Identification and Interaction Method, System, Electronic Device and Storage Medium Based on Electrostatic Signals

Abstract

The present invention relates to the technical field of electrostatic signal recognition, and discloses a recognition and interaction method, system, electronic device and storage medium based on electrostatic signals, including the following steps: Step S01: Divide the target object into regions and mark the charge amounts; Step S02: Collect the target data; Step S03: Preprocess the data set; Step S04: Analyze the data in Step S03 to obtain the induced current under different actions; Step S05: Construct a deep neural network model to extract features from the preprocessed electrostatic signals in Step S03, and construct a soft mapping relationship between the electrostatic signal features and different motion postures; Step S06: Obtain the induced current value and electrostatic signal features of the person to be tested, identify the actions and motion postures of the person to be tested, and perform interaction; By providing Step S04 and Step S05, action recognition and motion posture judgment are completed based on electrostatic signals, the acquisition method is simple, the recognition effect is effectively improved, and the privacy of users is protected at the same time.

Description

Technical Field

[0001] The present invention relates to the technical field of electrostatic signal recognition, and more particularly to an identification interaction method, system, electronic device and storage medium based on electrostatic signals. Background Art

[0002] In recent years, important progress has been made in the research on human action recognition at home and abroad, but there are still many difficulties and challenges. At present, the vast majority of research work focuses on the human action recognition method based on visual image analysis. The visual image analysis method mainly uses image sensors, and the video images collected by it contain rich target information, which provides the possibility for action recognition. However, image information is often very sensitive to lighting conditions, perspective changes, object occlusion, etc., and is easily affected by environmental interference; at the same time, video images are not only complex to process, but also often involve user privacy issues.

[0003] The electrostatic signal detection technology is a technology that uses the change of the electrostatic field to sense the surrounding environment and detect and identify target actions. It can detect without contact, without the need to contact the target object, avoiding interference and misjudgment caused by contact; in addition, the electrostatic detection technology has the advantages of small working blind area, low system power consumption, and simple system structure; by identifying the characteristics of human electrostatic signals, action recognition can be completed. By detecting the changes in the electrostatic field and capacitance distributed by an object in the surrounding space environment through an electrostatic acquisition instrument, electrostatic signals can be obtained. The acquisition method of the electrostatic detection technology is simple, and the acquisition instrument is simple and small, effectively avoiding the problem of inaccurate measurement caused by errors in instrument installation.

[0004] Therefore, the present application proposes an identification interaction method and system based on electrostatic signals, which improve the recognition effect and protect the user's privacy by analyzing the characteristics of electrostatic signals. Summary of the Invention

[0005] In order to overcome the above-mentioned defects of the prior art, the present invention provides an identification interaction method, system, electronic device and storage medium based on electrostatic signals to solve the problems existing in the above background art.

[0006] The present invention provides the following technical solutions: An identification interaction method based on electrostatic signals, including the following steps:

[0007] Step S01: Divide the target object into regions and mark the charge amount; the target object is the human body surface of the person to be measured.

[0008] Step S02: Collect target data to form a data set: Collect the human charge data and electrostatic signals in different actions of n target objects at T time points to form a data set.

[0009] Step S03: Preprocess the data set to facilitate further feature extraction;

[0010] Step S04: Analyze the data after current detection in Step S03 to obtain the induced current under different actions;

[0011] Step S05: Construct a deep neural network model to extract features from the preprocessed electrostatic signals in Step S03, and construct a soft mapping relationship between the electrostatic signal features and different motion postures;

[0012] Step S06: Obtain the induced current value and electrostatic signal features of the person to be measured, identify the actions and motion postures of the person to be measured, and perform interaction.

[0013] Preferably, the target object is the human body surface of the person to be measured; the area is divided into the left hand, right hand, head, torso and clothing of the target object, denoted as 1, 2, 3, 4, 5 in sequence; the electric charge amounts of each area are marked as Q1, Q2, Q3, Q4, Q5 in sequence; the mutual capacitances of each area with respect to the surrounding space object p are C 1p 、C 2p 、C 3p 、C 4p 、C 5p respectively.

[0014] Preferably, the preprocessing method includes data interception, data noise reduction and debiasing; the data interception refers to the process of selecting part of the data from the original data set for processing and analysis; intercept the collected original data, that is, the electrostatic signals under different motion postures, to obtain signal inputs of the same length;

[0015] The method for data noise reduction is to perform noise reduction processing on the electrostatic signals under different postures by using the smooth function for smoothing processing. The formula of the smooth function is: y′ = smooth(y0, span, method), where y′ is the column vector after smoothing processing, and y0 is the column vector before smoothing processing;

[0016] The debiasing uses the dyaddown function, which performs binary sampling on the time series and extracts one element every other element to achieve the purpose of downsampling. The calculation formula of the dyaddown function is: y = dyaddown(x, EVENODD), where x is the time series and y is the electrostatic signal data after debiasing; when EVENODD is 0, even sampling is to be performed on the data, that is, sampling starts from the second element of x; when EVENODD is 1, odd sampling is performed on the data, that is, sampling starts from the first element of x.

[0017] Preferably, the electric charges of each region of the target object can be represented by an electric potential equation system: Among them, U1 is the electric potential of region 1, i.e., the left hand; U2 is the electric potential of region 2, i.e., the right hand; U3 is the electric potential of region 3, i.e., the head; U4 is the electric potential of region 4, i.e., the torso; U5 is the electric potential of region 5, i.e., the clothing; C 11 is the self-capacitance of region 1, C 22 is the self-capacitance of region 2, C 33 is the self-capacitance of region 3, C 44 is the self-capacitance of region 4, C 55 is the self-capacitance of region 5; C 15 is the mutual capacitance between region 1 and region 5, C 25 is the mutual capacitance between region 2 and region 5, C 35 is the mutual capacitance between region 3 and region 5, C 45 is the mutual capacitance between region 4 and region 5; C 15 = C 51 , C 25 = C 52 , C 35 = C 53 , C 45 = C 54 .

[0018] Preferably, the specific method for obtaining the induced current under different actions in step S04 is as follows:

[0019] Step S11: Place an induction electrode at a certain distance from the target object; denote the human body electric field as D r , and denote the induced electric field as D g ;

[0020] Step S12: Denote the normal electric field of any point O near the surface of the side of the induction electrode close to the target object i as D foi , and the calculation formula is: D foi = D ri + D gi , where D ri is the human body electric field of the i-th target object, and D gi is the induced electric field of the i-th target object; based on Gauss's theorem, obtain the charge density on the surface of the induction electrode at point O: ρ i = ε * D foi , where ρ i is the charge density on the surface of the induction electrode at any point O near the surface of the side of the i-th target object, and ε is the dielectric constant in the air; i = 1, 2, 3... n;

[0021] Step S13: Calculate the total charge carried on the side of the induction electrode close to the target object i, and the calculation formula is: Q i = ∫ρi dS = ∫εD foi dS, where Q i is the total charge carried on the side of the induction electrode close to the i-th target object, and S is the area of the induction electrode; the induced current generated on the induction electrode is: where I i is the induced current generated by the i-th target object on the induction electrode, ΔQ i is the change in the total charge of the i-th target object, and Δt is the time corresponding to the change in the total charge; i = 1, 2, 3... n;

[0022] Step S14: Statistically analyze the induced currents generated on the induction electrode under different actions of n target objects; in the action of picking up an item, the induced currents generated by n target objects on the induction electrode are sequentially marked as I 1a 、I 2a 、I 3a ……I na ; in the waving action, the induced currents generated by n target objects on the induction electrode are sequentially marked as I 1b 、I 2b 、I 3b ……I nb ; Take the average of the two sets of induced current data: where I a ′ is the average value of the induced currents generated by n target objects on the induction electrode in the action of picking up an item, where I b ′ is the average value of the induced currents generated by n target objects on the induction electrode in the waving action.

[0023] Preferably, the deep neural network model constructed in step S05 is a convolutional neural network model, including an input layer, a convolutional layer, a pooling layer, a Flatten layer, a fully connected layer, and an output layer;

[0024] The convolutional layer is mainly used to extract the features of the input signal, and the output of the convolutional layer is expressed as: Y = q(X * W + h), where Y is the output feature map of the convolutional layer, q(·) is the activation function, * is the convolution operation, X is the input signal sample, W is the weight coefficient, and h is the bias; the size of the output feature map can be expressed as: where U is the size of the output feature map, R is the size of the input feature map, M is the size of the convolution kernel, P is the zero-padding size, and B is the stride;

[0025] The pooling layer is mainly used to reduce the parameter order and feature dimension, and reduces the dimension of the feature map by using the output of the partial region feature map instead of the output of the entire region feature map;

[0026] The Flatten layer can flatten the input data, that is, convert multi-dimensional input data into one-dimensional data;

[0027] The fully connected layer is mainly used to output the final result and perform batch normalization;

[0028] The input layer is used to input the preprocessed electrostatic signal, and the output layer is used to output the electrostatic signal features.

[0029] Preferably, the specific method for constructing the soft mapping relationship between the electrostatic signal features and different motion postures in step S05 is as follows:

[0030] Define a two-dimensional matrix Mapping[M][N] to represent the soft mapping relationship, where M is the number of electrostatic signal features and N is the number of motion postures; analyze the electrostatic signal features of n target objects, and count the motion postures corresponding to the electrostatic signal features of n target objects, which is the analysis result.

[0031] The recognition and interaction system based on electrostatic signals includes a region division module, a data acquisition module, a data preprocessing module, an action recognition and analysis module, a motion posture recognition and analysis module, and a recognition and interaction module;

[0032] The region division module is used to divide the target object into regions and mark the charge amount; the target object is the human body surface of the person to be measured; the region division is the left hand, right hand, head, torso, and clothing of the target object, denoted as 1, 2, 3, 4, 5 in sequence; the charge amounts of each region are marked as Q1, Q2, Q3, Q4, Q5 in sequence; the mutual capacitances of each region with the surrounding space object p are C 1p 、C 2p 、C 3p 、C 4p 、C 5p ;

[0033] The data acquisition module is used to collect target data to form a data set: collect the human charge data of n target objects under different actions and the electrostatic signals under different motion postures within T time points to form a data set;

[0034] The data preprocessing module is used to preprocess the data set to facilitate further feature extraction;

[0035] The action recognition and analysis module is used to analyze the data after current detection in the data preprocessing module to obtain the induced current under different actions;

[0036] The motion posture recognition and analysis module is used to construct a deep neural network model to extract features from the preprocessed electrostatic signals in step S03, and construct a soft mapping relationship between the electrostatic signal features and different motion postures;

[0037] The recognition and interaction module is used to obtain the induced current value and electrostatic signal features of the person to be measured, recognize the actions and motion postures of the person to be measured, and perform interactions.

[0038] An electronic device, comprising:

[0039] One or more processors;

[0040] A storage device on which one or more programs are stored;

[0041] When the one or more programs are executed by the one or more processors, the one or more processors implement an identification and interaction method based on electrostatic signals.

[0042] A computer-readable storage medium on which executable instructions are stored, and when the instructions are executed by a processor, the processor implements an identification and interaction method based on electrostatic signals.

[0043] The technical effects and advantages of the present invention:

[0044] By providing step S04 and step S05, the present invention is conducive to analyzing the data after current detection in step S03 to obtain the induced current under different actions; constructing a deep neural network model to extract features from the preprocessed electrostatic signals in step S03, and constructing a soft mapping relationship between the electrostatic signal features and different motion postures; the electrostatic induction current signal contains human motion information, analyzing the induced current can identify the corresponding human motion, and at the same time extracting and mapping the electrostatic signal features can obtain the motion posture corresponding to the electrostatic signal features, completing action recognition and motion posture judgment based on electrostatic signals, with a simple acquisition method, effectively avoiding the total amount of inaccurate measurement caused by errors in instrument installation, effectively improving the recognition effect, and protecting the privacy of users. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] Figure 1 It is a flowchart of the identification and interaction method based on electrostatic signals of the present invention.

[0046] Figure 2 It is a structural diagram of the identification and interaction system based on electrostatic signals of the present invention.

[0047] Figure 3 It is a schematic diagram of the electronic device according to Embodiment 3 of the present invention.

[0048] Figure 4 It is a schematic diagram of the storage medium according to Embodiment 4 of the present invention. Detailed implementation manners

[0049] The technical solutions in the present invention will be clearly and completely described below in conjunction with the accompanying drawings in the present invention. In addition, the forms of the various structures described in the following embodiments are merely examples. The identification and interaction method, system, electronic device, and storage medium based on electrostatic signals involved in the present invention are not limited to the various structures described in the following embodiments. All other embodiments obtained by those of ordinary skill in the art without creative efforts fall within the scope of protection of the present invention.

[0050] Embodiment 1:

[0051] As Figure 1 shown, this embodiment provides an identification and interaction method based on electrostatic signals, including the following steps:

[0052] Step S01: Divide the target object into regions and mark the charge amounts; the target object is the human body surface of the person to be measured; the regional division is the left hand, right hand, head, torso, and clothing of the target object, denoted as 1, 2, 3, 4, 5 in sequence; the charge amounts of each region are marked as Q1, Q2, Q3, Q4, Q5 in sequence; the mutual capacitances of each region with respect to the surrounding space object p are C 1p 、C 2p 、C 3p 、C 4p 、C 5p ;

[0053] Step S02: Collect target data to form a data set: Collect the human body charge data and electrostatic signals of n target objects under different actions at T time points and different motion postures to form a data set; the different actions include but are not limited to picking up items and waving; the different motion states include but are not limited to running, walking, and jumping; the charge data is the induction electrode area, induction electric field, human body electric field, and charge change amount; the target data is the human body charge data and electrostatic signals;

[0054] Step S03: Preprocess the data set to facilitate further feature extraction; Preprocessing the electrostatic signals to a certain extent can obtain more discriminative electrostatic signal features; the preprocessing methods include data interception, data noise reduction, and debiasing; Detect the current of the charge data, that is, the generated charge change amount, and convert it into a voltage signal for signal conditioning such as amplification and filtering, and convert the weak induced charge amount into a signal that can be processed;

[0055] The data interception refers to the process of selecting partial data from the original dataset for processing and analysis; intercepting the collected original data, i.e., the static electricity signals under different motion postures, to obtain signal inputs of the same length, maximizing the retention of the same segments for each motion posture as much as possible to reduce the training error, and selectively retaining the collected signals with larger errors to reduce overfitting in training; data interception is of great significance in data processing and analysis. When dealing with large-scale data, appropriately intercepting the dataset can reduce the data volume, reduce the algorithm calculation and storage costs, improve the algorithm efficiency, and can also filter out some abnormal data and missing data to improve the accuracy and quality of subsequent analysis and modeling;

[0056] Step S04: Analyze the data after current detection in step S03 to obtain the induced current under different actions;

[0057] Step S05: Construct a deep neural network model to extract features from the preprocessed static electricity signals in step S03, and construct a soft mapping relationship between the static electricity signal features and different motion postures;

[0058] Step S06: Obtain the induced current value and static electricity signal features of the person to be measured, identify the actions and motion postures of the person to be measured, and perform interaction.

[0059] In this embodiment, it should be specifically noted that the method for data denoising in step S03 is to perform denoising processing on the static electricity signals under different postures by using the smooth function. The formula of the smooth function is: y′ = smooth(y0, span, method), where y′ is the column vector after smooth processing, and y0 is the column vector before smooth processing; the smooth function is a denoising processing function that smooths the column vector y through an average filter and returns the value of the processed column vector after denoising the data. The main parameter span of this function fixes the window width of the moving average filter, and method specifies the method of smoothing. The methods of smoothing include the moving average method, local regression, and Savitzky-Golay filtering;

[0060] The debiasing adopts the dyaddown function, which performs binary sampling on the time series, extracting one element every other element to achieve the purpose of downsampling. The calculation formula of the dyaddown function is: y = dyaddown(x, EVENODD), where x is the time series and y is the debiased electrostatic signal data; when EVENODD is 0, even sampling is to be performed on the data, that is, sampling starts from the second element of x; when EVENODD is 1, odd sampling is performed on the data, that is, sampling starts from the first element of x; due to the influence of the residual static electricity potential of the human body, the static electricity signal voltage value is not the base value when at rest, and the average value of the first m points needs to be taken as the bias amount, that is, the average value of the human body static electricity potential within 1 s of standing still before the human body moves, so that the static electricity signal voltage value approaches 0 when at rest; where m = 100.

[0061] In this embodiment, it should be specifically noted that the electric charge amounts of each region of the target object can be represented by an electric potential equation system: Among them, U1 is the electric potential of region 1, i.e., the left hand, U2 is the electric potential of region 2, i.e., the right hand, U3 is the electric potential of region 3, i.e., the head, U4 is the electric potential of region 4, i.e., the torso, and U5 is the electric potential of region 5, i.e., the clothing; C 11 is the self-capacitance of region 1, C 22 is the self-capacitance of region 2, C 33 is the self-capacitance of region 3, C 44 is the self-capacitance of region 4, C 55 is the self-capacitance of region 5; C 15 is the mutual capacitance between region 1 and region 5, C 25 is the mutual capacitance between region 2 and region 5, C 35 is the mutual capacitance between region 3 and region 5, C 45 is the mutual capacitance between region 4 and region 5; C 15 = C 51 , C 25 = C 52 , C 35 = C 53 , C 45 = C 54 ;

[0062] When the human body does not come into contact and separate from the outside, the electric field intensity generated by the actions of prominent parts such as the hands will become the main factor affecting the human body's electric field, and the hand movement information can be obtained from the change of the induced electrostatic field.

[0063] In this embodiment, it should be specifically noted that the specific method for obtaining the induced current under different actions in step S04 is:

[0064] Step S11: Place an induction electrode at a certain distance from the target object. Affected by the human body's electrostatic field, the surface charges of the induction electrode are redistributed, and these charges generate an induced electric field opposite to the external electric field inside the electrode. When the human body is stationary, it can be considered that the capacitance values of each region remain unchanged. The induced electric field on the induction electrode is the same as the intensity of the human body's electric field and in the opposite direction, and there is no induced current in the induction system circuit. When an action occurs, the capacitance value between the moving human body or part and the surrounding environment changes, the charges carried by the human body are redistributed, the electrostatic field around the human body changes, the induced electric field of the detection electrode changes accordingly, and an induced current is generated in the induction system circuit; Denote the human body's electric field as D r , denote the induced electric field as D g ;

[0065] Step S12: Denote the normal electric field at any point O near the surface on the side of the induction electrode close to the target object i as D foi , and the calculation formula is: D foi = D ri + D gi , where D ri is the human body's electric field of the i-th target object, and D gi is the induced electric field of the i-th target object; Based on Gauss's theorem, obtain the surface charge density of the induction electrode at point O: ρ i = ε * D foi , where ρ i is the surface charge density of the induction electrode at any point O near the surface on one side of the i-th target object, and ε is the dielectric constant in the air; i = 1, 2, 3... n;

[0066] Step S13: Calculate the total amount of charge carried on the side of the induction electrode close to the target object i. The calculation formula is: Q i = ∫ρ i dS = ∫εD foi dS, where Q i is the total amount of charge carried on the side of the induction electrode close to the i-th target object, and S is the area of the induction electrode; When the human body's electric field changes, the total amount of charge on the induction electrode changes, and thus the induced current generated on the induction electrode is: where I i is the induced current generated by the i-th target object on the induction electrode, ΔQ i is the change in the total amount of charge of the i-th target object, and Δt is the time corresponding to the change in the total amount of charge; i = 1, 2, 3... n; t = 1, 2, 3... T;

[0067] Step S14: Statistically analyze the induced currents generated on the induction electrodes under different actions of n target objects; under the action of picking up an item, the induced currents generated by the n target objects on the induction electrodes are sequentially marked as I 1a 、I 2a 、I 3a ……I na ; under the action of waving, the induced currents generated by the n target objects on the induction electrodes are sequentially marked as I 1b 、I 2b 、I 3b ……I nb ; Take the mean of the two groups of induced current data: where, I a ′ is the mean value of the induced currents generated by the n target objects on the induction electrodes under the action of picking up an item, where, I b ′ is the mean value of the induced currents generated by the n target objects on the induction electrodes under the action of waving.

[0068] In this embodiment, it should be specifically noted that the deep neural network model constructed in step S05 is a convolutional neural network model, including an input layer, a convolutional layer, a pooling layer, a Flatten layer, a fully connected layer, and an output layer;

[0069] The convolutional layer is mainly used to extract the features of the input signal and is the core part of the convolutional neural network. It uses discrete convolution operations. The output of the convolutional layer is expressed as: Y = q(X * W + h), where Y is the output feature map of the convolutional layer, q(·) is the activation function, * is the convolution operation, X is the input signal sample, W is the weight coefficient, and h is the bias; the size of the output feature map can be expressed as: where, U is the size of the output feature map, R is the size of the input feature map, M is the size of the convolution kernel, P is the zero-padding size, and B is the stride;

[0070] The pooling layer is mainly used to reduce the parameter order of magnitude and feature dimension. It is usually used after the convolutional layer. The pooling layer reduces the dimension of the feature map by using the output of the partial region feature map instead of the output of the entire region feature map. The pooling layer reduces the number of parameters required during training by performing downsampling operations on the larger feature map output by the convolutional layer;

[0071] The Flatten layer can flatten the input data, that is, convert the multi-dimensional input data into one-dimensional data. The output signal obtained after the input signal sample is processed by the convolutional neural network or the recurrent neural network is multi-dimensional, while the input of the fully connected layer should be one-dimensional signal. Therefore, the Flatten layer can be applied between the convolutional layer and the fully connected layer to filter the signal;

[0072] The fully connected layer is mainly used to output the final result and perform batch normalization processing. The batch normalization processing avoids the change of the sample feature distribution through the method of transformation and reconstruction, and can normalize the output data of the neural network;

[0073] The input layer is used to input the preprocessed electrostatic signal, and the output layer is used to output the electrostatic signal features.

[0074] In this embodiment, it should be specifically noted that the specific method for constructing the soft mapping relationship between the electrostatic signal features and different motion postures in step S05 is as follows:

[0075] Define a two-dimensional matrix Mapping[M][N] to represent the soft mapping relationship, where M is the number of electrostatic signal features and N is the number of motion postures; analyze the electrostatic signal features of n target objects, and count the motion postures corresponding to the electrostatic signal features of n target objects, which is the analysis result; set the corresponding element value in the Mapping[M][N] matrix to 1 according to the analysis result; for example, if the electrostatic signal feature 1 corresponds to the motion posture 2, then set Mapping[1][2].

[0076] In this embodiment, it should be specifically noted that after obtaining the induced current and electrostatic signal features of the person to be measured in step S06, output the actions and motion postures of the person to be measured for interaction, and select the one closest to the induced current of the person to be measured as the action of the person to be measured; for example, if the induced current of the person to be measured is closest to I a ′, then identify the action of the person to be measured as picking up an item. If the induced current of the person to be measured is closest to I b ′, then identify the action of the person to be measured as waving.

[0077] Embodiment 2:

[0078] As Figure 2 shown, this embodiment provides an identification and interaction system based on electrostatic signals, including a region division module, a data acquisition module, a data preprocessing module, an action recognition and analysis module, a motion posture recognition and analysis module, and an identification and interaction module;

[0079] The region division module is used to divide the target object into regions and mark the charge amount; the target object is the human body surface of the person to be measured; the region division is the left hand, right hand, head, torso, and clothing of the target object, denoted as 1, 2, 3, 4, 5 in sequence; the charge amounts of each region are marked as Q1, Q2, Q3, Q4, Q5 in sequence; the mutual capacitances of each region with the surrounding space object p are C 1p 、C 2p 、C 3p 、C 4p 、C 5p ;

[0080] The data acquisition module is used to acquire target data to form a data set: acquire human charged data of n target objects under different actions and electrostatic signals under different motion postures within T time points to form a data set;

[0081] The data preprocessing module is used to preprocess the data set to facilitate further feature extraction;

[0082] The action recognition and analysis module is used to analyze the data after current detection in the data preprocessing module to obtain the induced current under different actions;

[0083] The motion posture recognition and analysis module is used to construct a deep neural network model to extract features from the preprocessed electrostatic signals in step S03, and construct a soft mapping relationship between the electrostatic signal features and different motion postures;

[0084] The recognition and interaction module is used to obtain the induced current value and electrostatic signal features of the person to be tested, recognize the actions and motion postures of the person to be tested, and perform interactions.

[0085] Embodiment 3:

[0086] As Figure 3 shown, this embodiment provides an electronic device 500, including a bus 501, one or more CPUs 502, a ROM 503, a RAM 504, a communication port 505 connected to a network, an input / output 506, a hard disk 507, etc. The storage device in the electronic device 500, such as the ROM 503 or the hard disk 507, can store the recognition and interaction method / system based on electrostatic signals provided by this application; further, the electronic device 500 may further include a user interface 508; Figure 3 The architecture shown is only exemplary, and when implementing different devices, one or more components in the shown electronic device can be omitted according to actual needs Figure 3 shown in the electronic device.

[0087] Embodiment 4:

[0088] As Figure 4 shown, this embodiment provides a computer-readable storage medium 600, on which computer-readable instructions are stored; when the computer-readable instructions are run by a processor, the recognition and interaction method / system based on electrostatic signals according to the embodiments of this application described with reference to the above drawings can be executed; the storage medium 600 includes but is not limited to, for example, volatile memory and / or non-volatile memory; volatile memory may include, for example, random access memory (RAM) and cache memory, etc.; non-volatile memory may include, for example, read-only memory (ROM), hard disk, flash memory, etc.

[0089] In this embodiment, it should be specifically noted that the main difference between this embodiment and the prior art is that this embodiment includes steps S04 and S05. By analyzing the data detected by the current in step S03, the induced current under different actions is obtained; a deep neural network model is constructed to extract the features of the preprocessed electrostatic signal in step S03, and a soft mapping relationship between the electrostatic signal features and different motion postures is constructed; the electrostatic induction current signal contains human motion information. Analyzing the induced current can identify the corresponding human motion, and at the same time, extracting and mapping the electrostatic signal features can obtain the motion posture corresponding to the electrostatic signal features. Action recognition and motion posture judgment are completed based on the electrostatic signal. The acquisition method is simple, effectively avoiding the total amount of inaccurate measurement caused by errors in instrument installation, effectively improving the recognition effect, and protecting the privacy of users.

[0090] Finally: The above is only the preferred embodiment of the present invention and is not used to limit the present invention. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

[0091] The above is only the specific implementation manner of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art can easily think of changes or replacements within the technical scope disclosed in the present application, and all should be covered by the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the protection scope of the claims.

Claims

1. An identification and interaction method based on electrostatic signals, characterized in that: It includes the following steps: Step S01: Perform regional division and charge quantity marking on the target object; the target object is the human body surface of the person to be measured; The described area is divided into the left hand, right hand, head, torso, and clothing of the target object, denoted as 1, 2, 3, 4, and 5 in sequence; the electric charges of each area are marked as Q1, Q2, Q3, Q4, and Q5 in sequence; the mutual capacitances of each area with respect to the surrounding space object p are C 1p , C 2p , C 3p , C 4p , C 5p ; Step S02: Collect target data to form a data set: Collect the human body charge data of n target objects under different actions and the static electricity signals under different motion postures within T time points to form a data set; Step S03: Preprocess the data set to facilitate further feature extraction; Detect the current of the charge data, that is, the generated charge change amount, and convert it into a voltage signal for amplification and filtering to convert the weak induced charge quantity into a signal that can be processed; Step S04: Analyze the data after current detection in Step S03 to obtain the induced current under different actions; Step S05: Construct a deep neural network model to extract features from the preprocessed static electricity signals in Step S03, and construct a soft mapping relationship between the static electricity signal features and different motion postures; Step S06: Obtain the induced current value and static electricity signal features of the person to be measured, identify the actions and motion postures of the person to be measured, and perform interaction.

2. The recognition and interaction method based on electrostatic signals according to claim 1, wherein: The preprocessing methods include data interception, data denoising, and debiasing; Data interception refers to the process of selecting part of the data from the original data set for processing and analysis; Intercept the collected original data, that is, the static electricity signals under different motion postures, to obtain signal inputs of the same length; The method of data denoising is to use the smooth function to perform denoising processing on the static electricity signals under different postures. The formula of the smooth function is: y′ = smooth(y0, span, method), where y′ is the column vector after smooth processing, and y0 is the column vector before smooth processing; Debiasing uses the dyaddown function, which performs binary sampling on the time series and extracts one element every other element to achieve the purpose of downsampling. The calculation formula of the dyaddown function is: y = dyaddown(x, EVENODD), where x is the time series and y is the static electricity signal data after debiasing; When EVENODD is 0, even sampling should be performed on the data, that is, sampling starts from the second element of x; When EVENODD is 1, odd sampling is performed on the data, that is, sampling starts from the first element of x.

3. The identification and interaction method based on electrostatic signals according to claim 1, wherein: The electric charges of each region of the target object can be represented by the electric potential equations: Among them, U1 is the electric potential of region 1, i.e., the left hand; U2 is the electric potential of region 2, i.e., the right hand; U3 is the electric potential of region 3, i.e., the head; U4 is the electric potential of region 4, i.e., the torso; U5 is the electric potential of region 5, i.e., the clothing; C 11 is the self-capacitance of region 1, C 22 is the self-capacitance of region 2, C 33 is the self-capacitance of region 3, C 44 is the self-capacitance of region 4, C 55 is the self-capacitance of region 5; C 15 is the mutual capacitance between region 1 and region 5, C 25 is the mutual capacitance between region 2 and region 5, C 35 is the mutual capacitance between region 3 and region 5, C 45 is the mutual capacitance between region 4 and region 5; C 15 =C 51 , C 25 =C 52 , C 35 =C 53 , C 45 =C 54 .

4. The identification and interaction method based on electrostatic signals according to claim 1, wherein: The specific method for obtaining the induced current under different actions in Step S04 is: Step S11: Place an induction electrode at a certain distance from the target object; denote the human body electric field as D r , denote the induction electric field as D g ; Step S12: Denote the normal electric field at any point O near one side surface of the sensing electrode close to the target object i as D foi , and the calculation formula is: D foi = D ri + D gi , where D ri is the human body electric field of the i-th target object, and D gi is the induced electric field of the i-th target object; based on Gauss's theorem, the charge density on the surface of the sensing electrode at point O is obtained: ρ i = ε * D foi , where ρ i is the charge density on the surface of the sensing electrode at any point O near one side surface of the i-th target object, and ε is the dielectric constant in air; i = 1, 2, 3... n; Step S13: Calculate the total charge carried on the side of the induction electrode close to the target object i. The calculation formula is: Q i = ∫ρ i dS = ∫εD foi dS, where Q i is the total charge carried on the side of the induction electrode close to the i-th target object, and S is the area of the induction electrode; the induced current generated on the induction electrode is: where I i is the induced current generated by the i-th target object on the induction electrode, ΔQ i is the change in the total charge of the i-th target object, and Δt is the time corresponding to the change in the total charge; i = 1, 2, 3... n; Step S14: Statistically analyze the induced currents generated on the induction electrodes under different actions of n target objects; under the action of picking up an item, the induced currents generated by the n target objects on the induction electrodes are sequentially marked as I 1a 、I 2a 、I 3a ……I na ; under the action of waving, the induced currents generated by the n target objects on the induction electrodes are sequentially marked as I 1b 、I 2b 、I 3b ……I nb ; Take the mean values of the two sets of induced current data: where, I a ′ is the mean value of the induced currents generated by the n target objects on the induction electrodes under the action of picking up an item, where, I b ′ is the mean value of the induced currents generated by the n target objects on the induction electrodes under the action of waving.

5. The identification and interaction method based on electrostatic signals according to claim 1, wherein: The specific method for constructing the soft mapping relationship between the static electricity signal features and different motion postures in Step S05 is: Define a two-dimensional matrix Mapping[M][N] to represent the soft mapping relationship, where M is the number of static electricity signal features and N is the number of motion postures; Analyze the static electricity signal features of n target objects, and count the motion postures corresponding to the static electricity signal features of n target objects, which is the analysis result.

6. An identification and interaction system based on electrostatic signals, which is used to implement the method for identifying and interacting based on electrostatic signals according to any one of claims 1-5 above, and is characterized in that: It includes a regional division module, a data collection module, a data preprocessing module, an action recognition and analysis module, a motion posture recognition and analysis module, and a recognition and interaction module; The area division module is used to divide the target object into areas and mark the charge amounts; the target object is the human body surface of the person to be measured; the area division is the left hand, right hand, head, torso, and clothing of the target object, denoted as 1, 2, 3, 4, 5 in sequence; the charge amounts of each area are marked as Q1, Q2, Q3, Q4, Q5 in sequence; the mutual capacitances of each area with respect to the surrounding space object p are C 1p , C 2p , C 3p , C 4p , C 5p ; The data acquisition module is used to acquire target data to form a data set: acquire the human body charged data of n target objects under different actions and the static electricity signals under different movement postures at T time points to form a data set; The data preprocessing module is used to preprocess the data set to facilitate further feature extraction; perform current detection on the charged data, i.e., the generated charge change amount, and convert it into a voltage signal for amplification and filtering to convert the weak induced charge amount into a signal that can be processed; The action recognition and analysis module is used to analyze the data after current detection in the data preprocessing module to obtain the induced current under different actions; The movement posture recognition and analysis module is used to construct a deep neural network model to extract features from the preprocessed static electricity signals in step S03, and construct a soft mapping relationship between the static electricity signal features and different movement postures; The recognition and interaction module is used to obtain the induced current value and static electricity signal features of the person to be tested, identify the actions and movement postures of the person to be tested, and perform interaction.

7. An electronic device, comprising: One or more processors; A storage device having one or more programs stored thereon; When the one or more programs are executed by the one or more processors, the one or more processors implement the electrostatic signal-based recognition and interaction method according to any one of claims 1-5.

8. A computer-readable storage medium having executable instructions stored thereon, and when the instructions are executed by a processor, the processor implements the electrostatic signal-based recognition and interaction method according to any one of claims 1-5.

Images