CN109948686A - A kind of stroke recognition methods based on nine axis transducing signal statistical natures - Google Patents
A kind of stroke recognition methods based on nine axis transducing signal statistical natures Download PDFInfo
- Publication number
- CN109948686A CN109948686A CN201910187161.9A CN201910187161A CN109948686A CN 109948686 A CN109948686 A CN 109948686A CN 201910187161 A CN201910187161 A CN 201910187161A CN 109948686 A CN109948686 A CN 109948686A
- Authority
- CN
- China
- Prior art keywords
- signal
- feature
- stroke
- svm
- methods based
- 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.)
- Granted
Links
- 238000000034 method Methods 0.000 title claims abstract description 34
- 230000002463 transducing effect Effects 0.000 title claims abstract description 20
- 230000009182 swimming Effects 0.000 claims abstract description 11
- 230000011218 segmentation Effects 0.000 claims abstract description 10
- 210000000707 wrist Anatomy 0.000 claims abstract description 9
- 238000000605 extraction Methods 0.000 claims abstract description 7
- 230000008569 process Effects 0.000 claims abstract description 6
- 230000001133 acceleration Effects 0.000 claims abstract description 4
- 238000012549 training Methods 0.000 claims abstract description 4
- 238000012360 testing method Methods 0.000 claims description 7
- 238000004364 calculation method Methods 0.000 claims description 4
- 238000005457 optimization Methods 0.000 claims description 4
- 238000012545 processing Methods 0.000 claims description 4
- 239000000284 extract Substances 0.000 claims description 3
- 230000009471 action Effects 0.000 claims description 2
- 230000008901 benefit Effects 0.000 abstract description 3
- 230000007812 deficiency Effects 0.000 description 1
- 230000036541 health Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000012544 monitoring process Methods 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
Landscapes
- Measurement Of The Respiration, Hearing Ability, Form, And Blood Characteristics Of Living Organisms (AREA)
Abstract
The invention discloses a kind of stroke recognition methods based on nine axis transducing signal statistical natures, comprising: the acceleration and angular speed of wrist portion when the nine axis inertial sensors record human body swimming being mounted at wrist, and as measured signal;The measured signal for taking out a unit length carries out identification segmentation;The signal obtained to segmentation pre-processes, and carries out feature extraction;SVM model is trained, the feature of extraction is input in the SVM model after training and is classified, determines Modulation recognition result;According to Modulation recognition as a result, determining the section of the signal segment of the unit length to be taken out in measured signal, repeat the above steps.The present invention is classified using statistical nature, can have the advantages that computation complexity is low and accuracy rate is high in the case where smaller computing resource.
Description
Technical field
The present invention relates to human action identification field more particularly to a kind of strokes based on nine axis transducing signal statistical natures
Recognition methods.
Background technique
With the prevalence of Wrist wearable type smart machine, the human motion recognition method based on nine axle sensors is widely studied simultaneously
Gradually it is applied in human lives, and in terms of measuring of human health, the elderly's safety monitoring and wired home, plays
Increasingly important role.But smart machine has that cruising ability and calculating are limited, and therefore, the calculating of algorithm is opened
The problem of pin becomes research direction core with accuracy rate.
Summary of the invention
It is an object of the invention to overcome the deficiencies of the prior art and provide a kind of based on nine axis transducing signal statistical natures
Stroke recognition methods.The present invention is classified using statistical nature, can be had in the case where smaller computing resource and be calculated
The advantage that complexity is low and accuracy rate is high.
The purpose of the present invention can be achieved through the following technical solutions:
A kind of stroke recognition methods based on nine axis transducing signal statistical natures, specific steps include:
(1) acceleration and angular speed of wrist portion is as measured signal when obtaining the human body swimming of nine axle sensors acquisition;
(2) measured signal for taking out a unit length carries out identification segmentation;
(3) signal obtained to segmentation pre-processes, and carries out feature extraction;
(4) feature that step (3) are extracted is input in SVM model and is classified, determine Modulation recognition result;It is described
SVM model is the trained model of parameter;
(5) according to Modulation recognition as a result, determine measured signal in the unit length to be taken out signal segment section,
Repeat step (2)~(5).
Specifically, in the step (2), identify and be partitioned into measured signal segment first that there is the entire motion period
Signal;If a signal with the entire motion period in measured signal segment cannot be partitioned into, it is partitioned into a fixation
The long signal of window;The window is long should be greater than or equal to all strokes the maximum actuation period.
Specifically, in the step (3), divide obtained signal for described pair and pre-process are as follows: to every one-dimensional signal into
The processing of row bilinear interpolation, keeps the length of all signals consistent.
Specifically, in the step (3), the step of extracting feature to signal after pretreatment, includes:
(3-1) carries out size sequence to every one-dimensional signal, takes the value of its six branches position as feature 1;
(3-2) calculates mean value, variance and the energy of every one-dimensional signal, as feature 2;
(3-3) merges feature 1 and feature 2, and feature is normalized after merging to every one-dimensional signal, obtains being extracted
Feature.
Specifically, in the step (4), the step of being trained to SVM model parameter, includes:
(4-1) builds the identification sample set of the stroke based on nine axis transducing signal statistical natures;
(4-2) initializes SVM model;
(4-3) is trained the SVM model of initialization using built sample set and optimization cost function.
Further, the step of building of stroke identification sample set includes: in the step (4-1)
(4-1-1) tester wears nine axle sensors at wrist, successively carries out breaststroke, backstroke, freestyle swimming and butterfly stroke,
Synchronization video is recorded simultaneously;
(4-1-2) is split according to signal of the synchronization video to the nine axle sensors acquisition that tester wears, and is divided
The signal segment for providing the entire motion period puts on corresponding stroke label;
(4-1-3) carries out pretreatment to the signal segment with the entire motion period being partitioned into and statistical nature extracts;
It is described pretreatment and feature extracting method in step (3) pretreatment and feature extracting method it is identical;
Signal segment with label label is divided into training set and test set using leaving-one method by (4-1-4), from same
The signal segment used of tester should be grouped into identity set.
Further, in the step (4-2), SVM model initialization are as follows: the penalty coefficient of error items is set as 1, core
Type function is set as diameter as kernel function, kernel function coefficient equipment 0.0045.
Further, optimize the calculation formula of cost function in the step (4-3) are as follows:
αi>=0, i=1,2 ..., m
Wherein, αiFor required parameter, xiFor the statistical characteristics of sample i, yiFor the label of sample i, k () is core letter
Number, k (xi,xj) it is value of the statistical characteristics of sample i and sample j in kernel function.
Specifically, in the step (4), the step of SVM model is classified, includes:
(4-a) sets five class target sorting items, respectively sets the stroke of identification, i.e. breaststroke, backstroke, freestyle swimming, butterfly stroke,
And non-targeted movement, a five value classifiers are constructed using SVM;
(4-b) designs a SVM between any two sorting item, therefore five value classifiers need 10 SVM;
(4-c) tests the statistical nature of unknown sample with 10 SVM respectively, records each SVM output as a result, adopting
Final classification result is determined with ballot form.
The present invention compared to the prior art, have it is below the utility model has the advantages that
1, the present invention classifies to signal by using statistical nature, reduces the meter to smart machine to a certain extent
The requirement of calculation ability and resource, while accuracy rate can be made to be greatly improved as classifier using SVM.
2, the present invention is by being partitioned into basic processing unit of the swimming signal of complete cycle in signal as classifier, energy
It is enough to realize the accurately stroke classification of motion in real time.
Detailed description of the invention
Fig. 1 is a kind of flow chart of stroke recognition methods based on nine axis transducing signal statistical natures.
Specific embodiment
Present invention will now be described in further detail with reference to the embodiments and the accompanying drawings, but embodiments of the present invention are unlimited
In this.
Embodiment
It is as shown in Figure 1 a kind of flow chart of stroke recognition methods based on nine axis transducing signal statistical natures, it is specific to walk
Suddenly include:
(1) acceleration and angular speed of wrist portion is as measured signal when obtaining the human body swimming of nine axle sensors acquisition;
(2) measured signal for taking out a unit length carries out identification segmentation;
(3) signal obtained to segmentation pre-processes, and carries out feature extraction;
(4) feature that step (3) are extracted is input in SVM model and is classified, determine Modulation recognition result;It is described
SVM model is the trained model of parameter;
(5) according to Modulation recognition as a result, determine measured signal in the unit length to be taken out signal segment section,
Repeat step (2)~(5).
Specifically, it in the step (2), identifies and is partitioned into first letter with the entire motion period in signal segment
Number;If a signal with the entire motion period in signal segment cannot be partitioned into, it is partitioned into the long letter of fixed window
Number, window is long should be greater than or equal to all strokes the maximum actuation period, in the present embodiment, the long maximum actuation for taking all strokes of window
Period.
Specifically, in the step (3), divide obtained signal for described pair and pre-process are as follows: to every one-dimensional signal into
The processing of row bilinear interpolation, keeps the length of all signals consistent.
Specifically, in the step (3), to including: the step of signal extraction feature after pretreatment
(3-1) carries out size sequence to every one-dimensional signal, takes the value of its six branches position as feature 1;
(3-2) calculates mean value, variance and the energy of every one-dimensional signal, as feature 2;
(3-3) merges feature 1 and feature 2, and feature is normalized after merging to every one-dimensional signal, obtains being extracted
Feature.
Specifically, in the step (4), the step of being trained to SVM model parameter, includes:
(4-1) builds the identification sample set of the stroke based on nine axis transducing signal statistical natures;
(4-2) initializes SVM model;
(4-3) is trained the SVM model of initialization using built sample set and optimization cost function;
Further, the step of building of stroke identification sample set includes: in the step (4-1)
(4-1-1) tester wears nine axle sensors at wrist, successively carries out breaststroke, backstroke, freestyle swimming and butterfly stroke,
Synchronization video is recorded simultaneously;
(4-1-2) is split according to signal of the synchronization video to the nine axle sensors acquisition that tester wears, and is divided
The signal segment for providing the entire motion period puts on corresponding stroke label;
(4-1-3) carries out pretreatment to the signal segment with the entire motion period being partitioned into and statistical nature extracts;
It is described pretreatment and feature extracting method in step (3) pretreatment and feature extracting method it is identical.
Signal segment with label label is divided into training set and test set using leaving-one method by (4-1-4), from same
The signal segment used of tester should be grouped into identity set.
Further, in the step (4-2), SVM model initialization are as follows: the penalty coefficient of error items is set as 1, core
Type function is set as diameter as kernel function, kernel function coefficient equipment 0.0045.
Further, optimize the calculation formula of cost function in the step (4-3) are as follows:
αi>=0, i=1,2 ..., m
Wherein, αiFor required parameter, xiFor the statistical characteristics of sample i, yiFor the label of sample i, k () is core letter
Number, k (xi,xj) it is value of the statistical characteristics of sample i and sample j in kernel function.
Specifically, in the step (4), the step of SVM model is classified, includes:
(4-a) sets five class target sorting items, respectively sets the stroke of identification, i.e. breaststroke, backstroke, freestyle swimming, butterfly stroke,
And non-targeted movement, a five value classifiers are constructed using SVM;
(4-b) designs a SVM between any two sorting item, therefore five value classifiers need 10 SVM;
(4-c) tests the statistical nature of unknown sample with 10 SVM respectively, records each SVM output as a result, adopting
Final classification result is determined with ballot form.
The above embodiment is a preferred embodiment of the present invention, but embodiments of the present invention are not by above-described embodiment
Limitation, other any changes, modifications, substitutions, combinations, simplifications made without departing from the spirit and principles of the present invention,
It should be equivalent substitute mode, be included within the scope of the present invention.
Claims (9)
1. a kind of stroke recognition methods based on nine axis transducing signal statistical natures, which is characterized in that specific steps include:
(1) acceleration and angular speed of wrist portion is as measured signal when obtaining the human body swimming of nine axle sensors acquisition;
(2) measured signal for taking out a unit length carries out identification segmentation;
(3) signal obtained to segmentation pre-processes, and carries out feature extraction;
(4) feature that step (3) are extracted is input in SVM model and is classified, determine Modulation recognition result;The SVM mould
Type is the trained model of parameter;
(5) according to Modulation recognition as a result, determining the section of the signal segment of the unit length to be taken out in measured signal, repeatedly
Step (2)~(5).
2. a kind of stroke recognition methods based on nine axis transducing signal statistical natures according to claim 1, feature exist
In identifying and the signal with the entire motion period that is partitioned into signal segment first in the step (2);If cannot divide
A signal with the entire motion period in signal segment is cut out, then is partitioned into the long signal of fixed window.
3. a kind of stroke recognition methods based on nine axis transducing signal statistical natures according to claim 1, feature exist
In in the step (3), the signal that described pair of segmentation obtains is pre-processed are as follows: carries out bilinear interpolation to every one-dimensional signal
Processing, keeps the length of all signals consistent.
4. a kind of stroke recognition methods based on nine axis transducing signal statistical natures according to claim 1, feature exist
In in the step (3), to including: the step of signal extraction feature after pretreatment
(3-1) carries out size sequence to every one-dimensional signal, takes the value of its six branches position as feature 1;
(3-2) calculates mean value, variance and the energy of every one-dimensional signal, as feature 2;
(3-3) merges feature 1 and feature 2, and the feature after merging to every one-dimensional signal is normalized, and obtains extracting spy
Sign.
5. a kind of stroke recognition methods based on nine axis transducing signal statistical natures according to claim 1, feature exist
In in the step (4), the step of being trained to SVM model parameter includes:
(4-1) builds the identification sample set of the stroke based on nine axis transducing signal statistical natures;
(4-2) initializes SVM model;
(4-3) is trained the SVM model of initialization using built sample set and optimization cost function.
6. a kind of stroke recognition methods based on nine axis transducing signal statistical natures according to claim 5, feature exist
In the step of building of stroke identification sample set includes: in the step (4-1)
(4-1-1) tester wears nine axle sensors at wrist, successively carries out breaststroke, backstroke, freestyle swimming and butterfly stroke, simultaneously
Record synchronization video;
(4-1-2) is split according to signal of the synchronization video to the nine axle sensors acquisition that tester wears, and segmentation is provided
There is the signal segment in entire motion period, puts on corresponding stroke label;
(4-1-3) carries out pretreatment to the signal segment with the entire motion period being partitioned into and statistical nature extracts;It is described
Pretreatment and feature extracting method in step (3) pretreatment and feature extracting method it is identical;
Signal segment with label label is divided into training set and test set using leaving-one method by (4-1-4), comes from same test
The signal segment used of personnel should be grouped into identity set.
7. a kind of stroke recognition methods based on nine axis transducing signal statistical natures according to claim 5, feature exist
In, in the step (4-2), SVM model initialization are as follows: the penalty coefficient of error items is set as 1, kernel function type is set as radial
Kernel function, kernel function coefficient are set as 0.0045.
8. a kind of stroke recognition methods based on nine axis transducing signal statistical natures according to claim 5, feature exist
In the calculation formula of optimization cost function in the step (4-3) are as follows:
αi>=0, i=1,2 ..., m
Wherein, αiFor required parameter, xiFor the statistical characteristics of sample i, yiFor the label of sample i, k () is kernel function, k
(xi,xj) it is value of the statistical characteristics of sample i and sample j in kernel function.
9. a kind of stroke recognition methods based on nine axis transducing signal statistical natures according to claim 1, feature exist
In in the step (4), the step of SVM model is classified includes:
(4-a) sets five class target sorting items, respectively sets the stroke of identification, i.e. breaststroke, backstroke, freestyle swimming, butterfly stroke and non-
Target action constructs a five value classifiers using SVM;
(4-b) designs a SVM between any two sorting item, therefore five value classifiers need 10 SVM;
(4-c) tests the statistical nature of unknown sample with 10 SVM respectively, records each SVM output as a result, using throwing
Ticket form determines final classification result.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910187161.9A CN109948686B (en) | 2019-03-13 | 2019-03-13 | Swimming stroke identification method based on nine-axis sensing signal statistical characteristics |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910187161.9A CN109948686B (en) | 2019-03-13 | 2019-03-13 | Swimming stroke identification method based on nine-axis sensing signal statistical characteristics |
Publications (2)
Publication Number | Publication Date |
---|---|
CN109948686A true CN109948686A (en) | 2019-06-28 |
CN109948686B CN109948686B (en) | 2021-06-08 |
Family
ID=67008814
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201910187161.9A Expired - Fee Related CN109948686B (en) | 2019-03-13 | 2019-03-13 | Swimming stroke identification method based on nine-axis sensing signal statistical characteristics |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN109948686B (en) |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110705496A (en) * | 2019-10-11 | 2020-01-17 | 成都乐动信息技术有限公司 | Swimming posture identification method based on nine-axis sensor |
Citations (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN106237604A (en) * | 2016-08-31 | 2016-12-21 | 歌尔股份有限公司 | Wearable device and the method utilizing its monitoring kinestate |
CN107376247A (en) * | 2017-08-16 | 2017-11-24 | 广东远峰电子科技股份有限公司 | A kind of swimming exercise analysis method based on intelligent watch and the intelligent watch |
CN107704799A (en) * | 2017-08-10 | 2018-02-16 | 深圳市金立通信设备有限公司 | A kind of human motion recognition method and equipment, computer-readable recording medium |
US10007863B1 (en) * | 2015-06-05 | 2018-06-26 | Gracenote, Inc. | Logo recognition in images and videos |
CN108460322A (en) * | 2017-12-28 | 2018-08-28 | 惠州市德赛工业研究院有限公司 | A kind of stroke recognition methods and application |
CN109029432A (en) * | 2018-05-31 | 2018-12-18 | 华南理工大学 | A kind of human action detection and recognition methods based on six axis inertia sensing signals |
US10204289B2 (en) * | 2017-06-14 | 2019-02-12 | International Business Machines Corporation | Hieroglyphic feature-based data processing |
-
2019
- 2019-03-13 CN CN201910187161.9A patent/CN109948686B/en not_active Expired - Fee Related
Patent Citations (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US10007863B1 (en) * | 2015-06-05 | 2018-06-26 | Gracenote, Inc. | Logo recognition in images and videos |
CN106237604A (en) * | 2016-08-31 | 2016-12-21 | 歌尔股份有限公司 | Wearable device and the method utilizing its monitoring kinestate |
US10204289B2 (en) * | 2017-06-14 | 2019-02-12 | International Business Machines Corporation | Hieroglyphic feature-based data processing |
CN107704799A (en) * | 2017-08-10 | 2018-02-16 | 深圳市金立通信设备有限公司 | A kind of human motion recognition method and equipment, computer-readable recording medium |
CN107376247A (en) * | 2017-08-16 | 2017-11-24 | 广东远峰电子科技股份有限公司 | A kind of swimming exercise analysis method based on intelligent watch and the intelligent watch |
CN108460322A (en) * | 2017-12-28 | 2018-08-28 | 惠州市德赛工业研究院有限公司 | A kind of stroke recognition methods and application |
CN109029432A (en) * | 2018-05-31 | 2018-12-18 | 华南理工大学 | A kind of human action detection and recognition methods based on six axis inertia sensing signals |
Non-Patent Citations (2)
Title |
---|
BRANDON VICTOR ET AL.: "Continuous Video to Simple Signals for Swimming Stroke Detection With Convolutional Neural Networks", 《PROCEEDINGS OF THE IEEE CONFERENCE ON COMPUTER VISION AND PATTERN RECOGNITION (CVPR) WORKSHOPS》 * |
吴海龙: "基于加速度传感器的游泳监测系统分析与实现", 《中国优秀硕士学位论文全文数据库信息科技辑》 * |
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110705496A (en) * | 2019-10-11 | 2020-01-17 | 成都乐动信息技术有限公司 | Swimming posture identification method based on nine-axis sensor |
CN110705496B (en) * | 2019-10-11 | 2023-04-18 | 成都乐动信息技术有限公司 | Swimming posture identification method based on nine-axis sensor |
Also Published As
Publication number | Publication date |
---|---|
CN109948686B (en) | 2021-06-08 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN110245718A (en) | A kind of Human bodys' response method based on joint time-domain and frequency-domain feature | |
CN108446733A (en) | A kind of human body behavior monitoring and intelligent identification Method based on multi-sensor data | |
CN100415159C (en) | Dynamic characteristic analysis method of real-time tendency of heart state | |
CN109086698A (en) | A kind of human motion recognition method based on Fusion | |
CN106580282A (en) | Human body health monitoring device, system and method | |
WO2010083562A1 (en) | Activity detection | |
Ghazali et al. | Common sport activity recognition using inertial sensor | |
CN103886341A (en) | Gait behavior recognition method based on feature combination | |
CN110084286A (en) | A kind of human motion recognition method of sensor-based ECOC technology | |
CN111597991A (en) | Rehabilitation detection method based on channel state information and BilSTM-Attention | |
CN106127191B (en) | Brain electricity classification method based on WAVELET PACKET DECOMPOSITION and logistic regression | |
CN111914643A (en) | Human body action recognition method based on skeleton key point detection | |
CN107092861A (en) | Lower limb movement recognition methods based on pressure and acceleration transducer | |
CN109330584A (en) | Electrocardiosignal personal identification method and system based on dictionary learning and rarefaction representation | |
Xu et al. | Intelligent emotion detection method based on deep learning in medical and health data | |
CN108717548B (en) | Behavior recognition model updating method and system for dynamic increase of sensors | |
CN108985223A (en) | A kind of human motion recognition method | |
CN108958474A (en) | A kind of action recognition multi-sensor data fusion method based on Error weight | |
CN106643722A (en) | Method for pet movement identification based on triaxial accelerometer | |
CN110008847A (en) | A kind of stroke recognition methods based on convolutional neural networks | |
CN108008151A (en) | A kind of moving state identification method and system based on 3-axis acceleration sensor | |
Wenyuan et al. | From Lens to Prism: Device-free modeling and recognition of multi-part activities | |
CN110705599B (en) | Human body action recognition method based on online transfer learning | |
CN109271889A (en) | A kind of action identification method based on the double-deck LSTM neural network | |
CN109948686A (en) | A kind of stroke recognition methods based on nine axis transducing signal statistical natures |
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 | ||
CF01 | Termination of patent right due to non-payment of annual fee | ||
CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20210608 |