CN116264012A - Contactless falling direction detection method for short message alarm - Google Patents
Contactless falling direction detection method for short message alarm Download PDFInfo
- Publication number
- CN116264012A CN116264012A CN202211325313.5A CN202211325313A CN116264012A CN 116264012 A CN116264012 A CN 116264012A CN 202211325313 A CN202211325313 A CN 202211325313A CN 116264012 A CN116264012 A CN 116264012A
- Authority
- CN
- China
- Prior art keywords
- time
- short message
- falling
- message alarm
- spectrogram
- 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
- 238000001514 detection method Methods 0.000 title claims abstract description 23
- 238000000034 method Methods 0.000 claims abstract description 21
- 238000013527 convolutional neural network Methods 0.000 claims abstract description 14
- 238000000605 extraction Methods 0.000 claims abstract description 8
- 238000011176 pooling Methods 0.000 claims abstract description 8
- 238000007781 pre-processing Methods 0.000 claims abstract description 6
- 230000009471 action Effects 0.000 claims description 50
- 239000011159 matrix material Substances 0.000 claims description 10
- 238000013145 classification model Methods 0.000 claims description 7
- 230000006870 function Effects 0.000 claims description 5
- 238000005070 sampling Methods 0.000 claims description 5
- 230000003068 static effect Effects 0.000 claims description 5
- 238000001914 filtration Methods 0.000 claims description 3
- 238000013135 deep learning Methods 0.000 claims description 2
- 239000003651 drinking water Substances 0.000 claims 1
- 235000020188 drinking water Nutrition 0.000 claims 1
- 238000012163 sequencing technique Methods 0.000 claims 1
- 208000027418 Wounds and injury Diseases 0.000 abstract description 3
- 230000006378 damage Effects 0.000 abstract description 3
- 208000014674 injury Diseases 0.000 abstract description 3
- 230000000875 corresponding effect Effects 0.000 description 16
- 238000012549 training Methods 0.000 description 6
- 238000004364 calculation method Methods 0.000 description 5
- 230000008859 change Effects 0.000 description 3
- 238000010586 diagram Methods 0.000 description 2
- 230000035622 drinking Effects 0.000 description 2
- 230000004927 fusion Effects 0.000 description 2
- 230000036541 health Effects 0.000 description 2
- 230000008569 process Effects 0.000 description 2
- 238000012360 testing method Methods 0.000 description 2
- 238000012795 verification Methods 0.000 description 2
- XLYOFNOQVPJJNP-UHFFFAOYSA-N water Substances O XLYOFNOQVPJJNP-UHFFFAOYSA-N 0.000 description 2
- 238000009825 accumulation Methods 0.000 description 1
- 238000013528 artificial neural network Methods 0.000 description 1
- 230000006399 behavior Effects 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 230000005802 health problem Effects 0.000 description 1
- 230000004630 mental health Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000000149 penetrating effect Effects 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V40/00—Recognition of biometric, human-related or animal-related patterns in image or video data
- G06V40/10—Human or animal bodies, e.g. vehicle occupants or pedestrians; Body parts, e.g. hands
- G06V40/103—Static body considered as a whole, e.g. static pedestrian or occupant recognition
-
- 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
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/40—Extraction of image or video features
- G06V10/44—Local feature extraction by analysis of parts of the pattern, e.g. by detecting edges, contours, loops, corners, strokes or intersections; Connectivity analysis, e.g. of connected components
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/82—Arrangements for image or video recognition or understanding using pattern recognition or machine learning using neural networks
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02D—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN INFORMATION AND COMMUNICATION TECHNOLOGIES [ICT], I.E. INFORMATION AND COMMUNICATION TECHNOLOGIES AIMING AT THE REDUCTION OF THEIR OWN ENERGY USE
- Y02D30/00—Reducing energy consumption in communication networks
- Y02D30/70—Reducing energy consumption in communication networks in wireless communication networks
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Evolutionary Computation (AREA)
- Multimedia (AREA)
- General Health & Medical Sciences (AREA)
- Software Systems (AREA)
- Health & Medical Sciences (AREA)
- Artificial Intelligence (AREA)
- Computing Systems (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Life Sciences & Earth Sciences (AREA)
- Medical Informatics (AREA)
- Human Computer Interaction (AREA)
- Databases & Information Systems (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Data Mining & Analysis (AREA)
- Molecular Biology (AREA)
- General Engineering & Computer Science (AREA)
- Mathematical Physics (AREA)
- Emergency Alarm Devices (AREA)
- Radar Systems Or Details Thereof (AREA)
Abstract
The invention discloses a non-contact falling direction detection method for a short message alarm, which comprises the following steps: different human motion information is acquired by using a millimeter wave radar sensor; preprocessing radar original data to obtain a time-distance graph and a time-Doppler graph, and denoising the time-distance graph and the time-Doppler graph by adopting an image edge contour extraction algorithm; performing feature extraction on the denoised time-distance graph and time-Doppler graph by using a double-channel convolutional neural network, then fusing to obtain new features, and performing convolutional pooling and full-connection layer on the new features to obtain corresponding Data values; when the person to be tested falls, the short message alarm gives an alarm according to the corresponding Data value and provides falling direction information; the method can better protect the privacy of people, provide more detailed falling direction information, help medical staff to quickly locate the injured part and effectively prevent secondary injury during treatment.
Description
Technical Field
The invention relates to the field of human body action recognition, in particular to a non-contact falling direction detection method for a short message alarm.
Technical Field
With the increasing number of the old people in the empty nest in China, the health problem of the old people becomes a hot topic of social concern increasingly. In a home environment, accidental falls of the old seriously affect physical and mental health. Moreover, when the head of the old is injured, most of the old is damaged in a recessive way, the body of the old can be turned unconsciously after the old falls down to the ground, so that the falling direction of the old can not be judged when a therapist arrives at the site, the injured part can not be determined rapidly, and secondary injury can be caused in the process of delivering and treating. Therefore, how to quickly and accurately detect the falling behavior in real time and provide key falling direction information for medical staff is a key problem to be solved in the health care and health management of the old, and has positive significance in building safe and intelligent home life.
Currently, fall detection techniques can be divided into three categories: detection technology based on wearable equipment, detection technology based on computer vision and detection technology based on millimeter wave radar. Fall detection by wearable devices generally refers to detection of human body posture movement signals by wearing accelerometers, pressure sensors, and the like. However, the detection person is required to wear the equipment all the time, and the method has the problems of wearing comfort and the like, and has certain limitations. The fall detection technology based on computer vision is most popular, simple and convenient, has high detection accuracy, is unfavorable for privacy protection, and has poor detection effect in a low-light environment. The fall detection technology based on the millimeter wave radar not only can protect the privacy of a user, but also has no strict limit on the environment, and along with the development of advanced learning in the past, the accuracy is higher and higher, and the fall detection technology can be applied to different scenes such as bathroom scenes, public toilets, malls and the like. The existing fall detection technology of the millimeter wave radar is used for detecting whether a human body falls down to obtain a final result, and does not provide detailed fall direction information for subsequent rescue.
Disclosure of Invention
The invention aims to: in order to overcome the defects in the prior art, a non-contact falling direction detection method for a short message alarm is provided, so that whether a person to be detected falls down in a room can be accurately detected under an environment with weak light and falling direction information is provided.
The technical scheme is as follows: in order to achieve the above object, the present invention provides a method for detecting a contactless falling direction for a short message alarm, comprising the following steps:
step 1: acquiring different action information of a tested person in a tested space through a millimeter wave radar, obtaining ADC sampling data, and extracting receiving antenna data;
step 2: preprocessing the acquired radar original signals, and obtaining corresponding time-distance spectrograms and time-Doppler spectrograms through speed dimension Fourier transform and time dimension Fourier transform and through time accumulation;
step 3: denoising the obtained time-distance spectrogram and the time-Doppler spectrogram by using an image edge contour extraction algorithm to obtain a new feature map;
step 4: establishing a two-channel convolutional neural network model, respectively inputting the denoised time-distance spectrogram and the denoised time-Doppler spectrogram obtained in the step 3 into the two-channel convolutional neural network, predicting action information through the trained human action classification model, and outputting a Data value corresponding to the action type;
step 5: the upper computer sends the obtained Data value to a short message alarm, and the short message alarm realizes a corresponding function according to the Data value;
further, the types of human body actions in the step 1 are designed into 6 types, namely, falling forwards, falling backwards, squatting, walking, standing up and sitting down, and are completed by 6 different volunteers. The method for collecting the human body action data comprises the following steps: only one volunteer is in the tested space, so that the millimeter wave radar is separated from the volunteer by a horizontal distance of 3m, the height of the millimeter wave radar is the same as that of the chest of the volunteer, and the 5 volunteers respectively complete 6×100 actions and total 5×6×100 action data. The designed forward fall motion can be divided into: the person to be tested falls forward when approaching the radar direction and falls forward when being far away from the radar direction; the designed fall-back actions can be divided into: the person to be measured falls backward when approaching the radar direction and falls backward when keeping away from the radar direction.
Further, the method for preprocessing the radar raw data in the step 2 is as follows: firstly, filtering static clutter of radar echo signals in a human body action radar echo signal data set by using an MTI filter, then carrying out fast Fourier transform on the data to obtain distance distribution information, accumulating the distance distribution information into a time-distance spectrogram along with time, and finally obtaining a time-Doppler spectrogram by using short-time Fourier transform on the time-distance distribution matrix.
Further, the method for denoising the time-distance spectrogram and the time-doppler spectrogram in the step 3 by using an image edge contour extraction algorithm comprises the following steps: finding out and defining the high power density region, sorting the points with power density greater than the threshold value to the high power density region, threshold value P th The calculation formula of (2) is as follows:
where a.epsilon.0, 1 is a predefined parameter, max { P }, andthe maximum and average power density values for all points in the two feature maps are chosen to represent the boundaries of the high power density region at each time point, the top point at the top and the bottom point at the bottom.
Further, the method for establishing the two-channel neural network in the step 4 is as follows: constructing a convolutional neural network containing two channels by using a Pytorch deep learning frame, taking a denoised time-distance spectrogram and a denoised time-Doppler spectrogram as input, merging the two extracted features after pooling the two features by a convolutional layer, retaining the time-distance feature and the time-Doppler feature simultaneously by the new feature after the merging, and obtaining a human body action classification result at an output layer through calculation of the convolutional layer, the pooling layer and a full-connection layer. The training method of the human body action classification model comprises the following steps: constructing two data sets by using the denoised time-distance spectrogram and the denoised time-Doppler spectrogram, and combining the two data sets according to a ratio of 6:2:2, dividing the training set, the verification set and the test set according to the proportion, training and storing the best model.
Further, the step 5 specifically includes: the millimeter wave radar sensor is connected with the short message alarm through Matlab, the millimeter wave radar collects action information of a person to be tested and carries out action recognition by the upper computer, the upper computer sends a Data value corresponding to falling, when the Data value corresponds to falling action, the short message alarm sends corresponding falling direction information to a guardian and gives an alarm reminding, when the upper computer sends other Data values, the short message alarm also sends state information of the person to be tested to the guardian, but does not give an alarm at the moment, and the guardian can conveniently inquire the state information of the person to be tested at any time.
The invention adopts the millimeter wave radar sensor to collect human body action information, the trained action classification model is called by the upper computer to identify human body actions and is transmitted to a Data value of the short message alarm, and the short message alarm realizes corresponding functions.
According to the invention, the millimeter wave radar is utilized to collect human motion information, on one hand, the characteristics of the millimeter wave radar in all weather and all weather determine that the millimeter wave radar can work normally under the conditions of weak light and even no light; on the other hand, the millimeter wave radar has the characteristics of strong penetrating power and the like, and the invention can still continue to work when interference exists and equipment is shielded by obstacles.
The invention is realized by a millimeter wave radar sensor module, an upper computer module and a short message alarm, wherein the radar sensor module is used for detecting the action of a person to be detected and generating action information; the upper computer module is connected with the millimeter wave module and the short message alarm, and is used for receiving human motion information transmitted by the millimeter wave radar sensor module and the trained convolutional neural network model, identifying the motion category and returning a Data value, and the short message alarm sends a short message alarm to a guardian and transmits specific falling direction information when the Data value corresponds to falling motion through identifying different Data values.
The beneficial effects are that: compared with the prior art, the invention uses the millimeter wave radar sensor and the convolutional neural network to detect the falling direction, has all-weather working capacity and strong anti-interference capacity in the whole day, can detect the falling action of a human body under the dark condition, can also send corresponding falling direction information to a guardian on the basis of detecting the falling action, quickly positions the injured part, prevents secondary injury from being formed in the process of treatment, and has good market prospect.
Drawings
FIG. 1 is a hardware configuration diagram of the present invention
FIG. 2 is an overall flow chart of fall direction detection in the present invention
FIG. 3is a flowchart of an image edge contour extraction algorithm according to the present invention
FIG. 4 is a graph of a two-channel convolutional neural network model in accordance with the present invention
FIG. 5 is a diagram of a short message alarm according to the present invention
Detailed Description
The present invention is further illustrated in the accompanying drawings and detailed description which are to be understood as being merely illustrative of the invention and not limiting of its scope, and various modifications of the invention, which are equivalent to those skilled in the art upon reading the invention, will fall within the scope of the invention as defined in the appended claims.
The invention provides a non-contact falling direction detection method for a short message alarm, which is shown in fig. 1, and the hardware structure of the non-contact falling direction detection method comprises a millimeter wave radar sensor module, an upper computer module and a short message alarm module. The millimeter wave radar module is used for collecting action information of a person to be detected, the upper computer module is connected with the millimeter wave radar module and the short message alarm module, when the millimeter wave radar detects the action of the person to be detected, a corresponding Data value is generated through a trained convolutional neural network, the corresponding Data value is sent to the short message alarm, and the short message alarm realizes a corresponding function.
Based on the above hardware structure, the invention provides a contactless fall direction detection method for a short message alarm, as shown in fig. 2, which specifically includes the following steps:
s1: 6 human body action types are designed: fall forward, fall backward, squat down, walk, drink water, sit down and be done by 5 different volunteers. The method for designing the human body action type comprises the following steps: the millimeter wave radar is separated from the volunteers by a horizontal distance of 3m, the height of the millimeter wave radar is level with the chest of the volunteers, and 5 volunteers successively complete 6×100 actions for 5×6×100 action data.
Acquiring human body action information by utilizing a millimeter wave radar to obtain radar original echo data, and extracting the radar original echo data to receiving antenna data:
in the example, an IWR6843ISK+DCA1000 high-speed data acquisition card is adopted, a transmitting antenna and two receiving antennas are arranged for acquiring human body action information, the number of sampling points is 128, the number of chirp is 128, the frame length is 60, and the acquired actions are finally stored in a computer in the form of bin files.
S2: preprocessing radar original data to obtain a time-distance spectrogram and a time-Doppler spectrogram, and specifically comprising the following steps:
a1: reading a bin file of radar original echo data by using a Matlab mathematical calculation tool to obtain a matrix of the number of receiving antennas multiplied by the total number of sampling points, extracting data of the receiving antennas, and recombining the extracted data into a data matrix of the number of sampling points multiplied by the number of chirp multiplied by the number of frames;
a2: performing MTI static filtering treatment on the obtained data matrix, and removing a static target in a measured space;
a3: performing fast Fourier transform on the processed data to obtain distance distribution information;
a4: accumulating the distance distribution information on a time axis to obtain a time-distance spectrogram;
a5: obtaining a time-Doppler spectrogram by using short-time Fourier transform on the time-distance spectrogram, and taking a model from the obtained time-Doppler spectrogram, wherein the forward falling close to the radar direction and the forward falling far from the radar direction of a person to be detected have the same time Doppler change trend, and combining the forward falling into forward falling; the backward falling close to Lei Dafang is the same as the backward falling of the tested person away from the radar direction in time Doppler change trend, and the time Doppler change trend is combined into the backward falling.
S3: denoising the time-distance spectrogram and the time-Doppler spectrogram by using an image edge contour extraction algorithm, and finding out a high-power density region and defining the boundary of the high-power density region. Ordering points with power density greater than a threshold value P in a high power density region th The calculation formula of (2) is as follows:
proved by experimental results, a is selected to be 0.6, and the noise in the whole data set is minimum. max { P }, sumAnd respectively selecting the maximum and average power density values of all points in the two feature maps, and selecting the top point at the upper part and the bottom point at the lower part to represent the boundary of the high power density region in each time point to obtain a denoised data set.
S4: the method for establishing the double-channel convolutional neural network comprises the following steps: and constructing a convolutional neural network containing two channels by using a Pytorch framework, extracting the characteristics of the denoised time-distance spectrogram and the denoised time-Doppler spectrogram by using a convolutional layer and a pooling layer respectively for each channel, fusing the extracted characteristics to obtain new characteristics, and classifying human actions after sequentially passing through the convolutional layer, the pooling layer and the full-connection layer.
In the embodiment, the two-channel convolutional neural network model is trained and stored in an environment of an lntel (R) Core (TM) i7-10700K processor + NVIDIA GeForce RTX 2080Ti+Pytorch1.10, and motion information is predicted through the trained motion classification model and a Data value corresponding to a motion class is output.
The training method of the human body action classification model comprises the following steps: respectively constructing data sets by using the denoised time-distance spectrogram and the denoised time-Doppler spectrogram, and carrying out 6 on the two data sets: 2:2, dividing the training set, the verification set and the test set according to the proportion, training and storing the best model.
S5: the millimeter wave radar sensor is connected with the short message alarm through the Mtalab, a person to be tested makes different actions against the millimeter wave radar sensor, the millimeter wave radar acquires human body action information and uploads the human body action information to the Mtalab, the Matlab calls a trained human body action classification model and predicts corresponding action categories, each action category corresponds to a Data value, the Matlab transmits the Data value to the short message alarm, and the short message alarm sends corresponding action information to a guardian according to the Data value.
In this embodiment, the Data values of the flags are 0,1, 2, 3, 4, 5, and 6, respectively, corresponding to different operations. The Data value defaults to 0, and the short message alarm does not perform any operation; when the Data value is 1, the short message alarm sends forward falling information to the guardian and generates an alarm prompt; when the Data value is 2, the short message alarm sends backward falling information to the guardian and generates an alarm; when the Data value is 3, the short message alarm sends squat information to the guardian; when the Data value is 4, the short message alarm sends walking information to the guardian; when the Data value is 5, the short message alarm sends water drinking information to the guardian; when the Data value is 6, the short message alarm sends sitting information to the guardian; when no person exists in the tested space, the Data returns to the default value of 0.
According to the above method procedure, the following is summarized with reference to fig. 3 to 5, respectively:
as shown in fig. 3, echo data acquired by millimeter waves are put into Matlab, and the maximum power density max { P } and the average power density of elements in the time-distance data matrix and the time-doppler data matrix of the echo data are calculated respectivelyThreshold P th The calculation formula of (2) is as follows:
the value of a is chosen to minimize noise interference in the overall data set, and in each time point, the top point at the top and the bottom point at the bottom are chosen to represent the boundaries of the high power density region, thereby obtaining denoised time-distance and time-doppler maps.
As shown in fig. 4, firstly, the denoised time-distance graph and the denoised time-doppler graph are convolutionally pooled to extract time-distance features and time-doppler features, secondly, feature fusion is performed on the extracted two features, convolution pooling is performed again on new features obtained after fusion, and a human body action classification result is obtained after the full connection layer.
As shown in fig. 5, the invention designs 6 human actions to control the related functions of the short message alarm, namely, forward falling (Data 1), backward falling (Data 2), squatting (Data 3), walking (Data 4), drinking (Data 5) and sitting (Data 6), when the millimeter wave radar collects the action information of the tested person and the upper computer carries out action recognition, the upper computer sends the Data value corresponding to the falling (Data 1 or Data 2), at the moment, the short message alarm sends the corresponding falling direction information to the guardian and sends an alarm reminding, and when the upper computer sends other Data values (Data 3, data4, data5 and Data 6), the short message alarm also sends the state information of the tested person to the guardian, but does not send an alarm at the moment, so that the guardian can conveniently inquire the state information of the tested person at any time.
Claims (5)
1. The non-contact falling direction detection method for the short message alarm is characterized by comprising the following steps of:
step 1: transmitting frequency modulation continuous wave signals through a millimeter wave radar, receiving different action signals of a tested person, and preprocessing echo signals of the radar to obtain a time-distance spectrogram and a time-Doppler spectrogram;
step 2: denoising the time-distance spectrogram and the time-Doppler spectrogram by using an image edge contour extraction algorithm;
step 3: building a double-channel convolutional neural network, putting the time-distance spectrogram and the time-Doppler spectrogram after denoising in the step 3 into a trained human body action classification model to predict action information and output action categories to obtain corresponding Data values;
step 4: the upper computer sends the Data value to the short message alarm, and the short message alarm executes a corresponding function according to the Data value.
2. The method for detecting the direction of falling of the short message alarm according to claim 1, wherein the step 1 collects the actions of falling forward, falling backward, squatting, walking, drinking water and the like of the person to be detected, and the preprocessing of the collected data comprises the following steps:
a1: analyzing radar echo data, and arranging the data into a data matrix of sampling points, chirp numbers and frame numbers;
a2: performing MTI static filtering treatment on the obtained data matrix, and removing a static target in a measured space;
a3: performing fast Fourier transform on the processed data to obtain distance distribution information;
a4: accumulating the distance distribution information on a time axis to obtain a time-distance spectrogram;
a5, obtaining a time-Doppler spectrogram by using short-time Fourier transform on the time-distance spectrogram;
a6: taking a model of the time-Doppler spectrogram, and combining forward falling actions when approaching the radar direction and forward falling actions when departing from the radar direction into forward falling; the backward falling motion when approaching the radar direction and the backward falling motion when departing from the radar direction are combined into backward falling.
3. The method for detecting the falling direction of the short message alarm according to claim 1, wherein the method using the image edge contour extraction algorithm in the step 2 is as follows: respectively selecting high power density regions in the time-distance matrix and the time-Doppler matrix, defining boundaries of the high power density regions, sequencing points with power density larger than a threshold value in the high power density regions, and determining a threshold value P th Is calculated as follows:
4. The method for detecting the non-contact falling direction of the short message alarm according to claim 1, wherein the method for constructing the two-channel convolutional neural network in the step 3is as follows: under a Pytorch deep learning framework, a convolutional neural network containing two channels is built, each channel respectively uses a convolutional layer and a pooling layer to extract features in a denoised time-distance spectrogram and a time-Doppler spectrogram, the extracted features are fused to obtain new features, and the new features are subjected to the pooling layer and the full-connection layer of the convolutional layer again to obtain corresponding Data values.
5. The method for detecting the direction of falling without contact for a short message alarm according to claim 1, wherein the step 4 is specifically: and (3) the upper computer sends the Data value obtained in the step (3) to a short message alarm, and when the tested person falls, the short message alarm displays falling direction information according to the corresponding Data value and sends out alarm reminding.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202211325313.5A CN116264012A (en) | 2022-10-28 | 2022-10-28 | Contactless falling direction detection method for short message alarm |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202211325313.5A CN116264012A (en) | 2022-10-28 | 2022-10-28 | Contactless falling direction detection method for short message alarm |
Publications (1)
Publication Number | Publication Date |
---|---|
CN116264012A true CN116264012A (en) | 2023-06-16 |
Family
ID=86722796
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202211325313.5A Pending CN116264012A (en) | 2022-10-28 | 2022-10-28 | Contactless falling direction detection method for short message alarm |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN116264012A (en) |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN117310646A (en) * | 2023-11-27 | 2023-12-29 | 南昌大学 | Lightweight human body posture recognition method and system based on indoor millimeter wave radar |
-
2022
- 2022-10-28 CN CN202211325313.5A patent/CN116264012A/en active Pending
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN117310646A (en) * | 2023-11-27 | 2023-12-29 | 南昌大学 | Lightweight human body posture recognition method and system based on indoor millimeter wave radar |
CN117310646B (en) * | 2023-11-27 | 2024-03-22 | 南昌大学 | Lightweight human body posture recognition method and system based on indoor millimeter wave radar |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN107462258B (en) | Step counting method based on mobile phone three-axis acceleration sensor | |
US20200166610A1 (en) | Detection method, detection device, terminal and detection system | |
CN103337132B (en) | Tumble detection method for human body based on 3-axis acceleration sensor | |
CN103886715B (en) | A kind of tumble detection method for human body | |
Ahanathapillai et al. | Preliminary study on activity monitoring using an android smart‐watch | |
CN105528859A (en) | Nursing intelligent alarm system based on human body falling down detection technology | |
CN105595990A (en) | Intelligent terminal device for evaluating and distinguishing quality of electrocardiosignal | |
CN104462744B (en) | Suitable for the data quality control method of cardiovascular remote supervision system | |
CN103049741A (en) | Foot-to-ground acting force-based gait feature extraction method and gait identification system | |
CN105788172B (en) | Indoor safeguard and supervision for the aged system and method based on Internet of Things | |
CN107290741A (en) | Combine the indoor human body gesture recognition method apart from time-frequency conversion based on weighting | |
CN103577836A (en) | Method and model system for establishing human body tumbling detection model | |
Yao et al. | Fall detection system using millimeter-wave radar based on neural network and information fusion | |
CN110456320A (en) | A kind of ULTRA-WIDEBAND RADAR personal identification method based on free space gait temporal aspect | |
CN114038012A (en) | Fall detection method and system based on millimeter wave radar and machine learning | |
CN107105092A (en) | A kind of human body tumble recognition methods based on dynamic time warping | |
CN111208508A (en) | Motion quantity measuring method and device and electronic equipment | |
CN105243285A (en) | Big data health forecast system | |
CN116264012A (en) | Contactless falling direction detection method for short message alarm | |
CN113963192A (en) | Fall detection method and device and electronic equipment | |
CN114814832A (en) | Millimeter wave radar-based real-time monitoring system and method for human body falling behavior | |
CN115474901A (en) | Non-contact living state monitoring method and system based on wireless radio frequency signals | |
Ahn et al. | A digital twin city model for age-friendly communities: Capturing environmental distress from multimodal sensory data | |
CN115422962A (en) | Gesture and gesture recognition method and device based on millimeter wave radar and deep learning algorithm | |
CN110693466A (en) | Rehabilitation evaluation device based on remote control induction technology and implementation method thereof |
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 |