Gesture recognition control system and control method

By establishing and deploying gesture state models, identifying and distinguishing gesture commands from different people, the problem that existing systems cannot distinguish unconscious or malicious commands is solved, and the accuracy and automation of intelligent control are improved.

CN118942150AInactive Publication Date: 2024-11-12CHENGDU NORMAL UNIV
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202410928276.X
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-07-11
Publication Date
2024-11-12
Estimated Expiration
Not applicable · inactive patent

AI Technical Summary

Technical Problem

The existing gesture recognition control system cannot distinguish between unconscious commands or malicious commands when performing gesture control, resulting in erroneous operations and defects in the degree of intelligent control.

Method used

By collecting and preprocessing gesture status data from different people, establishing and verifying gesture status models, using recurrent neural networks to identify gesture instructions, deploying models for real-time identification, and adjusting devices through rotation, displacement and swing drives to reduce misoperation.

Benefits of technology

It improves the accuracy of gesture recognition and the perfection of intelligent control, reduces misoperation, and enhances the automation and controllability of the system.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118942150A_ABST
    Figure CN118942150A_ABST
Patent Text Reader

Abstract

The invention discloses a gesture recognition control system and method, and belongs to the technical field of gesture recognition control systems. Comprising the steps of collecting gesture state data of different persons, collecting video data of gesture instructions of the different persons, performing preprocessing of frame extraction and size adjustment on the gesture state data of the different persons, establishing gesture state models of the different persons by utilizing the gesture state data of the different persons, and performing training through the gesture state data of the different persons. The method comprises the following steps: training gesture state models of different people, adjusting according to a result, deploying the trained and verified gesture state models of the different people into an actual system, identifying gesture instructions of a user in real time, adjusting according to prediction, reducing misoperation, and predicting effective gestures of different people through setting the gesture state models of the different people. Therefore, the intelligent control degree is further improved under the condition that misoperation is caused due to unconscious instructions or malicious instructions, and the effect and the experience feeling are improved.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present application relates to the technical field of gesture recognition control systems, and more specifically, to a gesture recognition control system and control method. Background Art

[0002] With the development of intelligent technology, home intelligence has also entered daily life, and with the advancement of intelligence, voice gesture command control has gradually improved.

[0003] The document with the prior art publication number CN103035115B provides a home appliance control system and method based on gesture recognition of an intelligent terminal. The device is used to receive infrared signals from home appliances through a wireless AP device, convert the infrared signals from home appliances into control signals and send them; convert the received control signals into infrared signals, and control the operation of home appliances with infrared signals; the intelligent terminal is used to obtain home appliance control signals, extract gesture signals, generate and store pairing data of home appliance control signals and gesture signals in a learning state; in a working state, when a gesture signal is detected, it is matched with the stored pairing data to obtain a corresponding control signal and send it to the wireless AP device. The device also includes a database, which includes an intelligent terminal database and a server database, for storing pairing data of infrared signals and control signals of home appliances. The intelligent terminal is also used to call up a virtual keyboard interface and obtain key signals on the virtual keyboard interface in a learning state, and generate key signals, control signals and gesture signal pairing data.

[0004] Although the above-mentioned prior art solution can achieve the relevant beneficial effects through the structure of the prior art, it still has the following defects: when the device performs gesture control, it is unable to distinguish between unconscious instructions or malicious instructions, resulting in erroneous operations, and the degree of intelligent control has certain defects.

[0005] In view of this, we propose a gesture recognition control system. Summary of the Invention

[0006] 1. Technical problems to be solved

[0007] The purpose of this application is to provide a control method for a gesture recognition control system, which solves the technical problem that the device in the above-mentioned background technology cannot distinguish between unconscious instructions or malicious instructions when performing gesture control, thereby causing erroneous operations and having certain defects in the degree of intelligent control, thereby achieving technical effects.

[0008] 2. Technical solution

[0009] The technical solution of this application provides a gesture recognition control system, including

[0010] Collect gesture status data of different people and video data of gesture commands of different people;

[0011] Preprocessing of gesture state data of different people, including frame extraction and size adjustment;

[0012] Establish gesture state model training for different people, and use gesture state data of different people to establish gesture state model for different people and train it through gesture state data of different people;

[0013] Evaluation and adjustment of gesture state models for different people: Use gesture state data validation sets for different people to verify and evaluate the trained gesture state models for different people, evaluate the accuracy of the models in recognizing different gesture commands, and make adjustments based on the results;

[0014] Deployment of gesture state models for different people: Deploy the trained and verified gesture state models for different people into the actual system for real-time recognition of user gesture commands;

[0015] Execute equipment and make adjustments based on predictions to reduce erroneous operations.

[0016] By adopting the above technical solution and setting gesture state models for different people, it is possible to predict effective gestures for different groups of people, thereby preventing unconscious or malicious commands and misoperations, and further improving the level of intelligent control, thereby enhancing the effect and experience.

[0017] As an optional solution of the technical solution of this application document, the gesture state data of different people are collected by a camera set on one side of the lamp holder or set in the room to collect data;

[0018] The collected gesture state data of different people are marked with action instructions, including invalid instructions such as unconscious instructions or malicious instructions.

[0019] As an optional solution to the technical solution of the present application, the gesture state data of different persons are pre-processed, and frames are extracted from the gesture state data videos of different persons, and the extraction frequency is set as needed, wherein the frame extraction rate of the video is calculated by using the following formula: [\text{Frame Extraction Rate}=\frac{\text{Number of Extracted Frames}}{\text{Total Number of Frames}}\times 100%] to determine how many frames are extracted from the video for subsequent processing and analysis;

[0020] Calculate the total length of the video in seconds: [\text{Video Length}=\frac{\text{Total Number of Frames}}{\text{Frame Rate}}] where the frame rate indicates the number of frames per second in the video.

[0021] Understanding the usage of different gestures and calculating the gesture frequency: [\text{Gesture Frequency}=\frac{\text{Number of Occurrences of Gesture}}{\text{Video Length}}] helps quantitatively analyze the collected video data, allowing the model to better understand the usage and characteristics of gesture commands.

[0022] As an optional solution to the technical solution of this application document, the gesture state model training of different people is established, a recurrent neural network architecture is selected, and the gesture state models of different people are trained using labeled data sets. During the training process, the gesture state models of different people will learn the characteristics and patterns of gesture instructions, so that different gestures can be accurately recognized.

[0023] As an optional solution of the technical solution of the present application document, the gesture state model evaluation adjustment of different people is used to evaluate the accuracy: [\text{Accuracy}=\frac{\text{Number of Correct Predictions}}{\text{Total Number of Predictions}}\times100%]

[0024] Precision evaluation: [\text{Precision}=\frac{\text{True Positives}}{\text{True Positives}+\text{False Positives}}]

[0025] Evaluation of recall: [\text{Recall}=\frac{\text{True Positives}}{\text{True Positives}+\text{False Negatives}}]

[0026] The F1 score comprehensive evaluation: [F1=2\times\frac{\text{Precision}\times\text{Recall}}{\text{Precision}+\text{Recall}}] provides a comprehensive understanding of the model's performance in terms of accuracy, precision, and recall in recognizing different gesture commands.

[0027] As an optional solution of the technical solution of this application document, the execution device includes

[0028] lamp holder;

[0029] A control box, the control box being rotatably driven and arranged outside the lamp holder;

[0030] A lamp stand, the lamp stand being driven to swing outside the control box by displacement;

[0031] A lampshade, which is rotatably driven by swinging and is arranged on the outer end of the lamp frame;

[0032] An energy-saving lamp is arranged inside a lampshade.

[0033] This application also discloses a method for using the aforementioned gesture recognition control system, comprising the following steps:

[0034] S1. First, collect video data of gesture commands performed by different people. Use a camera installed on the side of the lamp holder or in the room to collect data. The collected gesture state data of different people are marked as action commands, including invalid commands such as unconscious commands or malicious commands.

[0035] S2. At this time, the gesture state data of different people are pre-processed by frame extraction and size adjustment, and frames are extracted from the gesture state data videos of different people. The extraction frequency is set as needed, wherein the frame extraction rate of the video is calculated by using the following formula: [\text{Frame Extraction Rate}=\frac{\text{Number of ExtractedFrames}}{\text{Total Number of Frames}}\times 100%] to determine how many frames are extracted from the video for subsequent processing and analysis; the total length of the video is calculated in seconds: [\text{Video Length}=\frac{\text{Total Number of Frames}}{\text{Frame Rate}}] wherein the frame rate represents the number of frames contained in the video per second; the usage of different gestures is understood, and the gesture frequency is calculated: [\text{GestureFrequency}=\frac{\text{Number of Occurrences of Gesture}}{\text{VideoLength}}] helps to quantitatively analyze the collected video data, allowing the model to better understand the usage and characteristics of gesture commands;

[0036] S3. At this point, different gesture state models are established using gesture state data from different people. A recurrent neural network architecture is selected and the gesture state models from different people are trained using labeled datasets. During the training process, the gesture state models from different people will learn the characteristics and patterns of gesture commands, thereby being able to accurately recognize different gestures.

[0037] S4. At this time, the gesture state data validation set of different people is used to verify and evaluate the trained gesture state model of different people, evaluate the accuracy of the model in recognizing different gesture commands, and make adjustments based on the results. Accuracy evaluation: [\text{Accuracy}=\frac{\text{Number of Correct Predictions}}{\text{Total Number of Predictions}}\times100%] Precision evaluation: [\text{Precision}=\frac{\text{True Positives}}{\text{True Positives}+\text{False Positives}}] Recall evaluation: [\text{Recall}=\frac{\text{True Positives}}{\text{True Positives}+\text{False Negatives}}] Comprehensive evaluation through F1 score: [F1=2\times\frac{\text{Precision}\times\text{Recall}}{\text{Precision}+\text{Recall}}] fully understand the model's performance in accuracy, precision, and recall in recognizing different gesture commands;

[0038] S5. At this point, the trained and verified gesture state models of different people are deployed to the actual system for real-time recognition of user gesture commands.

[0039] S6. At this time, the energy-saving lamp is turned on and off and its direction is adjusted according to the prediction to reduce erroneous operation.

[0040] 3. Beneficial effects

[0041] One or more technical solutions provided in the technical solution of this application have at least the following technical effects or advantages:

[0042] 1. This application can predict the effective gestures of different groups of people by setting up gesture state models for different people, thereby preventing unconscious or malicious commands and misoperations, further improving the level of intelligent control and enhancing the effect and experience.

[0043] 2. This application uses a rotation drive setting to allow the energy-saving lamp to rotate to other angles for lighting, so that it can adapt to multiple directions and improve the degree of automation.

[0044] 3. This application uses a displacement-driven setting to determine different heights and internal and external positions in the vertical direction, which increases the controllable routes and makes it easier to find a suitable placement method.

[0045] 4. This application uses a swing drive setting to allow the energy-saving lamp to set its rotation angle, thereby adjusting the angle of light and further increasing the range. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] Figure 1 This is a schematic diagram of the structure of a gesture recognition control system disclosed in a preferred embodiment of the present application;

[0047] Figure 2 This is a schematic diagram of the overall structure of a gesture recognition control system disclosed in a preferred embodiment of the present application;

[0048] Figure 3 A bottom view of the overall structure of a gesture recognition control system disclosed in a preferred embodiment of the present application;

[0049] Figure 4 This is a disassembled diagram of the rotation drive structure of the gesture recognition control system disclosed in a preferred embodiment of the present application;

[0050] Figure 5 This is a disassembled diagram of the displacement drive structure of the gesture recognition control system disclosed in a preferred embodiment of the present application;

[0051] Figure 6 This is a disassembled diagram of the swing drive structure of the gesture recognition control system disclosed in a preferred embodiment of the present application;

[0052] Explanation of the numbers in the figure: 1. Lamp holder; 2. Control box; 21. Rotating table; 22. Rotating groove; 23. Rotating gear; 24. Driving gear; 25. Machine cavity; 26. First motor; 3. Lamp holder; 31. Limit rod; 32. Circumferential rod; 33. Connecting block; 34. Limit rod; 35. Connecting rod; 36. Second motor; 4. Lampshade; 41. Limit groove; 42. Third motor; 5. Energy-saving lamp. DETAILED DESCRIPTION

[0053] The present application is further described in detail below with reference to the accompanying drawings.

[0054] Reference Figure 1 , the embodiment of the present application provides a gesture recognition control system, including

[0055] Collect gesture status data of different people and video data of gesture commands of different people;

[0056] Preprocessing of gesture state data of different people, including frame extraction and size adjustment;

[0057] Establish gesture state model training for different people, and use gesture state data of different people to establish gesture state model for different people and train it through gesture state data of different people;

[0058] Evaluation and adjustment of gesture state models for different people: Use gesture state data validation sets for different people to verify and evaluate the trained gesture state models for different people, evaluate the accuracy of the models in recognizing different gesture commands, and make adjustments based on the results;

[0059] Deployment of gesture state models for different people: Deploy the trained and verified gesture state models for different people into the actual system for real-time recognition of user gesture commands;

[0060] Execute equipment and make adjustments based on predictions to reduce erroneous operations.

[0061] By setting gesture state models for different people, effective gestures of different groups of people can be predicted, thereby eliminating unconscious or malicious commands and misoperations, further improving the level of intelligent control and enhancing the effect and experience.

[0062] Reference Figure 1 and Figure 2 , the embodiment of the present application provides a gesture recognition control system, which collects gesture state data of different people, and collects data through a camera set on one side of the lamp holder 1 or set in the room;

[0063] The collected gesture state data of different people are marked with action instructions, including invalid instructions such as unconscious instructions or malicious instructions.

[0064] Preprocessing gesture state data of different people: extracting frames from gesture state data videos of different people, setting the extraction frequency as needed, wherein the frame extraction rate of the video is calculated by using the following formula: [\text{FrameExtractionRate}=\frac{\text{Number of Extracted Frames}}{\text{Total Number of Frames}}\times 100%] to determine how many frames are extracted from the video for subsequent processing and analysis;

[0065] Calculate the total length of the video in seconds: [\text{Video Length}=\frac{\text{Total Number of Frames}}{\text{Frame Rate}}] where the frame rate indicates the number of frames per second in the video.

[0066] Understanding the usage of different gestures and calculating the gesture frequency: [\text{Gesture Frequency}=\frac{\text{Number of Occurrences of Gesture}}{\text{Video Length}}] helps quantitatively analyze the collected video data, allowing the model to better understand the usage and characteristics of gesture commands.

[0067] Establish gesture state model training for different people, select the recurrent neural network (RNN) architecture, and use labeled data sets to train gesture state models for different people. During the training process, the gesture state models of different people will learn the characteristics and patterns of gesture instructions, so that they can accurately recognize different gestures.

[0068] Evaluation and adjustment of gesture state models for different people, accuracy evaluation: [\text{Accuracy}=\frac{\text{Number of Correct Predictions}}{\text{Total Number of Predictions}}\times 100%]

[0069] Precision evaluation: [\text{Precision}=\frac{\text{True Positives}}{\text{True Positives}+\text{False Positives}}]

[0070] Evaluation of recall: [\text{Recall}=\frac{\text{True Positives}}{\text{True Positives}+\text{False Negatives}}]

[0071] The F1 score comprehensive evaluation: [F1=2\times\frac{\text{Precision}\times\text{Recall}}{\text{Precision}+\text{Recall}}] provides a comprehensive understanding of the model's performance in terms of accuracy, precision, and recall in recognizing different gesture commands.

[0072] Execution equipment, including

[0073] Lamp holder 1;

[0074] The control box 2 is rotatably driven and arranged outside the lamp holder 1;

[0075] The lamp stand 3 is driven to swing by displacement and is arranged outside the control box 2;

[0076] The lampshade 4 is rotatably mounted on the outer end of the lamp frame 3 by swinging;

[0077] The energy-saving lamp 5 is arranged inside the lampshade 4 .

[0078] Reference Figure 2 and Figure 3 , the embodiment of the present application provides a gesture recognition control system, including

[0079] Lamp holder 1;

[0080] The control box 2 is rotatably driven and arranged outside the lamp holder 1;

[0081] The lamp stand 3 is driven to swing by displacement and is arranged outside the control box 2;

[0082] The lampshade 4 is driven and rotated by swinging and is arranged on the outer end of the lamp frame 3;

[0083] The energy-saving lamp 5 is arranged inside the lampshade 4 .

[0084] Reference Figure 4 , the embodiment of the present application provides a gesture recognition control system, the rotation drive includes a rotating platform 21, the rotating platform 21 is rotatably arranged inside a rotating groove 22 inside the lamp holder 1;

[0085] A rotating gear 23 is fixedly mounted on the outer wall of the rotating platform 21. A driving gear 24 is meshedly connected to the outer wall of the rotating gear 23. The driving gear 24 is rotatably mounted inside a machine cavity 25 opened inside the lamp holder 1.

[0086] A first motor 26 is further provided inside the machine cavity 25 , and an output end of the first motor 26 is coaxially and fixedly connected to the driving gear 24 .

[0087] The output end of the first motor 26 rotates to drive the driving gear 24, and the driving gear 24 engages with the rotating gear 23, which drives the rotating table 21 on the control box 2 to rotate. At this time, the energy-saving lamp 5 is rotated and changed position. Through the setting of the rotation drive, the energy-saving lamp 5 can be rotated to other angles for lighting, so that it can adapt to multiple directions and improve the degree of automation.

[0088] Reference Figure 5 The embodiment of the present application provides a gesture recognition control system, wherein the displacement drive includes a limit rod 31 fixedly arranged inside the control box 2, a circumferential rod 32 rotatably arranged in the middle of the limit rod 31, and a connecting block 33 rotatably arranged at one end of the circumferential rod 32 via a pin shaft, and one side of the connecting block 33 is rotatably connected to the outer wall of the lamp holder 3 rod via a pin shaft;

[0089] A limiting rod 34 is rotatably provided on the other side of the connecting block 33 via a pin, and one end of the limiting rod 34 is rotatably connected to one end of the limiting rod 31 via a pin;

[0090] A connecting rod 35 is rotatably provided on one side of the limiting rod 31 through a pin, and the outer end of the connecting rod 35 is rotatably connected to the outer wall of the lamp frame 3 rod through a pin;

[0091] A second motor 36 is fixedly installed inside the control box 2 . The output shaft of the second motor 36 passes through the interior of the limiting rod 31 and extends to the outside and is connected and fixed to the middle of the circumferential rod 32 .

[0092] The rotation of the output shaft of the second motor 36 drives the circumferential rod 32 to perform circular motion. At this time, the circumferential rod 32 will pull the connecting block 33 to reciprocate. At this time, the connecting block 33 will pull the lamp holder 3 to cooperate with the limit rod 34 and the connecting rod 35 to allow the energy-saving lamp 5 to swing at an angle. Through the displacement drive setting, the energy-saving lamp 5 can be determined at different heights and inside and outside positions in the vertical direction, which increases the controllable route and makes it easier to find a suitable placement method.

[0093] Reference Figure 6 , the embodiment of the present application provides a gesture recognition control system, wherein the swing drive includes a limit groove 41 provided on the outer wall of the lampshade 4, and the rod of the lamp holder 3 is rotatably arranged inside the limit groove 41;

[0094] A third motor 42 is disposed inside the limiting groove 41 , and an output shaft of the third motor 42 is fixedly connected to a rod of one of the lamp holders 3 .

[0095] The output shaft of the third motor 42 rotates to allow the lampshade 4 to rotate with the energy-saving lamp 5. Through the swing drive setting, the energy-saving lamp 5 can be set to rotate at a certain angle, thereby adjusting the angle of light and further increasing the range.

[0096] The present invention provides a gesture recognition control system, including the following steps:

[0097] S1. First, collect video data of gesture commands performed by different people. A camera is installed on one side of the lamp holder 1 or in the room to collect data. The collected gesture state data of different people are marked with action commands, including invalid commands such as unconscious commands or malicious commands.

[0098] S2. At this time, the gesture state data of different people are pre-processed by frame extraction and size adjustment, and frames are extracted from the gesture state data videos of different people. The extraction frequency is set as needed, wherein the frame extraction rate of the video is calculated by using the following formula: [\text{Frame Extraction Rate}=\frac{\text{Number of ExtractedFrames}}{\text{Total Number of Frames}}\times 100%] to determine how many frames are extracted from the video for subsequent processing and analysis; the total length of the video is calculated in seconds: [\text{Video Length}=\frac{\text{Total Number of Frames}}{\text{Frame Rate}}] wherein the frame rate represents the number of frames contained in the video per second; the usage of different gestures is understood, and the gesture frequency is calculated: [\text{GestureFrequency}=\frac{\text{Number of Occurrences of Gesture}}{\text{VideoLength}}] helps to quantitatively analyze the collected video data, allowing the model to better understand the usage and characteristics of gesture commands;

[0099] S3. At this point, different gesture state models are established using gesture state data from different people. The gesture state models are trained using the gesture state data from different people. A recurrent neural network (RNN) architecture is selected and the labeled dataset is used to train the gesture state models for different people. During the training process, the gesture state models for different people will learn the characteristics and patterns of gesture commands, thereby being able to accurately recognize different gestures.

[0100] S4. At this time, the gesture state data validation set of different people is used to verify and evaluate the trained gesture state model of different people, evaluate the accuracy of the model in recognizing different gesture commands, and make adjustments based on the results. Accuracy evaluation: [\text{Accuracy}=\frac{\text{Number of Correct Predictions}}{\text{Total Number of Predictions}}\times100%] Precision evaluation: [\text{Precision}=\frac{\text{True Positives}}{\text{True Positives}+\text{False Positives}}] Recall evaluation: [\text{Recall}=\frac{\text{True Positives}}{\text{True Positives}+\text{False Negatives}}] Comprehensive evaluation through F1 score: [F1=2\times\frac{\text{Precision}\times\text{Recall}}{\text{Precision}+\text{Recall}}] fully understand the model's performance in accuracy, precision, and recall in recognizing different gesture commands;

[0101] S5. At this point, the trained and verified gesture state models of different people are deployed to the actual system for real-time recognition of user gesture commands.

[0102] S6. At this time, the energy-saving lamp 5 is turned on and off and its direction is adjusted according to the prediction to reduce erroneous operation.

[0103] When the gesture recognition control system is needed, first, video data of gesture commands performed by different people are collected, and data is collected by a camera set on one side of the lamp holder 1 or set in the room. The collected gesture state data of different people are marked with action commands, including invalid commands such as unconscious commands or malicious commands. At this time, the gesture state data of different people are pre-processed by frame extraction and size adjustment, and frames are extracted from the gesture state data videos of different people. The extraction frequency is set as needed, wherein the frame extraction rate of the video is calculated by using the following formula: [\text{Frame ExtractionRate}=\frac{\text{Number of Extracted Frames}}{\text{Total Number of Frames}}\times 100%] to determine how many frames are extracted from the video for subsequent processing and analysis; the total length of the video is calculated in seconds: [\text{Video Length}=\frac{\text{Total Number of Frames}}{\text{Frame Rate}}] where the frame rate represents the number of frames per second in the video. To understand the usage of different gestures, the gesture frequency is calculated: [\text{Gesture Frequency}=\frac{\text{Number of Occurrences of Gesture}}{\text{Video Length}}] This helps to quantitatively analyze the collected video data, allowing the model to better understand the usage and characteristics of gesture commands. At this time, gesture state models of different people are established using gesture state data of different people. Through the gesture state data of different people, a recurrent neural network (RNN) architecture is selected, and the gesture state models of different people are trained using labeled data sets. During the training process, the gesture state models of different people will learn the characteristics and patterns of gesture commands, so that they can accurately recognize different gestures. At this time, the gesture state data verification set of different people is used to verify and evaluate the trained gesture state models of different people, evaluate the accuracy of the model in recognizing different gesture commands, and make adjustments based on the results.Accuracy evaluation: [\text{Accuracy}=\frac{\text{Number of Correct Predictions}}{\text{Total Number of Predictions}}\times 100%] Precision evaluation: [\text{Precision}=\frac{\text{True Positives}}{\text{True Positives}+\text{False Positives}}] Recall evaluation: [\text{Recall}=\frac{\text{True Positives}}{\text{True Positives}+\text{FalseNegatives}}] is comprehensively evaluated through the F1 score: [F1=2\times\frac{\text{Precision}\times\text{Recall}}{\text{Precision}+\text{Recall}}] to fully understand the performance of the model in recognizing different gesture commands, such as accuracy, precision, and recall. At this time, the trained and verified gesture state models of different people are deployed to the actual system for real-time recognition of user gesture commands. At this time, the switch and direction of the energy-saving lamp 5 are adjusted according to the prediction to reduce misoperation. Therefore, during the adjustment, the output end of the first motor 26 rotates to drive the drive gear 24, and the drive The gear 24 meshes with the rotating gear 23, which drives the rotating platform 21 on the control box 2 to rotate, allowing the energy-saving lamp 5 to rotate and change position. The rotation of the output shaft of the second motor 36 drives the circumferential rod 32 to perform a circular motion. The circumferential rod 32 will pull the connecting block 33 to reciprocate. At this time, the connecting block 33 will pull the lamp holder 3 to cooperate with the limit rod 34 and the connecting rod 35 to limit, thereby allowing the energy-saving lamp 5 to swing. Through the setting of the displacement drive, the energy-saving lamp 5 can be determined at different heights and inside and outside positions in the vertical direction, increasing the controllable route and making it easier to find a suitable placement method. The output shaft of the third motor 42 rotates to allow the lampshade 4 to rotate with the energy-saving lamp 5. Through the setting of the swing drive, the energy-saving lamp 5 can set the rotation angle, thereby adjusting the angle of the light, further increasing the range.

Claims

1. A gesture recognition control system, Features: Include: Collect gesture state data of different people, and collect video data of different people performing gesture commands; Preprocessing of gesture state data of different people, including frame extraction and size adjustment of gesture state data of different people; Establish gesture state model training for different people, and use gesture state data of different people to establish gesture state models for different people and train them through gesture state data of different people; Evaluation and adjustment of gesture state models for different people: Use gesture state data validation sets for different people to verify and evaluate the trained gesture state models for different people, evaluate the accuracy of the models in recognizing different gesture commands, and make adjustments based on the results; Deployment of gesture state models for different people: deploy the trained and verified gesture state models for different people into the actual system for real-time recognition of user gesture commands; Execute equipment and make adjustments based on predictions to reduce erroneous operations.

2. The gesture recognition control system according to claim 1, Features: The data on gesture states of different persons are collected by using a camera arranged on one side of the lamp holder (1) or arranged in the room; The collected gesture state data of different people are marked with action instructions, including invalid instructions such as unconscious instructions or malicious instructions.

3. The gesture recognition control system according to claim 1, Features: The gesture state data of different persons are preprocessed, frames are extracted from the gesture state data videos of different persons, and the extraction frequency is set as needed, wherein the frame extraction rate of the video is calculated by using the following formula: [\text{Frame Extraction Rate}=\frac{\text{Numberof Extracted Frames}}{\text{Total Number of Frames}}\times 100%] to determine how many frames are extracted from the video for subsequent processing and analysis; Calculate the total length of the video in seconds: [\text{Video Length}=\frac{\text{TotalNumber of Frames}}{\text{Frame Rate}}], where the frame rate indicates the number of frames per second in the video. Understand the usage of different gestures and calculate the gesture frequency: [\text{Gesture Frequency}=\frac{\text{Number of Occurrences of Gesture}}{\text{Video Length}}] helps to quantitatively analyze the collected video data, allowing the model to better understand the usage and characteristics of gesture commands.

4. The gesture recognition control system according to claim 1, Features: The gesture state model training for different people is established, a recurrent neural network (RNN) architecture is selected, and the gesture state models for different people are trained using labeled data sets. During the training process, the gesture state models for different people will learn the characteristics and patterns of gesture instructions, so as to accurately recognize different gestures.

5. The gesture recognition control system according to claim 1, Features: The gesture state model evaluation adjustment of different people is used to evaluate the accuracy: [\text{Accuracy}=\frac{\text{Number of CorrectPredictions}}{\text{Total Number of Predictions}}\times 100%] For precision evaluation: [\text{Precision}=\frac{\text{True Positives}}{\text{True Positives}+\text{False Positives}}] Evaluation of recall: [\text{Recall}=\frac{\text{True Positives}}{\text{True Positives}+\text{False Negatives}}] Through the comprehensive evaluation of the F1 score: [F1=2\times\frac{\text{Precision}\times\text{Recall}}{\text{Precision}+\text{Recall}}], we can fully understand the performance of the model in terms of accuracy, precision, and recall in recognizing different gesture commands.

6. The gesture recognition control system according to claim 1, Features: The execution device comprises Lamp holder (1); A control box (2), the control box (2) being rotatably arranged outside the lamp holder (1) by means of a rotational drive; A lamp stand (3), wherein the lamp stand (3) is arranged outside the control box (2) in a swinging manner by means of displacement driving; A lampshade (4), wherein the lampshade (4) is rotatably arranged at the outer end of the lamp frame (3) by swinging drive; An energy-saving lamp (5), wherein the energy-saving lamp (5) is arranged inside a lampshade (4).

7. A control method for a gesture recognition control system according to any one of claims 1 to 6, It is characterized in that The following steps are involved: S1. First, video data of gesture commands performed by different people are collected by using a camera arranged on one side of the lamp holder (1) or arranged in the room to collect data, and action commands are marked on the collected gesture state data of different people, including invalid commands such as unconscious commands or malicious commands; S2. At this time, the gesture state data of different people are preprocessed by frame extraction and size adjustment, and frames are extracted from the gesture state data videos of different people. The extraction frequency is set as needed, wherein the frame extraction rate of the video is calculated by using the following formula: [\text{Frame Extraction Rate}=\frac{\text{Number of Extracted Frames}}{\text{Total Number of Frames}}\times 100%] to determine how many frames are extracted from the video for subsequent processing and analysis; the total length of the video is calculated in seconds: [\text{Video Length}=\frac{\text{Total Number of Frames}}{\text{Frame Rate}}] wherein the frame rate indicates the number of frames contained in the video per second; the usage of different gestures is understood, and the gesture frequency is calculated: [\text{Gesture Frequency}=\frac{\text{Number of Occurrences of Gesture}}{\text{Video Length}}] helps to quantitatively analyze the collected video data, so that the model can better understand the usage and characteristics of gesture commands; S3. At this time, the gesture state models of different people are established by using the gesture state data of different people. The gesture state models of different people are trained by using the gesture state data of different people, and the recurrent neural network (RNN) architecture is selected. The gesture state models of different people are trained by using the labeled data sets. During the training process, the gesture state models of different people will learn the characteristics and patterns of gesture instructions, so as to accurately recognize different gestures. S4. At this time, the gesture state data verification set of different people is used to verify and evaluate the trained gesture state model of different people, evaluate the accuracy of the model in recognizing different gesture instructions, and make adjustments based on the results. Accuracy evaluation: [\text{Accuracy}=\frac{\text{Number of Correct Predictions}}{\text{TotalNumber of Predictions}}\times100%] Precision evaluation: [\text{Precision}=\frac{\text{True Positives}}{\text{True Positives}+\text{False Positives}}] Recall evaluation: [\text{Recall}=\frac{\text{True Positives}}{\text{True Positives}+\text{False Negatives}}] Comprehensive evaluation through F1 score: [F1=2\times\frac{\text{Precision}\times\text{Recall}}{\text{Precision}+\text{Recall}}] fully understand the performance of the model in recognizing different gesture commands, such as accuracy, precision and recall rate; S5. At this point, the trained and verified gesture state models of different people are deployed to the actual system for real-time recognition of user gesture commands; S6. At this time, the energy-saving lamp (5) is switched on and off and its direction is adjusted according to the prediction to reduce erroneous operation.