CN106502398B - A kind of semantization activity recognition method based on Multi-view Integration study - Google Patents
A kind of semantization activity recognition method based on Multi-view Integration study Download PDFInfo
- Publication number
- CN106502398B CN106502398B CN201610918275.2A CN201610918275A CN106502398B CN 106502398 B CN106502398 B CN 106502398B CN 201610918275 A CN201610918275 A CN 201610918275A CN 106502398 B CN106502398 B CN 106502398B
- Authority
- CN
- China
- Prior art keywords
- body movement
- semantization
- sequence
- simple body
- activity
- 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
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F3/00—Input arrangements for transferring data to be processed into a form capable of being handled by the computer; Output arrangements for transferring data from processing unit to output unit, e.g. interface arrangements
- G06F3/01—Input arrangements or combined input and output arrangements for interaction between user and computer
- G06F3/011—Arrangements for interaction with the human body, e.g. for user immersion in virtual reality
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/23—Clustering techniques
- G06F18/232—Non-hierarchical techniques
- G06F18/2321—Non-hierarchical techniques using statistics or function optimisation, e.g. modelling of probability density functions
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N20/00—Machine learning
Abstract
A kind of semantization activity recognition method learnt based on acceleration transducer and Multi-view Integration, is included the following steps: that (1) is based on simple body movement descriptive semantics activity, constructs simple body movement characteristic view;(2) it is based on potential theme distribution descriptive semantics activity, constructs potential theme distribution characteristic view;(3) Cooperative Study is carried out to a variety of views based on semi-supervised technology, and learning outcome is merged to obtain semantization activity recognition model.The present invention is based on multiple view descriptive semantics activities, improve the generalization ability and adaptability of identification model;Based on Cooperative Study technology using unlabeled data training identification model, the problem of mark sample deficiency is overcome.
Description
Technical field
The present invention relates to machine learning and human-computer interaction technology, and in particular to one kind is based on acceleration transducer and multiple view
The semantization activity recognition method of integrated study.
Background technique
The activity of user is to understand one of user context and the most important information of demand, and acceleration transducer is with sensitive
The advantages such as degree is high, power consumption is low.Therefore, the activity recognition based on acceleration transducer be general fit calculation and field of human-computer interaction most
One of important research contents.It is living that the current activity recognition research based on acceleration transducer has focused largely on simple body
In dynamic (such as walk, run, standing) identification.Compared with simple body movement, semantization activity, which refers to, has a meal, works, doing shopping
Complicated number of storage tanks produced per day.Semantization activity can provide richer user context information, while identify that difficulty is bigger.
Existing semantization activity recognition method mainly has following a few classes:
(1) it is used in model layer and identifies similar method with simple body movement, richer feature is introduced in characteristic layer.
For example,V.Mirchevska, V.Janko et al. are in " Recognition of high-level
In activities with a smartphone " (international conference UbiComp 2015:1453-1461) from GPS, microphone,
Complicated feature is extracted in the multiple sensors such as acceleration transducer, biosensor for training semantization activity recognition mould
Type.
(2) a series of combination that semantization activity is regarded as to simple body movements is identified semantic using hierarchical model
Change activity.For example, L.Liu, Y.Peng, M.Liu et al. are in " Sensor-based human activity recognition
system with a multilayered model using time series shapelets”(Knowledge-Based
Systems 90 (2015): 138-152) it conforms to the principle of simplicity to identify semanteme in unmarried body active sequences based on Time Series Matching algorithm in
Change activity.
However, there are the following problems for existing semantization recognition methods:
(1) be based on the activity of single view descriptive semantics: single view is difficult to adapt to different semantization activities not on the same day
Complexity under normal living environment, therefore exist and be easy to be influenced by noise data, be difficult to cover all semantization activity change rule
The problems such as rule.
(2) need largely to have mark sample training model: semantization identification model needs largely to have mark sample to be instructed
Practice.However, due to complexity of semantization activity itself, user, which is difficult to provide in daily life, enough has mark sample.
Summary of the invention
In order to overcome, identification model generalization ability and the adaptability of existing semantization recognition methods are poor, need largely to have
The deficiency for marking sample training model, the present invention provides a kind of raising identification model generalization ability and adaptability, available
The limited semantization activity recognition side learnt based on acceleration transducer and Multi-view Integration for having mark sample training model
Method.
The technical solution adopted by the present invention to solve the technical problems is:
A kind of semantization activity recognition method learnt based on acceleration transducer and Multi-view Integration, the semantization are living
Dynamic recognition methods the following steps are included:
(1) it is based on simple body movement descriptive semantics activity, constructs simple body movement characteristic view, steps are as follows:
(1-1) simple body movement identification model training: a simple body movement training set is given, i.e., is largely labelled with
Simple body movement type, the acceleration information sequence that length is w, firstly, being extracted from each acceleration information sequence all kinds of
Temporal signatures and frequency domain character form motion feature vector;Then, it is based on the simple body movement type mark of motion feature vector sum
Note, training obtain simple body movement identification model;
(1-2) simple body movement sequence generates: to each semantization active samples, i.e. the acceleration that a length is W
Data sequence is spent, wherein W > w, firstly, being divided into the data window that multiple sizes are w, forms data window sequence;So
Afterwards, above-mentioned motion feature vector is extracted from each data window, and is inputted the simple body movement identification that training obtains
Model obtains simple body movement recognition result;Finally, converting simple body movement sequence for data window sequence;
(1-3) simple body movement characteristic view building: firstly, extracting simple body from each simple body movement sequence
Body active characteristics, including following three types:
Set feature: the ratio of every kind of simple body movement type frequency of occurrence and simple body movement sequence length is calculated
Value;
Sequence signature: firstly, the multiple simple bodies of same type continuously occurred all in simple body movement sequence are living
Dynamic pressure is condensed to 1, obtains compressing simple body movement sequence;Then, length is excavated from the simple body movement sequence of compression
The all sequences mode for being M for 2 to length;Finally, calculating each sequence pattern pressure corresponding to simple body movement sequence
The number occurred in the unmarried body active sequences of breviaty;
Temporal characteristics: firstly, calculating all single duration of every kind of simple body movement type;Then, it calculates every
Mean value, intermediate value and the standard deviation of kind simple body movement type single duration;
Then, it is based on above-mentioned simple body movement feature construction feature vector, and movable as descriptive semanticsization
Simple body movement characteristic view;
(2) it is based on potential theme distribution descriptive semantics activity, constructs potential theme distribution characteristic view, steps are as follows:
(2-1) acceleration information window sequence: to each semantization active samples, being divided into multiple sizes is
The data window of w forms data window sequence;Then, above-mentioned motion feature vector is extracted from each data window, and to fortune
Dynamic feature vector is normalized;
(2-2) data window clusters sequence and generates: firstly, based on the Euclidean distance metric data window between motion feature vector
Distance between mouthful, clusters data window, so that the corresponding data window cluster of each data window;Then, by data
Series of windows is converted into data window cluster sequence;
(2-3) potential theme distribution characteristic view building: firstly, data window cluster is regarded as " word ", by data window
Cluster sequence regards " document " as, excavates potential theme based on LDA algorithm, and obtains " theme " distribution of " document ";Then, it is based on
" theme " distribution of " document " obtains the probability vector that data window sequence includes different potential themes, and as description language
The movable potential theme distribution characteristic view of justiceization;
(3) Cooperative Study is carried out to two kinds of characteristic views based on semi-supervised technology, and learning outcome is merged to obtain
Semantization activity recognition model.
Further, given to have mark semantization active samples collection L and without mark semantization activity sample in the step (3)
The step of this collection U, training semantization activity recognition model, is as follows:
(3-1) Training: firstly, being that all samples construct simple body in L based on simple body movement characteristic view
Body active eigenvector, and based on semantization Activity Type mark and simple body movement feature vector training identification model SM;
It then, is that all samples construct potential theme distribution feature vector in L, and are based on semanteme based on potential theme distribution characteristic view
Change Activity Type mark and potential theme distribution feature vector training identification model TM;
(3-2) semi-supervised training: being every class semantization firstly, being identified based on identification model SM to samples all in U
The highest n sample of recognition confidence is picked out in activity, using recognition result as its mark, is obtained pseudo- mark sample set and is put into
L;Then, samples all in U are identified based on identification model TM, picks out recognition confidence most for every class semantization activity
N high sample obtains pseudo- mark sample set and is put into L using recognition result as its mark;
(3-3) algorithm iteration: if sample size is insufficient in U or the number of iterations is more than specified threshold, SM and TM are exported, instead
It, then turn to step (3-1);
(3-4) Model Fusion: to there is each sample in mark semantization active samples collection L, SM and TM pairs are used respectively
It is identified, is obtained SM and TM and is identified that it is every movable probability of class semantization, and then obtains 2 probability vectors;Then, will
This 2 probability vectors and semantization Activity Type mark construct new sample set NL as new sample;Finally, based on NL, adopting
Final semantization activity recognition model FM is obtained with the training of Logistic Regression algorithm.
Further, in the step (1-1), simple body movement identification model is obtained using the training of C4.5 algorithm.
Further, simple from compression based on Apriori algorithm during extracting sequence signature in the step (1-3)
Excavated in body movement sequence length be 2 be M to length all sequences mode.
In the step (2-2), data window is clustered based on K-Medoids algorithm.
Beneficial effects of the present invention are mainly manifested in: 1, being based on the activity of multiple view descriptive semantics, improve identification model
Generalization ability and adaptability.2, mark has been overcome using unlabeled data training identification model based on Cooperative Study technology
Infuse the problem of sample deficiency.
Detailed description of the invention
Fig. 1 is the flow chart of the semantization activity recognition method learnt based on acceleration transducer and Multi-view Integration;
Fig. 2 is the flow chart of simple body movement characteristic view building;
Fig. 3 is the flow chart of potential theme distribution characteristic view building;
Fig. 4 is the flow chart of semantization activity recognition model training.
Specific embodiment
The invention will be further described below in conjunction with the accompanying drawings.
Referring to Fig.1~Fig. 4, a kind of semantization activity recognition side learnt based on acceleration transducer and Multi-view Integration
Method, the semantization recognition methods the following steps are included:
(1) it is based on simple body movement descriptive semantics activity, constructs simple body movement characteristic view.
(2) it is based on potential theme distribution descriptive semantics activity, constructs potential theme distribution characteristic view.
(3) Cooperative Study is carried out to two kinds of characteristic views based on semi-supervised technology, and learning outcome is merged to obtain
Semantization activity recognition model.
Referring to Fig. 2, in the step (1), the detailed step for constructing simple body movement characteristic view is as follows:
(1-1) simple body movement identification model training: a given simple body movement training set (is largely labelled with
Simple body movement type, the acceleration information sequence that length is w), firstly, being extracted from each acceleration information sequence all kinds of
Temporal signatures (including: mean value, standard deviation, interquartile range, energy) and frequency domain character (including: frequency amplitude, frequency domain entropy) are formed
Motion feature vector.Then, it based on the simple body movement type mark of motion feature vector sum, is obtained using the training of C4.5 algorithm
Simple body movement identification model.
(1-2) simple body movement sequence generates: to each semantization active samples (i.e. acceleration that a length is W
Data sequence is spent, wherein W > w), firstly, being divided into the data window that multiple sizes are w, form data window sequence.So
Afterwards, above-mentioned motion feature vector is extracted from each data window, and is inputted the simple body movement identification that training obtains
Model obtains simple body movement recognition result.Finally, converting simple body movement sequence for data window sequence.
(1-3) simple body movement characteristic view building: firstly, extracting simple body from each simple body movement sequence
Body active characteristics, including following three types:
Set feature: the ratio of every kind of simple body movement type frequency of occurrence and simple body movement sequence length is calculated
Value.
Sequence signature: firstly, the multiple simple bodies of same type continuously occurred all in simple body movement sequence are living
Dynamic pressure is condensed to 1, obtains compressing simple body movement sequence;Then, Apriori algorithm is based on from the simple body movement sequence of compression
Excavated in column length be 2 be M to length all sequences mode;Finally, calculating each sequence pattern in simple body movement
The number occurred in the simple body movement sequence of compression corresponding to sequence.
Temporal characteristics: firstly, calculating all single duration of every kind of simple body movement type;Then, it calculates every
Mean value, intermediate value, the standard deviation of kind simple body movement type single duration.
Then, it is based on above-mentioned simple body movement feature construction feature vector, and movable as descriptive semanticsization
Simple body movement characteristic view.
Referring to Fig. 3, in the step (2), the detailed step for constructing potential theme distribution characteristic view is as follows:
(2-1) acceleration information window sequence: to each semantization active samples (i.e. acceleration that a length is W
Spend data sequence), it is divided into the data window that multiple sizes are w, forms data window sequence.Then, from each data
Above-mentioned motion feature vector is extracted in window, and motion feature vector is normalized.
(2-2) data window clusters sequence and generates: firstly, based on the Euclidean distance metric data window between motion feature vector
Distance between mouthful, clusters data window based on K-Medoids algorithm, so that each data window corresponds to a data window
Cluster.Then, data window cluster sequence is converted by data window sequence.
(2-3) potential theme distribution characteristic view building: firstly, data window cluster is regarded as " word ", by data window
Cluster sequence regards " document " as, excavates potential theme based on LDA algorithm, and obtains " theme " distribution of " document ".Then, it is based on
" theme " distribution of " document " obtains the probability vector that data window sequence includes different potential themes, and as description language
The movable potential theme distribution characteristic view of justiceization.
It is given to have mark semantization active samples collection L and without mark semantization activity sample in the step (3) referring to Fig. 4
The detailed step of this collection U, training semantization activity recognition model are as follows:
(3-1) Training: firstly, being that all samples construct simple body in L based on simple body movement characteristic view
Body active eigenvector, and based on semantization Activity Type mark and simple body movement feature vector training identification model SM.
It then, is that all samples construct potential theme distribution feature vector in L, and are based on semanteme based on potential theme distribution characteristic view
Change Activity Type mark and potential theme distribution feature vector training identification model TM.
(3-2) semi-supervised training: being every class semantization firstly, being identified based on identification model SM to samples all in U
The highest n sample of recognition confidence is picked out in activity, using recognition result as its mark, obtains pseudo- mark sample set USM,n×S
And it is put into L (wherein, the quantity that S is semantization Activity Type).Then, samples all in U are known based on identification model TM
Not, the highest n sample of recognition confidence is picked out for every class semantization activity, using recognition result as its mark, obtains puppet
Mark sample set UTM,n×SAnd it is put into L.
(3-3) algorithm iteration: if sample size is insufficient in U or the number of iterations is more than specified threshold, SM and TM are exported.Instead
It, then turn to step (3-1).
(3-4) Model Fusion: to there is each sample in mark semantization active samples collection L, SM and TM pairs are used respectively
It is identified, is obtained SM and TM and is identified that it is every movable probability of class semantization, and then obtains 2 probability vectors (wherein,
PSM,ikThe probability that sample i is semantization Activity Type k, P are identified for SMTM,ikIdentify that sample i is semantization Activity Type k for TM
Probability, 1≤k≤S).Then, this 2 probability vectors and semantization Activity Type mark are constructed new as new sample
Sample set NL.Finally, obtaining final semantization activity recognition based on NL, using the training of Logistic Regression algorithm
Model FM.
Claims (4)
1. a kind of semantization activity recognition method based on Multi-view Integration study, it is characterised in that: the semantization activity is known
Other method the following steps are included:
(1) it is based on simple body movement descriptive semantics activity, constructs simple body movement characteristic view, steps are as follows:
(1-1) simple body movement identification model training: a simple body movement training set is given, i.e., is largely labelled with simple
Body movement type, the acceleration information sequence that length is w, firstly, extracting all kinds of time domains from each acceleration information sequence
Feature and frequency domain character form motion feature vector;Then, it is marked based on the simple body movement type of motion feature vector sum,
Training obtains simple body movement identification model;
(1-2) simple body movement sequence generates: to each semantization active samples, i.e. the acceleration degree that a length is W
According to sequence, wherein W > w forms data window sequence firstly, being divided into the data window that multiple sizes are w;Then, from
Above-mentioned motion feature vector is extracted in each data window, and is inputted the simple body movement identification model that training obtains,
Obtain simple body movement recognition result;Finally, converting simple body movement sequence for data window sequence;
(1-3) simple body movement characteristic view building: firstly, it is living to extract simple body from each simple body movement sequence
Dynamic feature, including following three types:
Set feature: the ratio of every kind of simple body movement type frequency of occurrence and simple body movement sequence length is calculated;
Sequence signature: firstly, by the multiple simple body movement pressures of same type continuously occurred all in simple body movement sequence
It is condensed to 1, obtains compressing simple body movement sequence;Then, from compress excavated in simple body movement sequence length be 2 to
Length is all sequences mode of M;Finally, it is simple to calculate the compression corresponding to simple body movement sequence of each sequence pattern
The number occurred in body movement sequence;
Temporal characteristics: firstly, calculating all single duration of every kind of simple body movement type;Then, every kind of letter is calculated
Mean value, intermediate value and the standard deviation of unmarried body Activity Type single duration;
Then, it is based on above-mentioned simple body movement feature construction feature vector, and movable simple as descriptive semanticsization
Body movement characteristic view;
(2) it is based on potential theme distribution descriptive semantics activity, constructs potential theme distribution characteristic view, steps are as follows:
(2-1) acceleration information window sequence: to each semantization active samples, being divided into multiple sizes is w's
Data window forms data window sequence;Then, above-mentioned motion feature vector is extracted from each data window, and to movement
Feature vector is normalized;
(2-2) data window clusters sequence and generates: firstly, based between the Euclidean distance metric data window between motion feature vector
Distance clusters data window, so that the corresponding data window cluster of each data window;Then, by data window
It is Sequence Transformed to cluster sequence for data window;
(2-3) potential theme distribution characteristic view building: firstly, regarding data window cluster as " word ", data window is clustered
Sequence regards " document " as, excavates potential theme based on LDA algorithm, and obtains " theme " distribution of " document ";Then, based on " text
" theme " distribution of shelves " obtains the probability vector that data window sequence includes different potential themes, and as descriptive semantics
Change movable potential theme distribution characteristic view;
(3) Cooperative Study is carried out to two kinds of characteristic views based on semi-supervised technology, and learning outcome is merged to obtain semanteme
Change activity recognition model;It is given to have mark semantization active samples collection L and without mark semantization active samples collection U, training semanteme
The step of changing activity recognition model is as follows:
(3-1) Training: firstly, being that the simple body of all samples buildings is living in L based on simple body movement characteristic view
Dynamic feature vector, and based on semantization Activity Type mark and simple body movement feature vector training identification model SM;Then,
It is that all samples construct potential theme distribution feature vector in L, and are lived based on semantization based on potential theme distribution characteristic view
Dynamic type mark and potential theme distribution feature vector training identification model TM;
(3-2) semi-supervised training: firstly, being identified based on identification model SM to samples all in U, for every class semantization activity
The highest n sample of recognition confidence is picked out, using recognition result as its mark, pseudo- mark sample set is obtained and is put into L;So
Afterwards, samples all in U are identified based on identification model TM, it is highest to pick out recognition confidence for every class semantization activity
N sample obtains pseudo- mark sample set and is put into L using recognition result as its mark;
(3-3) algorithm iteration: if sample size is insufficient in U or the number of iterations is more than specified threshold, exporting SM and TM, conversely,
Then turn to step (3-1);
(3-4) Model Fusion: to have mark semantization active samples collection L in each sample, respectively using SM and TM to its into
Row identification obtains SM and TM and identifies that it is every movable probability of class semantization, and then obtains 2 probability vectors;Then, by this 2
A probability vector and semantization Activity Type mark construct new sample set NL as new sample;Finally, based on NL, using
The training of Logistic Regression algorithm obtains final semantization activity recognition model FM.
2. the semantization activity recognition method as described in claim 1 based on Multi-view Integration study, it is characterised in that: described
In step (1-1), simple body movement identification model is obtained using the training of C4.5 algorithm.
3. the semantization activity recognition method as described in claim 1 based on Multi-view Integration study, it is characterised in that: described
In step (1-3), during extracting sequence signature, excavated from the simple body movement sequence of compression based on Apriori algorithm
Length be 2 be M to length all sequences mode.
4. the semantization activity recognition method as described in claim 1 based on Multi-view Integration study, it is characterised in that: described
In step (2-2), data window is clustered based on K-Medoids algorithm.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201610918275.2A CN106502398B (en) | 2016-10-21 | 2016-10-21 | A kind of semantization activity recognition method based on Multi-view Integration study |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201610918275.2A CN106502398B (en) | 2016-10-21 | 2016-10-21 | A kind of semantization activity recognition method based on Multi-view Integration study |
Publications (2)
Publication Number | Publication Date |
---|---|
CN106502398A CN106502398A (en) | 2017-03-15 |
CN106502398B true CN106502398B (en) | 2019-01-29 |
Family
ID=58319386
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201610918275.2A Active CN106502398B (en) | 2016-10-21 | 2016-10-21 | A kind of semantization activity recognition method based on Multi-view Integration study |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN106502398B (en) |
Families Citing this family (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109492678B (en) * | 2018-10-24 | 2021-11-23 | 浙江工业大学 | App classification method integrating shallow layer learning and deep learning |
TWI707249B (en) * | 2018-11-27 | 2020-10-11 | 美律實業股份有限公司 | System and method for generating label data |
US11455383B2 (en) * | 2019-04-30 | 2022-09-27 | TruU, Inc. | Supervised and unsupervised techniques for motion classification |
CN112988124B (en) * | 2021-05-10 | 2021-07-30 | 湖南高至科技有限公司 | Multi-view platform-independent model system |
CN117251770B (en) * | 2023-11-17 | 2024-02-13 | 北京新兴科遥信息技术有限公司 | Method for identifying low-utility land |
Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101389004A (en) * | 2007-09-13 | 2009-03-18 | 中国科学院自动化研究所 | Moving target classification method based on on-line study |
CN104331431A (en) * | 2014-10-22 | 2015-02-04 | 浙江中烟工业有限责任公司 | Mobile application ordering method of situational awareness |
CN105956614A (en) * | 2016-04-25 | 2016-09-21 | 中国科学院上海高等研究院 | Time series semantization predicting method and time series semantization predicting system |
-
2016
- 2016-10-21 CN CN201610918275.2A patent/CN106502398B/en active Active
Patent Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101389004A (en) * | 2007-09-13 | 2009-03-18 | 中国科学院自动化研究所 | Moving target classification method based on on-line study |
CN104331431A (en) * | 2014-10-22 | 2015-02-04 | 浙江中烟工业有限责任公司 | Mobile application ordering method of situational awareness |
CN105956614A (en) * | 2016-04-25 | 2016-09-21 | 中国科学院上海高等研究院 | Time series semantization predicting method and time series semantization predicting system |
Non-Patent Citations (2)
Title |
---|
《人体运动序列数据的语义化分析方法研究》;徐常有;《中国优秀硕士学位论文全文数据库 信息科技辑》;20120715;全文 * |
《语义化多角色可变形运动模型研究》;陈秋迪;《中国优秀硕士学位论文全文数据库 信息科技辑》;20150315;全文 * |
Also Published As
Publication number | Publication date |
---|---|
CN106502398A (en) | 2017-03-15 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN106502398B (en) | A kind of semantization activity recognition method based on Multi-view Integration study | |
Zhang et al. | Human activity recognition based on motion sensor using u-net | |
Su et al. | Activity recognition with smartphone sensors | |
Serpush et al. | Wearable sensor-based human activity recognition in the smart healthcare system | |
CN105320944B (en) | A kind of human body behavior prediction method based on human skeleton motion information | |
Zhang et al. | Human activity recognition with HMM-DNN model | |
Ko et al. | Using dynamic time warping for online temporal fusion in multisensor systems | |
Liu et al. | Research on human activity recognition based on active learning | |
CN105574510A (en) | Gait identification method and device | |
Thakur et al. | Convae-lstm: Convolutional autoencoder long short-term memory network for smartphone-based human activity recognition | |
CN105609116B (en) | A kind of automatic identifying method in speech emotional dimension region | |
Wen et al. | Adaptive activity learning with dynamically available context | |
CN104700100A (en) | Feature extraction method for high spatial resolution remote sensing big data | |
Zhao et al. | Local domain adaptation for cross-domain activity recognition | |
Garcia-Ceja et al. | An improved three-stage classifier for activity recognition | |
Huan et al. | Human complex activity recognition with sensor data using multiple features | |
CN107341471B (en) | A kind of Human bodys' response method based on Bilayer condition random field | |
Zeng et al. | Adaptive activity recognition with dynamic heterogeneous sensor fusion | |
Permatasari et al. | Inertial sensor fusion for gait recognition with symmetric positive definite Gaussian kernels analysis | |
Badshah | Sensor-based human activity recognition using smartphones | |
CN105631462A (en) | Behavior identification method through combination of confidence and contribution degree on the basis of space-time context | |
CN105740815B (en) | A kind of Human bodys' response method based on depth recurrence stratified condition random field | |
Wang et al. | Human activity recognition using smart phone embedded sensors: A linear dynamical systems method | |
Uslu et al. | A segmentation scheme for knowledge discovery in human activity spotting | |
Gerych et al. | Gan for generating user-specific human activity data from an incomplete training corpus |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
C06 | 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 | ||
TR01 | Transfer of patent right |
Effective date of registration: 20191126 Address after: Room 506-2, Block E, building 1, 1378 Wenyi West Road, Cangqian street, Yuhang District, Hangzhou City, Zhejiang Province Patentee after: Hangzhou smart strategy Technology Co., Ltd Address before: 310014, Zhejiang City, No. 18 Chao Wang Road, Zhejiang University of Technology Patentee before: Zhejiang University of Technology |
|
TR01 | Transfer of patent right |