CN106370180B - Inertial sensor initial position recognition methods based on dynamic time warping algorithm - Google Patents
Inertial sensor initial position recognition methods based on dynamic time warping algorithm Download PDFInfo
- Publication number
- CN106370180B CN106370180B CN201610739199.9A CN201610739199A CN106370180B CN 106370180 B CN106370180 B CN 106370180B CN 201610739199 A CN201610739199 A CN 201610739199A CN 106370180 B CN106370180 B CN 106370180B
- Authority
- CN
- China
- Prior art keywords
- inertial sensor
- initial position
- signal
- dynamic time
- time warping
- 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
- G01—MEASURING; TESTING
- G01C—MEASURING DISTANCES, LEVELS OR BEARINGS; SURVEYING; NAVIGATION; GYROSCOPIC INSTRUMENTS; PHOTOGRAMMETRY OR VIDEOGRAMMETRY
- G01C21/00—Navigation; Navigational instruments not provided for in groups G01C1/00 - G01C19/00
- G01C21/10—Navigation; Navigational instruments not provided for in groups G01C1/00 - G01C19/00 by using measurements of speed or acceleration
- G01C21/12—Navigation; Navigational instruments not provided for in groups G01C1/00 - G01C19/00 by using measurements of speed or acceleration executed aboard the object being navigated; Dead reckoning
- G01C21/16—Navigation; Navigational instruments not provided for in groups G01C1/00 - G01C19/00 by using measurements of speed or acceleration executed aboard the object being navigated; Dead reckoning by integrating acceleration or speed, i.e. inertial navigation
Landscapes
- Engineering & Computer Science (AREA)
- Radar, Positioning & Navigation (AREA)
- Remote Sensing (AREA)
- Automation & Control Theory (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Measurement Of The Respiration, Hearing Ability, Form, And Blood Characteristics Of Living Organisms (AREA)
Abstract
The present invention discloses a kind of inertial sensor initial position recognition methods based on dynamic time warping algorithm, including signal acquisition: inertial sensor being fixed on to the initial position of concrete application requirement, then acquires the original signal of inertial sensor in a period of time；Signal Pretreatment: denoising and rejecting abnormalities point, to filter out the system noise in original signal collection process and reject in communication and generated exceptional value in transmission process；Calculating similarity is carried out to pretreated signal based on dynamic time warping algorithm；Export the recognition result of inertial sensor initial position；Signal and correct initial position signal when being in initial position to inertial sensor using dynamic time warping algorithm carry out similarity measurements quantization, to identify and judge to the correctness of inertial sensor initial position；This method can be adapted for the time series of Length discrepancy, and can be adapted for the presence of the case where offset on a timeline, while not need acquisition early period great amount of samples.
Description
Technical field
The inertial sensor initial position recognition methods based on dynamic time warping algorithm that the present invention relates to a kind of is a kind of
Knowledge method for distinguishing can be carried out to the spatial attitude of inertial sensor and initial position automatically.
Background technique
In recent years, with the development of the technologies such as internet, Internet of Things, MEMS and wearable device, inertial sensor unit
(Inertial Measurement Unit, IMU) has obtained applying extensively in all trades and professions.Specific to medical treatment
Field, inertial sensor unit health monitoring, tumble early warning, movement capture and identification, in terms of have it is a large amount of
Research.
However, the experimental results show that the initial position of inertial sensor influences whether subsequent data handling procedure,
The result even applied.For example: in the application of wearable device, if the initial position of inertial sensor wears mistake,
Then may result in mistake as a result: the wrong report of tumble event, action recognition mistake etc..Therefore, to inertial sensor
Before data are analyzed, need that the initial position of inertial sensor is judged and identified.
Currently, general method includes two classes: (one) utilizes template matching method, for example: euclidean distance method, related coefficient
Method etc. measures the similitude between the initial position of inertial sensor and the initial position of " goldstandard ".However such method is logical
It is often only applicable to isometric time series, and sensitive to the offset of sequence on a timeline；(2) artificial intelligence and machine are used
The methods of study, for example: artificial neural network, support vector machines, extreme learning machine, decision tree, random forests algorithm etc..It is sorry
, many adjustable parameters are generally comprised in such method, and in order to establish the good identification model of Generalization Capability, early period
Need to acquire a large amount of sample data.
In conclusion in the application process of inertial sensor unit, in order to simplify the calculation of subsequent data analysis and processing
Method, while in order to avoid the interference of unnecessary factor, before the use, it is necessary to the initial position of inertial sensor unit into
Row identification.In addition, the initial position of inertial sensor unit can not be observed by the naked eye under the scene of some remote applications
When, it is a kind of can automatically to the initial position of inertial sensor unit carry out know method for distinguishing be then particularly important.
Therefore, it is necessary to design a kind of new inertial sensor initial position identification side based on dynamic time warping algorithm
Method, to solve the above technical problems.
Summary of the invention
The problem of for background technique, the object of the present invention is to provide a kind of based on dynamic time warping algorithm
Inertial sensor initial position recognition methods, can be adapted for the time series of Length discrepancy, and can be adapted on a timeline
The case where in the presence of offset, while not needing acquisition early period great amount of samples.
The technical scheme of the present invention is realized as follows: a kind of inertial sensor based on dynamic time warping algorithm is initial
Location recognition method, comprising the following steps: S1, signal acquisition: inertial sensor unit is fixed or is worn on first and is specifically answered
With desired initial position, the original signal of inertial sensor in a period of time is then acquired；S2, Signal Pretreatment: denoising and
Rejecting abnormalities point, to filter out the system noise in original signal collection process and reject original signal in communication and transmission process
Generated exceptional value；S3, calculating similarity is carried out to pretreated signal based on dynamic time warping algorithm；S4, output
Recognition result: output utilizes the calculated shortest path value of time wrapping algorithm.
In the above-mentioned technical solutions, the inertial sensor includes 3 axle accelerations, 3 axis gyroscopes and 3 axis magnetometers.
In the above-mentioned technical solutions, the system noise includes that high-frequency random noises, Hz noise noise and low frequency are artificial
Introduce noise.
In the above-mentioned technical solutions, the Signal Pretreatment includes smothing filtering, Wavelet Denoising Method, low-pass filtering, high pass filter
Wave, bandpass filtering, median filtering and derivative filtering.
In the above-mentioned technical solutions, the derivative filtering includes single order, second order or higher order.
It in the above-mentioned technical solutions, include preset more when the progress similarity calculation based on dynamic time warping algorithm
Inertial sensor initial position signal template library under a different mode.
The present invention is based on the inertial sensor initial position recognition methods of dynamic time warping algorithm, including signal acquisition,
Signal Pretreatment carries out similarity calculation and output recognition result to pretreated signal based on dynamic time warping algorithm,
The signal of signal and correct initial position when it utilizes dynamic time warping algorithm to be in initial position to inertial sensor
Similarity measurements quantization is carried out, to identify and judge to the correctness of inertial sensor initial position.With conventional method
It compares, this method can be adapted for the time series of Length discrepancy, and can be adapted for the presence of the case where offset on a timeline, together
When do not need early period acquisition great amount of samples.
Detailed description of the invention
Fig. 1 is that the present invention is based on the signals of the inertial sensor initial position recognition methods process of dynamic time warping algorithm
Figure；
Fig. 2 is dynamic time warping algorithm basic schematic diagram；
Fig. 3 is inertial sensor wearing position schematic diagram；
Fig. 4 is DTW shortest path schematic diagram when inertial sensor is worn on forearm；
Fig. 5 is DTW shortest path schematic diagram when inertial sensor is worn on upper arm.
Specific embodiment
Following will be combined with the drawings in the embodiments of the present invention, and technical solution in the embodiment of the present invention carries out clear, complete
Site preparation description, it is clear that the described embodiment is only a part of the embodiment of the present invention, instead of all the embodiments.Based on this
Embodiment in invention, every other reality obtained by those of ordinary skill in the art without making creative efforts
Example is applied, shall fall within the protection scope of the present invention.
As shown in Figure 1, a kind of inertial sensor initial position based on dynamic time warping algorithm of the present invention is known
Other method, comprising:
Step: S1, signal acquisition.Inertial sensor unit is fixed or is worn on first the initial bit of concrete application requirement
It sets, then utilizes the original signal of corresponding signal acquisition software collection inertial sensor interior for a period of time.Here inertia
Sensor generally includes 3 axle accelerations, 3 axis gyroscopes and 3 axis magnetometers, and is pointed out that proposed by the invention
In method, signal acquisition process does not need to limit the type of inertial sensor unit and signal acquisition software, i.e. hardware and software
Type, while also do not need limit sample frequency size.As long as original inertial sensor signal can be exported.
Step: S2, Signal Pretreatment.In order to eliminate the influence of additional interference factor, need to collected original signal
It is pre-processed.The step mainly includes two submodules: denoising and rejecting abnormalities point.The purpose of denoising mainly filters out signal
System noise in collection process, i.e. high-frequency random noises, Hz noise noise, low frequency are artificially introduced noise；And rejecting abnormalities
The purpose of point is to reject signal in communication and exceptional value caused by packet loss in transmission process.It, should in specific case study on implementation
The including but not limited to following common several method of the concrete methods of realizing of step: smothing filtering, Wavelet Denoising Method, low pass filtered
Wave, high-pass filtering, bandpass filtering, median filtering, derivative filtering, wherein derivative filtering includes single order, second order or higher order etc..
Step: calculating similarity S3, is carried out to pretreated signal based on dynamic time warping algorithm.The step is whole
The core of a recognizer, the input of the step is in addition to also wrapping by above-mentioned steps S1 and S2 treated signal to be identified
Include the inertial sensor initial position signal template library under preset multiple and different modes.Different mode mentioned herein refers to
The different application scenarios of inertial sensor.For example: when inertial sensor is worn on arm, usually requiring that initial position is hand
Arm naturally droops state, and some axis in X, Y and Z axis of inertial sensor is consistent with acceleration of gravity direction.
In this step, dynamic time warping (Dynamic Time Warping, DTW) algorithm is that one kind passes through calculating
The Time Warp path of minimum cost carries out the method for measuring similarity of matching mapping come the form to time series, its quilt earliest
Processing applied to voice signal.
Basic principle based on dynamic time warping algorithm as shown in Fig. 2, for example, it is assumed that be respectively there are two length
WithTime seriesWith.Distance matrixByWithIt is European
Square composition of distance, it may be assumed that
In distance matrixOne paths of middle searchingSo thatWithSimilitude most
That is, height willWithAfter being extended and being shortened, the distance of two time serieses is most short.
In view of pathIn element need to meet the conditions such as boundary constraint, monotonicity and continuity constraint, therefore, can
To carry out the solution of optimal path by dynamic programming method.Define cumulant matrixRecord shortest path,
That is:
Step: S4, output recognition result.The output of the step is to utilize the calculated shortest path of dynamic time warping algorithm
Diameter value R, R is smaller to show that the similarity between the template signal under signal to be identified and designated mode i.e. " goldstandard " signal is got over
It is high.By setting certain threshold value, that is, it can determine that whether current inertial sensor position meets the requirements, i.e., whether wear mistake
Or it is abnormal.Here threshold value needs the signal source different according to different application scenarios model selections, i.e., acceleration transducer or
The different directions (X-axis or Y-axis or Z axis) of gyroscope or magnetometer and determine.
Specific embodiment is carried out to the present invention below in conjunction with Fig. 3 and Fig. 4 and Fig. 5 to analyze:
In the present embodiment, acquire respectively and complete 4 of shoulder elbow section movements, i.e., shoulder joint is anteflexion, shoulder abduction,
Upper extremity exercise inertial sensor data before forearm revolves when supination, hand touching lumbar vertebrae, wherein default inertial sensor is worn on forearm
For " goldstandard " position, and inertial sensor is worn on the position that upper arm is mistake.Inertial sensor module in the present embodiment
Sample rate is 40Hz.
Before data acquisition, two inertial sensors are first worn on to the forearm and upper arm of patient's hemiplegia side respectively, such as Fig. 3 institute
Show.
Fig. 4 and Fig. 5 is described when inertial sensor module is worn on different positions, i.e. forearm and when upper arm, shoulder joint
The shortest path figure that DTW algorithm between the motor message of anteflexion movement and " goldstandard " motor message obtains.It can be with from figure
Find out, when the wearing position of inertial sensor module is correct, as shown in figure 4, its DTW shortest path R is 5.8337, when used
Property sensor module wearing mistake position when, as shown in figure 5, its DTW shortest path R be 11.6908, hence it is evident that it can be seen that
DTW shortest path R when wearing position is correct is less than the DTW shortest path R of wearing mistake, i.e., and 5.8337 < 11.6908.
The following table 1 is listed in detail when inertial sensor under the movements of four shoulder elbow sections is worn on forearm and upper arm respectively
DTW shortest path result.It can visually see from table, when the initial position difference of inertial sensor, utilize DTW algorithm
There are significant differences for calculated shortest path.In the present embodiment, when the threshold value L of shortest path is set as 7, and to inertia
The initial position of sensor identified, available 100% accuracy.
DTW shortest path when inertial sensor is worn on forearm and upper arm respectively under the movement of 1 four shoulder elbow sections of table
As a result
| Denomination of dive | Forearm DTW shortest path | Upper arm DTW shortest path |
| Shoulder joint is anteflexion | 5.8337 | 11.6908 |
| Shoulder abduction | 4.4367 | 9.8612 |
| Supination before forearm revolves | 6.1095 | 13.6624 |
| Hand touches lumbar vertebrae | 5.2193 | 12.6413 |
The present invention is based on the inertial sensor initial position recognition methods of dynamic time warping algorithm, including signal acquisition,
Signal Pretreatment carries out similarity calculation and output recognition result to pretreated signal based on dynamic time warping algorithm,
The signal of signal and correct initial position when it utilizes dynamic time warping algorithm to be in initial position to inertial sensor
Similarity measurements quantization is carried out, to identify and judge to the correctness of inertial sensor initial position.With conventional method
It compares, this method can be adapted for the time series of Length discrepancy, and can be adapted for the presence of the case where offset on a timeline, together
When do not need early period acquisition great amount of samples.
The foregoing is merely illustrative of the preferred embodiments of the present invention, is not intended to limit the invention, all in essence of the invention
Within mind and principle, any modification, equivalent replacement, improvement and so on be should all be included in the protection scope of the present invention.
Claims (6)
1. the inertial sensor initial position recognition methods based on dynamic time warping algorithm, it is characterised in that: including following step
It is rapid:
S1, signal acquisition: inertial sensor is fixed or is worn on first the initial position of concrete application requirement, then acquire
The original signal of inertial sensor in a period of time；
S2, Signal Pretreatment: denoising and rejecting abnormalities point, to filter out the system noise and rejecting in original signal collection process
Original signal is in communication and generated exceptional value in transmission process；
S3, calculating similarity is carried out to pretreated signal based on dynamic time warping algorithm；
S4, output recognition result: output utilizes the calculated shortest path value of time wrapping algorithm, by setting certain threshold
Value, determines whether current inertial sensor position meets the requirements, the difference that the threshold value is selected according to different application scenarios
Signal source and determine.
2. the inertial sensor initial position recognition methods according to claim 1 based on dynamic time warping algorithm,
Be characterized in that: the inertial sensor includes 3 axis accelerometers, 3 axis gyroscopes and 3 axis magnetometers.
3. the inertial sensor initial position recognition methods according to claim 1 based on dynamic time warping algorithm,
Be characterized in that: the system noise includes that high-frequency random noises, Hz noise noise and low frequency are artificially introduced noise.
4. the inertial sensor initial position recognition methods according to claim 1 based on dynamic time warping algorithm,
Be characterized in that: the Signal Pretreatment includes smothing filtering, Wavelet Denoising Method, low-pass filtering, high-pass filtering, bandpass filtering, intermediate value
Filtering and derivative filtering.
5. the inertial sensor initial position recognition methods according to claim 4 based on dynamic time warping algorithm,
Be characterized in that: the derivative filtering can be single order, second order or higher order.
6. the inertial sensor initial position recognition methods according to claim 1 based on dynamic time warping algorithm,
It is characterized in that: including the inertia biography under preset multiple and different modes when carrying out similarity calculation based on dynamic time warping algorithm
Sensor initial position signal template library.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610739199.9A CN106370180B (en) | 2016-08-29 | 2016-08-29 | Inertial sensor initial position recognition methods based on dynamic time warping algorithm |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610739199.9A CN106370180B (en) | 2016-08-29 | 2016-08-29 | Inertial sensor initial position recognition methods based on dynamic time warping algorithm |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN106370180A CN106370180A (en) | 2017-02-01 |
| CN106370180B true CN106370180B (en) | 2019-06-04 |
Family
ID=57904301
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201610739199.9A Active CN106370180B (en) | 2016-08-29 | 2016-08-29 | Inertial sensor initial position recognition methods based on dynamic time warping algorithm |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN106370180B (en) |
Families Citing this family (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109239456B (en) * | 2018-08-03 | 2020-12-25 | 福州大学 | Harmonic tracing method based on dynamic programming time series similarity algorithm |
| CN109157201A (en) * | 2018-08-13 | 2019-01-08 | 广州喜梁门科技有限公司 | A kind of system and its control method of bracelet equipment |
| CN110380390B (en) * | 2019-07-22 | 2021-05-14 | 西南交通大学 | High-voltage direct-current transmission line protection method based on traveling wave waveform similarity |
| CN114600417B (en) * | 2019-11-01 | 2024-04-26 | 株式会社电装 | Synchronization device, synchronization method, and storage device storing synchronization program |
| CN115868935B (en) * | 2022-12-16 | 2024-03-08 | 临沂中科睿鹤智慧科技有限公司 | Wearable-based portable balance evaluation device |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102427765A (en) * | 2009-05-20 | 2012-04-25 | 皇家飞利浦电子股份有限公司 | Sensing device for detecting a wearing position |
| CN103942456A (en) * | 2014-05-08 | 2014-07-23 | 北京睿仁医疗科技有限公司 | Measuring method and device capable of automatically judging wear position |
| CN103984416A (en) * | 2014-06-10 | 2014-08-13 | 北京邮电大学 | Gesture recognition method based on acceleration sensor |
| CN104434119A (en) * | 2013-09-20 | 2015-03-25 | 卡西欧计算机株式会社 | Body information obtaining device and body information obtaining method |
| CN105046215A (en) * | 2015-07-07 | 2015-11-11 | 中国科学院上海高等研究院 | Posture and behavior identification method without influences of individual wearing positions and wearing modes |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US7558622B2 (en) * | 2006-05-24 | 2009-07-07 | Bao Tran | Mesh network stroke monitoring appliance |
-
2016
- 2016-08-29 CN CN201610739199.9A patent/CN106370180B/en active Active
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102427765A (en) * | 2009-05-20 | 2012-04-25 | 皇家飞利浦电子股份有限公司 | Sensing device for detecting a wearing position |
| CN104434119A (en) * | 2013-09-20 | 2015-03-25 | 卡西欧计算机株式会社 | Body information obtaining device and body information obtaining method |
| CN103942456A (en) * | 2014-05-08 | 2014-07-23 | 北京睿仁医疗科技有限公司 | Measuring method and device capable of automatically judging wear position |
| CN103984416A (en) * | 2014-06-10 | 2014-08-13 | 北京邮电大学 | Gesture recognition method based on acceleration sensor |
| CN105046215A (en) * | 2015-07-07 | 2015-11-11 | 中国科学院上海高等研究院 | Posture and behavior identification method without influences of individual wearing positions and wearing modes |
Also Published As
| Publication number | Publication date |
|---|---|
| CN106370180A (en) | 2017-02-01 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN106370180B (en) | Inertial sensor initial position recognition methods based on dynamic time warping algorithm | |
| CN106228200B (en) | Action identification method independent of action information acquisition equipment | |
| US20200000379A1 (en) | Device and method for classifying the activity and/or counting steps of a user | |
| Kau et al. | A smart phone-based pocket fall accident detection, positioning, and rescue system | |
| CN107358248B (en) | Method for improving falling detection system precision | |
| CN105943013A (en) | Heart rate detection method and device, and intelligent wearable device | |
| Gjoreski et al. | Accelerometer data preparation for activity recognition | |
| CN110532898A (en) | A kind of physical activity recognition methods based on smart phone Multi-sensor Fusion | |
| Rigas et al. | Tremor UPDRS estimation in home environment | |
| Ali et al. | DeepHAR-Net: A novel machine intelligence approach for human activity recognition from inertial sensors | |
| Ghobadi et al. | A robust automatic gait monitoring approach using a single IMU for home-based applications | |
| CN105760819A (en) | Daily activity recognition method based on acceleration signal | |
| Ghobadi et al. | Foot-mounted inertial measurement unit for activity classification | |
| Samà et al. | Gait identification by means of box approximation geometry of reconstructed attractors in latent space | |
| IT201900019037A1 (en) | PERFECTED METHOD FOR DETECTING A WRIST TILT GESTURE AND ELECTRONIC UNIT AND WEARABLE ELECTRONIC DEVICE IMPLEMENTING THE SAME | |
| Deep et al. | Leveraging multi-view learning for human anomaly detection in industrial internet of things | |
| CN107270934A (en) | A kind of swimming information monitoring recognition methods based on wearable intelligent equipment | |
| Dimitrakopoulos et al. | Tremor quantification through event-based movement trajectory modeling | |
| CN112633467A (en) | Human behavior recognition method based on cat eye connection improved LSTM model | |
| Yunfang et al. | Survey of human posture recognition based on wearable device | |
| CN115731602A (en) | Human body activity recognition method, device, equipment and storage medium based on topological representation | |
| Gjoreski | A method for combining classical and deep machine learning for mobile health and behavior monitoring | |
| Abd Aziz et al. | Fall detection system: signal analysis in reducing false alarms using built-in tri-axial accelerometer | |
| CN113591561A (en) | Method and system for identity recognition based on spectrogram | |
| TW201729038A (en) | Method for controlling electronic equipment and wearable device |
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 |