Control system and method for identifying behaviors and actions of people by using electric heating towel rack

Abstract

The invention provides a human behavior and action recognition control system and method for an electric heating towel rack, and solves the problems of single falling detection mode and the like, the system comprises a main rack body with a built-in electric heating element, the lower end of the main rack body is connected with a control box, a sensing element is installed in the control box, and the sensing element is connected with a control module through a data processing module. The control module is connected with an alarm module and a communication module, and the sensing element is a radar sensor, an infrared sensor and an ultrasonic sensor. The method has the advantages of high detection precision, good applicability and the like.

Description

Technical Field

[0001] The invention belongs to the technical field of fall alarm systems, and in particular relates to a human behavior recognition and control system for an electric towel rack. Background Art

[0002] China has now entered an aging society, and the biggest worry is that no one knows when an accident happens. Falling is a sudden, involuntary, and unintentional change in body position, and falls of vulnerable groups such as the elderly can bring great danger. Especially in small and private places such as bathrooms, once an accident occurs, outsiders cannot enter, and more serious problems may occur. At present, there are voice recognition and alarm methods in this field, but they cannot solve the unconscious behavior after involuntary falls. In addition, the existing fall behavior detection methods are relatively single, and their detection accuracy cannot be guaranteed and the detection effect is greatly affected by the environment.

[0003] In order to solve the shortcomings of the existing technology, people have conducted long-term exploration and proposed various solutions. For example, a Chinese patent document discloses a voice recognition bathroom fall alarm system [202111195203.7], which includes a sound collection terminal, a control computing platform, a hospital management platform and a nurse station call platform; the sound collection terminal is installed in the bathroom of each ward, the sound collection terminal collects the sound in the bathroom, and analyzes the sound to obtain the corresponding voiceprint information, the sound collection terminal is provided with a feature code, and the sound collection terminal sends the acquired voiceprint information and feature code to the control computing platform; the control computing platform includes a central processing unit, a voiceprint information playback module, a communication module, an input module, a comparison module, an alarm module, a voiceprint database, a voiceprint storage module and a sound collection terminal position storage module; the present invention relates to the field of fall alarm technology.

[0004] The above solution solves the problem of alarm after a fall to a certain extent, but the solution still has many shortcomings, such as the relatively single detection method. Summary of the invention

[0005] The purpose of the present invention is to solve the above problems and provide a human behavior recognition control system for an electric towel rack with a reasonable design and multiple detection modes.

[0006] Another object of the present invention is to provide a method for controlling a human behavior recognition of an electric heated towel rack with high detection accuracy in order to solve the above-mentioned problem.

[0007] To achieve the above-mentioned purpose, the present invention adopts the following technical solutions: an electric towel rack for human behavior recognition and control system, including a main frame with a built-in electric heating element, a control box connected to the lower end of the main frame, a sensor element installed in the control box, the sensor element is connected to the control module through a data processing module, the control module is connected to an alarm module and a communication module, and the sensor element is selected from radar sensors, infrared sensors and ultrasonic sensors.

[0008] A method for controlling an electric towel rack to recognize human behavior. The above-mentioned electric towel rack to recognize human behavior control system comprises the following steps:

[0009] S1: Build database;

[0010] S2: pre-processing radar data, infrared data and ultrasonic data respectively;

[0011] S3: Perform data fusion and build a fall detection model;

[0012] S4: Test the effect of falling behavior recognition.

[0013] In the above-mentioned method for controlling the electric towel rack to recognize human behavior, step S1 includes the following steps:

[0014] S11: Classify the behavior of the actor into walking, standing still, squatting, raising hands, lying prone and falling, and then classify different ages, heights and weights to establish a behavior action database;

[0015] S12: Collect data from each person using a radar sensor, an infrared sensor, and an ultrasonic sensor, and process the data using a time window to obtain normal behavior data and abnormal behavior data.

[0016] In the above-mentioned method for controlling the electric towel rack to recognize human behavior, the time window in step S12 processes the data as follows:

[0017] S121: Determine the time window size, and set the action time window to 0.5-5s;

[0018] S122: Select an initial window, starting from when the person enters the designated space, and select a number of data points in the initial time period as the initial window;

[0019] S123: sliding the time window, and re-sorting the data in the time window from high to low in terms of the height of the actors;

[0020] S124: Perform aggregation calculation on the data in each time window and output the aggregation result.

[0021] In the above-mentioned method for controlling the electric towel rack to recognize human behavior, step S2 uses the EMD algorithm to perform signal decomposition on the radar data, and then performs Hilbert spectrum analysis on the decomposition result.

[0022] In the above-mentioned method for controlling the electric towel rack to recognize human behavior, step S2 pre-processes the infrared data by the following steps:

[0023] S21: Establish a nonlinear model of infrared signals;

[0024] S22: Select irradiance φ1 as the calibration point, and the output response V of N infrared sensors i (φ1) is averaged to obtain The response output V of the infrared sensor to any irradiance i (φ) is adjusted to

[0025] S23: Set the infrared sensor's response V i Adjusted for pan response V s , and then get Calibration correction factor a i Non-uniformity correction of infrared signals in is the actual response of the ith infrared sensor, is the corrected response value;

[0026] S24: The corrected data is input into the database.

[0027] In the above-mentioned method for controlling the electric towel rack to recognize human behavior, step S2 pre-processes the ultrasonic data by the following steps:

[0028] S25: inputting the ultrasound image and determining the external matrix of the area where the actor is located;

[0029] S26: Obtain the coordinates of the diagonal endpoints of the region according to the matrix (X min , Y min )、(X max , Y max ), the center point of the area is L c The coordinates (X c , Y c )for

[0030] S27: Center point L c Cut the image at the center to get the ROI area.

[0031] In the above-mentioned method for controlling the electric towel rack to recognize human behavior, step S3 includes the following steps:

[0032] S31: Gaussian process regression is used to regress the data of radar sensor, infrared sensor and ultrasonic sensor respectively to obtain the mean value μ of each data GPRi and variance

[0033] S32: Introduce expert experience to correct the data of radar sensors, infrared sensors, and ultrasonic sensors to determine the fidelity of each data

[0034] S33: Combine the uncertainties of each data to obtain the total uncertainty μ Ti and The following algorithm is used to fuse the total uncertainty of each data:

[0035] μ Ti (x) = μ GPRi (x),

[0036] S34: The data of radar sensor, infrared sensor and ultrasonic sensor are assumed to be relatively independent, and their weights are inversely proportional to the data uncertainty variance. The specific expression is as follows:

[0037]

[0038] In the above-mentioned method for controlling the electric towel rack to recognize human behavior, step S3 includes the following steps:

[0039] S35: inputting grayscale images of the radar sensor, the infrared sensor and the ultrasonic sensor;

[0040] S36: Use the GhostNet network to extract features and preliminarily classify the grayscale images;

[0041] S37: Use the sklearn model to predict the classification results to obtain the final prediction results.

[0042] In the above-mentioned method for controlling the electric towel rack to recognize human behavior, step S4 includes the following steps:

[0043] S41: Build an experimental scene. The experimental scene uses a closed environment and selects multiple height positions to assemble sensor elements respectively;

[0044] S42: Setting scenes with or without obstacles, the person enters different scenes and performs regular actions, unconventional actions and falling actions respectively, and the sensor elements collect data respectively;

[0045] S43: Establish a statistical table of detection results, set an accuracy threshold and calculate the recognition accuracy. If the recognition accuracy is lower than the specified threshold, rebuild the fall detection model.

[0046] Compared with the existing technology, the advantages of the present invention are: the control system uses a variety of sensor elements mounted on the electric towel rack to detect the actions of people in closed spaces such as bathrooms, and sends help and alarm signals when a fall occurs; radar data, infrared data and ultrasonic data are fused, and a corresponding detection model is constructed, so that the control system can be applied to complex monitoring environments; during experimental verification, the control system is assembled at different heights, and the detection model can be rebuilt according to changes in the environmental layout to eliminate blind spots in behavioral action recognition. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] Figure 1 It is a structural schematic diagram of the main frame of the present invention;

[0048] Figure 2 It is a structural principle diagram of the control box of the present invention;

[0049] Figure 3 It is a schematic diagram of the method of the present invention;

[0050] As shown in the figure, the main frame 1, the control box 2, the sensor element 3, the data processing module 4, the control module 5, the alarm module 51, and the communication module 52. DETAILED DESCRIPTION

[0051] The present invention is further described in detail below with reference to the accompanying drawings and specific embodiments.

[0052] like Figure 1-3As shown, a control system for identifying human behavior and action of an electric towel rack includes a main frame 1 with a built-in electric heating element. The electric heating element adopts hot blowing and is arranged at the lower end of the main frame 1. A windshield is installed at the upper end of the main frame 1 to ensure the concentration of hot blowing. A control box 2 is connected to the lower end of the main frame 1. A sensor element 3 is installed in the control box 2. The sensor element 3 is connected to the control module 5 through a data processing module 4. The control module 5 is connected to an alarm module 51 and a communication module 52. The sensor element 3 uses a radar sensor, an infrared sensor and an ultrasonic sensor. The radar sensor uses a high-precision millimeter wave radar, which can continuously transmit and receive millimeter wave signals at a high frequency and perform a 360-degree full-scale scan of the surrounding environment. By analyzing the reflected signal, the position, movement speed and trajectory of the person can be accurately obtained. The infrared sensor uses a pyroelectric infrared sensor, which is extremely sensitive to the specific wavelength infrared emitted by the human body, can monitor the presence of the human body in real time, and judge the distance and proximity between the person and the towel rack according to the change of infrared intensity. The ultrasonic sensor emits ultrasonic waves and receives echoes, using the propagation time and speed of sound waves in the air to calculate the distance between a person and the towel rack. Multiple ultrasonic sensors work together to more comprehensively sense a person's position and movements.

[0053] Specifically, the data processing module 4 in the present application receives the raw data from the sensor element 3, performs noise reduction, feature extraction and pattern recognition on the data to accurately identify the specific behavior of the person, and updates and stores the processed data in real time to provide timely and accurate decision-making basis for the control module 5. In addition to performing action recognition processing, the control module 5 also needs the heating power, temperature setting and working mode switching of the electric towel rack. The communication module 52 uses a 4G / 5G unit for remote monitoring and alarm.

[0054] A method for controlling human behavior by using an electric towel rack is different from a conventional wearable fall detection method. It uses sensors arranged in a closed space to capture motion. The non-contact recognition method does not affect the daily behavior of the person. The method specifically includes the following steps:

[0055] S1: Building a database. In addition to data collection for a single person in a closed scene, it is also necessary to consider data collection for multiple people in a closed scene. The corresponding sensor element 3 also needs to be adjusted in multiple directions.

[0056] S2: Preprocess the radar data, infrared data and ultrasonic data respectively. This step mainly ensures the integrity of subsequent data fusion. Specifically, it is necessary to remove the erroneous data and fill in the missing data. Since the collected data is also mixed with random noise, filtering and noise reduction processing is often required;

[0057] S3: Perform data fusion, feature extraction and classification to build a fall detection model;

[0058] S4: Verify the effect of fall behavior recognition. The verification method mainly uses setting thresholds until the specified detection accuracy is achieved.

[0059] Specifically, step S1 includes the following steps:

[0060] S11: Classify the behavior of the actor into walking, standing still, squatting, raising hands, lying prone and falling, and then classify different ages, heights and weights to establish a behavior action database;

[0061] S12: Collect data from each person using a radar sensor, an infrared sensor, and an ultrasonic sensor, and process the data using a time window to obtain normal behavior data and abnormal behavior data.

[0062] In-depth, the time window in step S12 processes the data as follows:

[0063] S121: Determine the time window size, and set the action time window to 0.5-5s;

[0064] S122: Select an initial window, starting from when the person enters the designated space, and select a number of data points in the initial time period as the initial window;

[0065] S123: sliding the time window, and re-sorting the data in the time window from high to low in terms of the height of the actors;

[0066] S124: Perform aggregation calculation on the data in each time window and output the aggregation result.

[0067] Furthermore, step S2 mainly processes the point cloud data collected by the radar sensor, uses the EMD algorithm to perform signal decomposition on the radar data, and then performs Hilbert spectrum analysis on the decomposition result.

[0068] Specifically, it is necessary to obtain all the extreme values ​​of the radar signal sequence, and obtain the peak / trough fitting curve through spline interpolation of the extreme values ​​of the signal sequence, that is, the envelope of the radar signal; then the extreme value curves are averaged to obtain the average envelope, and the average envelope is subtracted from the original signal to obtain the intermediate signal, and then the intrinsic mode function is determined. In the entire signal data segment, the number of extreme points and the number of zero-crossing points are equal, and the average value of the upper envelope formed by the maximum point and the lower envelope formed by the minimum point at any time is zero. The intermediate signal that meets the above conditions is the intrinsic mode function component of the original radar signal.

[0069] Furthermore, since the control system is installed in a closed environment such as a bathroom, the non-uniformity in the space causes a large difference in the response rate of the infrared sensor. If some infrared sensors are in a blind spot, they cannot sense the change in light intensity. In addition, at different times, the sensitivity of the infrared sensor to the change in light intensity is also different. The response rate of multiple drying sensors in the control system changes nonlinearly. Therefore, step S2 uses the following steps to pre-process the infrared data:

[0070] S21: Establish a nonlinear model of infrared signals;

[0071] S22: Select irradiance φ1 as the calibration point, and the output response V of N infrared sensors i (φ1) is averaged to obtain The response output V of the infrared sensor to any irradiance i (φ) is adjusted to

[0072] S23: Set the infrared sensor's response V i Adjusted for pan response V s , and then get Calibration correction factor a i Non-uniformity correction of infrared signals in is the actual response of the ith infrared sensor, is the corrected response value;

[0073] S24: The corrected data is input into the database.

[0074] In addition, step S2 uses the following steps to pre-process the ultrasonic data:

[0075] S25: inputting the ultrasound image and determining the external matrix of the area where the actor is located;

[0076] S26: Obtain the coordinates of the diagonal endpoints of the region according to the matrix (X min , Y min )、(X max , Y max ), the center point of the area is L c The coordinates (X c , Y c )for

[0077] S27: Center point L c Cut the image at the center to get the ROI area.

[0078] Meanwhile, step S3 includes the following steps:

[0079] S31: Gaussian process regression is used to regress the data of radar sensor, infrared sensor and ultrasonic sensor respectively to obtain the mean value μ of each data GPRi and variance

[0080] S32: Introduce expert experience to correct the data of radar sensors, infrared sensors, and ultrasonic sensors to determine the fidelity of each data

[0081] S33: Combine the uncertainties of each data to obtain the total uncertainty μ Ti and The following algorithm is used to fuse the total uncertainty of each data:

[0082] μ Ti (x) = μ GPRi (x),

[0083] S34: The data of radar sensor, infrared sensor and ultrasonic sensor are assumed to be relatively independent, and their weights are inversely proportional to the data uncertainty variance. The specific expression is as follows:

[0084]

[0085] It can be seen that step S3 includes the following steps:

[0086] S35: inputting grayscale images of the radar sensor, the infrared sensor and the ultrasonic sensor;

[0087] S36: Use the GhostNet network to extract features and preliminarily classify the grayscale images;

[0088] S37: Use the sklearn model to predict the classification results to obtain the final prediction results.

[0089] Obviously, step S4 includes the following steps:

[0090] S41: Building an experimental scene, wherein a closed environment is selected for the experimental scene, and sensor elements 3 are respectively installed at multiple height positions;

[0091] S42: Setting scenes with or without obstacles, the person enters different scenes and performs regular actions, unconventional actions and falling actions respectively, the sensor element 3 collects data respectively, and counts falling as a positive class and other actions as a negative class during detection;

[0092] S43: Establish a statistical table of detection results, set an accuracy threshold and count the recognition accuracy. If the recognition accuracy is lower than the specified threshold, rebuild the fall detection model. For the detection model, TP, TN, FP and FN are used to represent the number of actual positive class predicted positive class, actual negative class predicted negative class, actual negative class predicted positive class and actual positive class predicted negative class respectively. The performance of the detection model is as follows:

[0093]

[0094] Sensitivity indicates detection sensitivity, and Specificity indicates detection false alarm prevention rate. During the detection process, it is necessary to ensure that the detection sensitivity and the detection false alarm prevention rate are not lower than the specified threshold.

[0095] To sum up, the principle of this embodiment is that the electric towel rack is installed in a closed space such as a bathroom, and its sensor element 3 uses multiple sensing methods such as radar, infrared and ultrasonic waves to collect the behavior of the actor, and multiple data are integrated with each other to ensure the detection sensitivity. The control module 5 relies on the fall detection model to determine whether the actor falls, and thus sends an alarm signal in time.

[0096] The specific embodiments described herein are merely examples of the spirit of the present invention. Those skilled in the art may make various modifications or additions to the specific embodiments described or replace them in similar ways, but they will not deviate from the spirit of the present invention or exceed the scope defined by the appended claims.

[0097] Although the terms such as main frame 1, control box 2, sensor element 3, data processing module 4, control module 5, alarm module 51, communication module 52 are used more frequently in this article, the possibility of using other terms is not excluded. These terms are used only to more conveniently describe and explain the essence of the present invention; interpreting them as any additional limitation is contrary to the spirit of the present invention.

Claims

1. A human behavior recognition control system for an electric towel rack, comprising a main frame (1) with a built-in electric heating element, the lower end of the main frame (1) is connected to a control box (2), a sensor element (3) is installed in the control box (2), the sensor element (3) is connected to a control module (5) through a data processing module (4), the control module (5) is connected to an alarm module (51) and a communication module (52), characterized in that: The sensor element (3) is selected from radar sensors, infrared sensors and ultrasonic sensors.

2. A method for controlling an electric towel rack to recognize human behavior, using the electric towel rack to recognize human behavior control system as claimed in claim 1, characterized in that: The steps include: S1: Build database; S2: pre-processing radar data, infrared data and ultrasonic data respectively; S3: Perform data fusion and build a fall detection model; S4: Test the effect of falling behavior recognition.

3. The method for controlling the electric towel rack to recognize human behavior according to claim 2, characterized in that: The step S1 comprises the following steps: S11: Classify the behavior of the actor into walking, standing still, squatting, raising hands, lying prone and falling, and then classify different ages, heights and weights to establish a behavior action database; S12: Collect data from each person using a radar sensor, an infrared sensor, and an ultrasonic sensor, and process the data using a time window to obtain normal behavior data and abnormal behavior data.

4. The method for controlling the electric heated towel rack to recognize human behavior according to claim 3, characterized in that: In step S12, the time window processes the data as follows: S121: Determine the time window size, and set the action time window to 0.5-5s; S122: Select an initial window, starting from when the person enters the designated space, and select a number of data points in the initial time period as the initial window; S123: sliding the time window, and re-sorting the data in the time window from high to low in terms of the height of the actors; S124: Perform aggregation calculation on the data in each time window and output the aggregation result.

5. The method for controlling the electric heated towel rack to recognize human behavior according to claim 2, characterized in that: The step S2 uses the EMD algorithm to perform signal decomposition on the radar data, and then performs Hilbert spectrum analysis on the decomposition result.

6. The method for controlling the electric heated towel rack to recognize human behavior according to claim 2, characterized in that: The step S2 uses the following steps to pre-process the infrared data: S21: Establish a nonlinear model of infrared signals; S22: Select irradiance φ1 as the calibration point, and the output response V of N infrared sensors i (φ1) is averaged to obtain The response output V of the infrared sensor to any irradiance i (φ) is adjusted to S23: Set the infrared sensor's response V i Adjusted for pan response V s , and then get Calibration correction factor a i Non-uniformity correction of infrared signals in is the actual response of the ith infrared sensor, is the corrected response value; S24: The corrected data is input into the database.

7. The method for controlling the electric towel rack to recognize human behavior according to claim 2, characterized in that: The step S2 uses the following steps to pre-process the ultrasonic data: S25: inputting the ultrasound image and determining the external matrix of the area where the actor is located; S26: Obtain the coordinates of the diagonal endpoints of the region according to the matrix (x min ,y min )、(x max ,y max ), the center point of the area is L c The coordinates (X c , Y c )for S27: Center point L c Cut the image at the center to get the ROI area.

8. The method for controlling the electric heated towel rack to recognize human behavior according to claim 2, characterized in that: The step S3 comprises the following steps: S31: Gaussian process regression is used to regress the data of radar sensor, infrared sensor and ultrasonic sensor respectively to obtain the mean value μ of each data GPRi and variance S32: Introduce expert experience to correct the data of radar sensors, infrared sensors, and ultrasonic sensors to determine the fidelity of each data S33: Combine the uncertainties of each data to obtain the total uncertainty μ Ti and The following algorithm is used to fuse the total uncertainty of each data: S34: The data of radar sensor, infrared sensor and ultrasonic sensor are assumed to be relatively independent, and their weights are inversely proportional to the data uncertainty variance. The specific expression is as follows:

9. The method for controlling the electric heated towel rack to recognize human behavior according to claim 2, characterized in that: The step S3 comprises the following steps: S35: inputting grayscale images of the radar sensor, the infrared sensor and the ultrasonic sensor; S36: Use the GhostNet network to extract features and preliminarily classify the grayscale images; S37: Use the sklearn model to predict the classification results to obtain the final prediction results.

10. The method for controlling the electric heated towel rack to recognize human behavior according to claim 2, characterized in that: The step S4 comprises the following steps: S41: Building an experimental scene, wherein the experimental scene is a closed environment, and multiple height positions are selected to respectively assemble sensor elements (3); S42: Setting a scene with or without obstacles, and the person enters different scenes to perform conventional actions, unconventional actions, and falling actions, and the sensor element (3) collects data respectively; S43: Establish a statistical table of detection results, set an accuracy threshold and calculate the recognition accuracy. If the recognition accuracy is lower than the specified threshold, rebuild the fall detection model.

Images