Unlock instant, AI-driven research and patent intelligence for your innovation.

Deep network adaptive step size estimation method and device based on acceleration sensor

An acceleration sensor and self-adaptive step size technology, which can be used in measuring devices, instruments, measuring distances, etc., and can solve problems such as difficulty in measuring leg length and trouble in entering height.

Active Publication Date: 2019-11-19
北京羲和科技有限公司
View PDF1 Cites 0 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

[0003] The purpose of the present invention is to provide two kinds of deep network adaptive step size estimation methods and devices based on acceleration sensors, aiming to solve the problem that the existing physical model for estimating the step size is difficult to measure the leg length and the input height is troublesome

Method used

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
View more

Image

Smart Image Click on the blue labels to locate them in the text.
Viewing Examples
Smart Image
  • Deep network adaptive step size estimation method and device based on acceleration sensor
  • Deep network adaptive step size estimation method and device based on acceleration sensor

Examples

Experimental program
Comparison scheme
Effect test

specific Embodiment 1

[0066] Such as figure 1 As shown, the embodiment of the present invention provides an acceleration sensor-based deep network adaptive step size estimation method, including:

[0067] Collect the acceleration data of the walker through the acceleration sensor;

[0068] Obtain the height data of the walker;

[0069] The acceleration data and height data of the walker are trained to obtain a depth autoencoder; the depth autoencoder includes an input layer, a hidden layer and an output layer for inputting acceleration data and outputting height distribution data, and the height distribution data Including height interval data and probability distribution data of height interval data;

[0070] Obtain the acceleration data of the new walker, predict the height distribution data corresponding to the acceleration data of the walker through the depth autoencoder, and obtain the height interval data of the walker according to the height distribution data of the walker;

[0071] Accor...

specific Embodiment 2

[0077] An embodiment of the present invention provides an acceleration sensor-based deep network adaptive step size estimation method, including:

[0078] Obtain the height data of the walkers, and group the walkers according to the height data;

[0079] Gather the acceleration data of each group of walkers by an acceleration sensor, and divide the gained data into a training set, a verification set, and a test set; the acceleration data includes the components of the acceleration of gravity in the three-axis direction of the acceleration sensor;

[0080] Divide the components of the acceleration of gravity in the three-axis direction of the acceleration sensor by the constant of the acceleration of gravity to obtain the preprocessed acceleration data;

[0081] Train the acceleration data in the preprocessed training set and verification set, and establish a single-layer automatic decoder. The single-layer automatic decoder includes an input layer, a hidden layer, an output la...

specific Embodiment 3

[0087] Such as figure 2 As shown, the embodiment of the present invention provides a deep network adaptive step size estimation device based on an acceleration sensor, including:

[0088] Acquisition module for:

[0089] Collect the acceleration data of the walker through the acceleration sensor;

[0090] Obtain the height data of the walker;

[0091] Training modules for:

[0092] The acceleration data and height data of the walker are trained to obtain a depth autoencoder; the depth autoencoder includes an input layer, a hidden layer and an output layer for inputting acceleration data and outputting height distribution data, and the height distribution data Including height interval data and probability distribution data of height interval data;

[0093] Prediction module for:

[0094] Obtain the acceleration data of the new walker, predict the height distribution data corresponding to the acceleration data of the walker through the depth autoencoder, and obtain the he...

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to View More

PUM

No PUM Login to View More

Abstract

The invention provides a self-adaptive step length estimating method and device for a deep network based on an acceleration sensor. The self-adaptive estimating method is characterized in that step length is estimated in a self-adaptive mode by combining input body height with a three-axis acceleration signal by acquiring data based on the acceleration sensor and deeply learning and extracting characteristics. According to the method and the device disclosed by the invention, practical body height tested every time can be obtained in the self-adaptive mode, leg length does not need to be measured, and body height input is unnecessary, so that the method is convenient and accurate, and is not limited by personal privacy; high-reliability step length is estimated by estimating body height and integrating a reasonable physical model; error range can be given by body height estimated every time, and error range of the whole step length can be given by integrating the error range of the physical model, so that requirements, on errors, of follow-up technologies such as Kalman filtering are met.

Description

technical field [0001] The present invention relates to the technical field of adaptive step size estimation based on acceleration sensors, in particular to two acceleration sensor-based deep network adaptive step size estimation methods and devices. Background technique [0002] The adaptive step size estimation model is mostly in the PDR process. In the actual PDR process, the PDR is often corrected through the positioning of Bluetooth and wifi, and the reliability of the PDR itself plays a huge role in supplementing the fusion positioning. At present, there are static methods and dynamic methods for calculating the step length. The static method is k=0.413 for women and k=0.415 for men; most dynamic methods need to input leg length / height information, combined with dynamic evaluation of each step length Compared with the static model, the reliability and repeatability of the physical model are better. However, for most physical models, height information or leg length in...

Claims

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to View More

Application Information

Patent Timeline
no application Login to View More
Patent Type & Authority Patents(China)
IPC IPC(8): G01C22/00
CPCG01C22/006
Inventor 徐枫李金贵杜华睿
Owner 北京羲和科技有限公司