R＝(1/|S|2)∑xi,xj∈SI(xi,xj)
if it is a continuous variable, the maximum correlation:
maxDF,DF＝(1/|S|)∑xi∈SF(xi,c)
wherein: dFIs a correlation, (x)iAnd c) is the statistic of F;
F(xiand c) is F statistic, the minimum redundancy is:
minRc,R＝(1/|S|2∑xi,xj∈Sc(xi,xj)
wherein: rcFor redundancy, c (x)i,xj) Is composed ofA correlation function;
then, integrating the maximum correlation and the minimum redundancy, and adopting addition integration, then:
maxΦ(D,R),Φ＝D–R
in practice, an incremental search method is used for finding approximately optimal characteristics; suppose we have a set of features Sm-1Our task is to derive the remaining features X-Sm-1Finding the mth feature, and enabling phi to be maximum by selecting the feature; the incremental algorithm optimizes the following conditions:
maxxj∈X-Sm-1[I(xj,c)-[∑xi∈Xm-1I(xj,xi)]/(m-1)]
in the invention, the characteristics are continuous variables, the adopted objective function is additive integration, the first 12 characteristics are selected aiming at pure sleep, the 12 characteristics of the pure sleep are respectively Min (minimum value of RR interphase), Var (variance of RR interphase), Renyi (entropy of Shannon information), SDNN (standard deviation of normal sinus rhythm), RMSSD (root Mean square of difference of adjacent RR interphase), Mean (Mean value of RR interphase), SDSD (root Mean square of difference of adjacent RR interphase), CV (variation degree of RR sequence), Max (maximum value of RR interphase), vLF (energy value of extremely low frequency band), LF (energy value of low frequency band) and HF energy of high frequency band; for the sleep conversion stage, the selected characteristics are r _ Mean value, r _ Min minimum value, the ratio of the front and rear characteristic values of r _ TP, r _ CV variation coefficient, r _ vLF extremely low frequency band energy value, r _ LF low frequency band energy value, r _ var variance, r _ SDNN standard deviation, r _ Max maximum value, r _ Renyi Shannon information entropy, the ratio of r _ LF/HF low frequency energy value to high frequency energy value, r _ HF high frequency band energy value, r _ RMSSD difference root-Mean-square, r _ SDSD difference standard deviation and Sam _ Entroy sample entropy;
r _ Mean represents the Mean of RR intervals over a period of time;
the r _ Min minimum represents the minimum value of the RR intervals over a period of time;
r _ TP represents the ratio of the characteristic values before and after the phase transition;
the r _ CV coefficient of variation represents the degree of sequence variation of RR intervals over time;
r _ vLF extremely low frequency band energy value, research shows that the frequency band reflects the long-term regulation mechanism related to humoral factor, body temperature regulation and the like;
r _ LF low-frequency band energy values, which are shown by related researches to reflect the joint regulation of sympathetic nerves and parasympathetic nerves;
the r _ var variance represents the variance of RR intervals over a period of time;
the r _ SDNN standard deviation represents the mean value of the square sum of the difference values of the heartbeat interval and the heartbeat interval mean value, and the value of the SDNN reflects the total power of the heart rate time series signal from the aspect of mathematics and physical theory;
r _ Max maximum represents the maximum value of RR interval for a period of time;
r _ Renyi Shannon information entropy expresses RR interval complexity;
the ratio of the r _ LF/HF low-frequency energy value to the high-frequency energy value is related to the sympathetic nerve and the vagus nerve, and the mutual change relationship of the two is reflected;
r _ HF high frequency band energy value, the band energy reflects parasympathetic activity and reflects respiratory change of human body;
r _ RMSSD difference root mean square, representing the difference root mean square of adjacent RR intervals over a period of time;
r _ SDSD difference standard deviation, root mean square of difference between adjacent RR intervals over a period of time;
sam _ Entroy sample entropy, representing the complexity of the RR interval population;
the second part of the sleep staging model is specifically divided into two layers of sleep models, wherein the first layer carries out secondary classification on pure sleep and transition stages, the second layer carries out sectional classification and fragment classification on the pure sleep and carries out coarse-grained classification and fine-grained classification on stage transition;
the first layer performs a specific process of classifying pure sleep and stage transition into two categories:
and D, analyzing stage conversion under different time scales D (domain), wherein the time scale is 1-10 minutes, and analyzing the classification result of the stage conversion and the pure sleep under D:
s1, dividing the wake period, the light sleep period, the deep sleep period and the rapid eye movement period of the sleep stages into pure sleep, dividing the conversion of different sleep stages into stage conversion, and performing secondary classification;
s2, performing segmentation by using a sliding window, aiming at constructing samples of pure sleep P and stage transition T, and classifying which are pure sleep and which are stage transition by using a classifier, wherein the time scale is 1-10 minutes respectively;
further analysis for pure sleep was performed from two directions:
firstly, directly classifying the pure sleep of the sections on the basis of two classifications of pure sleep and stage transition, wherein the length of the section is uncertain, the section is divided into the pure sleep by default on the basis of the two classifications of the sleep stage into the pure sleep and the stage transition, and other pure sleep is not arranged in the middle of the section, as shown in figure 2, the label is removed, and the rest is the classification of different pure sleep; secondly, the segmented pure sleep is further divided into different slices in equal length, then the slices are classified, as shown in the attached figure 3, the marked pure sleep is removed, the equal length segmentation is further carried out, and then the slice classification is carried out, so that the defect that the pure sleep is classified by mistake due to inaccurate stage conversion classification in the first step can be overcome;
classifying the stage conversion classified in the first step, wherein fine-grained analysis is to classify all the stage conversions, namely, wakefulness conversion to light sleep WL, wakefulness conversion to deep sleep WD, wakefulness conversion to fast eye movement period WR, light sleep conversion to wakefulness LW, light sleep conversion to deep sleep LD, light sleep conversion to fast eye movement period LR, deep sleep conversion to wakefulness DW, deep sleep conversion to light sleep DL, deep sleep conversion to fast eye movement period DR, fast eye movement period conversion to wakefulness RW, fast eye movement period conversion to light sleep RL, fast eye movement period conversion to deep sleep RD, and 12 stage conversions in total, as shown in figure 4;
classifying the coarse granularity and the fine granularity of the stage conversion: coarse-grained classification, classifying the stage transition into 6 classes, namely, wakefulness transition to light sleep W-L, light sleep transition to wakefulness L-W, light sleep transition to deep sleep L-D, deep sleep transition to light sleep D-L, light sleep transition to fast eye movement period L-R, fast eye movement period transition to light sleep R-L, deep sleep transition to fast eye movement period D-R, fast eye movement period transition to deep sleep R-D; fine-grained classification, which divides the stage transition into 12 classes, namely, wakefulness to light sleep WL, wakefulness to deep sleep WD, wakefulness to fast eye movement WR, light sleep to wakefulness LW, light sleep to deep sleep LD, light sleep to fast eye movement LR, deep sleep to wakefulness DW, deep sleep to light sleep DL, deep sleep to fast eye movement DR, fast eye movement RW, fast eye movement to light sleep RL, fast eye movement RD to deep sleep RD;
considering the probability problem that the stage transition occurs in the sleeping process of a person, according to the statistical analysis of a large amount of sleep data, the transitions of W and L, L and D, and L and R account for more than 80% of the stage transition, and therefore, the identification of 6 stage transitions with a large occurrence probability is mainly performed, i.e., the transition of arousal to light sleep W-L, the transition of light sleep to arousal L-W, the transition of light sleep to deep sleep L-D, the transition of deep sleep to light sleep D-L, the transition of light sleep to fast eye movement period L-R, the transition of fast eye movement period to light sleep R-L, the transition of deep sleep to fast eye movement period D-R, and the transition of fast eye movement period to deep sleep R-D, respectively;
the third part is that the process of completing the sleep stage by using the trained model comprises the following steps:
s1, selecting a support vector machine classification method and an integrated learning classifier classification method in machine learning to carry out sleep stage model training;
s2, constructing a sleep stage identification model based on a hidden Markov decoding process, and describing a specific modeling process in detail as follows:
1) set of states S ═ S1，S2，S3，S4A sleep stage, namely { waking, light sleep, deep sleep, rapid eye movement };
2) obtaining initial state probability distribution, and taking value as the proportion occupied by each sleep stage of 456 individual sleep1＝0.2641，Π2＝0.4527,Π3＝0.1322，Π4＝0.1510；
3) Determining an observation sequence O ═ { O ] according to the heart rate feature set G1，O2，O3，O4{ Min, Var, Renyi, SDNN, RMSSD, Mean, SDSD, CV, Max, vLF, LF, HF };
4) obtaining a sequence of states Q ═ Q1，q2，q3，…,qtAnd q isi∈{S1，S2，S3，S4}；
5) Performing probability statistics on sleep stage transition by using training set data to obtain a state transition probability matrix, wherein the state transition probability matrix A is { a ═ aij},aij＝P(qt+1＝Sj|qt＝Si) I is more than or equal to 1 and j is less than or equal to 4, wherein aijIs in a state of SjHas a probability and a state of SiThe ratio of the probabilities of (a);
6) performing probability statistics on HRV characteristics corresponding to the sleep stage to obtain an observation probability distribution matrix:
B＝{bj(k)},bj(k)＝P(ot＝vk|qt＝Sj) J is more than or equal to 1 and less than or equal to 4, vk is the t-th characteristic O of the observation sequence OtThe actual value of (c);
7) the model is denoted as λ ═ (pi, a, B);
8) in the problem, we pay attention to the hidden state in the markov model, that is, the problem of learning observable HRV characteristics and researching sleep stages, so that a Viterbi algorithm in the hidden markov model decoding process is adopted to solve the most likely hidden state sequence, that is, the sleep stage sequence, according to an observation sequence O and a hidden markov model λ ═ (pi, a, B);
and S3, using an SVM classifier to divide stage conversion and pure sleep, further using an HMM classifier to finish pure sleep classification, then using an AdaBoost classifier to finish coarse-grained classification and fine-grained classification of the stage conversion, and finally finishing sleep staging.
The foregoing shows and describes the general principles, essential features, and advantages of the invention. It will be understood by those skilled in the art that the present invention is not limited to the embodiments described above, which are described in the specification and illustrated only to illustrate the principle of the present invention, but that various changes and modifications may be made therein without departing from the spirit and scope of the present invention, which fall within the scope of the invention as claimed. The scope of the invention is defined by the appended claims and equivalents thereof.
Claims (9)
1. A sleep staging method based on HRV is characterized by specifically comprising the following steps:
the method comprises the following steps: smoothing the sleep stage signal to smooth out the stage transition which is impossible to occur during sleep and storing the original sleep stage information;
step two: extracting RR intervals from the ECG signal subjected to the smoothing treatment in the step one;
step three: performing heart rate variability feature extraction by using the RR intervals extracted in the second step;
step four: and (4) training a sleep staging model through the features extracted in the step three to finish sleep staging.
2. An HRV-based sleep staging method as claimed in claim 1 wherein the sleep stages of step one include 4 pure sleep stages, namely light sleep L, deep sleep D, wake W, fast eye movement period R; the sleep stages further include 12 stage transitions, namely, wakefulness transition to light sleep WL, wakefulness transition to deep sleep WD, wakefulness transition to fast eye movement period WR, light sleep transition to wakefulness LW, light sleep transition to deep sleep LD, light sleep transition to fast eye movement period LR, deep sleep transition to wakefulness DW, deep sleep transition to light sleep DL, deep sleep transition to fast eye movement period DR, fast eye movement period transition to wakefulness RW, fast eye movement period transition to light sleep RL, and fast eye movement period transition to deep sleep RD.
3. The HRV-based sleep staging method as claimed in claim 1, wherein the specific process of extracting RR intervals from ECG signals in step two is:
s1, removing noise in an ECG signal by using wavelet transform;
s2, detecting R wave peak points based on a sliding window to obtain an R wave horizontal coordinate;
s3, subtracting the coordinate of the former peak point from the coordinate of the next continuous peak point of the obtained abscissa of the peak point to obtain an RR interval;
s4, performing error detection and omission detection of RR intervals by using a criterion based on 3 sigma.
4. The HRV-based sleep staging method according to claim 1, wherein the heart rate variability feature extraction on RR intervals of step three includes extraction of three-dimensional features of time domain, frequency domain and nonlinear domain.
5. A HRV-based sleep staging method according to claim 1, wherein the specific step of selecting the appropriate heart rate variability signature of step four includes:
s1, dividing a sleep process into a sleep transition stage and a pure sleep stage, and classifying the sleep transition stage and the pure sleep stage;
s2, classifying sleep transition stages in sections: the last pure sleep T of the sleep transition stagelatterAnd previous pure sleep TfommerThe characteristic ratio of (a) is used as a new characteristic of the sleep transition stage and the pure sleep stage classification;
s3, carrying out section classification on the pure sleep stages: taking the previous sleep stage and the night-long sleep cycle of the sleep stage as classification characteristics of pure sleep;
and S4, aiming at the time domain, the frequency domain and the nonlinear domain characteristics of the heart rate in the sleep conversion stage and the pure sleep stage, selecting the characteristics by using a maximum correlation minimum redundancy algorithm.
6. The HRV-based sleep staging method according to claim 5, wherein the specific process of feature selection using the maximum correlation minimum redundancy algorithm at S4 is:
(1) taking the time domain characteristic of the heart rate in the stage as a time dimension, the frequency domain characteristic as an energy dimension and the nonlinear characteristic as complexity measurement;
(2) given two random variables x and y, their probability density functions are p (x), p (y), p (x, y), then the mutual information is:
I(x；y)＝∫∫p(x,y)log[p(x,y)/p(x)p(y)]dxdy
the goal is to find a solution containing m { x }iA feature subset S of the features, if a discrete variable, the maximum correlation is calculated as:
maxD(S,c),D＝(1/|S|)∑xi∈SI(xi,c)
wherein: x is the number ofiIs the ith feature, c is a category variable, and S is a feature subset;
the minimum redundancy is then:
minR(S,c),R＝(1/|S|2)∑xi,xj∈SI(xi,xj)
wherein: c (x)i,xj) Is a correlation function;
if it is a continuous variable, the maximum correlation:
maxDF,DF＝(1/|S|)∑xi∈SF(xi,c)
wherein: dFIs a correlation, (x)iAnd c) is the statistic of F;
the minimum redundancy is:
minRc,Rc＝(1/|S|2∑xi,xj∈Sc(xi,xj)
wherein: rcRedundancy is adopted;
(3) then, integrating the maximum correlation and the minimum redundancy, and adopting addition integration, then:
maxΦ(D,R),Φ＝D–R
(4) finding approximately optimal characteristics by using an incremental search method; suppose we have a set of features Sm-1Our task is to derive the remaining features X-Sm-1Finding the mth feature, and enabling phi to be maximum by selecting the feature; the incremental algorithm optimizes the following conditions:
maxxj∈X-Sm-1[I(xj,c)-[∑xi∈Xm-1I(xj,xi)]/(m-1)]
(5) aiming at a pure sleep stage, selecting the first 12 characteristics, wherein the 12 characteristics of the pure sleep are respectively the minimum value of an RR interphase, the variance of the RR interphase, the Shannon information entropy, the standard deviation of a normal sinus rhythm, the root mean square of the difference value of adjacent RR interphase, the mean value of the RR interphase, the root mean square of the difference value of adjacent RR interphase, the variation degree of an RR sequence, the maximum value of the RR interphase, the energy value of an extremely-low frequency band, the energy value of a low-frequency band and the energy of an HF high-frequency band; aiming at the sleep conversion stage, r _ Mean, r _ Min minimum, ratio of characteristic values before and after r _ TP, r _ CV variation coefficient, r _ vLF extremely low frequency band energy value, r _ LF low frequency band energy value, r _ var variance, r _ SDNN standard deviation, r _ Max maximum, r _ Renyi Shannon information entropy, ratio of r _ LF/HF low frequency band energy value to high frequency energy value, r _ HF high frequency band energy value, r _ RMSSD difference root-Mean-square, r _ SDSD difference standard deviation and Sam _ Entroy sample entropy are selected.
7. The HRV-based sleep staging method as claimed in claim 1, wherein the sleep staging model of step four is specifically divided into two layers of sleep models, the first layer classifies pure sleep and stage transitions, the second layer classifies segments and slices of pure sleep, and classifies coarse and fine granularity of stage transitions.
8. The HRV-based sleep staging method as claimed in claim 1, wherein the concrete step of completing sleep staging using the trained model in step four includes:
s1, selecting a support vector machine classification method and an integrated learning classifier classification method in machine learning to carry out sleep stage model training;
s2, constructing a sleep stage identification model based on a hidden Markov decoding process;
and S3, using an SVM classifier to divide stage conversion and pure sleep, further using an HMM classifier to finish pure sleep classification, then using an AdaBoost classifier to finish coarse-grained classification and fine-grained classification of the stage conversion, and finally finishing sleep staging.
9. The HRV-based sleep staging method according to claim 8, wherein the step S2 of using the hidden markov decoding process to construct the sleep stage identification model includes the following specific steps:
1) set of states S ═ S1，S2，S3，S4A sleep stage, namely { waking, light sleep, deep sleep, rapid eye movement };
2) obtaining initial state probability distribution, and taking value as the proportion occupied by each sleep stage of 456 individual sleep1＝0.2641，Π2＝0.4527,Π3＝0.1322，Π4＝0.1510；
3) Determining an observation sequence O ═ { O ] according to the heart rate feature set G1，O2，O3，O4{ Min, Var, Renyi, SDNN, RMSSD, Mean, SDSD, CV, Max, vLF, LF, HF };
4) obtaining a sequence of states Q ═ Q1，q2，q3，…,qtAnd q isi∈{S1，S2，S3，S4In which Q represents the state sequence, QtFor the last state, qiIs one of t states;
5) performing probability statistics on sleep stage transition by using training set data to obtain a state transition probability matrix, wherein the state transition probability matrix A is { a ═ aij},aij＝P(qt+1＝Sj|qt＝Si) I is more than or equal to 1 and j is less than or equal to 4, wherein aijIs in a state of SjHas a probability and a state of SiThe ratio of the probabilities of (a);
6) performing probability statistics on HRV characteristics corresponding to the sleep stage to obtain an observation probability distribution matrix:
B＝{bj(k)},bj(k)＝P(ot＝vk|qt＝Sj),1≤j≤4,vkfor the t-th feature O of the observation sequence OtActual value of bj(k) Being elements of a probability distribution matrix, i.e. in state SjNext, the probability of occurrence of the tth feature;
7) the model is denoted as λ ═ (pi, a, B);
8) and solving the most possible hidden state sequence, namely the sleep stage sequence according to the observation sequence O and the hidden Markov model lambda (pi, A and B).
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201911412135.8A CN111067503A (en) | 2019-12-31 | 2019-12-31 | Sleep staging method based on heart rate variability |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201911412135.8A CN111067503A (en) | 2019-12-31 | 2019-12-31 | Sleep staging method based on heart rate variability |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN111067503A true CN111067503A (en) | 2020-04-28 |
Family
ID=70321157
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201911412135.8A Pending CN111067503A (en) | 2019-12-31 | 2019-12-31 | Sleep staging method based on heart rate variability |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN111067503A (en) |
Cited By (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112244773A (en) * | 2020-10-15 | 2021-01-22 | 上海我乐科技有限公司 | Sleep quality monitoring device and method and mattress |
| CN112353401A (en) * | 2020-10-19 | 2021-02-12 | 燕山大学 | Staged regulation and control method based on physiological state evaluation |
| CN112426131A (en) * | 2020-11-22 | 2021-03-02 | 华东理工大学 | Sleep stage automatic interpretation method based on step-by-step clustering model |
| CN112617761A (en) * | 2020-12-31 | 2021-04-09 | 湖南东晟南祥智能科技有限公司 | Sleep stage staging method for self-adaptive multipoint generation |
| CN112674779A (en) * | 2020-12-25 | 2021-04-20 | 西华大学 | Automatic sleep staging method based on ECG (electrocardiogram) signals |
| CN112842266A (en) * | 2020-12-31 | 2021-05-28 | 湖南东晟南祥智能科技有限公司 | Sleep stage identification method based on human body monitoring sleep data |
| CN113180596A (en) * | 2021-04-07 | 2021-07-30 | 中山大学 | Non-contact sleep analysis method and device and storage medium |
| CN115192850A (en) * | 2022-08-15 | 2022-10-18 | 福建工程学院 | Wake-up system based on heart rate variability |
| CN115844335A (en) * | 2023-01-29 | 2023-03-28 | 广东工业大学 | Sleep stage staging method and system based on feature overlapping and generalized decision forest |
| CN117357073A (en) * | 2023-12-07 | 2024-01-09 | 北京清雷科技有限公司 | Sleep stage method and device based on GMM-HMM model |
Citations (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20050054940A1 (en) * | 2003-04-23 | 2005-03-10 | Almen Adam J. | Apparatus and method for monitoring heart rate variability |
| US20080167565A1 (en) * | 2007-01-09 | 2008-07-10 | Timo Laitio | Method and Arrangement for Obtaining Diagnostic Information of a Patient |
| CN103717125A (en) * | 2011-05-18 | 2014-04-09 | V视股份有限公司 | System and method for determining sleep and sleep stages of a person |
| US20150190086A1 (en) * | 2014-01-03 | 2015-07-09 | Vital Connect, Inc. | Automated sleep staging using wearable sensors |
| CN105496363A (en) * | 2015-12-15 | 2016-04-20 | 浙江神灯生物科技有限公司 | Method for classifying sleep stages on basis of sleep EGG (electroencephalogram) signal detection |
| CN106419869A (en) * | 2016-08-24 | 2017-02-22 | 电子科技大学 | Real-time sleep staging detection method based on piezoelectric sensor and device for realizing method |
| US20170055898A1 (en) * | 2015-08-28 | 2017-03-02 | Awarables, Inc. | Determining Sleep Stages and Sleep Events Using Sensor Data |
| CN106936526A (en) * | 2017-03-30 | 2017-07-07 | 西北工业大学 | A kind of contactless sleep stage device and method based on channel condition information |
| CN107708528A (en) * | 2015-06-25 | 2018-02-16 | 皇家飞利浦有限公司 | Apparatus and method for the physiological status of monitoring object |
| WO2018070935A1 (en) * | 2016-10-11 | 2018-04-19 | National University Of Singapore | Determining sleep stages |
| CN110074765A (en) * | 2019-04-29 | 2019-08-02 | 中山大学 | A kind of method of discrimination of SAHS screening and sleep stage |
-
2019
- 2019-12-31 CN CN201911412135.8A patent/CN111067503A/en active Pending
Patent Citations (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20050054940A1 (en) * | 2003-04-23 | 2005-03-10 | Almen Adam J. | Apparatus and method for monitoring heart rate variability |
| US20080167565A1 (en) * | 2007-01-09 | 2008-07-10 | Timo Laitio | Method and Arrangement for Obtaining Diagnostic Information of a Patient |
| CN103717125A (en) * | 2011-05-18 | 2014-04-09 | V视股份有限公司 | System and method for determining sleep and sleep stages of a person |
| US20150190086A1 (en) * | 2014-01-03 | 2015-07-09 | Vital Connect, Inc. | Automated sleep staging using wearable sensors |
| CN107708528A (en) * | 2015-06-25 | 2018-02-16 | 皇家飞利浦有限公司 | Apparatus and method for the physiological status of monitoring object |
| US20170055898A1 (en) * | 2015-08-28 | 2017-03-02 | Awarables, Inc. | Determining Sleep Stages and Sleep Events Using Sensor Data |
| CN105496363A (en) * | 2015-12-15 | 2016-04-20 | 浙江神灯生物科技有限公司 | Method for classifying sleep stages on basis of sleep EGG (electroencephalogram) signal detection |
| CN106419869A (en) * | 2016-08-24 | 2017-02-22 | 电子科技大学 | Real-time sleep staging detection method based on piezoelectric sensor and device for realizing method |
| WO2018070935A1 (en) * | 2016-10-11 | 2018-04-19 | National University Of Singapore | Determining sleep stages |
| CN106936526A (en) * | 2017-03-30 | 2017-07-07 | 西北工业大学 | A kind of contactless sleep stage device and method based on channel condition information |
| CN110074765A (en) * | 2019-04-29 | 2019-08-02 | 中山大学 | A kind of method of discrimination of SAHS screening and sleep stage |
Non-Patent Citations (6)
| Title |
|---|
| JUHA M. KORTELAINEN等: "《Sleep Staging Based on Signals Acquired Through Bed Sensor》", 《TRANSACTIONS ON INFORMATION TECHNOLOGY IN BIOMEDICINE》 * |
| 卢欣欣: "《基于Device-free 技术的被动式行为感知及应用研究》", 《信息科技辑》 * |
| 孙微: "《基于HRV分析的睡眠分期方法研究》", 《医药卫生科技辑》 * |
| 朱彬: "《基于心率变异性的睡眠分期研究》", 《信息科技辑》 * |
| 江朝晖: "《R-R间期分析与睡眠分期》", 《生物医学工程研究》 * |
| 王爱波: "《基于隐马尔可夫模型的心率变异性分析在睡眠分期中的应用》", 《医药卫生科技辑》 * |
Cited By (16)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112244773A (en) * | 2020-10-15 | 2021-01-22 | 上海我乐科技有限公司 | Sleep quality monitoring device and method and mattress |
| CN112244773B (en) * | 2020-10-15 | 2023-01-24 | 上海我乐科技有限公司 | Sleep quality monitoring device and method and mattress |
| CN112353401A (en) * | 2020-10-19 | 2021-02-12 | 燕山大学 | Staged regulation and control method based on physiological state evaluation |
| CN112353401B (en) * | 2020-10-19 | 2022-04-19 | 燕山大学 | Staged regulation and control method based on physiological state evaluation |
| CN112426131A (en) * | 2020-11-22 | 2021-03-02 | 华东理工大学 | Sleep stage automatic interpretation method based on step-by-step clustering model |
| CN112674779A (en) * | 2020-12-25 | 2021-04-20 | 西华大学 | Automatic sleep staging method based on ECG (electrocardiogram) signals |
| CN112617761B (en) * | 2020-12-31 | 2023-10-13 | 湖南正申科技有限公司 | Sleep stage staging method for self-adaptive focalization generation |
| CN112617761A (en) * | 2020-12-31 | 2021-04-09 | 湖南东晟南祥智能科技有限公司 | Sleep stage staging method for self-adaptive multipoint generation |
| CN112842266A (en) * | 2020-12-31 | 2021-05-28 | 湖南东晟南祥智能科技有限公司 | Sleep stage identification method based on human body monitoring sleep data |
| CN113180596A (en) * | 2021-04-07 | 2021-07-30 | 中山大学 | Non-contact sleep analysis method and device and storage medium |
| CN113180596B (en) * | 2021-04-07 | 2024-02-06 | 中山大学 | Non-contact sleep analysis method, device and storage medium |
| CN115192850A (en) * | 2022-08-15 | 2022-10-18 | 福建工程学院 | Wake-up system based on heart rate variability |
| CN115192850B (en) * | 2022-08-15 | 2023-12-15 | 福建工程学院 | Wake-up system based on heart rate variability |
| CN115844335A (en) * | 2023-01-29 | 2023-03-28 | 广东工业大学 | Sleep stage staging method and system based on feature overlapping and generalized decision forest |
| CN117357073A (en) * | 2023-12-07 | 2024-01-09 | 北京清雷科技有限公司 | Sleep stage method and device based on GMM-HMM model |
| CN117357073B (en) * | 2023-12-07 | 2024-04-05 | 北京清雷科技有限公司 | Sleep stage method and device based on GMM-HMM model |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111067503A (en) | Sleep staging method based on heart rate variability | |
| US20230181077A1 (en) | Machine differentiation of abnormalities in bioelectromagnetic fields | |
| CN107951485B (en) | Ambulatory ECG analysis method and apparatus based on artificial intelligence self study | |
| Goh et al. | Robust PPG motion artifact detection using a 1-D convolution neural network | |
| Mohebbian et al. | Fetal ECG extraction from maternal ECG using attention-based CycleGAN | |
| US8290575B2 (en) | Biomedical signal morphology analysis method | |
| US20100217144A1 (en) | Diagnostic and predictive system and methodology using multiple parameter electrocardiography superscores | |
| Shin et al. | Decision boundary-based anomaly detection model using improved AnoGAN from ECG data | |
| Almutairi et al. | Classification of Obstructive Sleep Apnoea from single-lead ECG signals using convolutional neural and Long Short Term Memory networks | |
| Bulusu et al. | Transient ST-segment episode detection for ECG beat classification | |
| Taloba et al. | Machine algorithm for heartbeat monitoring and arrhythmia detection based on ECG systems | |
| Sinha et al. | Identification and localization of myocardial infarction based on analysis of ECG signal in cross spectral domain using boosted SVM classifier | |
| Cai et al. | Real-time arrhythmia classification algorithm using time-domain ECG feature based on FFNN and CNN | |
| Dhyani et al. | Arrhythmia disease classification utilizing ResRNN | |
| Wang et al. | A novel measure of cardiopulmonary coupling during sleep based on the synchrosqueezing transform algorithm | |
| CN115024725A (en) | Tumor treatment aid decision-making system integrating psychological state multi-parameter detection | |
| Subramanyan et al. | A novel deep neural network for detection of Atrial Fibrillation using ECG signals | |
| Zhang et al. | Interpretable detection and location of myocardial infarction based on ventricular fusion rule features | |
| Hao et al. | Adaptive r-peak detection algorithm based on brown exponential smoothing model | |
| CN111278353A (en) | Method and system for detecting vital sign signal noise | |
| Zhang et al. | PPG-based Heart Rate Estimation with Efficient Sensor Sampling and Learning Models | |
| Mathunjwa et al. | Automatic IHR-based sleep stage detection using features of residual neural network | |
| Begum et al. | ECG sensor signal analysis to represent cases in a case-based stress diagnosis system | |
| Shilvya et al. | Obstructive Sleep Apnea Detection from ECG Signals with Deep Learning | |
| Tyagi et al. | A review of automated diagnosis of ECG arrhythmia using deep learning methods |
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 | ||
| RJ01 | Rejection of invention patent application after publication | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20200428 |