Looking for breakthrough ideas for innovation challenges? Try Patsnap Eureka!

Micro-Doppler radar human body action classification method of convolutional neural network

A technology of convolutional neural network and human action, which is applied in the direction of radio wave reflection/reradiation, instrumentation, computing, etc., and can solve the problems that the end-to-end learning advantages of the neural network cannot be fully utilized.

Active Publication Date: 2019-12-13
SHENZHEN UNIV
View PDF4 Cites 3 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

Therefore, these deep convolutional neural network methods are not end-to-end neural networks, because their input is still the spectrogram of the radar signal (not the original radar data), and cannot fully utilize the end-to-end learning advantages of the neural network.

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
  • Micro-Doppler radar human body action classification method of convolutional neural network
  • Micro-Doppler radar human body action classification method of convolutional neural network
  • Micro-Doppler radar human body action classification method of convolutional neural network

Examples

Experimental program
Comparison scheme
Effect test

specific Embodiment

[0062] Use Infineon's Sense2GoL Doppler radar to acquire micro-Doppler signals for human sleep movements and human daily movements. The radar chip used in Sense2GoL is BGT24LTR11, which is a silicon-germanium MMIC (single-chip microwave integrated circuit) for signal transmission and reception, and the operating frequency is from 24.05GHz to 24.25GHz. The Sense2GoL Doppler Radar has a maximum power of 10mW and it combines a receiver and a transmitter. RadarNet is trained using a server with 32G memory and NVIDIA GTX1080Ti graphics card.

[0063] Collect radar data for both experiments with a sampling frequency of 2 kHz for 3 s. The size of the STFT time window was set to 25.5 ms, and the overlapping time step was set to 6 ms. Correspondingly, the size of the one-dimensional convolution kernel L is set to 51 and the stride size is 39 in RadarNet. The number of kernels of a one-dimensional convolutional layer is set to 150, which corresponds to a STFT of 150 points. The enti...

Embodiment 1

[0064] Embodiment 1 Human body sleep action classification:

[0065] Doppler radar was used to collect data on three types of human sleep activity: (a) turning over, (b) hand movements and (c) head movements. The data set is derived from four individuals. Data collection scenarios such as Figure 8 shown. The radar is placed on one side of the bed, at a distance of about 20 cm from the bed, at the same level as the bed. For a more sensitive signal, the radar detection range is roughly located on the chest. The collection time for each action is 3 seconds. The distribution of action data is shown in Table 1. The original signal waveforms of three human sleep movements are as follows: Figure 9 shown.

[0066] Table 1 The categories and groups of human sleep movements

[0067]

[0068] In this experiment, STFTNet is compared with the described RadarNet using 5-fold cross-validation. The total training duration is 5,814 seconds for RadarNet and 4,237 seconds for STFTN...

Embodiment 2

[0074] Embodiment 2 Human body's daily action classification:

[0075] In order to verify the flexibility and robustness of the RadarNet of the present invention, another more complex experiment can be performed to classify the daily actions of the human body. Doppler radar was used to collect data on seven types of human daily activities: (a) running, (b) walking, (c) walking with a gun, (d) crawling, (e) walking boxing, (f) standing boxing, ( g) sit still. These types of actions are the same as those used in the study by Kim et al. [1]. This dataset collects data from 4 individuals. The postures of these seven movements are as follows: Figure 11 shown. The radar is placed at a height of 1.2 meters above the ground. Actions (a), (b), (c), (d) and (e) were collected by the experimenter moving forward from a distance of 5 m to the radar. The remaining two activities were collected at 0.5 meters from the radar. The collection time for each activity of (a), (b), (c), (d) ...

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 discloses a micro-doppler radar human body action classification method of convolutional neural network, which comprises an original data processing process and a deep convolutional neural network, and is characterized in that the output of the original data processing process is connected to the deep convolutional neural network; the original data processing process is a one-dimensional convolution process; wherein the one-dimensional convolution process comprises a first 1D convolution layer representing a real part, a second 1D convolution layer representing an imaginary part,a first merging layer, a second merging layer, a total merging layer, a mapping layer and an active layer; wherein the first merging layer after the first 1D convolution layer is used for calculatingthe square of a real part, and the second merging layer after the second 1D convolution layer is used for calculating the square of an imaginary part; wherein the total merging layer sums the squarevalue of the real part and the square value of the imaginary part, the mapping layer is used for calculating a normalized square root value, and the active layer has an arc tangent function for nonlinear transformation. The input of the network is an original radar signal, the output of the network is an action category, and the end-to-end learning advantage in the neural network is fully exerted.

Description

[technical field] [0001] The invention relates to human body action recognition technology, in particular to a convolutional neural network micro-Doppler radar human action classification method. [Background technique] [0002] Human activity classification can be used in many fields, such as sleep monitoring, elderly care, and anti-terrorism monitoring. Traditional methods for classifying everyday human activities are based on optical cameras. However, the camera surveillance system suffers from two limitations: one is its relatively low accuracy in dark environments, and the other is that it cannot be used to capture human activities behind obstacles such as walls or curtains. To overcome these limitations, many studies have proposed replacing cameras with micro-Doppler radars. [0003] Traditional algorithms for classifying radar signals require feature extraction from the raw signal and then feed the extracted features into one or more classifiers. These algorithms ba...

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
IPC IPC(8): G06K9/62G01S7/41G01S13/58
CPCG01S7/417G01S7/418G01S13/58G06F18/241
Inventor 叶文彬陈海权
Owner SHENZHEN UNIV
Who we serve
  • R&D Engineer
  • R&D Manager
  • IP Professional
Why Patsnap Eureka
  • Industry Leading Data Capabilities
  • Powerful AI technology
  • Patent DNA Extraction
Social media
Patsnap Eureka Blog
Learn More
PatSnap group products

Browse by: Latest US Patents, China's latest patents, Technical Efficacy Thesaurus, Application Domain, Technology Topic, Popular Technical Reports.

© 2024 PatSnap. All rights reserved.Legal|Privacy policy|Modern Slavery Act Transparency Statement|Sitemap|About US| Contact US: help@patsnap.com