CN115265531A - System for identifying motion state by using inertia measuring device - Google Patents
System for identifying motion state by using inertia measuring device Download PDFInfo
- Publication number
- CN115265531A CN115265531A CN202210812326.9A CN202210812326A CN115265531A CN 115265531 A CN115265531 A CN 115265531A CN 202210812326 A CN202210812326 A CN 202210812326A CN 115265531 A CN115265531 A CN 115265531A
- Authority
- CN
- China
- Prior art keywords
- data
- circuit
- motion state
- signal processing
- classification
- 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.)
- Pending
Links
- 230000001788 irregular Effects 0.000 claims abstract description 48
- 238000005259 measurement Methods 0.000 claims abstract description 44
- 238000012545 processing Methods 0.000 claims abstract description 43
- 238000007781 pre-processing Methods 0.000 claims abstract description 9
- 238000000034 method Methods 0.000 claims description 49
- 230000006870 function Effects 0.000 claims description 48
- 238000012549 training Methods 0.000 claims description 42
- 238000003466 welding Methods 0.000 claims description 17
- 230000001133 acceleration Effects 0.000 claims description 16
- 230000009471 action Effects 0.000 claims description 16
- 239000011159 matrix material Substances 0.000 claims description 11
- 238000001914 filtration Methods 0.000 claims description 10
- 238000010606 normalization Methods 0.000 claims description 10
- 230000009467 reduction Effects 0.000 claims description 7
- 229920002379 silicone rubber Polymers 0.000 claims description 7
- 238000002790 cross-validation Methods 0.000 claims description 6
- 238000000605 extraction Methods 0.000 claims description 5
- 238000006243 chemical reaction Methods 0.000 claims description 4
- 238000000354 decomposition reaction Methods 0.000 claims description 3
- 230000002093 peripheral effect Effects 0.000 claims description 3
- 230000003321 amplification Effects 0.000 claims description 2
- 238000003199 nucleic acid amplification method Methods 0.000 claims description 2
- 239000004945 silicone rubber Substances 0.000 claims description 2
- 238000012706 support-vector machine Methods 0.000 description 22
- 239000013598 vector Substances 0.000 description 14
- 238000001514 detection method Methods 0.000 description 13
- 230000008569 process Effects 0.000 description 13
- 230000000694 effects Effects 0.000 description 11
- 238000004422 calculation algorithm Methods 0.000 description 10
- 238000004364 calculation method Methods 0.000 description 9
- 238000013507 mapping Methods 0.000 description 9
- 238000012937 correction Methods 0.000 description 8
- 238000012360 testing method Methods 0.000 description 8
- 238000010586 diagram Methods 0.000 description 6
- 230000002829 reductive effect Effects 0.000 description 4
- 230000003068 static effect Effects 0.000 description 4
- 238000000513 principal component analysis Methods 0.000 description 3
- 230000035939 shock Effects 0.000 description 3
- 238000005476 soldering Methods 0.000 description 3
- XEEYBQQBJWHFJM-UHFFFAOYSA-N Iron Chemical compound [Fe] XEEYBQQBJWHFJM-UHFFFAOYSA-N 0.000 description 2
- 230000006399 behavior Effects 0.000 description 2
- 239000000969 carrier Substances 0.000 description 2
- 238000007635 classification algorithm Methods 0.000 description 2
- 238000010276 construction Methods 0.000 description 2
- 238000005034 decoration Methods 0.000 description 2
- 238000009826 distribution Methods 0.000 description 2
- 238000010801 machine learning Methods 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 238000005457 optimization Methods 0.000 description 2
- 238000009966 trimming Methods 0.000 description 2
- 238000012935 Averaging Methods 0.000 description 1
- LFQSCWFLJHTTHZ-UHFFFAOYSA-N Ethanol Chemical compound CCO LFQSCWFLJHTTHZ-UHFFFAOYSA-N 0.000 description 1
- ATJFFYVFTNAWJD-UHFFFAOYSA-N Tin Chemical compound [Sn] ATJFFYVFTNAWJD-UHFFFAOYSA-N 0.000 description 1
- 230000002159 abnormal effect Effects 0.000 description 1
- 238000009825 accumulation Methods 0.000 description 1
- 238000005267 amalgamation Methods 0.000 description 1
- 238000004458 analytical method Methods 0.000 description 1
- 238000005452 bending Methods 0.000 description 1
- 230000005540 biological transmission Effects 0.000 description 1
- 238000013145 classification model Methods 0.000 description 1
- 238000004140 cleaning Methods 0.000 description 1
- 239000011248 coating agent Substances 0.000 description 1
- 238000000576 coating method Methods 0.000 description 1
- 238000004891 communication Methods 0.000 description 1
- 230000001186 cumulative effect Effects 0.000 description 1
- 238000013524 data verification Methods 0.000 description 1
- 238000011156 evaluation Methods 0.000 description 1
- 238000004880 explosion Methods 0.000 description 1
- 238000011049 filling Methods 0.000 description 1
- 230000005021 gait Effects 0.000 description 1
- 230000005484 gravity Effects 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 230000002401 inhibitory effect Effects 0.000 description 1
- 229910052742 iron Inorganic materials 0.000 description 1
- 230000009191 jumping Effects 0.000 description 1
- 230000000670 limiting effect Effects 0.000 description 1
- 238000013178 mathematical model Methods 0.000 description 1
- 239000002184 metal Substances 0.000 description 1
- 229910052751 metal Inorganic materials 0.000 description 1
- 238000012544 monitoring process Methods 0.000 description 1
- 239000003921 oil Substances 0.000 description 1
- 230000000737 periodic effect Effects 0.000 description 1
- 238000003672 processing method Methods 0.000 description 1
- 238000010845 search algorithm Methods 0.000 description 1
- 230000008054 signal transmission Effects 0.000 description 1
Images
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
- G01C21/166—Mechanical, construction or arrangement details of inertial navigation systems
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 invention discloses a system for identifying a motion state by using an inertia measuring device. Wherein, this system includes: the inertial measurement unit is configured to acquire axial inertial information and acquire irregular motion data of a human body; the classification recognition device is configured to classify and recognize irregular motion data through a motion state recognition model trained in advance so as to determine the motion state of the human body; wherein, inertia measuring device includes: the single-axis sensors are arranged on a circuit board of the sensor circuit and used for respectively acquiring axial inertia information of the inertia measuring device on an X axis, a Y axis and a Z axis; the sensor circuits are arranged in an orthogonal mode, are positioned above the signal processing circuit, are respectively connected with the single-axis sensors and are used for receiving axial inertia information; the signal processing circuit is respectively connected with the plurality of sensor circuits through the flexible circuits and is used for preprocessing the axial inertia information to obtain preprocessed irregular motion data.
Description
Technical Field
The invention relates to the field of inertial navigation, in particular to a system for identifying a motion state by using an inertial measurement unit.
Background
The human body may have some irregular behaviors in the motion state in some special spaces, for example, creeping and squatting actions in narrow spaces. The identification of the irregular human motion state can be used in the fields of human safety monitoring, accident and disaster rescue and the like. Taking accident and disaster rescue as an example, people need to be rescued quickly and efficiently by identifying the motion state of a human body, and the rescue personnel are prevented from being injured by secondary disasters. However, the current irregular human motion state identification method is insufficient in real-time performance and accuracy, and cannot be applied to an actual rescue scene.
In view of the above problems, no effective solution has been proposed.
Disclosure of Invention
The embodiment of the invention provides a system for identifying a motion state by using an inertia measuring device, which at least solves the technical problem of inaccurate identification of the motion state of an irregular human body.
According to an aspect of an embodiment of the present invention, there is provided a system for identifying a motion state using an inertial measurement unit, including: the inertial measurement unit is configured to acquire axial inertial information and acquire irregular motion data of a human body; the classification recognition device is configured to classify and recognize the irregular motion data through a pre-trained motion state recognition model so as to determine the motion state of the human body; wherein the inertial measurement unit comprises: the single-axis sensors are arranged on a circuit board of the sensor circuit and are used for respectively acquiring the axial inertia information of the inertia measuring device on an X axis, a Y axis and a Z axis; a plurality of sensor circuits arranged orthogonally, located above the signal processing circuit, respectively connected to the plurality of single-axis sensors, and configured to receive the axial inertial information; the signal processing circuit is respectively connected with the plurality of sensor circuits through flexible circuits and is used for preprocessing the axial inertia information to obtain the preprocessed irregular motion data.
In the embodiment of the invention, the inertia measuring device with higher accuracy is adopted to collect and preprocess the irregular motion data of the human body, and the classification and identification device is used for identifying the irregular motion data, so that the technical effect of accurately identifying the motion state of the human body is realized, and the technical problem of inaccurate identification of the irregular motion state of the human body is solved.
Drawings
The accompanying drawings, which are included to provide a further understanding of the invention and are incorporated in and constitute a part of this application, illustrate embodiment(s) of the invention and together with the description serve to explain the invention without limiting the invention. In the drawings:
fig. 1 is a schematic configuration diagram of a system for recognizing a motion state using an inertial measurement unit according to an embodiment of the present invention;
FIG. 2 is a schematic diagram of another system for identifying a motion state using an inertial measurement unit, according to an embodiment of the invention;
FIG. 3 is a schematic diagram of an inertial measurement unit assembled in accordance with an embodiment of the invention;
FIG. 4 is a schematic diagram of a measurement circuit according to an embodiment of the present invention;
FIG. 5 is a schematic view of a connector structure according to an embodiment of the present invention;
FIG. 6 is a schematic diagram of a connector soldering circuit configuration according to an embodiment of the present invention;
FIG. 7 is a schematic diagram of a base structure according to an embodiment of the present invention;
FIG. 8 is a schematic view of a housing construction according to an embodiment of the present invention;
FIG. 9 is a side schematic view of a measurement circuit assembly according to an embodiment of the invention;
FIG. 10 is a forward schematic view of a measurement circuit assembly according to an embodiment of the invention;
fig. 11 is a flowchart of a method of identifying a motion state according to an embodiment of the present invention.
Detailed Description
In order to make the technical solutions of the present invention better understood, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments 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 of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
It should be noted that the terms "first," "second," and the like in the description and claims of the present invention and in the drawings described above are used for distinguishing between similar elements and not necessarily for describing a particular sequential or chronological order. It is to be understood that the data so used is interchangeable under appropriate circumstances such that the embodiments of the invention described herein are capable of operation in sequences other than those illustrated or described herein. Furthermore, the terms "comprises," "comprising," and "having," and any variations thereof, are intended to cover a non-exclusive inclusion, such that a process, method, system, article, or apparatus that comprises a list of steps or elements is not necessarily limited to those steps or elements expressly listed, but may include other steps or elements not expressly listed or inherent to such process, method, article, or apparatus.
Example 1
According to an embodiment of the present invention, there is provided a system for recognizing a motion state using an inertial measurement unit, as shown in fig. 1, the system including:
the inertial measurement unit 100 is configured to acquire axial inertial information and acquire irregular motion data of a human body; a classification recognition device configured to classify and recognize the irregular motion data through a pre-trained motion state recognition model to determine a motion state of the human body; wherein the inertial measurement unit 100 comprises: the single-axis sensors are arranged on a circuit board of the sensor circuit and are used for respectively acquiring the axial inertia information of the inertia measuring device on an X axis, a Y axis and a Z axis; the sensor circuits are arranged in an orthogonal mode, are positioned above the signal processing circuit, are respectively connected with the single-axis sensors and are used for receiving the axial inertia information; the signal processing circuit is respectively connected with the plurality of sensor circuits through flexible circuits and is used for preprocessing the axial inertia information to obtain the preprocessed irregular motion data.
In one exemplary embodiment, an integrated structure formed by the plurality of single-axis sensors, the plurality of sensor circuits, and the signal processing circuit is fixed to a housing of the inertial measurement unit by silicone rubber.
In an exemplary embodiment, the signal processing circuit includes a circuit board, wherein the plurality of single-axis sensors and peripheral electronic components thereof are distributed on a front surface of the circuit board, and no electronic component is disposed on a back surface of the circuit board, wherein the front surface faces an inside of the inertial measurement unit, and the bottom plate is used for fixing the signal processing circuit.
In an exemplary embodiment, the inertia measurement device further includes a connector and a connector welding circuit, wherein the connector welding circuit is connected with the signal processing unit through a welding lead, and a contact pin of the connector is directly connected and fixed with the connector welding circuit through welding.
In one exemplary embodiment, the signal processing circuit is further configured to: preprocessing the irregular motion data by: AD conversion, signal amplification and filtering, data normalization processing, feature extraction and dimension reduction processing.
In an exemplary embodiment, the classification and identification device is configured to find an optimal kernel function and parameters thereof in a cross validation manner according to different application scenarios, feature dimensions and data characteristics, perform supervised learning on training samples by using the kernel function, and calculate a classifier to obtain the motion state identification model.
In one exemplary embodiment, the classification identifying means is configured to: according to different application scenes, characteristic dimensions and data characteristics, an optimal kernel function and parameters thereof are searched by using an iterative grid search method, supervised learning is carried out on training samples by using the kernel function, and a classifier is calculated to obtain the motion state recognition model.
In an exemplary embodiment, the classification identifying means is further configured to: setting the search range of each parameter of the iterative grid search method, and dividing the search range into fixed grids; traversing each grid for cross validation, obtaining the recognition rate of each grid, and calculating the difference between the maximum recognition rate and the minimum recognition rate according to the recognition rate of each grid; and when the difference is smaller than a preset threshold, outputting a parameter combination corresponding to the current maximum recognition rate as a parameter of the kernel function, when the difference is larger than the preset threshold, resetting the search range, and executing the steps until the difference is smaller than the preset threshold.
In one exemplary embodiment, the signal processing circuit is further configured to: searching a maximum value H and a minimum value L of a measuring point in the action triaxial acceleration data; searching a maximum value max and a minimum value min in the three-dimensional angular speed; and performing normalization processing based on the maximum value H and the minimum value L of the measuring points and the maximum value max and the minimum value min of the three-dimensional angular velocity.
In one exemplary embodiment, the signal processing circuit is further configured to: and carrying out singular value decomposition on the training matrix to obtain the characteristic value and the singular value of the training matrix, obtaining the contribution rate of each component in the training matrix according to the principle component idea that the larger the singular value is, the more the information contained in the singular value is, further obtaining the accumulated contribution rate, and selecting the characteristic according to the contribution rate to carry out dimension reduction processing.
In this embodiment, the classification recognition device first preprocesses the collected irregular motion data by a machine learning method, converts the preprocessed irregular motion data into an input vector of the motion state recognition model, and determines the irregular human motion state according to the pre-trained motion state recognition model, so that the real-time performance and the accuracy of the irregular human motion state recognition are improved, and the technical problem of inaccurate irregular human motion state recognition is solved.
In addition, the embodiment also improves the inertia measuring device.
In the prior art, an inertia measuring device realizes inertia information measurement through acceleration and angular velocity information of sensitive carriers of an accelerometer and a gyroscope, in order to realize accurate measurement of triaxial inertia information of the carriers, a sensor needs to be axially arranged on each carrier for measurement, and the sensors in different axial directions are connected in a wire welding mode, so that the problems of complex assembly process, occupation of internal space and the like are caused.
Moreover, in the prior art, the sensor circuit is usually fixed to the measuring unit appearance structure by means of screw fixation, and the stress conditions of different screws cannot be completely consistent in the process of fastening the screws, so that assembly errors are easily generated, and the measuring circuit structure is even affected. Meanwhile, the measurement circuit is directly and rigidly connected with the shell of the measurement unit, the measurement circuit is easily damaged due to the influence of external impact, and the problems can influence the overall stability of the inertia measurement device and further influence the accuracy of data acquisition of the inertia measurement device.
This embodiment has improved inertia measuring device, and every axial inertia information of inertia measuring device all adopts independent single-axis sensor to measure, and each part sensor circuit orthogonal arrangement measures triaxial inertia information, adopts flexible circuit to be connected between sensor circuit and the signal processing circuit, when reducing the space and occupying, realizes the stable transmission of measuring circuit signal. In addition, the measuring circuit and the measuring unit shell are fixed through GD414 silicon rubber, and the shock resistance of the measuring circuit and the stability of the whole structure of the measuring unit are improved.
Through the improvement, different modules and different sensor circuits are connected through the pad contacts by adopting the flexible circuit, so that the occupation of system space can be reduced, the carrying by a human body is convenient, and the signal transmission stability is improved. In addition, the measuring circuit and the measuring unit shell are fixed through silicon rubber, so that the shock resistance of the measuring circuit and the stability of the whole structure of the measuring unit are improved, and the inertia measuring device can accurately acquire the motion data of a human body in special spaces, such as accident and disaster rescue sites.
Example 2
According to an embodiment of the present invention, there is also provided a system for identifying a motion state using an inertial measurement unit, as shown in fig. 2, the system including an inertial measurement unit 100 and a classification identifying unit 200. The inertial measurement unit in this embodiment has the same structure and function as those in embodiment 1, and will not be described again here.
The present embodiment is different from embodiment 1 in a classification recognition apparatus 200, which includes a model training module 202, and how the model training module performs model training will be described in detail below.
A Support Vector Machine (SVM) is a model for classifying data into two classes in principle, and is defined as a linear classifier that maximizes a classification interval in a feature space. The learning strategy is that the classification interval is maximized, and the classification hyperplane searching problem is converted into a convex quadratic programming problem to be solved. The SVM is a supervised machine learning algorithm, based on the minimum principle of structured risk, the confidence range and the experience risk are reduced to the minimum, so that the generalization capability of the SVM is improved, a better classification effect can be realized even under the condition of few training samples, the core of the SVM is how to find an optimal hyperplane capable of dividing the samples into two types, and the main idea is to popularize the linear separable problem in the plane to the high-dimensional classification problem
However, in daily life, the problems needing classification are various, for example, the problem of classification of motion states of human bodies, the required recognition actions include various common actions such as walking, running, standing, going upstairs, going downstairs, sitting and the like, for the multi-classification problem, the construction of the multi-classification SVM is mainly realized by combining a plurality of two-classification SVM, and the common methods include 1-v-1SVMs and 1-v-r SVMs. The 1-v-r SVMs algorithm is the earliest method for solving the problem of multi-classification by the SVM, and for a total classification problem of k categories, k sub-classifiers are constructed, the ith classifier takes the ith category as one category, and all other categories as the other categories to construct the classifier, for data to be classified, k times of classification operation are required to be carried out, the classification function value of each classifier is calculated, and the category corresponding to the largest function value is selected as the category to which the classifier belongs. When the categories are more, the training data sets corresponding to the single category and the rest categories have larger asymmetry, the situation of complex data distribution exists, the trained classifiers have more support vectors and larger classification calculation amount, and in addition, when the classification function values are repeated, the problem of inseparability occurs. Combining the above points, such multi-classification algorithm is rarely used in practical applications. The 1-v-1SVMs method is used in the embodiment of the application.
In one example, the principle of the 1-v-1SVMs classification method is as follows:
for a total of k classes of classification problems, a total of k classes of classification problems are constructed
For each sub-classifier, the training data of each sub-classifier only needs the data sets corresponding to the two classes, so the data complexity is less than that of the 1-v-r algorithm. And during data verification, traversing all the sub-classifiers, taking the classification result of each sub-classifier as a ticket of the class, counting the final ticket number of all the classes, and selecting the class with the most ticket number as the final classification result. For a classifier that distinguishes the ith and jth classes, the ith class sample is defined as a positive class sample and the jth class sample is defined as a negative class sample.
The SVM classification process is as follows:
(1) And extracting the characteristics of the sample data, and selecting an optimal subset.
(2) Each action characteristic set is divided into a training set and a testing set, and the action sample sets are combined pairwise (here, six action classifications are assumed), so that 15 combination modes are obtained, namely k1-k15.
(3) Starting from k1, using the training set and the test set as data sources, performing optimal parameter search by using a grid search algorithm, creating a sub-classifier by using optimal parameters, and recording the optimal recognition rate of the sub-classifier.
(4) The operation in step (3) is carried out for k2, k3, 823060, 8230and k15 respectively.
After high-dimensional conversion is carried out on the feature vectors through the kernel function, supervised learning is carried out on the training samples, and the functional formula of the classifier is calculated as follows:
wherein: alpha is alphaiIs a lagrange multiplier; k (x)iX) is a calculation sample point xiAnd x, wherein n represents n labeled training data participating in modeling, and i is the ith training data; α i is a Lagrangian multiplier; k (xi, x) represents the inner product between the computed training data xi and x; b is a constant; yi is the label corresponding to the training data xi.
Another kernel function will be described in detail below.
The method comprises the following steps of classifying nonlinear separable problems by using an SVM (support vector machine), firstly mapping linear inseparable data into a new linear separable data set through a certain nonlinear mapping relation, then classifying the linear problems, and assuming that the mapping relation is phi, the process is divided into two steps:
(1) And establishing a mapping relation phi, and mapping the data to a new data space.
(2) The data of the new data space is linearly classified.
(3) The classification function can then be expressed in the form:
after high-dimensional conversion is carried out on the feature vectors through the kernel function, supervised learning is carried out on the training samples, and the functional formula of the classifier is calculated as follows:
wherein alpha isiIs a lagrange multiplier; (phi (x)i) Phi (x)) is the calculated sample point xiInner product between x. Except that x is mapped first. Now, a method is found for simultaneously carrying out mapping and inner product, and simplifying the classification of linear inseparable problem into the same steps as linear separable problem, thereby establishing a nonlinear learning machine, and the method is used for establishing the nonlinear learning machineThe direct calculation method is the kernel function method mentioned in the embodiments of the present application. As used herein, a "kernel" is a function that satisfies the following relationship for all x and z:
k(x，z)＝(φ(x)，φ(z))
that is, the result obtained by the function k (x, z) and the inner product operation (Φ (x), Φ (z)) performed after mapping have the same effect, and generally, the mapping to the high-dimensional data is accompanied by dimension explosion growth, and the inner product operation performed on the high-dimensional data requires a large amount of calculation. To this end, the classification function for the linear indivisible problem can be expressed in the form:
the application of the kernel function solves the problem of linear inseparability, avoids the problem of a large amount of inner product operations caused by dimension increase in dimension mapping, has no same standard for kernel function selection, and generally finds the optimal kernel function and parameters thereof in a cross validation mode according to different application scenes, characteristic dimensions and data characteristics. Several commonly used kernel functions are introduced and the application scenarios thereof are analyzed, and finally, the performance of each kernel function on the action recognition is compared in an experimental manner.
Selecting a kernel function:
the method of the embodiment of the present application uses Radial Basis Function Kernel (RBF),
K(xi，xj)＝exp(-γxi-xj 2) When gamma > 0
In the formula: x is the number ofiAnd xjIs a feature vector; γ is a parameter of the kernel function.
The SVM obtained by the kernel function is a classifier of a corresponding radial basis function. The most critical point is that such kernel functions are functions that do not exhibit large deviations among all kernel functions.
The process of finding the kernel function by the grid search method will be described in detail below.
The selection of the kernel function needs to be determined according to the classification effect after being tried, different kernel functions have different parameters, and the parameters have important influence on the classification effect, so that the embodiment of the application needs to perform parameter optimization of each kernel function, perform comparison of the optimal effect, and select the most suitable kernel function. The application adopts a grid search method to select the kernel function.
The function parameter and the penalty factor c are searched by using an iterative grid search method. The specific process is as follows:
(1) And setting the sum of the search ranges of all the parameters, and dividing the search ranges into fixed lattices.
(2) And traversing each grid for cross validation, and recording the corresponding recognition rate.
(3) And calculating the difference value between the maximum recognition rate and the minimum recognition rate after traversing once, terminating the process if the difference value is less than 0.01, and outputting the parameter combination corresponding to the current maximum recognition rate. And (4) if the difference is larger than 0.01, jumping to the step (1), and iteratively subdividing the lattices corresponding to the current maximum recognition rate until the difference is smaller than 0.01.
The parameters of the kernel function will be described in detail below.
In the process, the penalty factor C and parameters in the kernel function are determined, and the accuracy of identification is improved by adjusting the parameters.
C is a penalty factor, i.e. tolerance to errors. The higher C indicates that the error is less tolerable and is easily overfitted. The smaller C, the easier it is to under-fit. If C is too large or too small, the generalization capability is poor, so that finding proper C plays an important role in optimizing the network.
gamma is a parameter of the RBF function after the function is selected as the kernel. The distribution of the data after being mapped to a new feature space is determined implicitly, the larger the gamma is, the fewer the support vectors are, and the smaller the gamma value is, the more the support vectors are. The number of support vectors affects the speed of training and prediction.
Example 3
According to an embodiment of the present invention, there is also provided a system for identifying a motion state using an inertial measurement unit, the system including the inertial measurement unit and a classification identification unit. The classification and identification device in this embodiment has the same structure and function as those in embodiments 1 and 2, and is not described here again.
The present embodiment is different from embodiment 1 in an inertia measurement device, as shown in fig. 3 to 10, which is composed of a Z-axis gyroscope and accelerometer combination circuit 1, an X-axis gyroscope circuit 2, a Y-axis gyroscope circuit 3, an X-axis accelerometer circuit 4, a Y-axis accelerometer circuit 5, a signal processing circuit 6, a connector 7, a connector soldering circuit 8, a bottom plate 9, and a housing 10, wherein the bottom plate is used for fixing the signal processing circuit and is combined with the housing to provide protection for the circuit.
The signal processing circuit is connected with other sensor circuits and the connector welding circuit through the flexible circuit to transmit signals. On Z axle gyroscope and accelerometer amalgamation circuit, X axle gyroscope circuit, Y axle gyroscope circuit, X axle accelerometer circuit and Y axle accelerometer circuit, sensing device and peripheral electron device all distribute at the circuit board openly, do not have electronic components in the back, avoid causing the circuit short circuit with measuring device metal casing contact, make things convenient for circuit board and measuring unit shell to pass through sticky fixation simultaneously.
The signal processing circuit consists of a Cortex-M7 kernel STM32H753 chip, a signal amplifier, a 16-bit high-precision AD converter and a serial port, and can realize complex functions of sensor compensation, data filtering, combined navigation parameter calculation and the like.
In the assembling process, whether the surfaces of the shell and the bottom plate are obviously scratched, bruised or burred or not is firstly observed visually, whether the thin surface of the base is obviously bent or not is judged, the shell and the bottom plate are cleaned by adding absolute ethyl alcohol into an ultrasonic cleaner, the cleaning time is not less than half an hour, and the surfaces are ensured to have no scrap iron and no oil stain.
And the signal processing circuit is fixed with the base plate screw hole (11) through two M1.2x4 screws. GD414 silicon rubber is evenly coated on the bottom inside the shell, the Z-axis gyroscope and accelerometer combined circuit is assembled on the bottom inside the shell, and the combined circuit is pressed to enable the back of the combined circuit to be tightly attached to the bottom inside the shell. And then uniformly coating GD414 silicon rubber on the side surface in the shell, assembling the X-axis gyroscope circuit, the Y-axis gyroscope circuit, the X-axis accelerometer circuit and the Y-axis accelerometer circuit on the inner wall of the shell, and tightly attaching the back surface of the shell to the side surface in the shell. The GD414 silicon rubber is adopted to fix the sensor circuit, so that the shock resistance of the sensor and the overall stability of the measuring unit can be effectively improved.
Then, original screws on two sides of the J30J-15ZKN-J connector are taken down, the contact pins are inserted into the holes of the welding circuit of the connector in an aligning mode, all the holes are tightly welded with the contact pins in a soldering tin welding mode, the connector and the welding circuit of the connector are inserted into the holes of the shell from inside to outside, and special screws, original elastic pads, gaskets and nuts are taken out to be fixed with the shell. And connecting the connector welding circuit with the corresponding pin of the signal processing circuit through a welding lead. And finally, bending the flexible circuit, closing the bottom plate and the signal processing circuit, and tightly connecting the bottom plate and the shell by eight M1.2x4 screws.
The process of preprocessing the acquired motion data by the signal processing circuit will be described in detail below.
And step 1, performing filtering processing.
Typically, the sequence of raw signal data (i.e., the acquired irregular motion data of the human body) is long and contains a lot of noise. Therefore, in order to extract feature values and improve the accuracy of the algorithm, the original signal data needs to be processed, and the method of moving median filtering is adopted in the present application. The moving average filtering algorithm is a typical linear filtering algorithm and has a good effect of inhibiting periodic noise interference. The main principle of the method is to realize the filtering of original data by a moving window mode and a neighborhood averaging method. Assuming that the input is X and the output is y, the calculation method of the moving average filter is shown as the following formula, where M is the window size of the moving average filter.
Where n is the window length and y (n) is the output of the filter calculation.
Because the system uses the acceleration data and the angular velocity data of the inertial sensor, the two data have different physical meanings and larger numerical value range difference, the final characteristic value range is also larger when the two data are respectively subjected to characteristic extraction, certain interference can be caused to an action classification algorithm based on the characteristic value in the follow-up process, and the characteristic with large numerical value has larger influence on a classification result. Therefore, normalization processing is required, and the specific algorithm steps are as follows:
(1) And searching the maximum value H and the minimum value L of the measuring point in the action triaxial acceleration data.
(2) And searching for the maximum value max and the minimum value min in the three-dimensional angular speed.
(3) And calculating the following formula for the triaxial angular velocity data respectively:
in the formula, L is a minimum value after normalization, H is a maximum value after normalization, ui represents an original data point, and vi represents a data point after normalization.
And step 3, feature extraction.
Theoretically, the more the feature values are, the more the features of the motion data can be expressed, but the excessive feature values also greatly increase the calculation amount, so that a proper feature value needs to be selected for the algorithm, and the following method is a common data feature extraction method.
Mean value:
where N is the number of action data points.
Standard deviation:
where N is the number of action data points, xi represents the ith input,
represents the average of the input values.
Root mean square:
where N is the number of action data points.
Principal Component Analysis (PCA) is a common dimension reduction method, and is an unsupervised dimension reduction processing method by applying a training matrix Dm×nSingular value decomposition is carried out to obtain the characteristic value lambdaiAnd singular value uiAccording to the principle component idea, the larger the singular value is, the more information is contained, and D is obtainedm×nThe contribution rate of each component in the Chinese character image is calculated, and the accumulated contribution rate is further calculated, and the characteristics are selected according to the contribution rate.
Where U and V are orthogonal matrices, Λ is a non-negative diagonal of m × n, let Λ = Diag [ λ 1, λ 2]The relationship between the eigenvalue λ i and the singular value ui is as follows: lambda [ alpha ]i＝(ui)2
λiThe feature root of the training matrix D is, and according to the principal component concept, the larger the singular value is, the more information it contains, so that the feature space composed of the first 1 principal components corresponds to the new feature space D':
D′m×l＝U(：，1：l)×Λl×l
u (: 1: l) is a matrix corresponding to the first 1 column vector, Λl×lThe diagonal matrix corresponding to the first 1 larger singular values. Thereby obtaining the contribution rate of each main component in D as Ci：
Where λ i and λ j denote the ith and j eigenvalues.
Accordingly, the contribution rate C is accumulated
Where m is the number of columns of the matrix and k is the number of rows.
Typically, the cumulative contribution of features is required to be over 90% to ensure that the selected features contain the vast majority of the information about motion. However, if the dimension is too small, the feature value is insufficient and the identification cannot be accurately performed, and if the dimension is too large, the problems of too large calculated amount, long identification time and the like can occur. Therefore, other suitable methods are needed to select features so that the dimension can be reasonably reduced and the recognition time can be shortened and the classification accuracy can be improved.
In this embodiment, the implementation steps of the system for identifying the motion state by the inertial measurement unit are as follows:
1. the inertia measuring device worn on the chest is used for collecting irregular motion states in environments such as narrow and shielded space, the collected data are classified, time stamps are marked, and the time stamps are transmitted to the classification and identification device through the wireless communication module.
2. And carrying out data processing and analysis on the acquired target walking step frequency and the sensor information, and marking corresponding labels on different behavior activities to form a data label pair set.
3. And constructing an irregular action recognition network model, and adjusting parameters according to recognition effects and the types of the actions.
4. And dividing the obtained data label pair set into a training set and a test set, wherein the training set is sent into the built network model for training, and the test set is used for evaluating the classification effect of the model after the training is finished.
5. And importing real-time data of the personnel of the wearable equipment into the trained model to perform online real-time classification and evaluation.
Example 4
The embodiment of the invention provides a system for identifying a motion state by using an inertial measurement unit and a method for identifying an irregular human motion state by using a classification identification unit, as shown in fig. 11, the method comprises the following steps:
step S402, according to the irregular motion data, data preprocessing is carried out, and preprocessed irregular motion data are obtained;
and S404, inputting the preprocessed irregular motion data into a motion state recognition model to obtain an irregular human motion state recognition type output by the motion state recognition model.
The motion state recognition model is a multi-classification Support Vector Machine (SVM) model for training based on the pre-acquired irregular motion data corresponding to the irregular human body motion state.
In one example, the multi-classification SVM model comprises a pair of one-top-one feature Vector Machines (1-v-1 SVMs or pairwise) models; the 1-v-1SVMs comprise a plurality of irregular human motion state two-classification SVM models which are obtained by calculation according to a first formula, wherein the first formula is as follows: n = K (K-1)/2; and N is the number of the irregular human motion state two-classification SVM models, and K is the total number of the irregular human motion states to be identified by the 1-v-1 SVMs.
In one example, the irregular motion data includes: acceleration data, angular velocity data; the preprocessing the data according to the irregular motion data to obtain preprocessed irregular motion data includes: carrying out data trimming according to the irregular motion data to obtain trimmed irregular motion data; wherein the data trimming comprises at least one of: processing abnormal data, processing repeated data and filling missing values; and carrying out data normalization according to the trimmed irregular motion data to obtain the preprocessed irregular motion data.
In one example, the acceleration data is acquired by an acceleration acquisition device, and the angular velocity is acquired by an angular velocity acquisition device; the acceleration acquisition device comprises an accelerometer; the angular velocity acquisition device comprises a gyroscope; the acceleration data is triaxial acceleration data determined according to a three-dimensional space position, and comprises: first acceleration data, second acceleration data, and third acceleration data; the angular velocity data is triaxial angular velocity data determined according to a three-dimensional spatial position, and includes: first angular velocity data, second angular velocity data, third angular velocity data.
In one example, the method further comprises: acquiring training data, wherein the training data comprises the irregular motion data which is marked and preprocessed by the data; and, dividing the training data into a training set and a test set; constructing a classifier model and defining a plurality of binary classification models; inputting the training set into the classifier, and establishing an initial classifier model by calling an instantiation method; inputting the test set into a current classifier model, and obtaining the accuracy of the current classifier model according to the output result of the current classifier model and the irregular human motion state type marks in the test set; and performing parameter optimization on model parameters of the classifier model according to the accuracy of the current classifier model by a grid search method until the multi-classification SVM model with the accuracy reaching a preset requirement is obtained by training.
In one example, the classifier model is described by a second formula, wherein the second formula is:
wherein n represents n labeled training data participating in modeling, and i is the ith training data; α i is a Lagrangian multiplier; k (xi, x) represents the inner product between the computed training data xi and x; b is a constant; yi is the label corresponding to the training data xi.
In one example, the grid search method includes: setting a plurality of parameters to be adjusted in the classifier model and a search range of penalty factors; determining a plurality of fixed numerical values in the search range to form a data grid; according to the data grid, traversing and setting the classifier model parameters and the penalty factors as parameters corresponding to the data grid, and recording the accuracy of the classifier model under different parameters; if the difference value between the maximum value and the minimum value in the accuracy of the classifier model is not greater than a preset value under the different parameters, determining the model parameter and the penalty factor corresponding to the obtained maximum accuracy as the value of the classifier model; if the difference value between the maximum value and the minimum value in the accuracy of the classifier model is larger than the preset value under different parameters, the grid search method is repeatedly executed according to the model parameter corresponding to the maximum accuracy and the penalty factor until the difference value is not larger than the preset value.
In one example, the six-axis inertia measurement device is fixedly connected to the foot of a human body, a zero-speed correction method can be adopted to estimate the system state vector error by combining with an extended Kalman filter, and then the strapdown resolved state vector is corrected to obtain the position, the speed and the posture of the personnel after filtering estimation. The zero-speed correction method can effectively inhibit error accumulation of the inertia measurement device caused by acceleration integral, but the inertia measurement device is required to be only arranged on the foot, and the foot is required to be static relative to the ground within a certain time so as to ensure timely correction of the state vector. The zero speed correction method algorithm can be roughly divided into two parts: zero-speed detection and zero-speed correction.
When a pedestrian walks, the motion of the foot can be divided into 2 modes of rest and motion. The zero speed detection mainly completes the task of identifying the static stage of the feet of a person and determining the static stage as a zero speed interval in a gait cycle. A section of observation value sequence is given, zero speed detection is carried out through the observation value sequence, and the accuracy of zero speed interval identification is one of key factors influencing the navigation effect of the ZUPT algorithm. The requirement of zero-velocity detection is to reduce the probability of misjudging the motion state as the zero-velocity state and improve the probability of accurately judging the zero velocity, because if the motion state is misjudged as the zero-velocity state, a new error is introduced into the navigation system, and if the static state is misjudged as the motion state, the correction times of the navigation system are reduced. The zero-speed detection formula by using the generalized likelihood ratio detection method is as follows:
in the formula: w is the size of a zero-speed detection window; znIs the sensor output within the zero velocity window,
Setting a detection threshold gamma for zero-speed detectionTZero-speed detection flag bit C for determining datakThe formula is as follows:
where T represents the test statistic for zero-speed detection, UnRepresenting the observed magnitude of zero-speed detection. The main task of the zero-speed correction is to utilize the zero speed of a zero-speed interval as an observed quantity, carry out filtering estimation on a state variable error through an extended Kalman filter, and then carry out strapdown inertial solution on the resultAnd (6) correcting. The system mathematical model is established based on a navigation error model, the inertial navigation updating is completed in time updating, the measurement updating is executed in a zero-speed interval, and the state variable error is fed back to a navigation updating result immediately after the measurement updating for error correction.
The foregoing is only a preferred embodiment of the present invention, and it should be noted that, for those skilled in the art, various modifications and decorations can be made without departing from the principle of the present invention, and these modifications and decorations should also be regarded as the protection scope of the present invention.
Claims (10)
1. A system for identifying a state of motion using an inertial measurement unit, comprising:
the inertial measurement unit is configured to acquire axial inertial information and acquire irregular motion data of a human body;
the classification recognition device is configured to classify and recognize the irregular motion data through a pre-trained motion state recognition model so as to determine the motion state of the human body;
wherein the inertial measurement unit comprises:
the single-axis sensors are arranged on a circuit board of the sensor circuit and are used for respectively acquiring the axial inertia information of the inertia measuring device on an X axis, a Y axis and a Z axis;
the sensor circuits are arranged in an orthogonal mode, are positioned above the signal processing circuit, are respectively connected with the single-axis sensors and are used for receiving the axial inertia information;
the signal processing circuit is respectively connected with the plurality of sensor circuits through flexible circuits and is used for preprocessing the axial inertia information to obtain the preprocessed irregular motion data.
2. The system of claim 1, wherein the unitary structure formed by the plurality of single-axis sensors, the plurality of sensor circuits, and the signal processing circuit is secured to the housing of the inertial measurement unit by silicone rubber.
3. The system of claim 1, wherein the plurality of single-axis sensors and their peripheral electronics are distributed on a front side of the circuit board, and wherein no electronics are disposed on a back side of the circuit board, wherein the front side is facing an interior of the inertial measurement unit.
4. The system of claim 1, wherein the inertial measurement unit further comprises a connector and a connector welding circuit, wherein the connector welding circuit is connected with the signal processing unit through a welding wire, and the contact pin of the connector is directly connected and fixed with the connector welding circuit through welding.
5. The system of any of claims 1-4, wherein the signal processing circuitry is further configured to: preprocessing the irregular motion data by: AD conversion, signal amplification and filtering, data normalization processing, feature extraction and dimension reduction processing.
6. The system according to any one of claims 1 to 4, wherein the classification and identification device is configured to find an optimal kernel function and parameters thereof by means of cross validation according to different application scenarios, feature dimensions and data characteristics, and perform supervised learning on training samples by using the kernel function, and calculate a classifier to obtain the motion state identification model.
7. The system according to any one of claims 1 to 4, wherein the classification recognition means is configured to: according to different application scenes, characteristic dimensions and data characteristics, an optimal kernel function and parameters thereof are searched by using an iterative grid search method, supervised learning is carried out on training samples by using the kernel function, and a classifier is calculated to obtain the motion state recognition model.
8. The system of claim 7, wherein the classification recognition device is further configured to:
setting the search range of each parameter of the iterative grid search method, and dividing the search range into fixed grids;
traversing each grid for cross validation, obtaining the identification rate of each grid, and calculating the difference between the maximum identification rate and the minimum identification rate according to the identification rate of each grid;
and when the difference is smaller than a preset threshold, outputting a parameter combination corresponding to the current maximum recognition rate as a parameter of the kernel function, when the difference is larger than the preset threshold, resetting the search range, and executing the steps until the difference is smaller than the preset threshold.
9. The system of claim 5, wherein the signal processing circuit is further configured to:
searching a maximum value H and a minimum value L of a measuring point in the action triaxial acceleration data;
searching a maximum value max and a minimum value min in the three-dimensional angular speed;
and carrying out normalization processing based on the maximum value H and the minimum value L of the measuring points and the maximum value max and the minimum value min of the three-dimensional angular velocity.
10. The system of claim 5, wherein the signal processing circuit is further configured to: and carrying out singular value decomposition on the training matrix to obtain the characteristic value and the singular value of the training matrix, obtaining the contribution rate of each component in the training matrix according to the principle component idea that the larger the singular value is, the more the information contained in the singular value is, further obtaining the accumulated contribution rate, and selecting the characteristic according to the contribution rate to carry out dimension reduction processing.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210812326.9A CN115265531A (en) | 2022-07-12 | 2022-07-12 | System for identifying motion state by using inertia measuring device |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210812326.9A CN115265531A (en) | 2022-07-12 | 2022-07-12 | System for identifying motion state by using inertia measuring device |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN115265531A true CN115265531A (en) | 2022-11-01 |
Family
ID=83764398
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210812326.9A Pending CN115265531A (en) | 2022-07-12 | 2022-07-12 | System for identifying motion state by using inertia measuring device |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115265531A (en) |
Citations (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103776448A (en) * | 2014-02-17 | 2014-05-07 | 刘超军 | Posture course reference system |
| CN104296746A (en) * | 2014-10-13 | 2015-01-21 | 苏州文智芯微系统技术有限公司 | Novel minitype inertial measurement unit assembly |
| CN104482938A (en) * | 2014-10-24 | 2015-04-01 | 中国兵器工业集团第二一四研究所苏州研发中心 | Low temperature co-fired ceramic (LTCC) technology micro inertial measurement unit (MIMU)-based three-axis structure and manufacturing technology thereof |
| CN107449420A (en) * | 2017-07-28 | 2017-12-08 | 湖北三江航天红峰控制有限公司 | A kind of satellite and the rocket integration inertial attitude sensor |
| CN107966144A (en) * | 2017-11-20 | 2018-04-27 | 北京航天控制仪器研究所 | A kind of assembly structure of the inertial measurement combination based on MEMS sensor |
| CN207456462U (en) * | 2017-10-31 | 2018-06-05 | 北京北斗星通导航技术股份有限公司 | A kind of micro inertial measurement unit and Inertial Measurement Unit detection device |
| CN111811504A (en) * | 2020-08-13 | 2020-10-23 | 中国兵器工业集团第二一四研究所苏州研发中心 | Laminated micro-inertia measuring unit under large-overload high-dynamic application environment |
| CN111854739A (en) * | 2020-02-18 | 2020-10-30 | 重庆市皓颖科技发展有限公司 | Small-size high-precision inertial measurement unit |
| CN112229400A (en) * | 2020-09-15 | 2021-01-15 | 北京自动化控制设备研究所 | Miniaturized micro-electromechanical gyro inertia/satellite combined navigation system |
| CN113536552A (en) * | 2021-06-30 | 2021-10-22 | 杭州电子科技大学 | Human body posture visual tracking system |
| CN215177786U (en) * | 2021-04-22 | 2021-12-14 | 上海傲世控制科技股份有限公司 | Optical fiber inertia measurement unit |
-
2022
- 2022-07-12 CN CN202210812326.9A patent/CN115265531A/en active Pending
Patent Citations (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103776448A (en) * | 2014-02-17 | 2014-05-07 | 刘超军 | Posture course reference system |
| CN104296746A (en) * | 2014-10-13 | 2015-01-21 | 苏州文智芯微系统技术有限公司 | Novel minitype inertial measurement unit assembly |
| CN104482938A (en) * | 2014-10-24 | 2015-04-01 | 中国兵器工业集团第二一四研究所苏州研发中心 | Low temperature co-fired ceramic (LTCC) technology micro inertial measurement unit (MIMU)-based three-axis structure and manufacturing technology thereof |
| CN107449420A (en) * | 2017-07-28 | 2017-12-08 | 湖北三江航天红峰控制有限公司 | A kind of satellite and the rocket integration inertial attitude sensor |
| CN207456462U (en) * | 2017-10-31 | 2018-06-05 | 北京北斗星通导航技术股份有限公司 | A kind of micro inertial measurement unit and Inertial Measurement Unit detection device |
| CN107966144A (en) * | 2017-11-20 | 2018-04-27 | 北京航天控制仪器研究所 | A kind of assembly structure of the inertial measurement combination based on MEMS sensor |
| CN111854739A (en) * | 2020-02-18 | 2020-10-30 | 重庆市皓颖科技发展有限公司 | Small-size high-precision inertial measurement unit |
| CN111811504A (en) * | 2020-08-13 | 2020-10-23 | 中国兵器工业集团第二一四研究所苏州研发中心 | Laminated micro-inertia measuring unit under large-overload high-dynamic application environment |
| CN112229400A (en) * | 2020-09-15 | 2021-01-15 | 北京自动化控制设备研究所 | Miniaturized micro-electromechanical gyro inertia/satellite combined navigation system |
| CN215177786U (en) * | 2021-04-22 | 2021-12-14 | 上海傲世控制科技股份有限公司 | Optical fiber inertia measurement unit |
| CN113536552A (en) * | 2021-06-30 | 2021-10-22 | 杭州电子科技大学 | Human body posture visual tracking system |
Non-Patent Citations (8)
| Title |
|---|
| 刘福朝: "低成本捷联微惯性测量单元设计与误差补偿", 《电子测试》, 25 March 2022 (2022-03-25), pages 28 - 30 * |
| 刘福朝: "微惯性测量单元设计及抗高过载性能分析", 《压电与声光》, vol. 44, no. 2, 20 April 2022 (2022-04-20), pages 289 - 293 * |
| 卞玉民: "MEMS惯性传感器现状与发展趋势", 《计测技术》, 28 August 2019 (2019-08-28), pages 50 - 56 * |
| 司玉仕: "基于惯性传感器的动作识别研究", 《中国优秀硕士学位论文全文数据库 信息科技辑》, 15 February 2019 (2019-02-15), pages 140 - 537 * |
| 张会清;许潇民;: "基于运动姿态识别的行人惯性导航算法", 传感器与微系统, no. 05, 8 June 2018 (2018-06-08), pages 123 - 127 * |
| 徐丽敏: "一种适用于跑步状态的惯性/零速/GPS室内外无缝组合导航定位方法", 《导航与控制》, 5 December 2019 (2019-12-05), pages 21 - 28 * |
| 汪宋兵: "基于UWB/MEMS组合的行人室内定位系统关键技术研究", 《中国优秀硕士学位论文全文数据库 工程科技Ⅱ辑》, 15 June 2020 (2020-06-15), pages 029 - 7 * |
| 熊伟民: "可穿戴式运动识别系统及其算法的研究与实现", 《中国优秀硕士学位论文全文数据库 信息科技辑》, 15 July 2019 (2019-07-15), pages 138 - 1064 * |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN105184325B (en) | Mobile intelligent terminal | |
| JP4992043B2 (en) | Action identification device, action identification system, and action identification method | |
| Panahandeh et al. | Continuous hidden Markov model for pedestrian activity classification and gait analysis | |
| CN109619762B (en) | Physiological sign analysis method and system based on step length information | |
| US20160061582A1 (en) | Scale estimating method using smart device and gravity data | |
| CN108334827B (en) | Gait identity authentication method based on intelligent shoe and intelligent shoe | |
| Jalloul et al. | Activity recognition using complex network analysis | |
| CN110674875A (en) | Pedestrian motion mode identification method based on deep hybrid model | |
| CN112057066A (en) | Heart rate detection method, wearable device and computer storage medium | |
| CN106405384A (en) | Simulation circuit health state evaluation method | |
| CN110533166B (en) | Indoor positioning method based on time-space fusion characteristics | |
| CN111860446A (en) | Detection system and method for unknown mode of satellite remote measurement time sequence data | |
| JP4830765B2 (en) | Activity measurement system | |
| CN111931616A (en) | Emotion recognition method and system based on mobile intelligent terminal sensor equipment | |
| CN107277222A (en) | User behavior state judging method based on mobile phone built-in sensors | |
| Khokhlov et al. | Design of activity recognition systems with wearable sensors | |
| Xie et al. | Meta-activity recognition: A wearable approach for logic cognition-based activity sensing | |
| Khatun et al. | Human activity recognition using smartphone sensor based on selective classifiers | |
| US10555689B1 (en) | CPS pressure based sensing system for symmetry measurements | |
| CN115265531A (en) | System for identifying motion state by using inertia measuring device | |
| CN109685111B (en) | Motion recognition method, computing system, intelligent device and storage medium | |
| CN111839527A (en) | Fall detection method and device and computer readable storage medium | |
| CN115127550A (en) | System for recognizing motion state of human body by using six-axis inertia measuring device | |
| CN116071783A (en) | Sheep reproductive health early warning system and method | |
| CN116188445A (en) | Product surface defect detection and positioning method and device and terminal equipment |
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 |