CN111207739A - Pedestrian walking zero-speed detection method and device based on GRU neural network - Google Patents
Pedestrian walking zero-speed detection method and device based on GRU neural network Download PDFInfo
- Publication number
- CN111207739A CN111207739A CN201811405861.2A CN201811405861A CN111207739A CN 111207739 A CN111207739 A CN 111207739A CN 201811405861 A CN201811405861 A CN 201811405861A CN 111207739 A CN111207739 A CN 111207739A
- Authority
- CN
- China
- Prior art keywords
- vector
- data
- gru
- gate
- speed detection
- 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
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
-
- 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
- G01C21/20—Instruments for performing navigational calculations
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/048—Activation functions
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
Abstract
The invention provides a pedestrian walking zero-speed detection method based on a GRU neural network, wherein the GRU comprises an update gate and a reset gate, and the method comprises the following steps: receiving and recording data x sent by IMU in real timet(ii) a Data x of current state inputtData h output from the hidden layer at the previous momentt‑1Outputting a value between 0 and 1 through an updating gate, and calculating to obtain an updating gate vector ut(ii) a Data xtAnd data ht‑1The sigmoid layer entering the reset gate outputs a value between 0 and 1, and a reset gate vector r is obtained through calculationtWhile the tanh layer creates a new candidate memory value vector
The gate vector u will be updatedtAs a weight vector, a candidate memory value vector
And the data h output by the hidden layer at the previous momentt‑1Obtaining output vector h of GRU unit by weighted averaget(ii) a Selecting different data xtAnd repeating the steps as an input vector, carrying out GRU training and verification to obtain an optimal GRU model, and carrying out zero walking speed detection and judgment on the acquired data.
Description
Technical Field
The invention relates to the technical field of pedestrian walking detection, in particular to a pedestrian walking zero-speed detection method and device based on a GRU neural network.
Background
In view of the requirements of various aspects such as volume, weight, power consumption, cost and the like, most pedestrian walking Navigation systems use a low-precision MEMS (Micro-Electro-Mechanical System) inertial sensor to complete the measurement of gait and motion acceleration of pedestrians, and provide accurate position Navigation service for pedestrians in GNSS (Global Navigation Satellite System) and other areas, under the shade of trees and even indoors where most radio Navigation systems have poor performance. Because the precision of the MEMS inertial device is low, and the inertial device is difficult to be corrected by satellite signals under the environment with weak GNSS signals, an effective solution is to install the MEMS inertial device on a shoe, and perform zero-speed correction by using the zero-speed state of the instep in the walking process of a human body, so as to achieve a long-time high-precision pure-inertia autonomous positioning result.
Then, according to the above, the accuracy of the zero-speed detection plays a crucial role in completing a higher positioning index by using a lower-performance MEMS inertial device. The existing zero-speed detector mainly judges the zero-speed state of the foot surface when a pedestrian walks according to a given unified threshold, data read by an Inertial Measurement Unit (IMU) is substituted into a motion model for calculation, and if the calculation result is lower than the threshold, the system considers that the current foot surface of the pedestrian is in a static state. However, the walking posture of the human has higher diversity, and the walking posture of the pedestrian has stronger complexity for some special road conditions (such as climbing stairs, ascending slopes, etc.), so that it is difficult to accurately judge the zero-speed walking state by using a fixed statistical threshold, which seriously restricts the walking navigation algorithm from achieving higher walking navigation accuracy of the pedestrian.
While a fixed threshold may provide near-optimal zero velocity detection for uniform motion types, human motion gestures are diverse and have some complexity, and fixed threshold detection fails when the user changes his motion type or intensity. Because the threshold given by the system is no longer optimal, the following two cases of zero-speed detection failure occur: 1) the threshold value is set to be too high, so that the output result of the motion model is lower than the threshold value, namely the instep is misjudged as a static state when moving; 2) the threshold is set too low, resulting in the output result of the motion model being higher than the threshold, i.e. the foot surface is still misjudged as a motion state in a static state. Both of the above two errors will cause the zero-speed-based walking navigation algorithm to generate a large accumulated error, and finally, the positioning accuracy is reduced.
Disclosure of Invention
The invention provides a method for completing a pedestrian gait zero-speed detection task by using a Recurrent Neural Network (RNN) Gated cyclic Unit (GRU) to replace the existing model-based zero-speed detector, which can adaptively learn the walking posture characteristics of a pedestrian, distinguish different motion states of the human body during walking, improve the detection precision of the zero-speed state, solve the technical problems and finally achieve higher walking navigation precision.
The technical scheme adopted by the invention is as follows:
a pedestrian walking zero-speed detection method based on a GRU neural network, wherein the GRU comprises an updating gate and a resetting gate, and the method comprises the following steps:
receiving and recording data x sent by IMU in real timet；
Data x of current state inputtData h output from the hidden layer at the previous momentt-1Through the update gateA value between 0 and 1 is obtained, and an updated gate vector u is obtained through calculationt；
Data xtAnd data ht-1The sigmoid layer entering the reset gate outputs a value between 0 and 1, and a reset gate vector r is obtained through calculationtWhile the tanh layer creates a new candidate memory value vector
The gate vector u will be updatedtAs a weight vector, a candidate memory value vector
And the data h output by the hidden layer at the previous momentt-1Obtaining output vector h of GRU unit by weighted averaget；
Selecting different data xtAnd repeating the steps as an input vector, carrying out GRU training and verification to obtain an optimal GRU model, and carrying out zero walking speed detection and judgment on the acquired data through the optimal GRU model.
Further, the IMU is fixed on the sole, and data x sent by the IMU is received and recorded in real time through the handheld devicet。
Further, the IMU is a 6-axis IMU device.
Further, the update gate vector utThe calculation formula is as follows:
ut＝σ(bu+Uuxt+Wuht-1)
wherein u istRepresenting the updated gate vector at time t, σ being the activation function, buOffset vector, U, representing the update gateuInput weight, W, representing an update gateuIndicating the cyclic weight of the update gate, xtData, h, from IMU at time tt-1Representing the data output by the hidden layer at time t-1.
Further, the reset gate vector rtThe calculation formula is as follows:
rt＝σ(br+Urxt+Wrht-1)
the candidate memory value vector
The calculation formula is as follows:
wherein r istRepresenting the reset gate vector at time t, brOffset vector, U, representing reset gaterRepresenting the input weight of the reset gate, WrA cyclic weight of the reset gate is represented,
representing candidate memory value vectors at time t, bsOffset vector, U, representing candidate memory cellsRepresenting input weights, W, of candidate memory cellssRepresents the round-robin weight of the candidate memory cell, and tanh represents the activation function.
Further, the output vector h of the GRU unittThe calculation formula is as follows:
wherein h istRepresenting the output vector of the GRU unit at time t.
Further, different data x are selectedtAnd (4) as an input vector, taking the ground zero-speed label as a true value, repeating the steps, and performing GRU training and verification.
Further, 150 groups of data were selected for GRU training and 50 groups of data were selected for GRU validation.
Further, the step length of each set of data was 400 meters.
The invention also provides a pedestrian walking zero-speed detection device based on the GRU neural network, which comprises:
a data receiving unit for receiving and recording the data x sent by the IMU in real timet；
Update gate vector calculation Unit, data x input at Current StatetData h output from the hidden layer at the previous momentt-1Outputting a value between 0 and 1 through an updating gate for calculating to obtain an updating gate vector ut；
A unit for calculating a reset gate vector and a candidate memory value vector, data xtAnd data ht-1The sigmoid layer entering the reset gate outputs a value between 0 and 1 for calculating to obtain a reset gate vector rtWhile the tanh layer creates a new candidate memory value vector
A GRU output vector calculation unit for updating the gate vector utAs a weight vector, a candidate memory value vector
And the data h output by the hidden layer at the previous momentt-1Obtaining output vector h of GRU unit by weighted averaget；
A pedestrian walking zero-speed detection unit for selecting different data xtAnd repeating the steps as an input vector, carrying out GRU training and verification to obtain an optimal GRU model, and carrying out zero walking speed detection and judgment on the acquired data through the optimal GRU model.
The invention also provides a memory, in which a computer program is stored, the computer program performing the steps of:
receiving and recording data x sent by IMU in real timet；
Data x of current state inputtData h output from the hidden layer at the previous momentt-1Outputting a value between 0 and 1 through an updating gate, and calculating to obtain an updating gate vector ut；
Data xtAnd data ht-1The sigmoid layer entering the reset gate outputs a value between 0 and 1, and a reset gate vector r is obtained through calculationtWhile the tanh layer creates new candidate tokensMemory value vector
The gate vector u will be updatedtAs a weight vector, a candidate memory value vector
And the data h output by the hidden layer at the previous momentt-1Obtaining output vector h of GRU unit by weighted averaget；
Selecting different data xtAnd repeating the steps as an input vector, carrying out GRU training and verification to obtain an optimal GRU model, and carrying out zero walking speed detection and judgment on the acquired data through the optimal GRU model.
The invention trains GRU by using ground truth-value zero-speed labels and a large amount of asynchronous data, so that the zero-speed detector can accurately judge the zero-speed state of the pedestrian in different motion states, thereby improving the positioning precision of a pedestrian navigation system.
Drawings
Fig. 1 is a pedestrian trajectory graph calculated by a pedestrian walking zero-speed detection algorithm based on a GRU neural network.
Fig. 2 is a projection diagram of a pedestrian track on a google map, which is calculated by a GRU neural network-based pedestrian walking zero-speed detection algorithm.
Fig. 3 is a structural diagram of a pedestrian walking zero-speed detection device based on a GRU neural network.
Detailed Description
The invention mainly utilizes the GRU neural network of the gate control cycle unit to deeply learn the zero speed characteristic of the gait of the pedestrian, thereby more accurately judging the zero speed state of the pedestrian in different states and different road conditions, improving the correction accuracy of IMU devices and finally improving the position precision of pedestrian navigation. The GRU is an important variant of Long and Short Term Memory neural network (LSTM), and the GRU is a combination of a forgetting gate of LSTM and an external input gate into an update gate (update gate), and mainly functions to control which information of the current input is retained, and the reset gate (reset gate) mainly functions to control the influence of the output at the previous moment on the current input. Since the GRU has only two control gates, the calculation speed of the GRU is much faster than LSTM.
The invention is further illustrated below with reference to the figures and examples.
The first embodiment is as follows:
the invention provides a pedestrian walking zero-speed detection method based on a GRU neural network, which specifically comprises the following steps:
step 1: the 6-axis IMU device is fixed on the sole, and meanwhile, the handheld device can receive and record data x sent by the IMU in real timet。
Step 2: current state input IMU data xtHidden layer output h from previous timet-1(when t is 1, ht-10) is output through the update gate, a value between 0 and 1 is output, where 0 represents completely discarded information and 1 represents completely retained information, and the calculation formula is shown in formula (1).
ut＝σ(bu+Uuxt+Wuht-1) (1)
Wherein u istRepresenting the updated gate vector at time t, σ being the activation function, buOffset vector, U, representing the update gateuInput weight, W, representing an update gateuIndicating the cyclic weight of the update gate. x is the number oftIMU6 axis sensor input vector, h, representing time tt-1Representing the hidden layer output vector at time t-1.
And step 3: x is the number oftAnd ht-1The sigmoid layer entering the reset gate outputs a value between 0 and 1, and the tanh layer creates a new candidate memory value vector
The calculation formulas are shown as formula (2) and formula (3).
rt＝σ(br+Urxt+Wrht-1) (2)
Wherein r istRepresenting the reset gate vector at time t, brOffset vector, U, representing reset gaterRepresenting the input weight of the reset gate, WrA cyclic weight of the reset gate is represented,
representing a vector of candidate values at time t, bsOffset vector, U, representing candidate memory cellsRepresenting input weights, W, of candidate memory cellssRepresents the round-robin weight of the candidate memory cell, and tanh represents the activation function.
And 4, step 4: the output h of the GRU unit is obtained by weighted average of the candidate memory vector and the output vector of the hidden layer at the previous moment by using the update gate as a weight vectortThe calculation formula is shown in formula (4).
Where U represents the update gate vector, r represents the reset gate vector, b represents the offset vector, U represents the input weight, and W represents the cyclic weight. x is the number oftRepresenting the input vector at time t, htRepresenting the output vector at time t.
And 5: and (4) taking the IMU data in the first step as an input vector of the algorithm, and taking a ground zero-velocity label as a true value to train the algorithm built in the second step to the fourth step. A total of 150 sets of data were trained on the above model, 50 sets of data were used for validation, each set of data having a step length of about 400 meters. After the training and verification steps are completed, all the weights in the GRU reach the best, and the model can be used for carrying out pedestrian walking zero-speed detection and judgment on newly acquired data.
Fig. 1 and 2 show the test results of the pedestrian walking zero speed detection method provided by the invention, and the results show that the pedestrian positioning precision is higher by the invention. The test is carried out on a runway of a standard stadium, an IMU device is worn on the foot of a tested person and walks for about 1km according to a fixed route, after the test, the track of the tested person is calculated according to a zero-speed detection algorithm and an inertial navigation algorithm provided by the patent, the error of the track is evaluated, and the error of the track in the horizontal direction is less than 4 m. Fig. 1 is a schematic diagram of a pedestrian trajectory route, and fig. 2 is a diagram of projecting a route trajectory on a google map.
Example two:
the invention also provides a pedestrian walking zero-speed detection device based on the GRU neural network, as shown in fig. 3, the device comprises:
a data receiving unit for receiving and recording the data x sent by the IMU in real timet；
Update gate vector calculation Unit, data x input at Current StatetData h output from the hidden layer at the previous momentt-1Outputting a value between 0 and 1 through an updating gate for calculating to obtain an updating gate vector ut；
A unit for calculating a reset gate vector and a candidate memory value vector, data xtAnd data ht-1The sigmoid layer entering the reset gate outputs a value between 0 and 1 for calculating to obtain a reset gate vector rtWhile the tanh layer creates a new candidate memory value vector
A GRU output vector calculation unit for updating the gate vector utAs a weight vector, a candidate memory value vector
And the data h output by the hidden layer at the previous momentt-1Obtaining output vector h of GRU unit by weighted averaget；
A pedestrian walking zero-speed detection unit for selecting different data xtAnd repeating the steps as an input vector, carrying out GRU training and verification to obtain an optimal GRU model, and carrying out zero walking speed detection and judgment on the acquired data through the optimal GRU model.
Example three:
the invention also provides a memory, in which a computer program is stored, the computer program performing the steps of:
receiving and recording data x sent by IMU in real timet；
Data x of current state inputtData h output from the hidden layer at the previous momentt-1Outputting a value between 0 and 1 through an updating gate, and calculating to obtain an updating gate vector ut；
Data xtAnd data ht-1The sigmoid layer entering the reset gate outputs a value between 0 and 1, and a reset gate vector r is obtained through calculationtWhile the tanh layer creates a new candidate memory value vector
The gate vector u will be updatedtAs a weight vector, a candidate memory value vector
And the data h output by the hidden layer at the previous momentt-1Obtaining output vector h of GRU unit by weighted averaget；
Selecting different data xtAnd repeating the steps as an input vector, carrying out GRU training and verification to obtain an optimal GRU model, and carrying out zero walking speed detection and judgment on the acquired data through the optimal GRU model.
The invention utilizes the GRU recurrent neural network to detect the gait zero speed of the pedestrian, and the method can effectively improve the accuracy of zero speed detection of the pedestrian in different states, reduce the accumulated error of IMU devices and improve the performance of an algorithm for positioning the pedestrian.
Although the present invention has been described with reference to the preferred embodiments, it is not intended to limit the present invention, and those skilled in the art can make variations and modifications of the present invention without departing from the spirit and scope of the present invention by using the methods and technical contents disclosed above.
Claims (11)
1. A pedestrian walking zero-speed detection method based on a GRU neural network is characterized in that the GRU comprises an update gate and a reset gate, and the method comprises the following steps:
receiving and recording data x sent by IMU in real timet；
Data x of current state inputtData h output from the hidden layer at the previous momentt-1Outputting a value between 0 and 1 through an updating gate, and calculating to obtain an updating gate vector ut；
Data xtAnd data ht-1The sigmoid layer entering the reset gate outputs a value between 0 and 1, and a reset gate vector r is obtained through calculationtWhile the tanh layer creates a new candidate memory value vector
The gate vector u will be updatedtAs a weight vector, a candidate memory value vector
And the data h output by the hidden layer at the previous momentt-1Obtaining output vector h of GRU unit by weighted averaget；
Selecting different data xtAnd repeating the steps as an input vector, carrying out GRU training and verification to obtain an optimal GRU model, and carrying out zero walking speed detection and judgment on the acquired data through the optimal GRU model.
2. The GRU neural network-based pedestrian walking zero-speed detection method as claimed in claim 1, wherein the IMU is fixed on the sole, and data x sent by the IMU is received and recorded in real time through a handheld devicet。
3. The GRU neural network-based pedestrian walking zero-speed detection method of claim 2, wherein the IMU is a 6-axis IMU device.
4. The GRU neural network-based pedestrian walking zero-speed detection method as claimed in claim 1, wherein the update gate vector utThe calculation formula is as follows:
ut＝σ(bu+Uuxt+Wuht-1)
wherein u istRepresenting the updated gate vector at time t, σ being the activation function, buOffset vector, U, representing the update gateuInput weight, W, representing an update gateuIndicating the cyclic weight of the update gate, xtData, h, from IMU at time tt-1Representing the data output by the hidden layer at time t-1.
5. The GRU neural network-based pedestrian walking zero-speed detection method as claimed in claim 4, wherein the reset gate vector rtThe calculation formula is as follows:
rt＝σ(br+Urxt+Wrht-1)
the candidate memory value vector
The calculation formula is as follows:
wherein r istRepresenting the reset gate vector at time t, brOffset vector, U, representing reset gaterRepresenting the input weight of the reset gate, WrA cyclic weight of the reset gate is represented,
represents the candidate memory value direction at time tAmount bsOffset vector, U, representing candidate memory cellsRepresenting input weights, W, of candidate memory cellssRepresents the round-robin weight of the candidate memory cell, and tanh represents the activation function.
6. The GRU neural network-based pedestrian walking zero-speed detection method as claimed in claim 5, wherein an output vector h of the GRU unittThe calculation formula is as follows:
wherein h istRepresenting the output vector of the GRU unit at time t.
7. The GRU neural network-based pedestrian walking zero-speed detection method as claimed in claim 6, wherein different data x are selectedtAnd (4) as an input vector, taking the ground zero-speed label as a true value, repeating the steps, and performing GRU training and verification.
8. The GRU neural network-based pedestrian walking zero-speed detection method as claimed in claim 7, wherein 150 groups of data are selected for GRU training, and 50 groups of data are selected for GRU verification.
9. The GRU neural network-based pedestrian walking zero-speed detection method as claimed in claim 8, wherein the walking length of each set of data is 400 meters.
10. A pedestrian walking zero-speed detection device based on a GRU neural network is characterized by comprising:
a data receiving unit for receiving and recording the data x sent by the IMU in real timet；
Update gate vector calculation Unit, data x input at Current StatetData h output from the hidden layer at the previous momentt-1Output one through the refresh gateA value between 0 and 1 for calculating an updated gate vector ut；
A unit for calculating a reset gate vector and a candidate memory value vector, data xtAnd data ht-1The sigmoid layer entering the reset gate outputs a value between 0 and 1 for calculating to obtain a reset gate vector rtWhile the tanh layer creates a new candidate memory value vector
A GRU output vector calculation unit for updating the gate vector utAs a weight vector, a candidate memory value vector htAnd the data h output by the hidden layer at the previous momentt-1Obtaining output vector h of GRU unit by weighted averaget；
A pedestrian walking zero-speed detection unit for selecting different data xtAnd repeating the steps as an input vector, carrying out GRU training and verification to obtain an optimal GRU model, and carrying out zero walking speed detection and judgment on the acquired data through the optimal GRU model.
11. A memory storing a computer program, the computer program performing the steps of:
receiving and recording data x sent by IMU in real timet；
Data x of current state inputtData h output from the hidden layer at the previous momentt-1Outputting a value between 0 and 1 through an updating gate, and calculating to obtain an updating gate vector ut；
Data xtAnd data ht-1The sigmoid layer entering the reset gate outputs a value between 0 and 1, and a reset gate vector r is obtained through calculationtWhile the tanh layer creates a new candidate memory value vector
The gate vector u will be updatedtAs weight vector, candidateVector of selected memory values
And the data h output by the hidden layer at the previous momentt-1Obtaining output vector h of GRU unit by weighted averaget；
Selecting different data xtAnd repeating the steps as an input vector, carrying out GRU training and verification to obtain an optimal GRU model, and carrying out zero walking speed detection and judgment on the acquired data through the optimal GRU model.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811405861.2A CN111207739A (en) | 2018-11-22 | 2018-11-22 | Pedestrian walking zero-speed detection method and device based on GRU neural network |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811405861.2A CN111207739A (en) | 2018-11-22 | 2018-11-22 | Pedestrian walking zero-speed detection method and device based on GRU neural network |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN111207739A true CN111207739A (en) | 2020-05-29 |
Family
ID=70783769
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201811405861.2A Pending CN111207739A (en) | 2018-11-22 | 2018-11-22 | Pedestrian walking zero-speed detection method and device based on GRU neural network |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN111207739A (en) |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112578419A (en) * | 2020-11-24 | 2021-03-30 | 南京邮电大学 | GPS data reconstruction method based on GRU network and Kalman filtering |
| CN113570129A (en) * | 2021-07-20 | 2021-10-29 | 武汉钢铁有限公司 | Method for predicting strip steel pickling concentration and computer readable storage medium |
| CN114019182A (en) * | 2021-11-04 | 2022-02-08 | 苏州挚途科技有限公司 | Zero-speed state detection method and device and electronic equipment |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US6473698B1 (en) * | 1999-04-22 | 2002-10-29 | University Of Louisville Research Foundation, Inc. | Method and apparatus for automated rolling leukocyte velocity measurement in vivo |
| CN106285960A (en) * | 2015-05-12 | 2017-01-04 | 北京汽车动力总成有限公司 | The computational methods of a kind of vehicle simulation parameter and device |
| CN106643715A (en) * | 2016-11-17 | 2017-05-10 | 天津大学 | Indoor inertial navigation method based on bp neural network improvement |
| CN108334677A (en) * | 2018-01-17 | 2018-07-27 | 哈尔滨工程大学 | A kind of UUV Realtime collision free planing methods based on GRU networks |
| CN108621159A (en) * | 2018-04-28 | 2018-10-09 | 首都师范大学 | A kind of Dynamic Modeling in Robotics method based on deep learning |
-
2018
- 2018-11-22 CN CN201811405861.2A patent/CN111207739A/en active Pending
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US6473698B1 (en) * | 1999-04-22 | 2002-10-29 | University Of Louisville Research Foundation, Inc. | Method and apparatus for automated rolling leukocyte velocity measurement in vivo |
| CN106285960A (en) * | 2015-05-12 | 2017-01-04 | 北京汽车动力总成有限公司 | The computational methods of a kind of vehicle simulation parameter and device |
| CN106643715A (en) * | 2016-11-17 | 2017-05-10 | 天津大学 | Indoor inertial navigation method based on bp neural network improvement |
| CN108334677A (en) * | 2018-01-17 | 2018-07-27 | 哈尔滨工程大学 | A kind of UUV Realtime collision free planing methods based on GRU networks |
| CN108621159A (en) * | 2018-04-28 | 2018-10-09 | 首都师范大学 | A kind of Dynamic Modeling in Robotics method based on deep learning |
Non-Patent Citations (3)
| Title |
|---|
| BRANDON WAGSTAFF等: "LSTM-Based Zero-Velocity Detection for Robust Inertial Navigation", 《2018 INTERNATIONAL CONFERENCE ON INDOOR POSITIONING AND INDOOR NAVIGATION (IPIN)》 * |
| JIHEON KANG等: "Smartphone-Based Traveled Distance Estimation Using Individual Walking Patterns for Indoor Localization", 《SENSORS (BASEL, SWITZERLAND)》 * |
| MINGYANG WANG等: "Human body and limb motion recognition via stacked gated recurrent units netwok", 《IET RADAR,SONAR AND NAVIGATION》 * |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112578419A (en) * | 2020-11-24 | 2021-03-30 | 南京邮电大学 | GPS data reconstruction method based on GRU network and Kalman filtering |
| CN112578419B (en) * | 2020-11-24 | 2023-12-12 | 南京邮电大学 | GPS data reconstruction method based on GRU network and Kalman filtering |
| CN113570129A (en) * | 2021-07-20 | 2021-10-29 | 武汉钢铁有限公司 | Method for predicting strip steel pickling concentration and computer readable storage medium |
| CN114019182A (en) * | 2021-11-04 | 2022-02-08 | 苏州挚途科技有限公司 | Zero-speed state detection method and device and electronic equipment |
| CN114019182B (en) * | 2021-11-04 | 2024-02-02 | 苏州挚途科技有限公司 | Zero-speed state detection method and device and electronic equipment |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN103968827B (en) | A kind of autonomic positioning method of wearable body gait detection | |
| CN1932524B (en) | Apparatus and method for detecting step in a personal navigator | |
| US10267646B2 (en) | Method and system for varying step length estimation using nonlinear system identification | |
| CN104296750B (en) | Zero speed detecting method, zero speed detecting device, and pedestrian navigation method as well as pedestrian navigation system | |
| CN111207739A (en) | Pedestrian walking zero-speed detection method and device based on GRU neural network | |
| CN113091738B (en) | Mobile robot map construction method based on visual inertial navigation fusion and related equipment | |
| CN108957510B (en) | Pedestrian seamless integrated navigation positioning method based on inertia/zero speed/GPS | |
| CN111024075B (en) | Pedestrian navigation error correction filtering method combining Bluetooth beacon and map | |
| CN110553643A (en) | pedestrian self-adaptive zero-speed updating point selection method based on neural network | |
| JP2022531679A (en) | Device positioning methods, devices, and devices | |
| US20160097861A1 (en) | Method and apparatus for location determination | |
| CN107636420A (en) | Technology for pedestrian's dead reckoning | |
| CN107490378A (en) | A kind of indoor positioning based on MPU6050 and smart mobile phone and the method for navigation | |
| KR20130059344A (en) | Method and system for detection of a zero velocity state of an object | |
| CN108426582B (en) | Indoor three-dimensional map matching method for pedestrians | |
| Zhao et al. | Pseudo-zero velocity re-detection double threshold zero-velocity update (ZUPT) for inertial sensor-based pedestrian navigation | |
| Wu et al. | Indoor positioning system based on inertial MEMS sensors: Design and realization | |
| CN103499354B (en) | Neyman-Pearson criterion-based zero speed detection method | |
| CN109540133B (en) | Self-adaptive gait division method and system based on micro-inertia technology | |
| CN109297484A (en) | A kind of pedestrian's autonomous positioning error correcting method for thering is gait to constrain | |
| CN110455284A (en) | A kind of pedestrian movement patterns' recognition methods and device based on MEMS-IMU | |
| CN107105404B (en) | Pedestrian indoor positioning method based on step length matching | |
| Chen et al. | Pedestrian positioning with physical activity classification for indoors | |
| Ping et al. | Pedestrian stride-length estimation based on bidirectional LSTM network | |
| CN111735478B (en) | LSTM-based pedestrian real-time navigation zero-speed detection method |
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 |