CN111062412B - Novel intelligent shoe intelligent recognition method for indoor pedestrian movement speed - Google Patents

Abstract

The invention discloses a novel intelligent shoe intelligent recognition method for indoor pedestrian movement speed, which comprises the following steps that firstly, an IMU inertial sensing unit is used for extracting pedestrian foot inertial sensing data; secondly, dividing the time-domain continuous inertial data according to steps by adopting an acceleration peak value dividing method; thirdly, extracting features of the inertia data of each step, inputting a dictionary learning algorithm, and performing model training to obtain a speed recognition model; step four, carrying out single-step division and feature extraction on the newly input data, and identifying speed information according to a speed identification model; fifthly, integrating the characteristics and the speed of a group of newly obtained inertial data with the existing model, and updating model parameters; and sixthly, transmitting information such as the speed, the step number and the like of the pedestrians to the mobile phone app or the notebook computer terminal through the communication module, and realizing the visualization of the indoor pedestrian movement speed. The invention considers the wearing convenience of the whole set of device, integrates the IMU, the MCU and the wireless communication module on the shoes, is convenient and comfortable, and can conveniently perform the work of identifying the movement speed of the pedestrians indoors.

Description

Novel intelligent shoe intelligent recognition method for indoor pedestrian movement speed

Technical Field

The invention relates to the field of indoor positioning and machine learning, in particular to a method for identifying the movement speed of an indoor pedestrian by utilizing intelligent shoes.

Background

Machine Learning (ML) has emerged in recent years as a multi-domain interdisciplinary discipline involving probability theory, statistics, approximation theory, convex analysis, and algorithm complexity theory. Machine learning mainly researches how a computer realizes simulation and learning of human behaviors, acquires new knowledge, and improves self performance by organizing and perfecting an existing knowledge structure. As the core of artificial intelligence (Artificial Intelligence), machine learning is the fundamental approach that computers have to intelligence, which is applied throughout various areas of artificial intelligence, such as data processing and analysis and prediction.

With the improvement of living conditions and the development of science and technology, people stay indoors for a longer time. Whether positioning in a market or rescuing in a fire scene, people can accurately know information such as real-time position and speed of the people. Due to the complexity of indoor environment and various electromagnetic interferences, the satellite positioning signals can have attenuation and interference problems indoors, and the indoor positioning requirements can not be met. The pedestrian dead reckoning (Pedestrian Dead Reckoning, PDR) system can well make up for the deficiency of satellite positioning in indoor environments. However, due to the accumulated offset error existing in the PDR system itself, we need to perform zero-speed correction algorithm (Zero Velocity Update, ZUPT) to eliminate the error, and the zero-speed update algorithm needs real-time speed information as the basis.

Based on these problems existing in indoor positioning, consider that when a pedestrian moves in an indoor environment, its movement speed is acquired in real time, and speed information is processed and stored. The walking or running patterns of different people are different, but the characteristics extracted from the inertial data of the feet of a specific person can well reflect the movement state and speed of the specific person. The recognition method can be universally applicable to different pedestrians, is convenient to wear, is simple and convenient to use, has high operation speed, can acquire the pedestrian movement speed information in real time, and feeds back the pedestrian movement speed information to a zero-speed correction algorithm and a PDR system to position the pedestrians indoors.

Disclosure of Invention

Aiming at the problems and the defects existing in the prior method, the invention provides a novel intelligent shoe intelligent recognition method for the indoor pedestrian movement speed, inertial data are collected through sensors on the intelligent shoe, and movement speed recognition is realized based on the characteristic extraction of inertial sensing data.

The invention discloses an intelligent recognition method of novel intelligent shoes for indoor pedestrian movement speed, which specifically comprises the following steps:

firstly, extracting inertial sensing data of feet of pedestrians by using an IMU inertial sensing unit;

secondly, dividing the time-domain continuous inertial data according to steps by adopting an acceleration peak value dividing method, and detecting

The peak values of (2) are divided stepwise, and the two norms of the x, y and z triaxial acceleration are +.>

Is defined as shown in formula (1):

uploading and storing the divided single-step data,

thirdly, extracting features of the inertia data of each step, inputting a dictionary learning algorithm, and performing model training to obtain a speed recognition model;

firstly, extracting features of inertia data which are subjected to segmentation, increasing and decreasing experiments on various statistical data derived from the inertia data, comparing the performance of a speed recognition model, and finally selecting 33-dimensional features;

the dictionary learning algorithm is used for acquiring more intrinsic characteristic representation so as to improve the accuracy of speed recognition; using

Represents training samples, wherein Y ₁ ,...,Y _C Training samples representing a total of class C, Y _i (i=1,.,. C.) is detachable into y ₁ ,...,y _N N training sample data in total, +.>

Meaning a linear space of dimension N x N, describing the dimension of the training sample matrix Y. The method is characterized in that divided single-step inertial data are taken as sample units and are characterized by refined 33-dimensional characteristics, a label is a speed group trunk used for training a running machine experiment of a speed recognition model, C is a speed class number, n is a characteristic dimension, and dictionary learning aims at learning a latent variable projection dictionary->

Projection coefficient matrix +.>

Wherein K is dictionary atomic weight, < >>

The dimension of the learning dictionary D is described as n K, which is defined by K n-dimensional vectors D _i (i=1,., K); />

The dimension of the projection coefficient matrix X is described as KXN, consisting of N K-dimensional vectors X _i (i=1,., N);

the objective function of the dictionary learning algorithm used is as shown in formula (2):

s.t.||d _i || ² ＝1,i＝1,...,K

because of the mutual iterative relationship of D, X, V and L, the maximum iterative number is set as T _max Before the maximum iteration number is reached, enabling the formula (2) to take a minimum value;

wherein Y is a training sample, D is a learning dictionary, X is a projection coefficient matrix, V is a coding coefficient matrix, L is a graph Laplacian matrix, tr is the trace operation of the matrix, alpha, beta and gamma are regularization parameters, and the constraint condition satisfied by the first half of the s.t. expression (2) is all D _i Is equal to d _i || ² The values of 1, i are all 1,..,

iterating for the ith time to obtain D ⁱ And X is ⁱ Giving formula (2) the next iteration of D ⁱ⁺¹ And X is ⁱ⁺¹ The method comprises the steps of carrying out a first treatment on the surface of the Through the maximum times T set by iteration _max Secondary to obtain the final product

And +.>

The two are used as a speed classifier in combination;

step four, carrying out single-step division and feature extraction on the newly input data, and identifying speed information according to a speed identification model;

fifthly, integrating the newly obtained set of characteristics and speed data with the existing model, updating model parameters, and improving the recognition performance of the model on the input speed data;

and sixthly, transmitting information such as the speed, the step number and the like of the pedestrians to the mobile phone app or the notebook computer terminal through the communication module, and realizing the visualization of the indoor pedestrian movement speed.

Compared with the prior art, the invention considers the wearing convenience of the whole device, integrates the IMU, the MCU and the wireless communication module on the shoe, is convenient and comfortable, and can conveniently perform the work of identifying the movement speed of the indoor pedestrians.

Drawings

FIG. 1 is a flowchart of an intelligent recognition method of the novel intelligent shoe to the indoor pedestrian movement speed;

FIG. 2 is a basic schematic block diagram of an intelligent recognition device for the speed of an indoor pedestrian;

FIG. 3 is a diagram of an embodiment of a smart shoe;

reference numerals: 1. wireless communication module 2, IMU,3, microprocessor MCU,4, inertial sensor (the device can be embedded inside the shoe by means of a zip fastener).

FIG. 4 is a single step data partitioning result of acceleration peaks;

fig. 5 is a speed recognition result.

Detailed Description

The technical scheme of the invention is described in detail below with reference to the accompanying drawings and examples.

FIG. 1 is a flowchart showing the whole intelligent recognition method of the novel intelligent shoe to the indoor pedestrian movement speed. The method specifically comprises the following steps:

firstly, extracting inertial sensing data of feet of pedestrians by using an IMU inertial sensing unit;

secondly, dividing the time-domain continuous inertial data according to steps by adopting an acceleration peak value dividing method;

by detecting

The peak values of (2) are divided stepwise, and the two norms of the x, y and z triaxial acceleration are +.>

Is defined as formula (1):

as shown in fig. 4, the horizontal axis is a sampling point, the vertical axis is the two-norm data of the triaxial acceleration calculated by the formula (1), the star-shaped mark represents the identified acceleration two-norm peak value, and the step result is marked by a dotted line; the divided single-step data is uploaded to a hundred-degree network disk;

thirdly, extracting features of the inertia data of each step, inputting a dictionary learning algorithm, and performing model training to obtain a speed recognition model;

firstly, extracting features of the inertia data which are subjected to segmentation, performing increasing and decreasing experiments on various statistical data derived from the inertia data through an ablation experiment (ablation experiment, which can be regarded as a control variable method) and a reading document, and comparing the performances of a speed recognition model, wherein the finally selected 33-dimensional features are shown in a table 1.

The dictionary learning algorithm is used for acquiring more intrinsic characteristic representation so as to improve the accuracy of speed recognition; using

Represents training samples, wherein Y ₁ ,...,Y _C Training samples representing a total of class C, Y _i (i=1,.,. C.) is detachable into y ₁ ,...,y _N N training sample data in total, +.>

Meaning a linear space of dimension N x N, describing the dimension of the training sample matrix Y. We have divided single-step inertial data as sample cells, characterized by refined 33-dimensional features, tags forSpeed group trial during running machine experiment for training speed recognition model, C is speed category number, n is feature dimension, dictionary learning is aimed at learning latent variable projection dictionary->

Projection coefficient matrix +.>

Wherein K is dictionary atomic weight, < >>

The dimension of the learning dictionary D is described as n K, which is defined by K n-dimensional vectors D _i (i=1,., K); />

The dimension of the projection coefficient matrix X is described as KXN, consisting of N K-dimensional vectors X _i (i=1,., N);

the objective function of the dictionary learning algorithm used is as shown in formula (2):

s.t.||d _i || ² ＝1,i＝1,...,K

because of the mutual iterative relationship of D, X, V and L, we set the maximum iterative number as T _max Before the maximum number of iterations is reached, the equation (2) is minimized. Wherein Y is a training sample, D is a learning dictionary, X is a projection coefficient matrix, V is a coding coefficient matrix, L is a graph Laplace matrix, tr is the trace operation of the matrix, alpha, beta and gamma are regularization parameters, and s.t. means that the first half of the formula (2) meets the constraint condition that all D _i Is equal to d _i || ² The value ranges of both are 1, i are 1.

Iterating for the ith time to obtain D ⁱ And X is ⁱ Given to equation (2), find D for the next iteration ⁱ⁺¹ And X is ⁱ⁺¹ The method comprises the steps of carrying out a first treatment on the surface of the By iteratively setting the maximum timesNumber T _max Secondary to obtain the final product

And +.>

The two are used as a speed classifier in combination;

step four, carrying out single-step division and feature extraction on the newly input data, and identifying speed information according to a speed identification model;

fifthly, integrating the newly obtained set of characteristics and speed data with the existing model, and updating the model to achieve higher recognition rate and better robustness;

and sixthly, transmitting information such as the speed, the step number and the like of the pedestrians to the mobile phone app or the notebook computer terminal through the communication module, and realizing the visualization of the indoor pedestrian movement speed.

As shown in fig. 2, a basic schematic block diagram of the intelligent recognition device for the speed of the pedestrians in the room is provided.

The correlation algorithm is specifically described as follows:

1. peak detection&Single step division. The method of the invention is based on the basis of accurately dividing steps based on inertial data, so the problem to be solved is to accurately divide steps without affecting the PDR system and the rest of the zero-speed updating algorithm. The invention selects the two norms of the triaxial acceleration data

As the basis of the step division, the division is more accurate and does not affect the subsequent steps because the data most likely is when the foot is stationary relative to the ground (i.e. the acceleration is the greatest, and the sole is most likely to be close to the ground) when the data reaches the peak value.

2. And (5) extracting characteristics. There are many statistical features that can be selected, but not every feature is suitable for the speed recognition job. By innovating and deleting the features, summarizing and summarizing, 33-dimensional statistical features (shown in table 1) are finally refined as feature expressions (dictionary learning algorithms) of each set of data.

3. Feature expression (dictionary learning algorithm). This section includes improvements in algorithms: the traditional machine learning algorithm comprises SVM (support vector machine), naive Bayes, K-neighbor algorithm and the like, improves the algorithm for training the recognition model and carrying out speed recognition, and selects the dictionary learning algorithm. The dictionary learning algorithm can obtain stronger robustness and more discriminative characteristic expression on sample data, and can improve the accuracy of recognition while shortening the time of the algorithm. The method has the advantages that the characteristics capable of reflecting the speed characteristics are selected, redundant characteristics are removed, and the reduction of the recognition rate and the occurrence of overfitting are prevented. A specific 33-dimensional feature overview is shown in table 1.

TABLE 1

The three steps are integrated, and meanwhile, the step number and the identified speed information are input into the mobile phone app or the notebook computer terminal through the wireless communication module, so that a user can conveniently acquire the motion information of the user in an indoor environment in real time.

As shown in FIG. 3, a diagram of an embodiment of a smart shoe is shown. On the basis of common sports shoes, a plurality of preset positions for configuring devices such as inertial sensors and the like are set, mark points are made, and only a designated data acquisition module, a data processing and storage module or a communication module is required to be installed at the designated positions during mass production, so that the cost is low. The embodiment comprises a wireless communication module 1, an IMU2, a microprocessor MCU 3 and an inertial sensor 4 to realize the intelligent recognition method of the novel intelligent shoe for the indoor pedestrian movement speed. The wearable recognition device is used for acquiring original inertial data through the foot inertial sensor 4, extracting features, and carrying out speed recognition in real time after training a recognition model. The required inertial sensing unit has higher portability degree, can be bound on shoes, has a wireless communication function, can detect the real-time speed of pedestrians, and then serves an indoor positioning system. The method provides convenience for people to acquire the position and the state of the people in the indoor environment in real time.

The present invention is not limited to the specific steps described above. The invention extends to any novel one, or any novel combination, or novel combination, of the features disclosed in this specification. In summary, the present description should not be construed as limiting the invention.

Claims (1)

1. The intelligent recognition method of the novel intelligent shoe for the indoor pedestrian movement speed is characterized by comprising the following steps of:

firstly, extracting inertial sensing data of feet of pedestrians by using an IMU inertial sensing unit;

secondly, dividing the time-domain continuous inertial data according to steps by adopting an acceleration peak value dividing method, and detecting

The peak values of (2) are divided stepwise, and the two norms of the x, y and z triaxial acceleration are +.>

Is defined as shown in formula (1):

uploading and storing the divided single-step data,

thirdly, extracting features of the inertia data of each step, inputting a dictionary learning algorithm, and performing model training to obtain a speed recognition model;

firstly, extracting features of inertia data which are subjected to segmentation, increasing and decreasing experiments on various statistical data derived from the inertia data, comparing the performance of a speed recognition model, and finally selecting 33-dimensional features;

the dictionary learning algorithm is used for acquiring more intrinsic characteristic representation so as to improve the accuracy of speed recognition; using

Represents training samples, wherein Y ₁ ,...,Y _C Training samples representing a total of class C, Y _i (i=1,.,. C.) is detachable into y ₁ ,...,y _N N training sample data in total, +.>

Means linear space of N x N dimension, describes dimension of training sample matrix Y, takes divided single-step inertial data as sample unit, and is characterized by refined 33-dimensional characteristics, the label is speed group trunk when running machine experiment is used for training speed recognition model, C is speed class number, N is characteristic dimension, dictionary learning is aimed at learning latent variable projection dictionary->

Projection coefficient matrix +.>

Wherein K is dictionary atomic weight, < >>

The dimension of the learning dictionary D is described as n K, which is defined by K n-dimensional vectors D _i (i=1,., K); />

The dimension of the projection coefficient matrix X is described as KXN, consisting of N K-dimensional vectors X _i (i=1,., N);

the objective function of the dictionary learning algorithm used is as shown in formula (2):

s.t.||d _i || ² ＝1,i＝1,...,K

due to the phases D, X, V, LSetting the maximum iteration number as T according to the mutual iteration relation _max Before the maximum iteration number is reached, enabling the formula (2) to take a minimum value;

wherein Y is a training sample, D is a learning dictionary, X is a projection coefficient matrix, V is a coding coefficient matrix, L is a graph Laplacian matrix, tr is the trace operation of the matrix, alpha, beta and gamma are regularization parameters, and the constraint condition satisfied by the first half of the s.t. expression (2) is all D _i Is equal to d _i || ² The values of 1, i are all 1,..,

iterating for the ith time to obtain D ⁱ And X is ⁱ Giving formula (2) the next iteration of D ⁱ⁺¹ And X is ⁱ⁺¹ The method comprises the steps of carrying out a first treatment on the surface of the Through the maximum times T set by iteration _max Secondary to obtain the final product

And +.>

The two are used as a speed classifier in combination;

step four, carrying out single-step division and feature extraction on the newly input data, and identifying speed information according to a speed identification model;

fifthly, integrating the newly obtained set of characteristics and speed data with the existing model, updating model parameters, and improving the recognition performance of the model on the input speed data;

and sixthly, transmitting the speed and step number information of the pedestrians to a mobile phone app or a notebook computer terminal through a communication module, and realizing the visualization of the indoor pedestrian movement speed.

Images