CN112861796B - Feature adaptive motion recognition method - Google Patents
Feature adaptive motion recognition method Download PDFInfo
- Publication number
- CN112861796B CN112861796B CN202110268443.9A CN202110268443A CN112861796B CN 112861796 B CN112861796 B CN 112861796B CN 202110268443 A CN202110268443 A CN 202110268443A CN 112861796 B CN112861796 B CN 112861796B
- Authority
- CN
- China
- Prior art keywords
- electrode
- electrodes
- missing
- existing
- features
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V40/00—Recognition of biometric, human-related or animal-related patterns in image or video data
- G06V40/10—Human or animal bodies, e.g. vehicle occupants or pedestrians; Body parts, e.g. hands
- G06V40/107—Static hand or arm
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/243—Classification techniques relating to the number of classes
- G06F18/24323—Tree-organised classifiers
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N20/00—Machine learning
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2218/00—Aspects of pattern recognition specially adapted for signal processing
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2218/00—Aspects of pattern recognition specially adapted for signal processing
- G06F2218/12—Classification; Matching
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- Artificial Intelligence (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Computation (AREA)
- Bioinformatics & Computational Biology (AREA)
- Software Systems (AREA)
- Life Sciences & Earth Sciences (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Evolutionary Biology (AREA)
- Computing Systems (AREA)
- Mathematical Physics (AREA)
- Medical Informatics (AREA)
- Human Computer Interaction (AREA)
- Multimedia (AREA)
- Measurement And Recording Of Electrical Phenomena And Electrical Characteristics Of The Living Body (AREA)
Abstract
The invention provides a characteristic self-adaptive action recognition method, which comprises the following steps: constructing a recognition model based on an initial data set, wherein the initial data set comprises initial characteristics of all electrodes in an existing electrode set corresponding to multiple times of physiological signal acquisition and labels corresponding to the multiple times of physiological signal acquisition; acquiring a feature set collected from a target to be identified; determining whether a feature from one or more electrodes in the existing electrode set is missing from the acquired feature set; in response to determining that features from one or more electrodes in the existing electrode set are missing, treating the one or more electrodes as one or more missing electrodes, supplementing features from the one or more missing electrodes in the acquired feature set; and taking the acquired feature set as the input of the recognition model to obtain an action recognition result. The feature self-adaptive motion recognition method provided by the invention supports the dynamic change of the feature space in the motion recognition process.
Description
Technical Field
The invention relates to the technical field of machine learning, in particular to a feature self-adaptive action recognition method.
Background
Conventional machine learning techniques typically assume that the data stream (including offline training and online prediction data) has a fixed feature space, however this assumption does not hold in many real-world scenarios. In fact, the increase and decrease of the feature space of the data stream may be caused by many reasons, such as the change of the feature space caused by the variability of the electrodes and the diversity of the recognition requirements in the physiological calculation, so that the recognition model constructed on the basis of the machine learning technology on the initial feature space is no longer suitable for the changed feature space.
Taking the motion recognition (more specifically, gesture recognition) shown in fig. 1 as an example, many factors (e.g., motion artifacts, hair and sweat on the skin surface, electrical noise from power lines and external power sources, etc.) may affect the quality of physiological signals (e.g., electromyographic signals) collected by the electrodes, and as time progresses, a case may occur in which physiological signals from some existing electrodes are absent (as indicated by the absence of some existing electrodes); in addition, as recognition needs or rehabilitation goals change, additional electrodes may need to be added to support these changes (see fig. 2). The absence and the addition of the electrodes can cause the absence and the addition of features extracted from physiological signals acquired by the electrodes, namely, the feature space changes, so that the existing recognition model constructed based on the initial feature space cannot adapt to the motion recognition after the change of the feature space.
To cope with dynamic changes in feature space, one solution that is currently common is to reconstruct the recognition model using the feature space changed data set. However, this solution has three limitations: firstly, the learned useful knowledge in the existing recognition model is ignored; secondly, a large amount of data is needed for reconstructing the identification model, and the requirement of the reconstructed identification model on the sample size can not be met by the increased data set easily; more importantly, this solution is difficult to adapt to the dynamic changes of the feature space.
Disclosure of Invention
In order to overcome the problems in the prior art, according to an embodiment of the present invention, there is provided a feature adaptive motion recognition method, including: constructing a recognition model based on an initial dataset, wherein the initial dataset comprises initial features from all electrodes in an existing electrode set corresponding to a plurality of physiological signal acquisitions and labels corresponding to the plurality of physiological signal acquisitions, the labels being used to indicate categories of actions; acquiring a feature set collected from a target to be identified; determining whether a feature from one or more electrodes in the existing electrode set is missing from the acquired feature set; in response to determining that a feature from one or more electrodes in the existing electrode set is missing from the acquired feature set, treating the one or more electrodes as one or more missing electrodes, supplementing the acquired feature set with features from the one or more missing electrodes; and taking the acquired feature set as the input of the recognition model to obtain an action recognition result.
The above method, wherein supplementing the acquired feature set with features from the one or more missing electrodes comprises: for each missing electrode of the one or more missing electrodes, selecting the non-missing electrode of the existing electrode set that is most relevant to the missing electrode, and supplementing the features from the missing electrode according to the features from the selected non-missing electrode in the acquired feature set. Wherein the non-missing electrode represents an electrode in the existing electrode set other than the one or more missing electrodes.
The above method may further comprise: for each electrode in the existing electrode set, constructing an electrode sketch of that electrode based on the initial data set; and after obtaining the action recognition result, updating the electrode sketch of each electrode in the existing electrode set. Wherein supplementing features from the missing electrode according to features from the selected non-missing electrode in the acquired feature set comprises:
the features from the missing electrode were calculated based on the following formula:
wherein i denotes that the missing electrode is the ith electrode in the existing electrode set, j denotes that the selected non-missing electrode is the jth electrode in the existing electrode set, t denotes the round corresponding to the acquired feature set, (o) denotes the existing electrode,
indicates the characteristic from the missing electrode, S i Electrode sketch showing the missing electrode, S j An electrode sketch indicating the selected non-missing electrode, based on the status of the electrode, and/or based on the status of the electrode>
Representing features from the selected non-missing electrodes in the acquired feature set; and
supplementing features from the missing electrode in the acquired feature set.
The above method may further comprise: for each missing electrode of the one or more missing electrodes, a degree of correlation between the missing electrode and each non-missing electrode of the existing electrode set is calculated. Wherein the degree of correlation between the two electrodes includes a pearson correlation coefficient between electrode sketches of the two electrodes and a spatial distance between the two electrodes.
The above method may further comprise: determining whether features from the newly added electrode exist in the acquired feature set; and, in response to determining that features from a newly added electrode are present in the acquired feature set, construct an incremental dataset based on the acquired feature set, update the identification model based on the incremental dataset, and add the newly added electrode to the existing electrode set.
In the above method, updating the recognition model based on the incremental dataset comprises:
calculating an evaluation value of each individual classifier in the recognition model based on:
wherein, S (h) m ) Representing the m-th individual classifier h in the recognition model m The evaluation value of (1); acc (h) m ) Represents h m The classification accuracy of (2); # feature m Representation construction h m The number of all features used;
representing the number of features from all electrodes in the existing electrode set, t indicatingThe turn corresponding to the acquired feature set;
selecting individual classifiers with a preset proportion from all the individual classifiers in the identification model from small to large according to the evaluation values as the individual classifiers to be updated; and
updating the individual classifier to be updated based on the incremental dataset.
In the above method, updating the individual classifier to be updated based on the incremental data set includes: calculating the impurity degree of leaf nodes which can be reached by the incremental dataset in the individual classifier to be updated; and if the impurity degree of the leaf node reaches a preset threshold value, splitting the leaf node.
In the above method, updating the individual classifier to be updated based on the incremental data set includes:
for each internal node in the individual classifier to be updated, calculating a divergence index corresponding to each candidate segmentation threshold value in the candidate segmentation threshold value set of the internal node based on the following formula:
wherein v represents an internal node v in the individual classifier to be updated; d v Representing a subset of the incremental dataset within which an internal node v can be reached; phi (v) represents the partitioning property of the internal node v; τ (v) represents the current segmentation threshold for internal node v; q L And Q R Respectively representing D when the current segmentation threshold is tau (v) v The label distribution in (1); q' L And Q' R Respectively, D when using the candidate segmentation threshold v The label distribution in (1); i D v I denotes D v The size of (d); i D L,τ L represents the subset size of the left child node which can reach the internal node v in the incremental data set when the current segmentation threshold is tau (v); | D R,τ L represents the subset size of the right child node which can reach the internal node v in the incremental data set when the current segmentation threshold is tau (v); 2JSD (P) 1 ,P 2 )＝KL(P 1 ||P avg )+KL(P 2 ||P avg ) And P is avg ＝(P 1 +P 2 ) (iv)/2, wherein KL (. Cndot.) represents the Kullback-Leibler divergence;
and selecting the candidate segmentation threshold corresponding to the maximum divergence index from the candidate segmentation threshold set to update the current segmentation threshold of the internal node.
The method may further include constructing an electrode sketch for the newly added electrode.
The above method may further comprise: after obtaining an action recognition result, selecting an individual classifier to be updated from the recognition model; and for the individual classifier to be updated, carrying out clipping operation on the internal node of the individual classifier.
The embodiment of the invention can adapt to the dynamic change of the characteristic space. In a real action recognition scene, the characteristic loss is usually unplanned and unexpected, so for the lost characteristic, the invention selects the non-lost electrode with the highest correlation degree with the lost electrode, and solves the characteristic from the lost electrode according to the characteristic from the non-lost electrode, so that the action recognition can be carried out by utilizing the existing recognition model under the condition of the characteristic loss. In addition, the characteristic increase is usually planned and expected, so the invention updates the recognition model on the basis of the new characteristics, so that the recognition model is more consistent with the actual motion recognition scene, thereby improving the motion recognition accuracy. The identification model is updated by adjusting the structure and/or the segmentation threshold, so that new knowledge is learned while useful knowledge learned by the identification model is not lost, and the requirement of the updated identification model on the sample size is met by the incremental data set. Further, the recognition model can be updated even under the condition that no new features are added, for example, useless internal nodes are reduced, and therefore the accuracy of action recognition is further improved.
Drawings
Example embodiments will be described in detail with reference to the accompanying drawings, which are intended to depict example embodiments and should not be construed as limiting the intended scope of the claims. The drawings are not to be considered as drawn to scale unless explicitly indicated.
FIG. 1 is a schematic diagram of incremental changes in electrodes over time in a gesture recognition application;
FIG. 2 is a schematic diagram of the incremental change in feature space over time according to one embodiment of the present invention;
FIG. 3 is a flow diagram of a feature adaptive motion recognition method according to one embodiment of the present invention;
FIG. 4 is a flow diagram of a feature adaptive motion recognition method according to another embodiment of the present invention;
FIG. 5 is a schematic diagram of an individual classifier according to one embodiment of the present invention;
fig. 6 (a) -6 (f) are schematic diagrams showing results obtained by performing a comparative experiment on the feature adaptive motion recognition method provided by the present invention and a conventional method.
Detailed Description
In order to make the objects, technical solutions and advantages of the present invention more apparent, the present invention is further described in detail by embodiments with reference to the accompanying drawings. It should be understood that the specific embodiments described herein are merely illustrative of the invention and do not limit the invention.
Before describing embodiments of the present invention, the symbols used in the embodiments and the corresponding definitions are first introduced, as shown in table 1:
TABLE 1
According to one embodiment of the invention, a feature-adaptive motion recognition method is provided, which is suitable for implementing motion recognition in the case of feature loss.
Referring to fig. 3, the feature adaptive motion recognition method includes: s11, constructing a recognition model based on an initial data set, wherein the initial data set comprises initial characteristics of all electrodes in an existing electrode set corresponding to multiple times of physiological signal acquisition and labels corresponding to the multiple times of physiological signal acquisition, and the labels are used for indicating the category of actions; s12, acquiring a feature set acquired from a target to be identified; s13, determining whether the acquired feature set lacks the features of one or more electrodes in the existing electrode set, obtaining the missing electrodes in the existing electrode set in response to the determination that the features of one or more electrodes in the existing electrode set are absent, and supplementing the features from the missing electrodes in the acquired feature set; s14, taking the acquired feature set as the input of an identification model to obtain an action identification result; and S15, if the action recognition is not finished, returning to the step S12, otherwise, finishing the method.
The feature-adaptive motion recognition method will be described below with reference to fig. 3. In particular, construction of the recognition model and continuous motion recognition based on the recognition model are described in rounds (also referred to as acquisition rounds) as time advances. Wherein the data collected in round 1 is used to identify the construction of the model; in 2,3, \8230, data are collected from objects to be recognized, respectively, in N rounds, and motion recognition is performed through the constructed recognition model based on the collected data.
And S11, constructing an identification model based on the initial data set. Step S11 includes the following substeps:
and S111, constructing an initial data set. Wherein the initial dataset includes initial features from all electrodes in the existing electrode set corresponding to the plurality of physiological signal acquisitions and labels corresponding to the plurality of physiological signal acquisitions, the labels indicating a category of action.
The electrode from which the physiological signal could be acquired in round 1 was taken as an existing electrode, and the array of a plurality of existing electrodes constituted an existing electrode set (in round 1, there were no missing electrodes and added electrodes). As shown in fig. 1, an array of existing electrodes may be arranged on the arm for acquiring physiological signals (e.g. myoelectric signals).
Specifically, constructing the initial data set includes:
n in round 1 using each existing electrode in the existing electrode set 1 The next physiological signal acquisition to obtain n 1 A sample of a physiological signal.
N acquired in round 1 for each existing electrode 1 Extracting d characteristics from each physiological signal sample in the physiological signal samples to obtain initial characteristics from the existing electrode in the 1 st round
Is n 1 X d matrix, i.e. <>
Wherein, (o) represents existing electrodes, 1 represents 1 st round, and i represents ith existing electrode in existing electrode set, and/or>
Representing the number of existing electrodes in the existing electrode set in the 1 st round;
from the initial features of all existing electrodes from the set of existing electrodes in round 1 and n acquired in round 1 1 Constructing an initial data set D by using labels corresponding to physiological signal samples 1 ，
Wherein +>
Indicates the initial characteristic of all existing electrodes from the existing electrode set in round 1, and ` Ks `>
Is->
Is based on a matrix of->
Wherein->
Representing the number of existing electrodes in the existing electrode set in the 1 st round; y is 1 Denotes n collected from round 1 1 Labels corresponding to individual physiological signal samples (labels indicating the category of action, e.g. gesture), Y 1 Is n 1 X c matrix, i.e. <>
Where c represents the total number of categories.
In the above, for each physiological signal sample acquired in the 1 st round for each existing electrode, d features are extracted therefrom. For example, 9 (d = 9) features as shown in table 2 may be extracted:
TABLE 2
And S112, constructing a random forest based on the initial data set as a recognition model H for motion recognition. Wherein the random forest comprises M decision trees (i.e., M individual classifiers), and each individual classifier is trained from an initial data set; each trained individual taxonomy classifier includes a root node, an internal node, and a leaf node, each internal node having a partitioning attribute (corresponding to a feature in the initial dataset) and a partitioning threshold for classifying datasets arriving at the internal node.
The result of motion recognition on a given sample x for a recognition model with M individual classifiers can be expressed as follows:
where H (x) represents the output of the recognition model for a given sample x,
represents the m-th individual classifier h m For the output of a given sample x on the c 'th class, M represents the total number of individual classifiers in the recognition model, 1 ≦ c' ≦ c, and c represents the total number of classes.
S113, constructing an initial electrode sketch of each existing electrode in the existing electrode set, wherein the electrode sketch is used for describing the characteristics of physiological signals collected by the corresponding electrode.
From the initial data set constructed in substep S111, the corresponding acquired data for each existing electrode in the set of existing electrodes in round 1 may be obtained
Wherein i denotes the i-th existing electrode in the existing electrode set, is present>
Indicates the number of existing electrodes in the existing electrode set in run 1, and as described above,
representing an initial characteristic from the ith existing electrode in round 1->
For the ith existing electrode in the existing electrode set, the current can be measured from D 1,i In
Middle selection of l i An electrode sketch S for constructing the existing electrode i So that (S) i ) T S i ≈(D 1,i ) T D 1,i And l i ＜＜n 1 . Wherein S is i Is 1 i X d matrix, i.e.
For example, the classification matrix sketch technique proposed by Xin Mu et al can be employed to select ÷ based on>
To construct an electrode sketch. By constructing an electrode sketch, a lower dimensional matrix (e.g., S) may be used i ) To approximate a higher-dimensional matrix (e.g. < >)>
)。
And S12, acquiring a feature set acquired from the target to be identified.
An existing electrode set for constructing a recognition model is arranged on a target to be recognized (for example, a human body or an animal whose action is to be recognized), and in addition, a newly added electrode can be arranged on the target to be recognized. N in the t (t ≧ 2) th round using electrodes arranged on the target to be recognized t Acquiring n of the electrodes in the t round t A sample of a physiological signal. It will be appreciated that in the t-th pass, some of the existing electrodes in the existing electrode set from which the recognition model was constructed may be missing for the reason (i.e. missing electrodes), resulting in failure to acquire features from the missing electrodes in the t-th pass.
For each of the plurality of electrodes (i.e., including the non-missing electrode and the added electrode in the existing electrode set, wherein the non-missing electrode refers to the electrode in the existing electrode set other than the missing electrode), n collected in the t-th pass from that electrode t D features are extracted from each physiological signal sample in the physiological signal samples to obtain the features from the electrode in the t round.
And constructing a feature set by using the features from the plurality of electrodes (namely, the non-missing electrodes and the newly added electrodes in the existing electrode set) in the t-th round to obtain the feature set collected from the target to be identified.
Step S13, determining whether the characteristics of one or more existing electrodes in the existing electrode set are lacked in the characteristic set acquired in the step S12, obtaining the lacked electrodes in the existing electrode set in response to determining that the characteristics of one or more electrodes in the existing electrode set are lacked, and supplementing the characteristics from the lacked electrodes in the acquired characteristic set. Step S13 includes the following substeps:
s131, determining whether the acquired feature set lacks features from one or more existing electrodes in the existing electrode set according to the physiological signals acquired in the t round for constructing the acquired feature set.
For example, a marker may be added to the physiological signal sent by the electrode to indicate from which electrode the signal came. From this identification, it can be determined from which existing electrode(s) the features from which are missing from the acquired feature set.
S132, in response to determining that the characteristics of one or more electrodes from the existing electrode set are missing, one or more missing electrodes from the existing electrode set are obtained.
For example, if the physiological signal acquired in the t-th pass does not contain an identification of some existing electrode(s), then that electrode is considered to be the missing electrode.
And S133, for each missing electrode, selecting an un-missing electrode with the largest correlation degree with the missing electrode from the existing electrode set.
Specifically, for each missing electrode, the following substeps S1331-1334 are performed:
and S1331, calculating correlation between the physiological signals acquired by the missing electrode and each non-missing electrode. The pearson correlation coefficient between the electrode sketches of the two electrodes is taken as the correlation between the physiological signals acquired by the two electrodes, and is shown as the following formula:
in the formula (2), the first and second groups,
electrode sketch S representing ith electrode i Drawing S with the jth electrode j A pearson correlation coefficient in between, i ≠ j, < ≠ j >>
Has a value range of [ -1,1]And->
The larger the value of (A) indicates that the correlation between the physiological signals acquired by the ith electrode and the jth electrode is higher; cov denotes covariance; e represents expectation; />
Electrode sketch S representing ith electrode i Is based on the mean value of>
Electrode sketch S representing ith electrode i Standard deviation of (d). It is to be understood that the calculation
And &>
And calculate->
And &>
In a similar manner, the ≧ calculation may be made with reference to equations (3) and (4)>
And &>
And S1332, calculating the space distance between the missing electrode and each non-missing electrode. Wherein the spatial distance between the two electrodes can be calculated according to:
wherein the content of the first and second substances,
indicating the spatial distance between the ith and jth electrodes, i ≠ j +>
Has a value range of (0, 1)]；pos(e i )＝<haxis i ,vaxis i >，pos(e j )＝<haxis j ,vaxis j >Respectively representing the coordinates of the ith electrode and the jth electrode on a two-dimensional space; i | pos (e) i ),pos(e j ) | | represents the euclidean distance between the ith electrode and the jth electrode.
In addition, for two electrodes in the same row or column of the electrode array (as shown in fig. 1), after the spatial distance between the two electrodes is calculated, a predetermined increment can be added to further improve the spatial position correlation between the two electrodes in the same row or column of the array. Preferably, the increment can be set to 1, i.e. 1
Wherein 1 is->
The upper limit value of (2).
And S1333, calculating distribution differences between the missing electrodes and each non-missing electrode according to the correlation between the physiological signals acquired by the electrodes and the space distance between the electrodes, wherein the distribution differences are used for measuring the correlation degree between the electrodes. Wherein the difference in distribution between the two electrodes can be calculated according to:
wherein DD (i.j) represents a distribution difference between the ith electrode and the jth electrode, i ≠ j; alpha and beta represent parameters, are regularization coefficients and are used for limiting the value distribution of DD (i.j) within a certain range; preferably, when α =1 and β =4, the value of DD (i.j) may be made to range from [0,1]；
Representing the correlation between the physiological signals acquired by the ith electrode and the jth electrode; />
Representing the spatial distance between the ith and jth electrodes. If the distribution difference between the two electrodes is larger, the two electrodes have more similar signal characteristics and closer spatial positions; conversely, if the distribution difference between the two electrodes is small, it indicates that the two electrodes have signal characteristics with large difference and far spatial positions.
S1334, selecting the non-missing electrodes corresponding to the maximum distribution difference, as shown in the following formula:
j＝arg max j' (DD(i,j')),j≠i (7)
and S134, for each missing electrode, searching the features from the selected non-missing electrodes in the t round in the feature set acquired in the step S12, and calculating the features from the missing electrodes in the t round according to the searched features.
Suppose that
Evaluation by means of a false inverse->
I.e., the features from the missing electrode in round t, as shown in the following equation:
wherein i represents that the missing electrode is the first electrode in the existing electrode set
A plurality of electrodes, j indicates whether the non-missing electrode selected for the missing electrode is a fifth or fifth electrode in the existing electrode set>
A plurality of electrodes>
Representing the number of existing electrodes in the existing electrode set in the t-th round; />
Indicates a characteristic of the i-th electrode from the existing set of electrodes in the t-th round, and ` H `>
Representing features from a jth electrode in the set of existing electrodes in the tth round; s i Electrode sketch representing the ith electrode in the existing electrode set, S j An electrode sketch representing the jth electrode in the existing electrode set.
S135. In the feature set acquired in step S12, the features from each missing electrode in the t-th round are supplemented.
And S14, taking the acquired feature set as the input of the recognition model to obtain an action recognition result.
And S15, if the action recognition is not finished, returning to the step S12, otherwise, finishing the method.
In step S15, if the operation recognition is not completed, the following processing is executed before returning to step S12:
updating the electrode sketch for each existing electrode in the existing electrode set, and let t = t + 1. Similar to constructing the initial electrode sketch, updating the electrode sketch for each existing electrode in the existing electrode set includes: obtaining the characteristics of each existing electrode from the existing electrode set in the tth round according to the acquired characteristic set; from the features from each existing electrode in the set of existing electrodes in the t-th pass and n acquired in the t-th pass t The label corresponding to each physiological signal sample constructs the corresponding acquired data of each existing electrode in the existing electrode set in the t round
Wherein it is present>
Representing features from the ith existing electrode in the tth round, based on the characteristics of the preceding electrode>
Indicating the number of existing electrodes in the existing electrode set in the t-th run, Y t Representing n collected in the t-th round t Labels corresponding to samples of physiological signals, Y t Is n t X c matrix, i.e. <>
c represents the total number of categories; for the ith existing electrode in the existing electrode set, the current can be measured from D t,i In (A)>
Middle selection of l i Line-to-line updating the electrode sketch S of the existing electrode i So that (S) i ) T S i ≈(D t,i ) T D t,i And l i ＜＜n t . Wherein S is i Is 1 i X d matrix, i.e. <>
The feature adaptive motion recognition method provided by the embodiment can realize motion recognition by using the existing recognition model under the condition of feature loss.
In the above embodiments, an electrode sketch is constructed and updated for each existing electrode in the existing electrode set, which can be used to calculate correlations between physiological signals acquired by the electrodes, as well as to calculate features from the missing electrodes. In other embodiments, the electrode sketch may not be constructed for the existing electrodes, for example, the spatial distance between the electrodes may be used to measure the distribution difference (i.e., the degree of correlation) between the electrodes; and for a missing electrode in the t (t ≧ 2) round, the feature from that missing electrode in the t round can be supplemented with the feature from that electrode in the previous round.
Furthermore, although in the above described embodiments pearson's correlation coefficient is used to calculate the correlation between physiological signals acquired by the electrodes, in other embodiments other linear correlation methods may be used to measure the correlation between variables; in addition, the spatial distance between the electrodes may also be calculated using the euclidian distance of the electrodes in the three-dimensional space.
According to an embodiment of the invention, a feature-adaptive motion recognition method is also provided. The method is suitable for realizing motion recognition under the condition of feature loss, and can update the recognition model under the condition of feature increase so as to enable the recognition model to better conform to the actual motion recognition scene.
Referring to fig. 4, the feature adaptive motion recognition method includes: s21, constructing an identification model based on an initial data set, wherein the initial data set comprises initial characteristics of all electrodes in an existing electrode set corresponding to multiple times of physiological signal acquisition and labels corresponding to the multiple times of physiological signal acquisition, and the labels are used for indicating the category of actions; s22, acquiring a feature set acquired from a target to be identified; s23, determining whether the acquired feature set lacks the features of one or more electrodes in the existing electrode set, obtaining the missing electrodes in the existing electrode set in response to the determination that the features of one or more electrodes in the existing electrode set are absent, and supplementing the features from the missing electrodes in the acquired feature set; s24, taking the acquired feature set as an input of an identification model to obtain an action identification result; s25, determining whether the obtained feature set has features from the newly added electrode or not, establishing an incremental data set in response to the determination that the features from the newly added electrode exist, updating the identification model based on the incremental data set, and adding the newly added electrode into the existing electrode set; and S26, returning to the S22 if the action identification is not finished, otherwise, finishing the method.
The feature adaptive motion recognition method will be described below with reference to fig. 4.
S21, constructing a recognition model based on the initial data set, comprising the following steps: constructing an initial dataset D for round 1 1 ，
The initial dataset comprises initial features from all electrodes in an existing electrode set corresponding to a plurality of physiological signal acquisitions and labels corresponding to the plurality of physiological signal acquisitions, the labels being used to indicate a category of action; constructing a random forest as an identification model for action identification based on the initial data set, wherein the random forest comprises M individual classifiers, each individual classifier comprises a root node, an internal node and a leaf node, and each internal node has a segmentation attribute and a segmentation threshold; and constructing an initial electrode sketch of each existing electrode in the existing electrode set.
And S22, acquiring a feature set acquired from the target to be identified.
The existing electrode set for constructing the recognition model is arranged on the target to be recognized, and in addition, the newly added electrodes can be arranged on the target to be recognized. Using electrodes arranged on the target to be identified at the t (t ≧ 2) th wheelIs subjected to n t Acquiring n of the electrodes (including the non-missing electrodes and the newly added electrodes in the existing electrode set) in the t round of physiological signal acquisition t And d features are extracted from each physiological signal sample.
Step S23, determining whether the feature set acquired in step S22 lacks features from one or more existing electrodes in the existing electrode set, obtaining a missing electrode in the existing electrode set in response to determining that the features from one or more electrodes in the existing electrode set are missing, and supplementing the features from the missing electrode in the acquired feature set.
And S24, taking the acquired feature set as the input of the recognition model to obtain an action recognition result.
And S25, determining whether the acquired feature set has features from the newly added electrodes or not, constructing an incremental data set in response to the determination that the features from the newly added electrodes exist, updating the identification model based on the incremental data set, and adding the newly added electrodes into the existing electrode set. Step S25 includes the following substeps:
s251, determining whether features from the newly added electrode (i.e., one or more electrodes outside the existing electrode set) exist in the acquired feature set according to the physiological signals acquired in the t-th round for constructing the acquired feature set.
For example, a marker may be added to the physiological signal sent by the electrode to indicate from which electrode the signal came. From this identification, it can be determined from which newly added electrode(s) features are present in the acquired feature set.
S252. In response to determining that the feature from the newly added electrode exists, the newly added electrode is determined. For example, the newly added electrode may be determined from an identification contained in the physiological signal.
And S253, constructing an increment data set under the condition that a newly added electrode exists.
Specifically, the t-th turn of incremental data set D is constructed based on the acquired feature set t ，
Wherein it is present>
Indicates that in the tth round all existing electrodes from the existing electrode set have been characterized as being present>
Indicates the number of features from all existing electrodes in the existing electrode set in the tth round, and ` H `>
Representing the number of existing electrodes in the existing electrode set in the t round; />
Means characteristics from all newly added electrodes in the tth round>
Represents the number of features from all newly added electrodes in the tth round, and->
Indicating the number of newly added electrodes in the t round; y is t Representing n collected in the t-th round t Label corresponding to a sample of a physiological signal, Y t Is n t X c matrix, i.e. [ means ] for>
And c represents the total number of categories.
And S254, updating the identification model based on the incremental data set. The method comprises the following substeps:
s2541, selecting an individual classifier to be updated in the recognition model.
The individual classifier is evaluated by its accuracy and all the features used to construct it, as shown in the following equation:
wherein, S (h) m ) Represents the m-th individual classifier h m M is more than or equal to 1 and less than or equal to M, wherein M represents the total number of the individual classifiers; acc (h) m ) Denotes h m The classification accuracy of (2); # feature m Represents construction h m The number of all features used;
representing the number of features from all existing electrodes in the existing electrode set at round t.
After the evaluation values of all the individual classifiers in the identification model are obtained, the individual classifiers are ranked according to the evaluation values from small to large, and the top M × δ individual classifiers are selected as the individual classifiers to be updated. Wherein δ represents a preset update ratio of the individual classifiers, and M represents the total number of the individual classifiers.
S2542, for each individual classifier to be updated, performing structure adjustment and/or segmentation threshold adjustment on the individual classifier based on the incremental data set.
A. Structural adjustment (FIDE _ s for short):
for each individual classifier to be updated, the impurity degree (e.g., measured by error rate, entropy, kini index, etc.) of a leaf node that can be reached by the incremental dataset in the individual classifier is calculated, and if the calculated impurity degree of the leaf node reaches a predetermined threshold (i.e., indicating that the leaf node is not pure), the splitting operation is performed on the leaf node.
B. Segmentation threshold adjustment (FIDE _ m for short):
and presetting a corresponding candidate segmentation threshold value set for each internal node in each individual classifier.
For each internal node in each individual classifier to be updated, calculating the divergence index corresponding to each candidate segmentation threshold value in the candidate segmentation threshold value set of the internal node by the following formula:
2JSD(P 1 ,P 2 )＝KL(P 1 ||P avg )+KL(P 2 ||P avg ) (11)
P avg ＝(P 1 +P 2 )/2 (12)
wherein v represents an internal node v in the individual classifier to be updated; d v Represents a subset of reachable nodes v in the incremental dataset; phi (v) represents the partitioning property of node v (corresponding to a feature in the feature space); τ (v) represents the current segmentation threshold for node v; q L And Q R Respectively representing D when the current segmentation threshold is tau (v) v The label distribution in (1); q' L And Q' R Respectively representing D when using the candidate segmentation threshold v The label distribution in (1); | D v I denotes D v The size of (d); i D L,τ L represents the subset size of the left child node which can reach the node v in the incremental data set when the current segmentation threshold is tau (v); i D R,τ L represents the subset size of the right child node which can reach the node v in the incremental data set when the current segmentation threshold is tau (v); p avg Represents the distribution P 1 And P 2 The mean value of (a); jensen-Shannon Divergence (JSD) is a symmetric extension of Kullback-Leibler Divergence (KL), JSD (P 1 ,P 2 ) For measuring two distributions P 1 And P 2 The distance of (c). In formula (10), JSD (Q' L ,Q L ) And JSD (Q' R ,Q R ) Respectively representing the JSD distances of the left and right subtrees of node v when it uses the candidate and current partitioning thresholds τ (v). In the formula (10), the divergence index has a value range of [0,1 ]]The larger the divergence index is, the more optimal the corresponding candidate segmentation threshold is.
After traversing all candidate partition thresholds in the candidate partition threshold set, selecting the candidate partition threshold corresponding to the maximum divergence index to update the current partition threshold of the node v.
The update process of the recognition model can be seen in algorithm 1.
S255, adding the newly added electrode into the existing electrode set
And S26, returning to the S22 if the action identification is not finished, otherwise, finishing the method.
In step S26, if the motion recognition is not finished, the electrode sketch of each existing electrode in the existing electrode set is also updated before returning to step S22, wherein the electrode sketch is constructed for the new feature and t = t + 1.
The feature adaptive motion recognition method provided by the embodiment can realize motion recognition by using the existing recognition model under the condition of feature loss, and update the recognition model on the basis of the new added features so that the recognition model is more consistent with the actual motion recognition scene, thereby improving the precision of motion recognition. The identification model is updated by adjusting the structure and/or the segmentation threshold, so that new knowledge is learned while useful knowledge learned by the identification model is not lost, and the requirement of the updated identification model on the sample size is met by the incremental data set.
In the above-described embodiment, the recognition model is updated as the feature increases. According to another embodiment of the invention, no matter whether the newly added electrode exists or not, the recognition model can be updated by reducing useless internal nodes after the action recognition result is obtained, so that the action recognition accuracy is further improved.
Specifically, after an action recognition result is obtained, constructing an incremental data set of the t-th turn based on the obtained feature set, and selecting an individual classifier to be updated in the recognition model; for each individual classifier to be updated, calculating a subtree penalty (i.e., evaluating the penalty value of the subtree having the internal node as the root node) and a leaf penalty (i.e., the penalty value if the internal node is clipped to a leaf node) for its internal node; nodes with a clipping sub-tree loss greater than a leaf loss or nodes unreachable by the incremental dataset.
Referring to fig. 5, assume that the classification label of the sample on the node v is 1 (same as the leaf node class determination mechanism, determined in the training process according to the majority voting method), and two child nodes v thereof l And v r The classification labels of (1) and (0) are respectively. <xnotran> , 10 v, {1,1,1,1,1,1,1,1,0,0}, 1 , , 20%; </xnotran> The 10 samples arrive at node v along a branch of node v l And v r <xnotran> , 5 , {1,1,1,1,1} {1,1,1,0,0}, {1,1,1,1,1} {0,0,0,0,0}, 0 3 , 0% 60%. </xnotran> At this time, for node v, its subtree loss is v l And v r The sum of the weighted losses of (a) is 5/10 × 0% +5/10 × 60% =30%; whereas if v is considered a leaf node, the loss is only 20%. The subtree loss is now greater than the leaf loss, and therefore the internal node v will be pruned away.
To verify the beneficial effects of the feature adaptive motion recognition method provided by the present invention, the method provided by the present invention will be compared with other existing methods in gesture recognition applications on three reference datasets (including CapgMyo, CSL-hdemg, and ninapor) below. An associative thinkstate 10G4S04E00 (Intel Core i 7-6700/169B DDR3) desktop is adopted in the experiment, and a development platform is Matlab R2018b. And the update ratio δ, the total number M of individual classifiers, and the number of candidate features of each node in the individual classifiers are set to 0.4, 100, and
wherein it is present>
Representing the number of features utilized to construct the individual classifier.
Table 3 shows the details of three reference data sets:
TABLE 3
CapgMyo: the myoelectricity is collected by using a high-density myoelectricity device (8 rows by 16 columns, sampling frequency 1000 hz) and is composed of three sub data sets (Capg) 1 、Capg 2 And Capg 3 ) Composition of the three subdata sets [ number of subjects, number of gestures, number of repetitions]Are respectively [18,8,10 ]]、[20,8,10]And [10,12,10]。
CSL-hdemg: using high density electromyography acquisition (8 rows x 24 columns, sampling frequency 2048 hz), CSL-hdemg contains 27 gesture data from 5 subjects, each gesture acquired in 5 different scenes, each scene being repeated 10 times.
NinaPro: ninaPro contains 8 sub-datasets, two representative sub-datasets of which are selected, including Nina 1 And Nina 2 。Nina 1 The acquisition equipment comprises 10 electrodes, and the sampling frequency is 100hz; nina 2 The acquisition device of (1) is an electrode array with 2 rows by 8 columns, the sampling frequency is 200hz, and the number of subjects, the number of gestures and the number of repetitions of two subdata sets]Are respectively [27,52,10 ]]And [18,8,6]。
For each electrode, its 9 time and frequency features were extracted (as shown in table 2). Due to Capg 1 、Capg 2 、Capg 3 、CSL-hdemg、Nina 1 、Nina 2 There were 128, 192, 10, 16 electrodes respectively, so 1152, 1512, 90 and 144 features were extracted on these 6 datasets respectively.
Other prior methods for comparative experiments are described below:
extreme random trees (ET): an ensemble learning method based on decision trees in which segmentation attributes and segmentation thresholds are selected randomly.
K-grouped random forest (K-striified random forest, KSRF): a grouping random forest method is used for grouping different characteristics and respectively constructing a decision tree model.
Improved feature incremental random forest (iFIRF): the original FIRF can only adapt to the characteristic increment scene, and the iFIRF is improved by adopting a characteristic missing value processing strategy.
Common feature missing value processing Method (MCFI): the reduced features are treated as feature deletions and padded with the most common features in the feature set. For added features, processing is done using methods similar to the present invention. Thus, there are two different variations of MCFI, including MCFI _ s and MCFI _ m.
In the ET and KSRF methods, all the initial data sets and the data streams in the incremental updating process are used to train the initial recognition model, so the ET and KSRF are the recognition models in the most ideal situation, i.e. theoretically, the effect of the method provided by the invention can only approach the ET and KSRF, but hardly exceed them.
In the comparative experiment, 10% of the total data was used as the training data set when the initial recognition model was constructed, and 80% of the total data was used as the data stream in the incremental update process, and 10% of the total data was used as the test data set. In terms of number of electrodes, capg when constructing the initial model 1 、Capg 2 、Capg 3 、CSL-hdemg、Nina 1 、Nina 2 The number of electrodes available in the data set was [28, 36,2,4, respectively](ii) a In each batch update process, there are [25, 32,2,3 respectively]Individual electrodes were added and 10% of the existing electrodes were randomly missing. In the parametric analysis experiments, 50% of the total data was used as the initial training data set, 40% of the total data was used as the data stream during the incremental update, and 10% of the total data was used as the test data set. In terms of the number of electrodes, when an initial model is constructed, 50% of all electrodes are used as available electrodes, the rest 50% of the electrodes are used as newly added electrodes, and 10% of the existing electrodes are randomly deleted. The settings in the parametric analysis experiment were the same as those in the comparative experiment except for the control parameters M and δ.
FIGS. 6 (a) -6 (f) are shown at Capg, respectively 1 、Capg 2 、Capg 3 、CSL-hdemg、Nina 1 、Nina 2 On the data set, the gesture recognition accuracy results of the present invention and other 5 existing methods, from 6 (a) -6 (f), the following conclusions can be drawn:
the identification precision of the method provided by the invention is close to the identification result under the 'ideal condition', namely the identification result of the KSRF, which proves the effectiveness of the method provided by the invention. In Capg 2 On the data set, the accuracy of using FIDE _ s is almost the same as using FIDE _ s and FIDE _ m at the same time (FIDE _ s + FIDE _ m are both simplified and represented as FIDE _ m in FIGS. 6 (a) -6 (f)); on the CSL-hdemg dataset, FIDE _ s even exceed the accuracy of KSRF on the last round, one possible explanation for this result is that the present invention is more concerned with learning new knowledge and therefore achieves better recognition results on the last round.
In most cases, the method provided by the invention is superior to other comparative methods. In addition to the KSRF method, the simultaneous use of FIDE _ s and FIDE _ m achieved superior results over the other comparison methods on 5 out of all 6 validation data sets, while FIDE _ s achieved superior results over the 4 out of 6 validation data sets.
ET is the simplest comparative method, which is described in Nina 1 The data set exceeds the method provided by the invention, and the optimal recognition result is obtained. One possible explanation is Nina 1 Only 10 discrete electrodes are included, and the method provided by the invention is more suitable for high-density electrode arrays.
The FIDE _ s method only adjusts the structure to realize the feature adaptation, and the FIDE _ s + FIDE _ m method simultaneously uses the expansion/cutting of the decision tree structure and the adjustment strategy of the segmentation threshold value to realize the feature adaptation. However, FIDE _ s outperforms FIDE _ s + FIDE _ m in most cases, indicating that the structure extension/clipping is sufficient for feature adaptation.
The initial recognition models of 5 methods, i.e. FIDE _ s, FIDE _ s + FIDE _ m, KSRF, MCFI _ s and MCFI _ m, are all the same, so the initial recognition accuracy of the 5 methods is the same.
It should be noted that some exemplary methods are depicted as flowcharts. Although a flowchart may describe the operations as being performed serially, it can be appreciated that many of the operations can be performed in parallel, concurrently, or with synchronization. In addition, the order of operations may be rearranged, for example, step S25 may be performed in synchronization with step S23 or before step S23. A process may terminate when an operation is completed, but may have additional steps not included in the figure or embodiments.
The above-described methods may be implemented by hardware, software, firmware, middleware, pseudocode, hardware description languages, or any combination thereof. When implemented in software, firmware, middleware or pseudo code, the program code or code segments to perform the tasks may be stored in a computer readable medium such as a storage medium, and a processor may perform the tasks.
It should be appreciated that the exemplary software-implemented embodiments are typically encoded on some form of program storage medium or implemented over some type of transmission medium. The program storage medium may be any non-transitory storage medium such as a magnetic disk (e.g., a floppy disk or a hard drive) or an optical disk (e.g., a compact disk read only memory or "CD ROM"), and may be read only or random access. Similarly, the transmission medium may be twisted wire pairs, coaxial cable, optical fiber, or some other suitable transmission medium known to the art.
Although the present invention has been described by way of preferred embodiments, the present invention is not limited to the embodiments described herein, and various changes and modifications may be made without departing from the scope of the present invention.
Claims (8)
1. A feature adaptive motion recognition method, the method comprising:
constructing a recognition model based on an initial dataset, wherein the initial dataset comprises initial features from all electrodes in an existing electrode set corresponding to a plurality of physiological signal acquisitions and labels corresponding to the plurality of physiological signal acquisitions, the labels being used to indicate categories of actions;
for each electrode in the existing electrode set, constructing an electrode sketch for that electrode based on the initial data set;
acquiring a feature set collected from a target to be identified;
determining whether a feature from one or more electrodes in the existing electrode set is missing from the acquired feature set;
in response to determining that a feature from one or more electrodes in the existing electrode set is missing from the acquired feature set, treating the one or more electrodes as one or more missing electrodes, supplementing the acquired feature set with features from the one or more missing electrodes; wherein supplementing features from the one or more missing electrodes in the acquired set of features comprises:
for each missing electrode of the one or more missing electrodes, selecting an undeleted electrode from the existing electrode set that is most relevant to the missing electrode, wherein the undeleted electrode represents an electrode in the existing electrode set other than the one or more missing electrodes, and calculating a characteristic from the missing electrode based on the following equation:
wherein i indicates that the missing electrode is the ith electrode in the existing electrode set, j indicates that the selected non-missing electrode is the jth electrode in the existing electrode set, t indicates the round corresponding to the acquired feature set, (o) indicates an existing electrode,
indicates the characteristic from the missing electrode, S i Electrode sketch showing the missing electrode, S j An electrode sketch indicating the selected non-missing electrode, based on the status of the electrode, and/or based on the status of the electrode>
Representing a set of features in said acquiredCombining features from the selected non-missing electrodes, and
supplementing features from the missing electrode in the acquired feature set;
taking the acquired feature set as the input of the recognition model to obtain an action recognition result; and
after obtaining the action recognition result, updating the electrode sketch of each electrode in the existing electrode set.
2. The method of claim 1, further comprising, for each missing electrode of the one or more missing electrodes, calculating a degree of correlation between the missing electrode and each non-missing electrode of the existing electrode set;
wherein the degree of correlation between the two electrodes includes a pearson correlation coefficient between electrode sketches of the two electrodes and a spatial distance between the two electrodes.
3. The method according to claim 1 or 2, characterized in that the method further comprises:
determining whether a feature from a newly added electrode is present in the acquired feature set; and
in response to determining that features from a newly added electrode are present in the acquired feature set, constructing an incremental dataset based on the acquired feature set, updating the recognition model based on the incremental dataset, and adding the newly added electrode to the existing electrode set.
4. The method of claim 3,
constructing the recognition model based on the initial dataset includes:
constructing a plurality of individual classifiers based on the initial dataset to form a random forest, the random forest being used as the recognition model; and
updating the recognition model based on the incremental dataset comprises:
calculating an evaluation value for each individual classifier in the recognition model based on:
wherein, S (h) m ) Representing the m-th individual classifier h in the recognition model m The evaluation value of (1); acc (h) m ) Represents h m The classification accuracy of (2); # feature m Represents construction h m The number of all features used;
representing the number of features from all electrodes in the existing electrode set, t indicating the turn to which the obtained feature set corresponds;
selecting individual classifiers with a preset proportion from all the individual classifiers in the identification model from small to large according to the evaluation values as the individual classifiers to be updated; and
updating the individual classifier to be updated based on the incremental dataset.
5. The method of claim 4, wherein updating the individual classifiers to be updated based on the incremental dataset comprises:
calculating the impurity degree of leaf nodes which can be reached by the incremental dataset in the individual classifier to be updated;
if the leaf node's impurity level reaches a predetermined threshold, then a splitting operation is performed on the leaf node.
6. The method of claim 4 or 5, wherein updating the individual classifiers to be updated based on the incremental dataset comprises:
for each internal node in the individual classifier to be updated, calculating a divergence index corresponding to each candidate segmentation threshold in the candidate segmentation threshold set of the internal node based on the following formula:
wherein v represents an internal node v in the individual classifier to be updated; d v Representing a subset of the incremental dataset within which an internal node v can be reached; phi (v) represents the partitioning property of the internal node v; τ (v) represents the current segmentation threshold for internal node v; q L And Q R Respectively representing D when the current segmentation threshold is tau (v) v The label distribution in (1); q' L And Q' R Respectively, D when using the candidate segmentation threshold v The label distribution in (1); i D v | denotes D v The size of (d); | D L,τ L represents the subset size of the left child node which can reach the internal node v in the incremental data set when the current segmentation threshold is tau (v); i D R,τ L represents the subset size of the right child node which can reach the internal node v in the incremental data set when the current segmentation threshold is tau (v);
and P is avg ＝(P 1 +P 2 ) (ii)/2, wherein KL (. Cndot.) represents the Kullback-Leibler divergence;
and selecting the candidate segmentation threshold corresponding to the maximum divergence index from the candidate segmentation threshold set to update the current segmentation threshold of the internal node.
7. The method of claim 3, further comprising:
and constructing an electrode sketch for the newly added electrode.
8. The method according to claim 1 or 2,
constructing the recognition model based on the initial dataset includes:
constructing a plurality of individual classifiers based on the initial dataset to form a random forest, the random forest being the recognition model; and
the method further comprises the following steps:
after obtaining an action recognition result, selecting an individual classifier to be updated from the recognition model; and
and for the individual classifier to be updated, clipping operation is carried out on the internal node of the individual classifier.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110268443.9A CN112861796B (en) | 2021-03-12 | 2021-03-12 | Feature adaptive motion recognition method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110268443.9A CN112861796B (en) | 2021-03-12 | 2021-03-12 | Feature adaptive motion recognition method |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN112861796A CN112861796A (en) | 2021-05-28 |
| CN112861796B true CN112861796B (en) | 2023-04-07 |
Family
ID=75994273
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110268443.9A Active CN112861796B (en) | 2021-03-12 | 2021-03-12 | Feature adaptive motion recognition method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN112861796B (en) |
Family Cites Families (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US8095210B2 (en) * | 2007-01-19 | 2012-01-10 | California Institute Of Technology | Prosthetic devices and methods and systems related thereto |
| US11113337B2 (en) * | 2016-09-08 | 2021-09-07 | Indian Institute Of Technology Bombay | Method for imputing missed data in sensor data sequence with missing data |
| CN107480697B (en) * | 2017-07-12 | 2020-04-03 | 中国科学院计算技术研究所 | Myoelectric gesture recognition method and system |
| CN108717548B (en) * | 2018-04-10 | 2020-09-15 | 中国科学院计算技术研究所 | Behavior recognition model updating method and system for dynamic increase of sensors |
| CN111967495A (en) * | 2020-07-06 | 2020-11-20 | 中国科学院计算技术研究所 | Classification recognition model construction method |
-
2021
- 2021-03-12 CN CN202110268443.9A patent/CN112861796B/en active Active
Also Published As
| Publication number | Publication date |
|---|---|
| CN112861796A (en) | 2021-05-28 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN107822622B (en) | Electrocardiogram diagnosis method and system based on deep convolutional neural network | |
| CN109671102B (en) | Comprehensive target tracking method based on depth feature fusion convolutional neural network | |
| CN109543727B (en) | Semi-supervised anomaly detection method based on competitive reconstruction learning | |
| CN111000553B (en) | Intelligent classification method for electrocardiogram data based on voting ensemble learning | |
| CN113326731A (en) | Cross-domain pedestrian re-identification algorithm based on momentum network guidance | |
| CN111125469B (en) | User clustering method and device of social network and computer equipment | |
| CN111860576A (en) | Endometrium tumor classification labeling method based on random forest | |
| Chen et al. | An effective feature selection scheme for healthcare data classification using binary particle swarm optimization | |
| KR20200018868A (en) | Method for Adaptive EEG signal processing using reinforcement learning and System Using the same | |
| CN111297327B (en) | Sleep analysis method, system, electronic equipment and storage medium | |
| CN112884149A (en) | Deep neural network pruning method and system based on random sensitivity ST-SM | |
| CN116524960A (en) | Speech emotion recognition system based on mixed entropy downsampling and integrated classifier | |
| CN112861796B (en) | Feature adaptive motion recognition method | |
| CN112465054B (en) | FCN-based multivariate time series data classification method | |
| CN116595465A (en) | High-dimensional sparse data outlier detection method and system based on self-encoder and data enhancement | |
| CN116738330A (en) | Semi-supervision domain self-adaptive electroencephalogram signal classification method | |
| CN108564167A (en) | The recognition methods of abnormal data among a kind of data set | |
| Kalkan et al. | Online feature selection and classification | |
| CN114398991A (en) | Electroencephalogram emotion recognition method based on Transformer structure search | |
| Bharathi et al. | The significance of feature selection techniques in machine learning | |
| Lim et al. | Analyzing deep neural networks with noisy labels | |
| CN113297935A (en) | Feature adaptive motion recognition system | |
| Suresh et al. | Uncertain data analysis with regularized XGBoost | |
| CN112861798B (en) | Classification recognition method based on physiological signals, medium and electronic equipment | |
| CN111738439A (en) | Artificial intelligence processing method and processor supporting online learning |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |