CN113422660B - Step number detection method based on wireless signals - Google Patents

Step number detection method based on wireless signals Download PDF

Info

Publication number
CN113422660B
CN113422660B CN202110525054.XA CN202110525054A CN113422660B CN 113422660 B CN113422660 B CN 113422660B CN 202110525054 A CN202110525054 A CN 202110525054A CN 113422660 B CN113422660 B CN 113422660B
Authority
CN
China
Prior art keywords
data
step number
walking
phase
amplitude
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.)
Active
Application number
CN202110525054.XA
Other languages
Chinese (zh)
Other versions
CN113422660A (en
Inventor
卢晓
闫书升
崔玮
李冰
王海霞
张治国
梁慧斌
李学艺
张华宇
聂君
宋诗斌
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Shandong University of Science and Technology
Original Assignee
Shandong University of Science and Technology
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by Shandong University of Science and Technology filed Critical Shandong University of Science and Technology
Priority to CN202110525054.XA priority Critical patent/CN113422660B/en
Publication of CN113422660A publication Critical patent/CN113422660A/en
Application granted granted Critical
Publication of CN113422660B publication Critical patent/CN113422660B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04B—TRANSMISSION
    • H04B17/00—Monitoring; Testing
    • H04B17/30—Monitoring; Testing of propagation channels
    • H04B17/309—Measuring or estimating channel quality parameters
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
    • G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F2218/00—Aspects of pattern recognition specially adapted for signal processing
    • G06F2218/02—Preprocessing
    • G06F2218/04—Denoising

Abstract

The invention discloses a wireless signal-based step number detection method, which comprises the steps of experimental environment deployment, data acquisition, data preprocessing, walking step number detection model establishment, model training and step number detection by utilizing a walking step number detection model. The invention utilizes the principle that the wireless signal can generate signal change along with the walking of people in the transmission process, and establishes the fingerprint database by extracting the amplitude and phase characteristics of CSI in the subcarrier of the wireless signal. And (3) carrying out noise reduction processing on amplitude and phase signals in the database, calibrating the phase signals, fusing the amplitude and the phase, converting data with different sizes generated in different periods into a format with the same size, and finally inputting the data into a predrnn network to train to obtain a final model. And collecting new walking data, inputting the new walking data into the trained model, comparing the predicted step number result with the actual step number in the experiment, and verifying the effect of the model. The experimental result shows that the method can detect the number of steps of the person walking indoors with high precision under the condition that other equipment is not needed.

Description

Step number detection method based on wireless signals
Technical Field
The invention relates to the field of step number detection, in particular to a step number detection method based on wireless signals.
Background
Walking is one of the most simple human activities that are fundamental. In recent years, with the popularization of smart devices (e.g., smart band, smart phone), walking step number detection is becoming more popular. There are two main methods for detecting and researching the number of steps, one is active detection, and the other is passive detection. The active detection is characterized in that a detected person needs to wear equipment or detect within a specified range, and the method mainly comprises the steps of based on a camera and based on a portable sensor. The passive detection has the characteristics that the passive detection does not need to be matched by a detected person deliberately, and cannot be restricted by a sight line environment, and the main methods are based on Bluetooth (Bluetooth), Ultra Wide Band (UWB), Infrared (infra) and wireless network (WIFI). In passive detection, a WIFI-based mode is different from others, and WIFI devices are widely applied to homes, offices and other public places along with the development of networks. And the WIFI signal has a wider infrared transmission range, is more accurate than Bluetooth, has low cost and a wide popularization range, so that many researchers research how to detect the step number by using the WIFI signal.
Through research in recent years, the WIFI signal is used for step number detection, which is different from a traditional acceleration sensing-based mode (such as a smart band). The acceleration sensor mode records the walking steps according to the change times of the speed of a person wearing the acceleration sensor when walking, and WIFI signals are signals which fluctuate when the person walks in an area covered by wireless signals, and Channel State Information (CSI) in the wireless signals is extracted. And analyzing the walking steps according to the change of the CSI. It is conceivable that the WIFI signal is easily affected by other factors such as arm swing, and the like, and these factors are difficult to be completely processed, so that the recognition accuracy is not good. But the accuracy is still acceptable if it is handled reasonably. At present, WIFI is utilized to summarize step number detection research and the defects exist:
WiWho is a detection method based on machine learning, considers that human walking gaits are the same, divides each step of a CSI signal by using a time window mode, and trains a decision forest classifier to detect the step number. The disadvantage of this method is that in reality the human gait is not fixed. Moreover, even if the gait is the same, since the wireless signal is affected differently depending on the position, each divided step is not necessarily accurate.
Wi-Rnn firstly processes the collected CSI signal, separates out an effective signal and calculates the step number by using a peak value detection method. The method has the disadvantages that errors are easy to exist in the separation of effective signals, and the accuracy of the step number is influenced by the selection of the peak value.
Disclosure of Invention
Aiming at the defects of the existing step number detection method, the invention provides a step number detection method based on wireless signals, which can more accurately realize step number detection.
The invention adopts the following technical scheme:
a step number detection method based on wireless signals comprises the following steps:
step 1: deploying an experimental environment;
selecting a laboratory as an experimental environment, deploying a wireless router in the laboratory as a transmitter of a wireless signal, and deploying a computer as a wireless signal receiver;
step 2: collecting data;
setting three data acquisition periods which are respectively T1, T2 and T3, wherein an experimenter sits on a stool before the acquisition starts, the experimenter stands up from the stool after the acquisition starts and enters the data acquisition period, and the experimenter walks in a normal walking mode by a designated step number p, and each step number acquires n groups of data; wherein, in the T1 data acquisition cycle, the acquired step number is p1Each datum is recorded as
Figure BDA0003065496950000021
In the T2 data acquisition period, the acquired step number is p2Each data is recorded as
Figure BDA0003065496950000022
In the T3 data acquisition period, the acquired step number is p3Each datum is recorded as
Figure BDA0003065496950000023
Building a fingerprint database
Figure BDA0003065496950000024
And step 3: preprocessing data;
step 3.1: extracting amplitude and phase from the fingerprint database S;
step 3.2: carrying out noise reduction processing on the obtained amplitude and phase signals by utilizing discrete wavelet transform;
discrete wavelet transformation is carried out on the amplitude and the phase by using a second-order Daubechies wavelet basis, wavelet expansion coefficients d (k) with different scales are obtained through decomposition, the wavelet expansion coefficients of k layers are reconstructed to obtain corresponding approximate signals a and detail signals d, and then the approximate signals a and the detail signals d are combined to obtain final de-noised amplitude and phase signals;
step 3.3: performing linear phase calibration on the phase subjected to noise reduction;
step 3.4: the amplitude after the noise reduction treatment is represented in a two-dimensional matrix form, a two-dimensional matrix with the same dimension as the amplitude is also generated after the phase calibration, and a new two-dimensional matrix is formed through data fusion;
and 4, step 4: establishing a walking step number detection model and training;
and 5: and detecting the step number by using a walking step number detection model.
Preferably, step 1 is specifically:
the size of the laboratory is 10m multiplied by 15m, and the model of the wireless router is commercial version TP-LINK WDR 6500;
the distance between the computer and the router is 9m, a Ubutu12.04LTS operating system is installed on the computer, an Intel 5300 network card is installed on the computer, the computer is externally connected with three antennas, and the ground height of each antenna is 1.2 m;
and configuring a kernel driver and a wireless network card of the computer to stabilize a 100MHZ signal transmission mode between the router and the computer.
Preferably, step 2 is specifically: t1 ═ 10s, T2 ═ 15s, T3 ═ 20 s;
wherein p is1Taking the value from 0 to 6 steps, acquiring n groups of data in each step, wherein n is 50, acquiring 7 multiplied by 50 data in the T1 data acquisition period, and recording each data as
Figure BDA0003065496950000025
Wherein p is2Taking the value from 0 to 11 steps, acquiring n groups of data in each step, wherein n is 50, acquiring 12 multiplied by 50 data in the T2 data acquisition period, and recording each data as
Figure BDA0003065496950000031
Wherein p is3Taking the value from 0 to 22 steps, acquiring n groups of data in each step, wherein n is 50, acquiring 23 multiplied by 50 data in the T3 data acquisition period, and recording each data as
Figure BDA0003065496950000032
Because the set frequency is 100MHz, the data format of each data collected in the T1 data collection period is [1000,30], wherein 30 is the number of channels, and 1000 is the number collected on each channel;
similarly, the data format of each data collected in the T2 data collection period is [1500,30 ]; the data format of each data collected during the T3 data collection cycle is [2000,30]
So as to construct fingerprint database
Figure BDA0003065496950000033
Preferably, the format of each datum in the new two-dimensional matrix after data fusion in step 3.4 is [500 × 30 ].
Preferably, step 4 specifically includes: selecting a Predrnn network as a feature extraction network;
the Predrnn network at time t is described as:
Figure BDA0003065496950000034
Figure BDA0003065496950000035
wherein, WoIs a weight vector, boIs a vector of the offset to the offset,
Figure BDA0003065496950000036
for the output of the l-th layer hidden unit at time t, XtIs an input to the computer system,
Figure BDA0003065496950000037
Figure BDA0003065496950000038
hidden information of the first layer unit at the time t-1, i is the number of layers, t is the time step, sigma is the data matrix division calculation, W1×1Represents a 1 × 1 dimensional convolution operation;
the features extracted by Predrnn are input to the fully-linked layer, which is described as:
hl+1＝g(Wlhl+bl)，l＝1,2,3...
wherein l is the number of layers, hlIs the input of the full connection layer, WlAnd blIs the weight vector and offset vector of the fully connected layer, hl+1Is the output of the full link layer;
after the walking step number detection model is built, a new two-dimensional matrix is used as input to train the walking step number detection model, a prediction result is obtained through feedforward operation and is matched with a real step number label to calculate errors, an ADAM function is selected by an optimizer, parameters are continuously adjusted in the training process, and the final walking step number detection model is established.
Preferably, step 5 specifically comprises:
collecting new test step number data to construct test fingerprint database SnewPreprocessing the data in the test fingerprint database to obtain a new walking step number test characteristic matrix, transmitting the walking step number test characteristic matrix into a trained walking step number detection model, and directly obtaining the walking step number of the test dataThe number of steps.
The step number detection method based on the wireless signals provided by the invention utilizes the principle that the wireless signals change along with the walking of people in the transmission process, and establishes the fingerprint database by extracting the amplitude and phase characteristics of CSI in subcarriers. And (3) carrying out noise reduction on the amplitude and phase signals in the database, calibrating the phase signals, fusing the amplitude and the phase, converting data with different sizes generated in different periods into a format with the same size, and finally inputting the data into a space-time prediction model predrnn to train to obtain a final model. And collecting new walking data, inputting the new walking data into the trained model, comparing the predicted step number result with the actual step number in the experiment, and verifying the effect of the model. The experimental result shows that the method provided by the invention can detect the number of steps of people walking indoors with high precision without wearing other equipment.
Drawings
FIG. 1 is a schematic diagram of the experimental environment of the present invention.
Fig. 2 is a schematic diagram of a system for wireless walking step number detection.
Fig. 3 is a structure of a step number detection model.
Fig. 4 is a waveform diagram of an unprocessed raw amplitude signal.
Fig. 5 is a waveform diagram of a wavelet-filtered amplitude signal.
Fig. 6 is a waveform diagram of an uncorrected raw phase signal.
Fig. 7 is a waveform diagram of a phase signal after linear correction.
FIG. 8 is a predrnn internal data transfer diagram.
Fig. 9 is a graph of the test 900 step identification error profiles.
FIG. 10 is a graph of the recognition error probability for 900 steps tested.
Detailed Description
The following description of the embodiments of the present invention will be made with reference to the accompanying drawings:
with reference to fig. 1 to 10, a method for detecting steps based on wireless signals includes the following steps:
step 1: deploying an experimental environment;
a laboratory was selected as the experimental environment, the size of the laboratory being 10m 15 m. Deploying a wireless router in a laboratory, wherein the wireless router is used as a transmitter of a wireless signal, and the model of the wireless router is TP-LINK WDR 6500; a computer is deployed as a wireless signal receiver; the distance between the computer and the router is 9m, a Ubutu12.04LTS operating system, an Intel 5300 network card and three external antennas are installed on the computer, so that the intensity of received wireless signals is guaranteed, and the ground height of the antennas is 1.2 m.
And configuring a kernel driver and a wireless network card of the computer to stabilize a 100MHZ signal transmission mode between the router and the computer.
Step 2: collecting data;
setting three data acquisition periods which are respectively T1, T2 and T3, wherein an experimenter sits on a stool before the acquisition starts, the experimenter stands up from the stool after the acquisition starts and enters the data acquisition period, and the experimenter walks in a normal walking mode by a designated step number p, and each step number acquires n groups of data; wherein, in the T1 data acquisition cycle, the acquired step number is p1Each data is recorded as
Figure BDA0003065496950000041
In the T2 data acquisition period, the acquired step number is p2Each data is recorded as
Figure BDA0003065496950000042
In the T3 data acquisition period, the acquired step number is p3Each data is recorded as
Figure BDA0003065496950000051
Building a fingerprint database
Figure BDA0003065496950000052
In one embodiment, T1 ═ 10s, T2 ═ 15s, T3 ═ 20 s;
wherein p is1Taking the value from 0 to 6 steps, acquiring n groups of data in each step, wherein n is 50, and then acquiring 7 multiplied by 50 data in the T1 data acquisition period, wherein each dataIs marked as
Figure BDA0003065496950000053
Wherein p is2Taking the value from 0 to 11 steps, acquiring n groups of data in each step, wherein n is 50, acquiring 12 multiplied by 50 data in the T2 data acquisition period, and recording each data as
Figure BDA0003065496950000054
Wherein p is3Taking the value from 0 to 22 steps, acquiring n groups of data in each step, and if n is 50, acquiring 23 multiplied by 50 data in the T3 data acquisition period, and recording each data as
Figure BDA0003065496950000055
Because the set frequency is 100MHz, the data format of each data acquired in the T1 data acquisition cycle is [1000,30], where 30 is the number of channels and 1000 is the number acquired on each channel;
similarly, the data format of each data collected in the T2 data collection period is [1500,30 ]; the data format of each data collected in the T3 data collection period is [2000,30]
So as to construct fingerprint database
Figure BDA0003065496950000056
And step 3: preprocessing data;
step 3.1: extracting amplitude and phase from data in a fingerprint database S;
step 3.2: carrying out noise reduction processing on the obtained amplitude and phase signals by using Discrete Wavelet Transform (DWT);
discrete wavelet transformation is carried out on the amplitude and the phase by using a second-order Daubechies wavelet basis, wavelet expansion coefficients d (k) with different scales are obtained through decomposition, the wavelet expansion coefficients of k layers are reconstructed to obtain corresponding approximate signals a and detail signals d, and then the approximate signals a and the detail signals d are combined to obtain final de-noised amplitude and phase signals;
step 3.3: performing linear phase calibration on the phase subjected to noise reduction;
the phase obtained after the processing of step 3.2 cannot be used directly, because the obtained phase information is confused due to hardware problems of commercial router devices, such as frequency offset, time delay, and the like, which cause errors. A phase calibration is required, here in a linear calibration.
The actual measured phase is expressed as:
Figure BDA0003065496950000057
in the formula
Figure BDA0003065496950000058
Expressed as the measured phase,. phi.represents the true phase, and m is expressed as the index ([ -58, -57,. cndot.,. 0,1,. cndot.,. 57,58 ·)]) And N is expressed as the number of frequency samples,
Figure BDA0003065496950000059
the clock offset is represented, the beta is represented by unknown phase offset, the Z is noise, and the noise is small and negligible in actual acquisition.
The linearity correction is performed by making α denote the slope of the reception phase:
Figure BDA0003065496950000061
let b denote the offset:
Figure BDA0003065496950000062
since the indices of the subcarriers are symmetric, therefore
Figure BDA0003065496950000063
Then b is represented as
Figure BDA0003065496950000064
In practice, the unknown elements exist
Figure BDA0003065496950000065
β makes it impossible to obtain a true phase, however:
Figure BDA0003065496950000066
can indicate the corrected phase of the jth sub-carrier and can eliminate
Figure BDA0003065496950000067
β。
Step 3.4: the amplitude after the noise reduction treatment is represented in a two-dimensional matrix form, a two-dimensional matrix with the same dimension as the amplitude is also generated after the phase calibration, and a new two-dimensional matrix is formed through data fusion;
because the shapes of the data in different periods are different, namely [1000 × 30], [1500 × 30], [2000 × 30], the shapes are also different after fusion, the data in different periods are converted into the data with the same shape and size, and the format of each data in a new two-dimensional matrix after data fusion is [500 × 30], and the data are used as input.
And 4, step 4: establishing a walking step number detection model and training;
predrnn is a network proposed in an article published on top meeting NIPS. Compared with the traditional RNN, the RNN has the advantages that spatial information can be lost in training due to the structural defects of the RNN, and meanwhile, the RNN cannot perform parallel operation and is long in training time. Compared with the traditional CNN network, the CNN emphasizes the spatial features, resulting in missing information in the time dimension. Step number detection is a series of continuous actions, including characteristics of two fields of time and space. Therefore, the characteristics are extracted from time and space by using Predrnn, the obtained characteristic information is richer, and the prediction effect is more obvious.
the Predrnn network at time t is described as:
Figure BDA0003065496950000068
Figure BDA0003065496950000069
wherein, WoIs a weight vector, boIs a vector of the offset to the offset,
Figure BDA00030654969500000612
for the output of the l-th layer hidden unit at time t, XtIs an input to the computer system that is,
Figure BDA00030654969500000610
Figure BDA00030654969500000611
hidden information of the first layer unit at the time t-1, i is the number of layers, t is the time step, sigma is the data matrix division calculation, W1×1Represents a 1 × 1 dimensional convolution operation;
the features extracted by Predrnn are input to the fully-linked layer, which is described as:
hl+1＝g(Wlhl+bl)，l＝1，2，3...
wherein l is the number of layers, hlIs the input of the full connection layer, WlAnd blIs the weight vector and offset vector of the fully connected layer, hl+1Is the output of the full link layer;
after the walking step number detection model is built, a new two-dimensional matrix is used as input to train the walking step number detection model, a prediction result is obtained through feedforward operation and is matched with a real step number label to calculate errors, an ADAM function is selected by an optimizer, parameters are continuously adjusted in the training process, and the final walking step number detection model is established.
And 5: and detecting the step number by using a walking step number detection model.
Collecting new test step number data to construct test fingerprint database SnewTo test fingerprint database dataAnd preprocessing the data to obtain a new walking step number test characteristic matrix, and transmitting the walking step number test characteristic matrix into the trained walking step number detection model to directly obtain the walking step number of the test data.
It is to be understood that the above description is not intended to limit the present invention, and the present invention is not limited to the above examples, and those skilled in the art may make modifications, alterations, additions or substitutions within the spirit and scope of the present invention.

Claims (5)

1. A wireless signal-based step number detection method is characterized by comprising the following steps:
step 1: deploying an experimental environment;
selecting a laboratory as an experimental environment, deploying a wireless router in the laboratory as a transmitter of a wireless signal, and deploying a computer as a wireless signal receiver;
and 2, step: collecting data;
setting three data acquisition periods which are respectively T1, T2 and T3, wherein an experimenter sits on a stool before the acquisition starts, the experimenter stands up from the stool after the acquisition starts and enters the data acquisition period, and the experimenter walks in a normal walking mode by a designated step number p, and each step number acquires n groups of data; wherein, in the T1 data acquisition cycle, the acquired step number is p1Each data is recorded as
Figure FDA0003596509730000011
In the T2 data acquisition period, the acquired step number is p2Each data is recorded as
Figure FDA0003596509730000012
In the T3 data acquisition period, the acquired step number is p3Each data is recorded as
Figure FDA0003596509730000013
Building a fingerprint database
Figure FDA0003596509730000014
And 3, step 3: preprocessing data;
step 3.1: extracting amplitude and phase from the fingerprint database S;
step 3.2: carrying out noise reduction processing on the obtained amplitude and phase signals by utilizing discrete wavelet transform;
discrete wavelet transformation is carried out on the amplitude and the phase by using a second-order Daubechies wavelet basis, wavelet expansion coefficients d (k) with different scales are obtained through decomposition, the wavelet expansion coefficients of the k layer are reconstructed, a corresponding approximate signal a and a detail signal d are obtained, and then the approximate signal a and the detail signal d are combined to obtain a final de-noised amplitude signal and a final de-noised phase signal;
step 3.3: performing linear phase calibration on the phase subjected to noise reduction;
step 3.4: the amplitude after the noise reduction treatment is represented in a two-dimensional matrix form, a two-dimensional matrix with the same dimension as the amplitude is also generated after the phase calibration, and a new two-dimensional matrix is formed through data fusion;
and 4, step 4: establishing a walking step number detection model and training;
the step 4 specifically comprises the following steps: selecting a Predrnn network as a feature extraction network;
the Predrnn network at time t is described as:
Figure FDA0003596509730000015
Figure FDA0003596509730000016
wherein, WoIs a weight vector, boIs a vector of the offset to the offset,
Figure FDA0003596509730000017
for the output of the l-th layer hidden unit at time t, XtIs an input to the computer system that is,
Figure FDA0003596509730000018
Figure FDA0003596509730000019
hidden information of the first layer unit at the time t-1, i is the number of layers, t is the time value, sigma is the data matrix division calculation, W1×1Represents a 1 × 1 dimensional convolution operation;
the features extracted by Predrnn are input to the fully-linked layer, which is described as:
hl+1＝g(Wlhl+bl)，l＝1,2,3…
wherein l is the number of layers, hlIs the input of the full connection layer, WlAnd blIs the weight vector and offset vector of the fully connected layer, hl+1Is the output of the full link layer;
after a walking step number detection model is built, a new two-dimensional matrix is used as input to train the walking step number detection model, a prediction result is obtained through feed-forward operation and is matched with a real step number label to calculate errors, an ADAM function is selected by an optimizer, parameters are continuously adjusted in the training process, and a final walking step number detection model is established;
and 5: and detecting the step number by using a walking step number detection model.
2. The method for detecting the number of steps based on the wireless signal according to claim 1, wherein the step 1 specifically comprises:
the size of the laboratory is 10m multiplied by 15m, and the model of the wireless router is commercial version TP-LINK WDR 6500;
the distance between the computer and the router is 9m, a Ubutu12.04LTS operating system is installed on the computer, an Intel 5300 network card is installed on the computer, the computer is externally connected with three antennas, and the ground height of each antenna is 1.2 m;
and configuring a kernel driver and a wireless network card of the computer to stabilize a 100MHZ signal transmission mode between the router and the computer.
3. The method for detecting the number of steps based on the wireless signal according to claim 1, wherein the step 2 is specifically: t1 ═ 10s, T2 ═ 15s, T3 ═ 20 s;
wherein p is1Taking the value from 0 to 6 steps, acquiring n groups of data in each step, wherein n is 50, acquiring 7 multiplied by 50 data in the T1 data acquisition period, and recording each data as
Figure FDA0003596509730000021
Wherein p is2Taking the value from 0 to 11 steps, acquiring n groups of data in each step, wherein n is 50, acquiring 12 multiplied by 50 data in the T2 data acquisition period, and recording each data as
Figure FDA0003596509730000022
Wherein p is3Taking the value from 0 to 22 steps, acquiring n groups of data in each step, wherein n is 50, acquiring 23 multiplied by 50 data in the T3 data acquisition period, and recording each data as
Figure FDA0003596509730000023
Because the set frequency is 100MHz, the data format of each data collected in the T1 data collection period is [1000,30], wherein 30 is the number of channels, and 1000 is the number collected on each channel;
similarly, the data format of each data collected in the T2 data collection period is [1500,30 ]; the data format of each data collected in the T3 data collection period is [2000,30]
So as to construct fingerprint database
Figure FDA0003596509730000024
4. The method of claim 1, wherein the format of each data in the new two-dimensional matrix after data fusion in step 3.4 is [500,30 ].
5. The method for detecting the number of steps based on the wireless signal according to claim 1, wherein the step 5 specifically comprises:
collecting new test step number data to construct test fingerprint database SnewAnd preprocessing the data in the test fingerprint database to obtain a new walking step number test characteristic matrix, and transmitting the walking step number test characteristic matrix into a trained walking step number detection model to directly obtain the walking step number of the test data.
CN202110525054.XA 2021-05-14 2021-05-14 Step number detection method based on wireless signals Active CN113422660B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202110525054.XA CN113422660B (en) 2021-05-14 2021-05-14 Step number detection method based on wireless signals

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202110525054.XA CN113422660B (en) 2021-05-14 2021-05-14 Step number detection method based on wireless signals

Publications (2)

Publication Number Publication Date
CN113422660A CN113422660A (en) 2021-09-21
CN113422660B true CN113422660B (en) 2022-07-19

Family

ID=77712401

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202110525054.XA Active CN113422660B (en) 2021-05-14 2021-05-14 Step number detection method based on wireless signals

Country Status (1)

Country Link
CN (1) CN113422660B (en)

Citations (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN110260883A (en) * 2019-07-22 2019-09-20 内蒙古大学 The energy saving step-recording method and pedometer of intelligent electronic device
CN112508819A (en) * 2020-12-17 2021-03-16 广州大学 Wrapping phase denoising method and system based on convolutional neural network and storage medium

Family Cites Families (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
AUPO951397A0 (en) * 1997-09-29 1997-10-23 Canon Information Systems Research Australia Pty Ltd A method for digital data compression
WO2017100706A1 (en) * 2015-12-09 2017-06-15 Origin Wireless, Inc. Method, apparatus, and systems for wireless event detection and monitoring
CN106802404B (en) * 2017-01-18 2018-10-16 深圳大学 A kind of human body indoor positioning automatic testing method and system
CN112147573A (en) * 2020-09-14 2020-12-29 山东科技大学 Passive positioning method based on amplitude and phase information of CSI (channel State information)
CN112423265B (en) * 2020-11-13 2022-04-12 武汉理工大学 CSI-based dual-cycle neural network shipborne environment indoor positioning method

Patent Citations (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN110260883A (en) * 2019-07-22 2019-09-20 内蒙古大学 The energy saving step-recording method and pedometer of intelligent electronic device
CN112508819A (en) * 2020-12-17 2021-03-16 广州大学 Wrapping phase denoising method and system based on convolutional neural network and storage medium

Also Published As

Publication number Publication date
CN113422660A (en) 2021-09-21

Similar Documents

Publication Publication Date Title
Hsieh et al. Deep learning-based indoor localization using received signal strength and channel state information
CN108629380B (en) Cross-scene wireless signal sensing method based on transfer learning
Khan et al. A deep learning framework using passive WiFi sensing for respiration monitoring
CN112036433B (en) CNN-based Wi-Move behavior sensing method
CN112147573A (en) Passive positioning method based on amplitude and phase information of CSI (channel State information)
CN108171278B (en) Motion pattern recognition method and system based on motion training data
CN110363246B (en) Fusion method of vegetation index NDVI with high space-time resolution
CN109444813A (en) A kind of RFID indoor orientation method based on BP and DNN amphineura network
CN114423034B (en) Indoor personnel action recognition method, system, medium, equipment and terminal
Xia et al. Person identification with millimeter-wave radar in realistic smart home scenarios
CN109658655A (en) A kind of passive intrusion detection method in interior based on wireless signal
CN114781463A (en) Cross-scene robust indoor tumble wireless detection method and related equipment
Zhang et al. CSI-based location-independent human activity recognition using feature fusion
CN113422660B (en) Step number detection method based on wireless signals
CN114818788A (en) Tracking target state identification method and device based on millimeter wave perception
Moshiri et al. CSI-based human activity recognition using convolutional neural networks
CN113743374A (en) Personnel identity identification method based on channel state information respiration perception
CN114757237B (en) Speed-independent gait recognition method based on WiFi signals
WO2018053912A1 (en) Method for real-time action recognition, and related bracelet and computing device
CN114639169B (en) Human motion recognition system based on attention mechanism feature fusion and irrelevant to position
CN112183586B (en) Human body posture radio frequency identification method for online multitask learning
CN110852266A (en) Gait feature extraction method based on wireless signals
CN112883905B (en) Human behavior recognition method based on coarse-grained time-frequency features and multi-layer fusion learning
CN115567871A (en) WiFi fingerprint indoor floor identification and position estimation method
CN113642457B (en) Cross-scene human body action recognition method based on antagonistic meta-learning

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
GR01 Patent grant
GR01 Patent grant