1. A transmission-type intelligent super-surface assisted fall detection method is characterized by comprising the following steps:

carrying out periodic switching configuration on the transmission type intelligent super surface;

acquiring receiving antenna signals of a multi-antenna receiver under the configuration of a transmission-type intelligent super surface of a current switching period;

extracting a feature vector according to the receiving antenna signal;

inputting the feature vector into a neural network of a coder-decoder to obtain a reconstructed feature vector;

determining a fall index from the feature vector and the reconstructed feature vector;

judging whether the falling index is larger than a set falling threshold value or not to obtain a first judgment result;

if the first judgment result shows that the falling index is larger than a set falling threshold value, determining that the person falls;

and if the first judgment result shows that the fall index is smaller than or equal to a set fall threshold, determining that the person is in a normal state.

2. The transmissive intelligent super-surface assisted fall detection method according to claim 1, further comprising, before the periodically switching configuration of the transmissive intelligent super-surface:

determining an optimal configuration of the transmissive intelligent super-surface;

and detecting personnel according to the receiving antenna signals under the optimal configuration.

3. The method for fall detection assisted by a transmissive intelligent super surface according to claim 2, wherein the determining the optimal configuration of the transmissive intelligent super surface specifically comprises:

initially configuring the transmission type intelligent super surface;

acquiring receiving antenna signals of a multi-antenna receiver under initial configuration of the transmission-type intelligent super-surface;

and performing optimal configuration according to the receiving antenna signals of the multi-antenna receiver under the initial configuration of the transmission-type intelligent super surface, and determining the optimal configuration of the transmission-type intelligent super surface.

4. The method for fall detection assisted by a transmissive intelligent super-surface according to claim 2, wherein the detecting the person according to the receiving antenna signal under the optimal configuration specifically comprises:

optimally configuring the transmission-type intelligent super surface;

acquiring a receiving antenna signal under the optimal configuration;

judging whether the variance of the receiving antenna signals under the optimal configuration is larger than a set threshold value in a set period or not to obtain a second judgment result;

if the second judgment result shows that the variance of the receiving antenna signals under the optimal configuration is larger than a set threshold value in a set period, determining that a person enters; and if the second judgment result shows that the variance of the receiving antenna signal under the optimal configuration is less than or equal to a set threshold value in a set period, determining that no person enters the current space to be tested and returning to the step of acquiring the receiving antenna signal under the optimal configuration.

5. The method of claim 2, wherein the pre-training process of the codec neural network specifically comprises:

and taking the characteristic vector as the input of the encoder-decoder neural network, taking the reconstructed characteristic vector as the output, taking the average error between the reconstructed characteristic vector and the input characteristic vector as a loss function, performing optimization training on the neural network, and determining the weight and the average fall index of the encoder-decoder neural network.

6. The method of claim 5, wherein the set fall threshold is an average fall index within a training data set of a set multiple.

7. The method for fall detection with assistance of a transmissive intelligent super-surface according to claim 6, wherein if the first determination result indicates that the fall index is smaller than or equal to a set fall threshold, the determining that the person is in a normal state further comprises:

judging whether the variance of the receiving antenna signals under the optimal configuration is smaller than a set threshold value in a set period to obtain a third judgment result;

if the third judgment result shows that the variance of the receiving antenna signals under the optimal configuration is smaller than a set threshold value in a set period, acquiring a corresponding feature vector when the personnel is in a normal state;

and updating the weight and the average fall index of the neural network of the coder-decoder according to the corresponding feature vector when the person is in a normal state.

8. A transmissive intelligent super-surface assisted fall detection system applying the method as claimed in any one of claims 1 to 7, the system comprising: the system comprises a transmission type intelligent super surface, a multi-antenna receiver and a computer;

the transmission type intelligent super surface is arranged on one side wall of a room close to an environmental signal source; the transmission type intelligent super surface comprises a transmission type super surface and a super surface intelligent control circuit; the transmissive meta-surface comprises a plurality of transmissive meta-material cells; a plurality of the transmission metamaterial units are laid on the side wall; the super-surface intelligent control circuit is used for controlling the configuration of the transmission type super-surface; the transmission type intelligent super surface is used for carrying out beam forming on the environment signal source; the multi-antenna receiver is used for converting the radio-frequency signals subjected to the beam forming into receiving antenna signals; the computer is used for receiving the receiving antenna signals and judging whether the person falls down or not according to the receiving antenna signals.

9. The transmissive intelligent super-surface assisted fall detection system according to claim 8, wherein the transmissive metamaterial unit comprises two surface layers arranged mirror-symmetrically; the surface layer comprises a diode, a copper printed circuit and a dielectric plate connected with the copper printed circuit; the diode connects the copper printed circuits of the two surface layers; the diode is also connected with the super-surface intelligent control circuit.

Background

The video-based fall detection system is a system which is characterized in that a computer is connected with a camera, images or videos are captured through the camera, the posture of a human body is extracted from the images or videos, and whether the human body falls or not is judged according to the posture of the human body.

The falling detection system based on the radar/WIFI is a system which utilizes the existing radar/WIFI transmitter and receiver and analyzes different influences of statistical characteristics caused by different posture actions of a human body on the received radar/WIFI signal, so that the posture of an analyst is analyzed and whether the person falls is judged.

However, since wireless signals propagated in a space are superimposed at a receiving end, information about the space contained in the received signals is mixed together and cannot be effectively extracted, and the dimension of the space information contained in the received signals is low, which results in that a fall detection system based on radar/WIFI needs to rely on the statistical characteristic change of the received signals caused by human body motion for identification, and the identification precision is insufficient.

The gesture recognition based on the intelligent super surface and the point cloud extraction related system are based on the beam forming capability of the reflection type intelligent super surface, the reflection coefficient in the space is extracted by using a wireless transceiver, and a supervised learning method is adopted, so that the system for recognizing the human body gesture or the reflector in the space is realized.

The gesture recognition system based on the reflective super-surface needs to add an additional signal emission source, and the complexity of the system is increased.

The fall detection system based on the sensor realizes the detection of the posture and the fall of a human body based on the position, the speed, the angular velocity, the heartbeat, the blood pressure and other information of the human body provided by the intelligent monitoring equipment worn by people. However, this system relies on the person wearing the smart sensor continuously, and fall detection is not possible for persons not wearing the sensor.

Therefore, there is a need for a method for high-precision fall detection in space without wearing an intelligent sensor and without an additional internal signal emission source.

Disclosure of Invention

The invention aims to provide a transmission-type intelligent super-surface assisted fall detection method and system, which can realize fall detection in space without an additional internal signal source and an intelligent sensor.

In order to achieve the purpose, the invention provides the following scheme:

a transmission-type intelligent super-surface assisted fall detection method comprises the following steps:

carrying out periodic switching configuration on the transmission type intelligent super surface;

acquiring receiving antenna signals of a multi-antenna receiver under the configuration of a transmission-type intelligent super surface of a current switching period;

extracting a feature vector according to the receiving antenna signal;

inputting the feature vector into a neural network of a coder-decoder to obtain a reconstructed feature vector;

determining a fall index from the feature vector and the reconstructed feature vector;

judging whether the falling index is larger than a set falling threshold value or not to obtain a first judgment result;

if the first judgment result shows that the falling index is larger than a set falling threshold value, determining that the person falls;

and if the first judgment result shows that the fall index is smaller than or equal to a set fall threshold, determining that the person is in a normal state.

Optionally, before the configuration of periodically switching the transmissive intelligent super-surface, the method further includes:

determining an optimal configuration of the transmissive intelligent super-surface;

and detecting personnel according to the receiving antenna signals under the optimal configuration.

Optionally, the determining the optimal configuration of the transmissive intelligent super-surface specifically includes:

initially configuring the transmission type intelligent super surface;

acquiring receiving antenna signals of a multi-antenna receiver under initial configuration of the transmission-type intelligent super-surface;

and performing optimal configuration according to the receiving antenna signals of the multi-antenna receiver under the initial configuration of the transmission-type intelligent super surface, and determining the optimal configuration of the transmission-type intelligent super surface.

Optionally, the detecting the person according to the receiving antenna signal under the optimal configuration specifically includes:

optimally configuring the transmission-type intelligent super surface;

acquiring a receiving antenna signal under the optimal configuration;

judging whether the variance of the receiving antenna signals under the optimal configuration is larger than a set threshold value in a set period or not to obtain a second judgment result;

if the second judgment result shows that the variance of the receiving antenna signals under the optimal configuration is larger than a set threshold value in a set period, determining that a person enters; and if the second judgment result shows that the variance of the receiving antenna signal under the optimal configuration is less than or equal to a set threshold value in a set period, determining that no person enters the current space to be tested and returning to the step of acquiring the receiving antenna signal under the optimal configuration.

Optionally, the pre-training process of the codec neural network specifically includes:

and taking the characteristic vector as the input of the encoder-decoder neural network, taking the reconstructed characteristic vector as the output, taking the average error between the reconstructed characteristic vector and the input characteristic vector as a loss function, performing optimization training on the neural network, and determining the weight and the average fall index of the encoder-decoder neural network.

Optionally, the set fall threshold is an average fall index within the training data set at a set multiple.

Optionally, if the first determination result indicates that the fall index is smaller than or equal to a set fall threshold, determining that the person is in a normal state further includes:

judging whether the variance of the receiving antenna signals under the optimal configuration is smaller than a set threshold value in a set period to obtain a third judgment result;

if the third judgment result shows that the variance of the receiving antenna signals under the optimal configuration is smaller than a set threshold value in a set period, acquiring a corresponding feature vector when the personnel is in a normal state;

and updating the weight and the average fall index of the neural network of the coder-decoder according to the corresponding feature vector when the person is in a normal state.

A transmissive intelligent super-surface assisted fall detection system applying a transmissive intelligent super-surface assisted fall detection method as claimed in any one of the above, the transmissive intelligent super-surface assisted fall detection system comprising: the system comprises a transmission type intelligent super surface, a multi-antenna receiver and a computer;

the transmission type intelligent super surface is arranged on one side wall of a room close to an external environment signal source; the transmission type intelligent super surface comprises a transmission type super surface and a super surface intelligent control circuit; the transmissive meta-surface comprises a plurality of transmissive meta-material cells; a plurality of the transmission metamaterial units are laid on the side wall; the super-surface intelligent control circuit is used for controlling the configuration of the transmission type super-surface; the transmission type intelligent super surface is used for carrying out beam forming on the environment signal source; the multi-antenna receiver is used for converting the radio-frequency signals subjected to the beam forming into receiving antenna signals; the computer is used for receiving the receiving antenna signals and judging whether the person falls down or not according to the receiving antenna signals.

Optionally, the transmission metamaterial unit comprises two surface layers arranged in mirror symmetry; the surface layer comprises a diode, a copper printed circuit and a dielectric plate connected with the copper printed circuit; the diode connects the copper printed circuits of the two surface layers; the diode is also connected with the super-surface intelligent control circuit.

According to the specific embodiment provided by the invention, the invention discloses the following technical effects:

according to the transmission-type intelligent super-surface assisted falling detection method and system provided by the invention, the transmission-type intelligent super-surface is used for carrying out beam forming on a signal transmitted by an environment signal source, carrying out characteristic extraction on a receiving antenna signal of a multi-antenna receiver, and then reconstructing by using a coder-decoder neural network, so that a falling coefficient is determined, and further whether personnel fall or not is determined. The invention can realize the fall detection in space without wearing an intelligent sensor or an additional internal signal source and directly utilizing a signal source of an external environment.

Drawings

In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings needed to be used in the embodiments will be briefly described below, and it is obvious that the drawings in the following description are only some embodiments of the present invention, and it is obvious for those skilled in the art to obtain other drawings without inventive exercise.

Fig. 1 is a flow chart of a transmission-type intelligent super-surface assisted fall detection method provided by the present invention;

fig. 2 is a flow chart of a transmission-type intelligent super-surface assisted fall detection method provided by the invention in practical application;

fig. 3 is a schematic diagram of a transmission-type intelligent super-surface assisted fall detection system provided by the invention.

Description of the symbols:

1-an ambient signal source; 2-a transmissive super-surface; 3-a computer; 4-super surface intelligent control circuit; 5-multiple antenna receiver.

Detailed Description

The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.

The invention aims to provide a transmission-type intelligent super-surface assisted fall detection method and system, which can realize fall detection in space without wearing an intelligent sensor or an additional internal signal source.

In order to make the aforementioned objects, features and advantages of the present invention comprehensible, embodiments accompanied with figures are described in further detail below.

As shown in fig. 1, the invention provides a transmission-type intelligent super-surface assisted fall detection method, which includes:

step 101: and carrying out periodic switching configuration on the transmission type intelligent super surface.

Step 102: and acquiring receiving antenna signals of the multi-antenna receiver under the configuration of the transmission-type intelligent super surface in the current switching period.

Step 103: extracting a feature vector according to the receiving antenna signal.

Step 104: and inputting the characteristic vector into a neural network of a coder-decoder to obtain a reconstructed characteristic vector.

Step 105: determining a fall index from the feature vector and the reconstructed feature vector.

Step 106: and judging whether the falling index is larger than a set falling threshold value or not to obtain a first judgment result. If the first judgment result indicates that the fall index is greater than a set fall threshold, executing step 107; if the first determination result indicates that the fall index is less than or equal to the set fall threshold, step 108 is executed. Wherein the set fall threshold is an average fall index in a set multiple of the training data set.

Step 107: it is determined that the person has fallen.

Step 108: it is determined that the person is in a normal state.

In practical applications, the step 108 is followed by:

and judging whether the variance of the receiving antenna signals under the optimal configuration is smaller than a set threshold value in a set period to obtain a third judgment result.

And if the third judgment result shows that the variance of the receiving antenna signals under the optimal configuration is smaller than a set threshold value in a set period, acquiring a corresponding feature vector when the personnel is in a normal state.

And updating the weight and the average fall index of the neural network of the coder-decoder according to the corresponding feature vector when the person is in a normal state.

In practical applications, before the step 101, the method further includes:

determining an optimal configuration of the transmissive intelligent super-surface. Wherein the determining the optimal configuration of the transmissive intelligent super-surface specifically comprises: and initially configuring the transmission type intelligent super surface. And acquiring receiving antenna signals of the multi-antenna receiver under the initial configuration of the transmission type intelligent super-surface. And performing optimal configuration according to the receiving antenna signals of the multi-antenna receiver under the initial configuration of the transmission-type intelligent super surface, and determining the optimal configuration of the transmission-type intelligent super surface.

And detecting personnel according to the receiving antenna signals under the optimal configuration.

Wherein, the detecting personnel according to the receiving antenna signal under the optimal configuration specifically comprises:

optimally configuring the transmission-type intelligent super surface; acquiring a receiving antenna signal under the optimal configuration; judging whether the variance of the receiving antenna signals under the optimal configuration is larger than a set threshold value in a set period or not to obtain a second judgment result; if the second judgment result shows that the variance of the receiving antenna signals under the optimal configuration is larger than a set threshold value in a set period, determining that a person enters; and if the second judgment result shows that the variance of the receiving antenna signal under the optimal configuration is less than or equal to a set threshold value in a set period, determining that no person enters the current space to be tested and returning to the step of acquiring the receiving antenna signal under the optimal configuration.

In practical applications, the pre-training process of the codec neural network specifically includes:

and taking the characteristic vector as the input of the encoder-decoder neural network, taking the reconstructed characteristic vector as the output, taking the average error between the reconstructed characteristic vector and the input characteristic vector as a loss function, performing optimization training on the neural network, and determining the weight and the average fall index of the encoder-decoder neural network. The average error is specifically an average of squares of absolute values of the differences.

In a practical scenario, as shown in fig. 2, the transmissive intelligent super-surface assisted fall detection method provided by the present invention includes the following three stages.

Transmission-type intelligent super-surface initial configuration optimization process

In the process of selecting the transmission type intelligent super-surface configuration, K configurations are randomly selected, and each configuration c is composed of random 0-1. The multi-antenna receiver records when the transmission-type intelligent super surface is in different configurations, N_RxThe signal received by the receiving antenna is recorded as vector y_k～(k∈[1，K]). To ensure that the beams generated between the configurations have significant differences, the computer will optimize the K configurations, minimizing y₁，...，y_kAnd (3) solving the following optimization problem of the transmission type intelligent super-surface configuration by using pairwise correlation values among the vectors:

wherein, c₁，...，c_KK configurations of the transmissive hyperplane are represented,represents y_kAre both [1, K ] K]Ordinal number between, and k' ≠ k. y is_k、y_k'denotes the configuration of the k and k' respectively, N_RxThe signals from the root receive antenna constitute a received signal vector.

K configurations determined by the transmission-type intelligent super-surface in the transmission-type intelligent super-surface initial configuration selection process are recorded asIn the case of no person in space, the multi-antenna receiver N is recorded_RxThe signals received by the root receiving antenna being vectors

Pre-training process

In the pre-training process, a plurality of different testers sequentially and independently enter a room to normally moveThe operation includes standing, walking, running, squatting and the like. In the process, the periodic sequential switching of the super-surface is configuredEach configuration lasts δ seconds, so each cycle lasts K δ seconds, which is denoted as a configuration switching cycle.

Within delta seconds of the k configuration duration in the t configuration switching period, the multi-antenna receiver N_RxThe collected baseband signals are transmitted to the computer according to the receiving antenna records, and the computer records the average value of the sampled signals and a plurality of statistical characteristic quantities including maximum-minimum value, variance and change rate.

Note that N is the number of K configurations in the tth configuration switching period_RxThe vector formed by the average value and the statistical characteristics of the received signals of the root antenna is a characteristic vector f of the t-th configuration switching period^(t)。

Collecting T data of configuration switching period in total, and collecting a data set with size of TEach data packet contains a feature vector f^(t)。

The computer can obtain an algorithm which can judge whether the person falls or not through data collected in real time for a parameter algorithm (such as a neural network algorithm and a support vector machine algorithm) by using a machine learning algorithm.

Considering that the data set generally only contains data of normal activities (the falling situation of a person is highly random and difficult to collect in advance), the collected data setIs a single-class data set and needs to use an unsupervised learning algorithm to perform anomaly detection.

In contrast, the invention adopts an Encoder-Decoder neural Network algorithm (Encoder-Decoder Network) which is suitable for single-class anomaly detection and is suitable for mass data, and adds an LSTM (Long-short term memory) core in the Encoder-Decoder neural Network algorithm, so that the Encoder-Decoder neural Network algorithm has the capability of extracting the time correlation of the feature vectors, and the neural Network can judge whether people fall or not according to the trend expressed in a plurality of switching periods by the characteristics of received signals, thereby enhancing the accuracy of fall detection.

As shown in FIG. 2, in the codec neural network algorithm, the dimensions of the network input layer and the network output layer are the dimensions of the feature vector, which is denoted as N_F. Two fully-connected layers with the node number decreasing are included in the middle and serve as encoders, one LSTM (Long-short term memory) layer with the S-length memory step number is used for extracting time sequence correlation between the feature vectors, and the two fully-connected layers with the node number increasing serve as decoders.

The input to the codec neural network algorithm is a feature vector f^(t)After the compression, abstraction, combination of time domain information and reduction of the middle layer, a reconstructed feature vector is output at the output layer and recorded asRecording all parameters in the neural network algorithm of the coder-decoder as vectors w, in the pre-training stage, the computer optimizes w through an optimizer, so that the collected data setThe average difference between the restored feature vector and the input feature vector is minimized. Namely, it is

Wherein, the optimal parameter vector obtained by solving is recorded as w^*. The pre-training process is complete.Representing the eigenvectors reconstructed by the encoder-decoder.

In addition, the codec neural network uses w^*When the parameter is recorded

Wherein the content of the first and second substances,the difference of the average feature vector reconstruction after optimization is reflected for the average fall index after training.

Condition monitoring process

And starting a normal running state monitoring process after pre-training. The system firstly judges whether a person enters the room or not, and after the person enters the room, the system starts a continuous monitoring stage until the person leaves the room.

Specifically, the method is divided into the following 4 stages:

1 initial idle phase

In the initial idle stage, the transmission type intelligent super-surface keeps c equal to 0, the multi-antenna receiver continuously receives signals in the space, the signals are transmitted to the computer, and the variance of the received signals in each delta time is calculated.

When the computer calculates the variance of the signal on one or more receiving antennas lasting N_InAnd when the period is larger than sigma, judging that the personnel enters, and enabling the system to enter a falling monitoring stage.

2 Fall monitoring phase

In this phase, the configuration of the transmissive smart super-surface is periodically switched in turnEach configuration lasts δ seconds.

In the t-th configuration switching period, transmitting the multipath receiving baseband signals of the multi-antenna receiver to a computer, and extracting to obtain a characteristic vector f^(t)。

Codec neural network usage optimized parameter w^*Is inputted f^(t)To obtainComputer fallReciprocal indexIf it is notJudging that the person falls down once in the t period; otherwise, the person is considered to be in a normal state. Where β > 1 is a multiple, which represents that the system considers a fall to occur when the current fall index is greater than what multiple of the post-training average fall index.

When the computer finds a falling event, the computer enters a falling event processing stage. Otherwise, entering the next switching period.

When N lasts_outIn each switching period, when the computer calculates that the signal variance on all the receiving antennas is smaller than sigma, the computer judges that the personnel leaves, and the computer control system enters an online training stage.

3 Fall event processing phase

When the computer judges that the person falls down, the computer prompts the person to detect and find the fall down on the display and requires the person to feed back whether the person falls down.

If the person falls in the feedback or no feedback exists within the time exceeding L, the computer reports that suspected persons in the external guardian fall to be helped through the network, and the computer enters an ending state after the fall event is processed.

When the person indicates that no fall has been detected, the system will return to the fall monitoring phase.

4 on-line training phase

In the on-line training stage, the system performs on-line training (adaptive online training) on the codec neural network according to the new non-fall data received in the continuous monitoring state, weights w, and the trained average fall indexAnd (6) updating. Therefore, the computer utilizes the latest data to further improve the accuracy of system identification.

After the on-line training phase is completed, the system enters an initial idle phase.

As shown in fig. 3, the transmission-type intelligent super-surface-assisted fall detection system provided by the present invention applies the transmission-type intelligent super-surface-assisted fall detection method as described above, and includes: a transmissive intelligent super-surface, a multi-antenna receiver 5 and a computer 3.

The transmission type intelligent super surface is arranged on one side wall of a room close to the environmental signal source 1; the transmission type intelligent super surface comprises a transmission type super surface 2 and a super surface intelligent control circuit 4; the transmission-type meta-surface 2 comprises a plurality of transmission meta-material units; a plurality of the transmission metamaterial units are laid on the side wall; the super-surface intelligent control circuit 4 is used for controlling the configuration of the transmission type super-surface 2; the transmission type intelligent super surface is used for carrying out beam forming on the environment signal source 1; the multi-antenna receiver 5 is configured to convert the radio frequency signal subjected to the beamforming into a receiving antenna signal; the computer 3 is used for receiving the receiving antenna signals and judging whether the person falls down according to the receiving antenna signals. The environment signal source 1 is an external environment signal source.

Wherein the transmission metamaterial unit comprises two surface layers which are arranged in mirror symmetry; the surface layer comprises a diode, a copper printed circuit and a dielectric plate connected with the copper printed circuit; the diode connects the copper printed circuits of the two surface layers; the diode is also connected with the super-surface intelligent control circuit 4.

In a practical scenario, consider two adjacent indoor rooms in a generalized sense (e.g., a living room and a bathroom in a traditional home, or a hallway and a bathroom in a hospital, a dressing room, etc.), where one room a is a public area and the other room B is relatively private. The intelligent super-surface assisted falling detection method and system are deployed in the environment, and the system can monitor whether people (especially the old, the patient and other objects needing extra care) entering a room B fall or not under the conditions that videos are not needed and no equipment is needed to be worn, and gives an alarm and prompts when the people fall, so that the safety of the people is protected, and the privacy of the people is protected.

The system comprises a transmissive intelligent super surface, a multi-antenna receiver 5 and a computer 3.

The transmission-type intelligent super surface comprises two parts, namely a transmission-type super surface 2 and a super surface intelligent control circuit 4. The transmission-type meta-surface 2 is formed by two-dimensionally and densely paving transmission-type meta-material units, and N transmission-type meta-material units are counted in total. The transmission-type metamaterial unit comprises two surface layers which are arranged in mirror symmetry, a copper printed circuit, a dielectric plate and a diode are arranged on each surface layer, the diodes are connected with the two separated copper printed circuits, and the on-off state of the diodes can be controlled through the magnitude of an applied voltage, so that the transmission coefficient of the transmission-type metamaterial unit to transmission electromagnetic waves is influenced.

The transmissive super-surface 2 has the ability to transmit radio frequency signals and can be changed by changing the voltage applied to its cells. And controlling the transmitted radio frequency signal so as to control the signal beam shape of the transmitted radio frequency signal.

In fig. 3, an electromagnetic signal emitted from an external environment signal source 1 passes through a transmissive super-surface 2 and is shaped into a signal beam. The state applied to the N transmissive metamaterial units is the vector c, which determines the beamforming of the transmissive super-surface for the transmissive signal, and the vector is the configuration of the super-surface. Each element in the vector c is an ordinal number from 0 to 1, representing the off-state and on-state of the diode, respectively.

The super-surface intelligent control circuit 4 consists of an FPGA and a stabilized voltage power supply. The super-surface intelligent control circuit 4 is connected to the electrode of the diode of each unit in the transmission-type super-surface 2, and controls the configuration of the transmission-type super-surface 2 by controlling the voltage on two sides of the variable capacitance diode. In addition, in order to simplify the control complexity of the transmissive super-surface 2, the control voltages of the adjacent S cells are made the same when the circuits are connected, and the cells are called as a macro cell. Thus, of the transmission typeThe configuration c of the super-surface 2 may be represented by N_macN/S macro-units. Wherein N is_macThe number of the transmission hyperplane macro-units.

The multi-antenna receiver 5 is composed of a plurality of omnidirectional antennas and a wireless signal transceiver, receives the radio frequency signals which are processed by beam forming through the transmission type intelligent super surface, interact with personnel in space and obtain personnel state information. The multi-antenna receiver 5 can simultaneously receive the radio frequency signals on the multiple antennas and convert them into baseband signals. The baseband signal will be transmitted to the computer 3 by wire for subsequent processing.

The computer 3 receives the baseband signal of the multi-antenna receiver 5, processes the baseband signal, extracts the state information of the personnel by using a machine learning algorithm, and judges whether the personnel falls down. If the person falls down, the computer 3 firstly prompts the internal person through the display to enable the person to feed back whether to fall down within a certain time, and if the person fails to feed back within the set time or the person feeds back to fall down, the computer 3 informs the rescuer of rescue.

In addition, the computer 3 performs online training on the machine learning algorithm by using the online learning algorithm according to the data collected when the person normally falls in the house, so that the effect of updating the algorithm in real time is realized, and the accuracy of the fall detection algorithm is optimized.

The transmission type intelligent super surface is equipment which can allow electromagnetic waves to penetrate through the super surface from one side to the other side and can realize beam forming on signals penetrating through the super surface, the existing transmission type super surface 2 is formed by two-dimensional dense paving of transmission type super material units, and N transmission type super material units are counted. The transmission-type metamaterial unit comprises two surface layers which are arranged in mirror symmetry, a copper printed circuit, a dielectric plate and a diode are arranged on each surface layer, the diodes are connected with two separated copper circuits, and the on-off state of the diodes can be controlled through the magnitude of an applied voltage, so that the transmission coefficient of the transmission-type metamaterial unit to transmission electromagnetic waves is influenced.

The transmission-type super surface 2 is used, so that the propagation environment of wireless signals can be flexibly reconstructed. By adopting different configurations at different times, intuitively, signals for detecting different spatial positions can be separated on a receiving end time domain, so that the dimensionality of spatial information contained in the received signals is increased, and the gesture recognition of the system has higher precision.

The invention has the characteristics of no need of wearing equipment by personnel, privacy protection, no influence by obstacles, higher accuracy when the people do not fall down, and real-time updating capability, and specifically comprises the following steps:

1) the fall detection system can realize monitoring without the need of wearing any wearable equipment by a target person.

2) The system senses by using radio frequency (electromagnetic signals near 3 GHz), and has the advantages of privacy protection and no barrier in comparison with traditional camera-based radio frequency fall detection.

3) Through the wave beam shaping of super surface of transmission to signal in the space to through the wave beam result of algorithm optimal control transmission super surface, can let target person's in the exploration space signal mutually more independent, thereby obtain more spatial information, compare the uncontrollable radio frequency gesture recognition system of traditional environment, have higher rate of accuracy.

4) Compared with the existing posture recognition system based on the reflective super-surface, the invention does not need an additional signal emission source, can utilize the existing signal source in the environment, and can coexist with other devices working at the same frequency in the environment. In addition, a uniquely designed unsupervised learning algorithm is adopted in the system, so that falling detection can be realized without depending on data collected under the falling condition of a target person, and the difficulty of data set collection is reduced; the statistical characteristics of the received signals under each configuration are utilized, the time domain connection of the signal characteristic vector sequence is extracted by adopting the LSTM layer, richer information is given to a neural network algorithm to judge the posture of the person, and the judgment accuracy is improved. Furthermore, the invention has the capability of on-line learning, and supports that the computer utilizes the latest data to update the algorithm in real time in the process of continuous operation of the equipment, thereby further improving the real-time accuracy of system identification.

The embodiments in the present description are described in a progressive manner, each embodiment focuses on differences from other embodiments, and the same and similar parts among the embodiments are referred to each other.

The principles and embodiments of the present invention have been described herein using specific examples, which are provided only to help understand the method and the core concept of the present invention; meanwhile, for a person skilled in the art, according to the idea of the present invention, the specific embodiments and the application range may be changed. In view of the above, the present disclosure should not be construed as limiting the invention.