CN113378966B - Mobile phone sensor data labeling method based on weak supervised learning - Google Patents
Mobile phone sensor data labeling method based on weak supervised learning Download PDFInfo
- Publication number
- CN113378966B CN113378966B CN202110713453.9A CN202110713453A CN113378966B CN 113378966 B CN113378966 B CN 113378966B CN 202110713453 A CN202110713453 A CN 202110713453A CN 113378966 B CN113378966 B CN 113378966B
- Authority
- CN
- China
- Prior art keywords
- mobile phone
- sensor data
- fuzzy
- phone sensor
- label
- 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
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
- G06F18/2415—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on parametric or probabilistic models, e.g. based on likelihood ratio or false acceptance rate versus a false rejection rate
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/047—Probabilistic or stochastic networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N7/00—Computing arrangements based on specific mathematical models
- G06N7/02—Computing arrangements based on specific mathematical models using fuzzy logic
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02D—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN INFORMATION AND COMMUNICATION TECHNOLOGIES [ICT], I.E. INFORMATION AND COMMUNICATION TECHNOLOGIES AIMING AT THE REDUCTION OF THEIR OWN ENERGY USE
- Y02D30/00—Reducing energy consumption in communication networks
- Y02D30/70—Reducing energy consumption in communication networks in wireless communication networks
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- General Physics & Mathematics (AREA)
- Artificial Intelligence (AREA)
- Life Sciences & Earth Sciences (AREA)
- Evolutionary Computation (AREA)
- General Engineering & Computer Science (AREA)
- Software Systems (AREA)
- Biomedical Technology (AREA)
- General Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- Health & Medical Sciences (AREA)
- Mathematical Physics (AREA)
- Computational Linguistics (AREA)
- Biophysics (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Probability & Statistics with Applications (AREA)
- Evolutionary Biology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Bioinformatics & Computational Biology (AREA)
- Algebra (AREA)
- Pure & Applied Mathematics (AREA)
- Mathematical Optimization (AREA)
- Mathematical Analysis (AREA)
- Computational Mathematics (AREA)
- Fuzzy Systems (AREA)
- Automation & Control Theory (AREA)
- Telephone Function (AREA)
Abstract
The invention discloses a mobile phone sensor data labeling method based on weak supervised learning, which comprises the following steps: step 1, establishing a labeling task, and generating a fuzzy query question corresponding to the labeling task, wherein the answer of the fuzzy query question is yes or no; step 2, sending a request for answering the fuzzy query question to an acquirer providing data, and acquiring a fuzzy label obtained by the acquirer answering the fuzzy query question and mobile phone sensor data related to the fuzzy label; step 3, training a two-class depth model by using the fuzzy label acquired in the step 2 and the associated mobile phone sensor data; and 4, processing the subsequently acquired mobile phone sensor data to be labeled through the trained two-classification depth model to deduce an accurate label of the mobile phone sensor data to be labeled, and labeling the mobile phone sensor data to be labeled by using the obtained accurate label. The method enables the fuzzy query of the mobile phone sensor to be simpler, simplifies the manual process in the process, and obtains accurate label data through the algorithm post-processing only through the fuzzy answer of the collector.
Description
Technical Field
The invention relates to the field of mobile computing, in particular to a mobile phone sensor data labeling method based on weak supervised learning.
Background
Machine learning and deep learning methods often rely on a large amount of labeled data, and acquiring such data sets requires enormous manpower and material resources, and is expensive. Especially for the mobile phone sensor data (including accelerometer, gyroscope, etc.), the data type is different from the picture and audio, and the annotator cannot directly give the corresponding label from the appearance of the value, waveform, etc. of the sensor data, for example, the mobile phone owner's ongoing activities (running, walking, going upstairs, etc.) corresponding to this data, which also greatly increases the annotation cost of such data.
The existing mobile phone sensor labeling modes mainly include the following modes:
(1) When a marker carries out a certain activity, the marker manually clicks the start and the end of the corresponding label on the mobile phone application, and the data to be marked are bound and marked by utilizing the timestamp.
(2) The marker is continuously in the monitored environment or the body carries the camera in the data acquisition process, and the later stage is handled camera video data and is beaten the mark note, then with the timestamp to its sensor data with correspond the label.
However, the existing labeling method often has higher requirements on the acquisition environment and the acquirer, and also requires the annotator to have certain professional knowledge, so that the method is difficult to acquire the data of the large-scale mobile phone sensor.
Disclosure of Invention
Aiming at the problems of the existing method, the invention aims to provide a mobile phone sensor data labeling method based on weak supervised learning, which can solve the problem that the existing mobile phone sensor data labeling method has higher requirements on the collected environment and a collector, and requires a certain professional knowledge of the label collector, so that the large-scale mobile phone sensor data collection is difficult to carry out.
The purpose of the invention is realized by the following technical scheme:
the embodiment of the invention provides a mobile phone sensor data labeling method based on weak supervised learning, which comprises the following steps:
step 1, establishing a labeling task, and generating a fuzzy query question corresponding to the labeling task, wherein the fuzzy query question is a question sentence containing acquisition duration and a motion type, and the answer of the fuzzy query question is yes or no;
step 2, sending a request for answering the fuzzy query question to an acquirer providing data by using the established labeling task and the corresponding fuzzy query question, and acquiring a fuzzy label obtained by the acquirer for answering the fuzzy query question and mobile phone sensor data related to the fuzzy label;
step 3, training a two-class depth model by using the fuzzy labels acquired in the step 2 and the mobile phone sensor data related to the fuzzy labels;
and 4, processing a fuzzy label obtained by a subsequently collected person to be collected and answering the fuzzy query question and mobile phone sensor data associated with the fuzzy label through the trained two-classification depth model to deduce an accurate label of the mobile phone sensor data associated with the fuzzy label, and labeling the mobile phone sensor data to be labeled by using the obtained accurate label.
According to the technical scheme provided by the invention, the novel marking method provided by the embodiment of the invention has the beneficial effects that:
by setting the fuzzy query question corresponding to the labeling task, the data labeling can be completed only by yes or no answer to the fuzzy query question, the labeling mode is simple, the requirement on professional knowledge of a person to be collected is low, the method is very friendly to the person to be collected, accurate labeling can be obtained after the collected mobile phone sensor data are processed, and the method is beneficial to improving the scale of mobile phone sensor data collection.
Drawings
In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings needed to be used in the description of the embodiments are briefly introduced below, and it is obvious that the drawings in the following description are only some embodiments of the present invention, and it is obvious for those skilled in the art to obtain other drawings based on the drawings without creative efforts.
Fig. 1 is a flowchart of a mobile phone sensor data annotation method based on weakly supervised learning according to an embodiment of the present invention;
fig. 2 is a flowchart of processing a binary classification depth model in the method according to the embodiment of the present invention.
Detailed Description
The technical solutions in the embodiments of the present invention are clearly and completely described below with reference to the specific contents of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments of the present invention without making any creative effort, shall fall within the protection scope of the present invention. Details which are not described in detail in the embodiments of the invention belong to the prior art which is known to the person skilled in the art.
Referring to fig. 1, an embodiment of the present invention provides a method for labeling mobile phone sensor data based on weak supervised learning, which is a friendly man-machine interaction mode, reduces the threshold and cost of a labeler of mobile phone sensor data, and is more suitable for scale data acquisition, and includes:
step 1, establishing a labeling task, and generating a fuzzy query question corresponding to the labeling task, wherein the fuzzy query question is a question sentence containing acquisition duration and a motion type, and the answer of the fuzzy query question is yes or no; an example of the fuzzy query question is "do you go through action a in K minutes? The question is simple to answer, does not need special knowledge, and is very friendly to the acquired person;
step 2, sending a request for answering the fuzzy query question to an acquirer providing data by using the established labeling task and the corresponding fuzzy query question, and acquiring a fuzzy label obtained by the acquirer for answering the fuzzy query question and mobile phone sensor data related to the fuzzy label;
step 3, training a two-class depth model by using the fuzzy labels acquired in the step 2 and the mobile phone sensor data related to the fuzzy labels;
and 4, processing a fuzzy label obtained by a subsequently collected person to be collected and answering the fuzzy query question and mobile phone sensor data associated with the fuzzy label through the trained two-classification depth model to deduce an accurate label of the mobile phone sensor data associated with the fuzzy label, and labeling the mobile phone sensor data to be labeled by using the obtained accurate label.
In step 1 of the method, the acquisition time contained in the fuzzy query question is a preset time;
in step 2, the fuzzy label obtained by answering the fuzzy query question includes: a tag labeled 1 corresponding to a yes answer and a tag labeled 0 corresponding to a no answer.
In step 3 of the method, the two-classification depth model extracts the features of the mobile phone sensor data in a data feature extraction mode, and establishes the association between the mobile phone sensor data and the fuzzy label according to the extracted features.
Referring to fig. 2, the binary depth model is a classifier composed of an input layer, a first layer of long-short time memory network, a second layer of long-short time memory network, a flat layer and a softmax activation function, which are connected in sequence; wherein the content of the first and second substances,
the first layer long-time memory network LSTM and the second layer long-time memory network LSTM form a time sequence characteristic extraction structure for extracting mobile phone sensor data related to the fuzzy label;
the Flatten layer and the softmax layer form a feature classifier for classifying the fuzzy labels.
In step 4 of the method, the trained two-classification depth model processes the fuzzy label obtained by the subsequently collected collector answering the fuzzy query question and the mobile phone sensor data associated with the fuzzy label in the following manner to deduce the accurate label of the mobile phone sensor data associated with the fuzzy label, and the method comprises the following steps:
by the formula S t ＝∑w ti f ti Calculating the time sequence characteristic importance score S of the mobile phone sensor data t Where w is the weight connecting the timing signature unit with the positive sample activation unit (i.e., the handset sensor data associated with the ambiguity label of 1),the weight is obtained by the trained two-classification depth model; f is a time sequence characteristic unit of the mobile phone sensor data in front of the softmax layer of the two-classification depth model, the size of f is t multiplied by k, t is the length of a time axis, and k is the characteristic length corresponding to each moment; i is the serial number of the subunit in each time sequence characteristic unit;
according to the obtained importance score S of the time sequence feature t Obtaining a heat map of the mobile phone sensor data with the fuzzy label as 1 correlation;
selecting an area meeting a preset activation value in the heat map as an activation area, taking the starting time and the ending time of the activation area as the accurate starting time and the accurate ending time of the mobile phone sensor data, and setting a corresponding accurate label for the mobile phone sensor data.
The processing mode of the two-classification depth model realizes that the target marking area is determined by utilizing the influence degree of the calculated time sequence characteristics on the model decision, and the activation area can be determined more accurately.
The region satisfying the preset activation value accounts for the importance score S of the time sequence characteristic t 30% of the area.
In the method, the positive sample unit refers to a segment of mobile phone sensor data associated with the fuzzy label of 1 (the answer is yes corresponding to the fuzzy label); conversely, the negative sample unit refers to a piece of mobile phone sensor data associated with the fuzzy label being 0 (the fuzzy label corresponding to no answer).
In the above method, the mobile phone sensor data refers to: accelerometer and gyroscope data of the mobile phone.
The method of the invention starts with two categories and can mark which accurate moments in a positive sample (i.e. the mobile phone sensor data associated with a fuzzy label of 1) are real corresponding motions. The fuzzy query of the mobile phone sensor is simpler, and accurate label data can be obtained only through the fuzzy answer of the collected person and the trained two-classification depth model processing.
The embodiments of the present invention are described in further detail below.
The embodiment of the invention provides a novel mobile phone sensor marking method, which mainly comprises the following steps:
step 1, constructing a labeling task and a fuzzy query problem corresponding to the labeling task; the fuzzy query question is a question sentence containing acquisition duration and a motion type, and the answer of the fuzzy query question is yes or no;
step 2, sending a request for answering the fuzzy query question to an acquirer providing data by using the established labeling task and the corresponding fuzzy query question, and acquiring a fuzzy label obtained by the acquirer for answering the fuzzy query question and mobile phone sensor data related to the fuzzy label;
step 3, training a two-classification depth model by using the data obtained in the step 2;
and 4, processing a fuzzy label obtained by the collector needing to be processed answering the fuzzy query question and mobile phone sensor data associated with the fuzzy label through the trained two-classification depth model, deducing an accurate label of the mobile phone sensor data in an interpretability manner, and finishing the accurate labeling of the mobile phone sensor data to be labeled.
In step 1, a data tagging task to be performed and a corresponding fuzzy query question are constructed, for example, if the tagging task is to acquire sensor data of activity a and an acquisition time range is set to k minutes, the corresponding fuzzy query question may be "asking for whether activity a has been performed within k minutes? ". In the labeling method, the collected person only needs to answer the simple question and answer of yes or no to the question; the fuzzy query question in the form or the similar form can be accurately answered without professional knowledge as long as the answer of the person to be acquired is yes or no, so that the difficulty of answering is reduced, the enthusiasm of answering is improved, and the data acquisition of the mobile phone sensor is facilitated;
step 2, obtaining a labeling task and a corresponding fuzzy query question of a person to be queried from step 1, randomly sending the labeling task to each person to be queried, wherein each person to be queried answers the fuzzy query question according to actual conditions and real information, if the answer is yes, a fuzzy label corresponding to the label 1 is obtained, the label 1 is abbreviated as follow, mobile phone sensor data associated with the label 1 is called a positive sample unit, if the answer is not yes, a fuzzy label corresponding to the label 0 is obtained, the label 0 is abbreviated as follow, and mobile phone sensor data associated with the label 0 is called a negative sample unit, and step 2 can obtain the label 0, the label 1 and mobile phone sensor data (the mobile phone sensor data can specify data of an accelerometer, a gyroscope and the like) collected in the first k minutes (k minutes before a timestamp and the time of submitting a question answer is taken as the timestamp), and then transmit the collected mobile phone sensor data to step 3 after packaging;
step 3, obtaining the mobile phone sensor data and the fuzzy label acquired in step 2, training a two-class depth model shown in fig. 2, wherein the two-class depth model is obtained by an input layer and two layers of long-time and short-time memory networks (LSTM), performing feature extraction and construction on the mobile phone sensor data, then performing feature flattening by a Flatten layer, and finally performing yes or no (namely 0 or 1) two-class classification by a classifier activated by softmax, so that after the step is completed, a mobile phone sensor data classifier constructed in fuzzy answer of the acquired person can be obtained;
step 4, using the mobile phone sensor data classifier constructed in the fuzzy answer of the acquired person obtained in step 3, and using interpretability to mine more accurate information from the fuzzy label, and using the trained two-classification depth model to process the mobile phone sensor data associated with the fuzzy label and acquired subsequently, when the trained two-classification depth model is used to process the mobile phone sensor data associated with the fuzzy label, the core idea is to determine the target labeling area by using the influence degree of the calculated time sequence characteristics on the decision of the two-classification depth model; the method specifically comprises the following steps: if the final time sequence feature before the softmax layer of the two-class depth model is f (the size is t × k, t is the time axis length, and k is the feature length corresponding to each time), the feature importance score of the timestamp data t is: s t ＝∑w ti f ti Where w is the weight connecting the feature unit and the positive sample activation unit, and i is the sequence number of the subunit in each time sequence feature unit; feature importance score S from derived timestamp data t t To obtain a 1-markSelecting a heat map of the associated handset sensor data, selecting a region of the heat map having a greater activation rate (i.e., a region satisfying a predetermined activation value, which may be a percentage value, e.g., S t 30% of the area) as the activation area, and the start and end times of the activation area as the accurate start and end times of the activity, thereby determining the accurate label of the mobile phone sensor data.
The method utilizes the interpretability of the weak supervised learning and the binary depth model, can carry out post-processing on the simple answer of the acquired person, further deduces a corresponding label which is more accurate to the acquired mobile phone sensor data, and is beneficial to improving the large-scale acquisition of the mobile phone sensor data which is not suitable for accurate marking.
The above description is only for the preferred embodiment of the present invention, but the scope of the present invention is not limited thereto, and any changes or substitutions that can be easily conceived by those skilled in the art within the technical scope of the present invention are included in the scope of the present invention. Therefore, the protection scope of the present invention should be subject to the protection scope of the claims.
Claims (6)
1. A mobile phone sensor data labeling method based on weak supervised learning is characterized by comprising the following steps:
step 1, establishing a labeling task, and generating a fuzzy query question corresponding to the labeling task, wherein the fuzzy query question is a question sentence containing acquisition duration and a motion type, and the answer of the fuzzy query question is yes or no;
step 2, sending a request for answering the fuzzy query question to an acquirer providing data by using the established labeling task and the corresponding fuzzy query question, and acquiring a fuzzy label obtained by the acquirer for answering the fuzzy query question and mobile phone sensor data related to the fuzzy label;
step 3, training a two-class depth model by using the fuzzy labels acquired in the step 2 and the mobile phone sensor data related to the fuzzy labels;
step 4, processing a fuzzy label obtained by a subsequently collected person to be collected answering the fuzzy query question and mobile phone sensor data associated with the fuzzy label through the trained two-classification depth model to deduce an accurate label of the mobile phone sensor data associated with the fuzzy label, and labeling the mobile phone sensor data to be labeled by using the obtained accurate label; the trained two-classification depth model processes a fuzzy label obtained by a subsequently collected person who answers the fuzzy query question and mobile phone sensor data associated with the fuzzy label to deduce an accurate label of the mobile phone sensor data associated with the fuzzy label according to the following modes, and the method comprises the following steps:
by the formula S t ＝∑w ti f ti Calculating the time sequence characteristic importance score S of the mobile phone sensor data t Wherein w is the weight connecting the timing characteristic unit and the positive sample activation unit; f is a time sequence characteristic unit of the mobile phone sensor data in front of the softmax layer of the two-classification depth model, the size of f is t multiplied by k, t is the length of a time axis, and k is the characteristic length corresponding to each moment; i is the serial number of the subunit in each time sequence characteristic unit;
according to the obtained importance scores S of the time sequence characteristics t Obtaining a heat map of the mobile phone sensor data with the fuzzy label as 1 correlation;
and selecting an area meeting a preset activation value in the heat map as an activation area, taking the start time and the end time of the activation area as the accurate start time and the accurate end time of the mobile phone sensor data, and setting a corresponding accurate label for the mobile phone sensor data.
2. The mobile phone sensor data labeling method based on the weak supervised learning as recited in claim 1, wherein in the step 1, the acquisition time included in the fuzzy query question is a preset time;
in step 2, the fuzzy label obtained by answering the fuzzy query question includes: a tag labeled 1 corresponding to answer result yes and a tag labeled 0 corresponding to answer result no.
3. The mobile phone sensor data labeling method based on weak supervised learning as claimed in claim 1, wherein in the step 3, the binary depth model performs feature extraction on the mobile phone sensor data through a data feature extraction mode, and establishes the association between the mobile phone sensor data and the fuzzy label according to the extracted features.
4. The mobile phone sensor data labeling method based on the weak supervised learning as recited in claim 1 or 3, wherein the two classification depth models are two classifiers consisting of an input layer, a first layer of long-term and short-term memory network, a second layer of long-term and short-term memory network, a Flatten layer and a softmax activation function which are connected in sequence; wherein, the first and the second end of the pipe are connected with each other,
the first layer long-time and short-time memory network LSTM and the second layer long-time and short-time memory network LSTM form a time sequence characteristic extraction structure for extracting mobile phone sensor data related to the fuzzy label;
the Flatten layer and the softmax layer form a feature classifier for classifying the fuzzy labels.
5. The method for labeling mobile phone sensor data based on weak supervised learning as recited in claim 1, wherein the region satisfying a preset activation value is the importance score S occupying the time series feature t 30% of the area.
6. The method for labeling mobile phone sensor data based on weakly supervised learning as claimed in any one of claims 1 to 3, wherein in the method, the mobile phone sensor data refers to: accelerometer and gyroscope data of the mobile phone.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110713453.9A CN113378966B (en) | 2021-06-25 | 2021-06-25 | Mobile phone sensor data labeling method based on weak supervised learning |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110713453.9A CN113378966B (en) | 2021-06-25 | 2021-06-25 | Mobile phone sensor data labeling method based on weak supervised learning |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN113378966A CN113378966A (en) | 2021-09-10 |
| CN113378966B true CN113378966B (en) | 2023-03-10 |
Family
ID=77579242
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110713453.9A Active CN113378966B (en) | 2021-06-25 | 2021-06-25 | Mobile phone sensor data labeling method based on weak supervised learning |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113378966B (en) |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111242101A (en) * | 2020-03-08 | 2020-06-05 | 电子科技大学 | Behavior identification method based on spatiotemporal context association |
| CN112287089A (en) * | 2020-11-23 | 2021-01-29 | 腾讯科技（深圳）有限公司 | Classification model training and automatic question-answering method and device for automatic question-answering system |
| CN112288034A (en) * | 2020-11-19 | 2021-01-29 | 江南大学 | Semi-supervised online anomaly detection method for wireless sensor network |
Family Cites Families (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US11696714B2 (en) * | 2019-04-24 | 2023-07-11 | Interaxon Inc. | System and method for brain modelling |
| CN111444342B (en) * | 2020-03-24 | 2021-12-10 | 湖南董因信息技术有限公司 | Short text classification method based on multiple weak supervision integration |
-
2021
- 2021-06-25 CN CN202110713453.9A patent/CN113378966B/en active Active
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111242101A (en) * | 2020-03-08 | 2020-06-05 | 电子科技大学 | Behavior identification method based on spatiotemporal context association |
| CN112288034A (en) * | 2020-11-19 | 2021-01-29 | 江南大学 | Semi-supervised online anomaly detection method for wireless sensor network |
| CN112287089A (en) * | 2020-11-23 | 2021-01-29 | 腾讯科技（深圳）有限公司 | Classification model training and automatic question-answering method and device for automatic question-answering system |
Non-Patent Citations (2)
| Title |
|---|
| LSTM and HMM Comparison for Home Activity Anomaly Detection;Soon-Chang Poh，and etc;《2019 IEEE 3rd Information Technology, Networking, Electronic and Automation Control Conference (ITNEC)》;20190606;第1564-1568页 * |
| 基于LSTM模型的人体情景多标签识别研究;王嘉强等;《青岛大学学报（工程技术版）》;20181130;第33卷(第4期);第40-44页 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN113378966A (en) | 2021-09-10 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108920648B (en) | Cross-modal matching method based on music-image semantic relation | |
| CN104094287A (en) | A method, an apparatus and a computer software for context recognition | |
| CN114419736B (en) | Experiment scoring method, system, equipment and readable storage medium | |
| CN106649694A (en) | Method and device for identifying user's intention in voice interaction | |
| CN104978884A (en) | Teaching system of preschool education profession student music theory and solfeggio learning | |
| CN113365147A (en) | Video editing method, device, equipment and storage medium based on music card point | |
| CN110418204B (en) | Video recommendation method, device, equipment and storage medium based on micro expression | |
| CN112541529A (en) | Expression and posture fusion bimodal teaching evaluation method, device and storage medium | |
| CN109446946A (en) | A kind of multi-cam real-time detection method based on multithreading | |
| CN111126280B (en) | Gesture recognition fusion-based aphasia patient auxiliary rehabilitation training system and method | |
| CN112766218B (en) | Cross-domain pedestrian re-recognition method and device based on asymmetric combined teaching network | |
| CN110019703A (en) | Data markers method and device, intelligent answer method and system | |
| CN109165881A (en) | Teachers' instruction method for evaluating quality based on big data | |
| CN109783613A (en) | One kind searching topic method and system | |
| Lu et al. | Deep multimodal learning for municipal solid waste sorting | |
| CN111027419A (en) | Method, device, equipment and medium for detecting video irrelevant content | |
| CN113378966B (en) | Mobile phone sensor data labeling method based on weak supervised learning | |
| CN109409423A (en) | A kind of image-recognizing method, device, terminal and readable storage medium storing program for executing | |
| CN109657527B (en) | Painting brush touch identification system and method | |
| CN110472655A (en) | A kind of marker machine learning identifying system and method for border tourism | |
| CN103744958A (en) | Webpage classification algorithm based on distributed computation | |
| CN104850225A (en) | Activity identification method based on multi-level fusion | |
| CN114492310B (en) | Text labeling method, text labeling device, electronic equipment and storage medium | |
| CN106373121A (en) | Fuzzy image identification method and apparatus | |
| CN109583277B (en) | Gender determination method of barefoot footprint based on CNN |
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 |