Self-adaptive environment regulation and control method and system for household appliances

Abstract

The invention relates to the technical field of intelligent control, in particular to a home appliance self-adaptive environment regulation and control method and system, and the method comprises the steps: carrying out the processing of collected original data, and extracting a feature value; gesture recognition is carried out on the preprocessed data based on a lightweight 1D-CNN algorithm, local features are extracted, and gesture categories are recognized after activation function ReLU and pooling layer processing; according to the recognized gesture type, mapping the gesture type to a corresponding control instruction; the control instruction is sent to the household appliance through Bluetooth or infrared signals; and the household appliance executes corresponding operation after receiving the control instruction. According to the invention, the optimized CNN is embedded into the STM32 to realize self-adaptive environment regulation and control of the household appliances, so that the household appliances can automatically adjust the operation states according to environment changes and user habits, the intelligent level and energy utilization efficiency of the household appliances are improved, more comfortable, convenient and personalized household experience is provided for users, and the user experience is improved. Therefore, the problem that the existing household appliance regulation and control technology cannot meet user requirements is solved.

Description

Technical Field

[0001] The present invention relates to the field of intelligent control technology, and in particular to a method and system for adaptively controlling the environment of household appliances. Background Art

[0002] In the traditional home appliance sector, common control methods include manual operation and simple automated control. Manual operation relies on the user directly touching the physical buttons or knobs on the appliance, such as turning the temperature control knob on an air conditioner or pressing the on / off button on a TV, to achieve basic control of the appliance. This method is simple and direct, but completely relies on manual labor, resulting in limited flexibility and convenience. Simple automated control uses tools such as timers and remote controls to achieve a certain degree of remote operation and timing. For example, an infrared remote control can be used to remotely change the TV channel and adjust the volume; a timer can be used to set the heating time of a water heater to operate during specific time periods.

[0003] With the development of technology, smart homes have gradually emerged, and some home appliances have begun to incorporate environmental control functions. They can sense environmental information using sensors such as temperature and humidity, and adjust according to simple preset rules. For example, smart air conditioners can automatically start or stop cooling and heating based on the temperature feedback from the indoor temperature sensor, maintaining the indoor temperature within the set range. Smart humidifiers can automatically turn on or off the humidification function based on the humidity value detected by the humidity sensor. However, these technologies still have significant shortcomings when it comes to coping with complex and changing environments and personalized user needs.

[0004] For example, they have poor adaptability to complex environments. The environmental control rules of traditional home appliances are often relatively simple and fixed, making it difficult to cope with diverse and dynamically changing environments. For example, the indoor environment's requirements for temperature and humidity vary greatly in different seasons and time periods, but existing home appliances are unable to flexibly and accurately adjust to these complex environmental changes. When factors such as the number of people indoors and the intensity of activity change, existing home appliances are unable to automatically sense and make corresponding adjustments. For example, during a large gathering, the indoor temperature and humidity will change due to the increase in the number of people and the intensification of activity, but ordinary air conditioners and humidifiers cannot adapt to these changes in time and provide the most suitable environmental parameters.

[0005] Precise control is difficult. The control accuracy of existing home appliances is limited, making it difficult to meet users' high demands for a comfortable environment. Taking temperature control as an example, many traditional air conditioners can only achieve a temperature control accuracy of ±1°C or even higher. This may affect user comfort in temperature-sensitive environments such as baby rooms and study rooms. Moreover, traditional home appliances often lack comprehensive perception and comprehensive analysis of environmental parameters during the control process, which can easily lead to over- or under-control. For example, they only control the temperature of a single point in the room, ignoring the temperature differences between different areas of the room, resulting in uneven indoor temperature distribution.

[0006] Individualized user needs are difficult to meet. Different users have different habits and functional requirements for home appliances, but most existing home appliances use fixed control modes and function settings, making them difficult to customize based on user needs. For example, some users like to set the air conditioner temperature to 26°C and the fan speed to the minimum before going to bed, while others prefer 24°C and a medium fan speed. Traditional home appliances struggle to meet these individualized settings. Furthermore, existing technologies struggle to effectively address the needs of specific scenarios, such as the integration of smart homes and office environments, and the special requirements of the elderly or disabled for convenient appliance operation. Summary of the Invention

[0007] The purpose of the present invention is to provide a method and system for adaptive environmental control of home appliances, aiming to solve the problem that existing home appliance control technologies cannot meet user needs.

[0008] To achieve the above objectives, in a first aspect, the present invention provides a method for adaptively controlling an environment of a household appliance, comprising the following steps:

[0009] Collect environmental data and IMU data;

[0010] Process the collected raw data and extract feature values;

[0011] The pre-processed data is used for gesture recognition based on a lightweight 1D-CNN algorithm. Local features are extracted and processed through activation functions (ReLU) and pooling layers to identify gesture categories.

[0012] According to the recognized gesture type, it is mapped to the corresponding control instruction;

[0013] Send control instructions to home appliances via Bluetooth or infrared signals;

[0014] After receiving the control instruction, the home appliance performs the corresponding operation.

[0015] Among them, in "Collecting environmental data and IMU data", environmental data and IMU data are collected through the MPU6050 sensor inside the controller and multiple temperature, humidity, and air pressure sensors distributed in the room.

[0016] The step of "processing the collected raw data and extracting characteristic values" includes the following steps:

[0017] The Kalman filter algorithm is used to filter the collected raw data;

[0018] Normalize the filtered data;

[0019] Extract the eigenvalues from the normalized data.

[0020] The following steps are included in the process of "performing gesture recognition on the preprocessed data based on the lightweight 1D-CNN algorithm, extracting local features, and identifying gesture categories after processing with the activation function ReLU and the pooling layer":

[0021] The convolution kernel slides on the data sequence, calculates the convolution result at each position, and obtains the feature map;

[0022] The feature map is nonlinearly transformed through the activation function ReLU function;

[0023] After average pooling in the pooling layer, the feature map is downsampled;

[0024] The extracted features are globally analyzed and decided through the fully connected layer to identify the category of the gesture.

[0025] In a second aspect, a home appliance adaptive environmental control system is provided, for use in the home appliance adaptive environmental control method described in the first aspect, comprising a data acquisition module, a data processing module, a gesture recognition module, a decision and mapping module, a signal transmission module, and a home appliance execution module, wherein the data acquisition module, the data processing module, the gesture recognition module, the decision and mapping module, the signal transmission module, and the home appliance execution module are connected in sequence;

[0026] The data acquisition module is used to collect environmental data;

[0027] The data processing module is used to filter and normalize the collected data and extract eigenvalues;

[0028] The gesture recognition module is used to calculate the pre-processed data and identify different gesture categories;

[0029] The decision and mapping module maps the identified gesture type to a corresponding control instruction;

[0030] The signal transmission module sends instructions to the smart home device via Bluetooth control or infrared control;

[0031] The household appliance execution module is used to receive instructions and respond to achieve adaptive control.

[0032] The present invention provides a method for adaptive environmental control of household appliances, comprising the following steps: collecting environmental data and IMU data; processing the collected raw data and extracting eigenvalues; performing gesture recognition on the pre-processed data based on a lightweight 1D-CNN algorithm, extracting local features, and identifying gesture categories after processing with an activation function ReLU and a pooling layer; mapping the identified gesture type to a corresponding control instruction; sending the control instruction to a household appliance via Bluetooth or infrared signals; and executing the corresponding operation after receiving the control instruction. The present invention implements adaptive environmental control of household appliances by embedding an optimized CNN into an STM32, enabling the household appliances to automatically adjust their operating status according to environmental changes and user habits, improving the intelligence level and energy efficiency of the household appliances, and providing users with a more comfortable, convenient, and personalized home experience, thereby solving the problem that existing household appliance control technologies cannot meet user needs. BRIEF DESCRIPTION OF THE DRAWINGS

[0033] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0034] Figure 1 This is a flow chart of a method for adaptively controlling the environment of household appliances provided by the present invention.

[0035] Figure 2 It is a flow chart for processing the collected raw data and extracting characteristic values.

[0036] Figure 3 This is a flowchart for performing gesture recognition on preprocessed data based on the lightweight 1D-CNN algorithm, extracting local features, and identifying gesture categories after processing through the activation function ReLU and pooling layer.

[0037] Figure 4 It is a schematic diagram of a household appliance adaptive environment control system provided by the present invention.

[0038] In the figure: 1-data acquisition module, 2-data processing module, 3-gesture recognition module, 4-decision-making and mapping module, 5-signal transmission module, 6-home appliance execution module. DETAILED DESCRIPTION

[0039] The following describes embodiments of the present invention in detail, examples of which are shown in the accompanying drawings, wherein the same or similar reference numerals throughout represent the same or similar elements or elements having the same or similar functions. The embodiments described below with reference to the accompanying drawings are exemplary and are intended to be used to explain the present invention, and are not to be construed as limiting the present invention.

[0040] See also Figures 1 to 3 In a first aspect, the present invention provides a method for adaptively controlling an environment of a household appliance, comprising the following steps:

[0041] S1 collects environmental data and IMU data;

[0042] Environmental data and IMU data are collected through the MPU6050 sensor inside the controller and multiple temperature, humidity, and air pressure sensors distributed in the room.

[0043] Specifically, environmental data is collected through the MPU6050 sensor installed inside the controller and multiple temperature, humidity, and air pressure sensors distributed in the room, and then wirelessly transmitted to the host for subsequent processing. This method uses multi-sensor fusion technology to combine data from the inertial measurement unit (IMU), air pressure sensor, and magnetometer to improve the accuracy of gesture recognition and reduce the measurement error of a single sensor. The IMU consists of an accelerometer and a gyroscope, which are used to measure linear acceleration and angular velocity. Because the IMU is susceptible to drift and cumulative errors, relying solely on the IMU for gesture recognition will lead to a decrease in accuracy after long-term use. The air pressure sensor measures ambient pressure and can be used to infer altitude changes, providing additional position information, which helps to distinguish gestures at different altitudes. The magnetometer measures the strength and direction of the Earth's magnetic field, which can provide azimuth information and reduce the drift of the IMU during rotational motion.

[0044] In order to increase the endurance of the controller, a low-power function is set. The low-power optimization triggers interrupts through the IMU and dynamically adjusts the sampling rate to achieve energy consumption management of sensor data acquisition and improve the endurance of the system. The IMU is in low-power mode in standby mode and only monitors acceleration changes. When it is detected that the acceleration exceeds the set threshold, an interrupt signal is triggered to wake up the microcontroller from low-power mode and enter the data acquisition and processing process. The acceleration threshold is set based on the system's need to distinguish between small movements and static states to avoid unnecessary power consumption. The judgment formula for triggering the interrupt is

[0045]

[0046] Among them, Δa is the acceleration change, and Δx, Δy, and Δz are the acceleration increments in the three-axis directions respectively. When Δa exceeds the set threshold, the IMU outputs an interrupt signal to notify the microcontroller to start data acquisition. Dynamic sampling rate control adjusts the sampling frequency of the IMU during the recognition process based on the gesture action characteristics. During intense exercise, the sampling rate is increased to obtain high-time resolution data and enhance the accuracy of motion recognition. When stationary or moving slowly, the sampling rate is reduced to reduce unnecessary data processing and power consumption. The strategy of dynamically adjusting the sampling rate determines the update frequency by detecting changes in angular velocity, and the expression is

[0047] f s =f base +k·|ω|

[0048] Among them, f s is the current sampling frequency, f base is the minimum sampling frequency, k is the adjustment coefficient, and ω is the current angular velocity. When the angular velocity is high, the sampling frequency is increased to ensure data integrity. When the angular velocity is low, the sampling frequency is reduced to reduce the computational burden and optimize power management. By using IMU-triggered interrupts and dynamic sampling rate adjustment, the system reduces energy consumption while maintaining recognition accuracy, improving the long-term stability of the gesture recognition system.

[0049] S2 processes the collected raw data and extracts eigenvalues;

[0050] S21 uses Kalman filter algorithm to filter the collected raw data;

[0051] Specifically, the host STM32 filters the collected raw data and uses algorithms such as Kalman filtering to remove noise interference in the data and improve the accuracy and stability of the data.

[0052] S22 performs normalization processing on the filtered data;

[0053] Specifically, normalization is then performed to map the data to a specific range, such as [0, 1], so that data of different dimensions are comparable, which facilitates subsequent analysis and processing.

[0054] Sensor fusion uses the extended Kalman filter algorithm for data optimization, and the state update process is expressed as

[0055]

[0056] in, is the current state estimate, is the state at the previous moment, u k is the control input, w k is the process noise. The observation update process is

[0057]

[0058] Among them, z k is the sensor observation value, is the measurement model, v k To reduce measurement noise, the Kalman filter performs a weighted fusion of the IMU, air pressure sensor, and magnetometer data, reducing individual sensor errors and improving the robustness of the gesture recognition system. The fused data is then fed into a CNN model for classification, improving the reliability and stability of feature extraction and ensuring recognition accuracy in various environments.

[0059] S23 extracts characteristic values from the normalized data.

[0060] Specifically, the characteristic values of the data are extracted, including time domain features (mean, variance, etc.) and frequency domain features (frequency components, etc.). These characteristic values represent the characteristics of gesture actions, reduce the amount of data, and improve processing efficiency.

[0061] S3 performs gesture recognition on pre-processed data based on a lightweight 1D-CNN algorithm, extracts local features, and identifies gesture categories after processing using ReLU activation functions and pooling layers.

[0062] The S31 convolution kernel slides on the data sequence, calculates the convolution result at each position, and obtains the feature map;

[0063] Specifically, a lightweight 1D-CNN algorithm is used to perform gesture recognition on preprocessed data. The lightweight 1D-CNN algorithm extracts local features of gestures by convolving the preprocessed data with multiple convolution kernels in a one-dimensional convolutional layer. The convolution kernel slides over the data sequence, calculating the convolution result at each position to generate a feature map.

[0064] The application of 1D-CNN in gesture recognition tasks has the advantages of low computational complexity and applicability to resource-constrained devices. Compared with traditional 2D-CNN, 1D-CNN focuses on feature extraction of one-dimensional time series data, reducing computational complexity and improving processing efficiency. 1D-CNN performs convolution operations on input data through a one-dimensional convolution kernel to extract local patterns in the time series. The convolution operation is calculated as follows:

[0065]

[0066] Among them, y i is the convolution output of the i-th position, x i+j represents the i+jth data point in the input data sequence, w jis the convolution kernel weight, k is the convolution kernel size, and b is the bias term. The convolution kernel slides along the time axis, calculating the features of the local area each time to capture the temporal information. In the 1D-CNN architecture, the nonlinear activation function ReLU is used to enhance the network's expressive power, and the expression is

[0067] f(x)=max(0,x)

[0068] Where x is the input value, and f(x) is the output after activation. ReLU effectively suppresses vanishing gradients and improves network training efficiency. The pooling layer uses average pooling to reduce the dimensionality of the convolutional output, improving feature robustness while reducing computational effort and storage requirements. The fully connected layer integrates the extracted features and performs final classification to determine the gesture category. Compared to 2D-CNN, 1D-CNN avoids additional dimensional computation and performs feature extraction only on the time axis, reducing computational effort and enabling efficient operation on resource-constrained embedded systems such as the STM32, while maintaining accurate and real-time gesture recognition.

[0069] Based on the 1D-CNN model, quantization and pruning optimization are also performed. The quantization and pruning methods can optimize the inference speed of the 1D-CNN model, making it run efficiently on embedded systems while reducing the consumption of computing resources. Quantization reduces the computational complexity and storage requirements by reducing the numerical representation precision in the model. Specifically, weight quantization and activation quantization are used to convert the original floating-point weights and activation values into low-bit integer representations. The quantization process can be expressed as

[0070] q = round(s x + z)

[0071] Where q is the quantized integer value, x is the original floating-point number, s is the scaling factor, and z is the zero-point offset. Quantization reduces memory usage and enables the use of fixed-point arithmetic on low-power devices, improving computational efficiency. Pruning further reduces computational and storage requirements by removing redundant parameters that have little impact on inference.

[0072] The weight pruning method sets low weight values to zero based on the importance index of the weight, and only retains important connections. The convolution calculation after pruning can be expressed as

[0073]

[0074] Among them, y i is the convolution output, x i+j is the input data, w j is the weight after pruning, b is the bias term, m jA binary mask indicates whether the weight should be retained. After pruning, some calculations are skipped, reducing computational effort and improving inference speed. Combining quantization and pruning methods reduces model storage overhead while maintaining gesture recognition accuracy. This improves inference efficiency on embedded devices like the STM32 and enables stable system operation in low-power environments.

[0075] S32 performs nonlinear transformation on the feature map through the activation function ReLU function;

[0076] Specifically, the ReLU function is used to activate the feature map and perform nonlinear transformation to enhance the expressive power of the model.

[0077] S33 performs average pooling in the pooling layer and downsamples the feature map;

[0078] Specifically, it then passes through the pooling layer and average pooling to downsample the feature map, reducing the data dimension and redundant information, and improving the robustness and generalization ability of the model.

[0079] S34 performs global analysis and decision-making on the extracted features through the fully connected layer to identify the category of the gesture.

[0080] Specifically, a fully connected layer performs global analysis and decision-making on the extracted features to identify the gesture category. Compared to traditional 2D-CNN, 1D-CNN is optimized for one-dimensional data, reducing computational effort and making it more suitable for running on the resource-limited STM32.

[0081] S4 maps the gesture to a corresponding control instruction according to the recognized gesture type;

[0082] Specifically, according to the recognized gesture type, it is mapped to a corresponding control instruction according to a preset mapping rule.

[0083] Users have their own habits, so there's a particular need for user-defined gestures. A self-learning system allows for user-defined gestures, combined with real-time data for training and adjustment, to improve gesture recognition accuracy. Users can define new gestures based on their personal needs and habits. The gesture data recorded by the system serves as training samples and is fed back into the model for retraining. The self-learning process is accomplished through an incremental learning algorithm. As users input new gestures, the system continuously updates the classifier, ensuring that recognition accuracy improves over time. The incremental learning algorithm uses the following formula for training and updating:

[0084] θ t+1 =θ t +η·▽L(θ t ,x t ,y t )

[0085] Here, θt is the model parameter, η is the learning rate, L(θt, xt, yt) is the loss function, xt is the current input sample, and yt is the label of the current sample. The system continuously updates the parameter θt to optimize the model based on the user's newly added gesture data, gradually improving the accuracy of gesture recognition with each step. Users can flexibly set gesture categories and action requirements, and the system optimizes model performance based on their operating habits. Through its self-learning function, the system continuously accumulates training data, enabling gesture recognition to not only adapt to individual needs but also effectively respond to environmental changes, providing accurate recognition services. Whenever a user defines a new gesture, the system collects data, conducts real-time training, and makes adjustments, improving the model's generalization and adaptability, ensuring high recognition rates in dynamic environments.

[0086] The self-learning system based on historical data improves the adaptability and accuracy of the gesture recognition system between different users through adaptive training and LSTM-based time series optimization. Traditional gesture recognition systems use fixed CNN models. However, the differences in gestures between users make it difficult for fixed models to cope with diverse gestures, so the recognition rate may drop. Through adaptive training, users can enter their own gesture data, and the system fine-tunes the parameters of the CNN model based on this data to adapt to the characteristics of different users. The adaptive training process is achieved through incremental learning, and new gesture data is used to adjust the weights of the existing model to optimize the model's performance. The update process of incremental learning can be expressed as

[0087]

[0088] Where θt is the current model parameter, η is the learning rate, L(θt,xt,yt) is the loss function, xt is the current input data, and yt is the corresponding gesture label. Through this update, the system can adjust to the gesture characteristics of different users, significantly improving personalized recognition capabilities.

[0089] In addition to adaptive training, LSTM-based time series optimization can also improve recognition accuracy. Gestures are usually expressed as a series of continuous data streams, not just static single-point data. Traditional CNN models can only extract local features and lack the ability to model time series information. To solve this problem, adding an LSTM (long short-term memory network) layer after CNN can effectively capture the temporal dependency of gestures. The LSTM model maintains long-term memory of historical information through memory cells and determines the transmission and update of information through a gating mechanism, thereby processing dynamic changes in time series. The optimization of LSTM is described by the following formula:

[0090] h t =LSTM(x t ,h t-1 ,c t-1 )

[0091] Here, ht is the current hidden state, xt is the current input, ht-1 is the previous hidden state, and ct-1 is the previous memory cell state. Through this time series modeling, LSTM can better capture the temporal characteristics of gestures, improving overall recognition accuracy. Combining adaptive training and LSTM-based optimization, the system can make personalized adjustments based on the user's historical gesture data, improving performance in different environments and for different users, and ensuring efficient and accurate gesture recognition.

[0092] In addition to the custom gesture function, another innovation is the ability to increase the adaptability of gesture control by combining voice. By combining environmental sensors and voice assistance, intelligent control of home appliances can be achieved. Environmental sensors such as temperature and humidity sensors, air pressure sensors, and light sensors monitor changes in the home environment in real time. The collected environmental data is processed and used to adjust the operating status of home appliances. For example, when the indoor temperature rises, the air conditioner will automatically adjust the temperature to ensure a comfortable ambient temperature. The system uses the following model to perform weighted fusion of environmental data to calculate a comprehensive control value:

[0093]

[0094] Where y is the control output, xi is the input value of the i-th environmental sensor, wi is the weight of the sensor, and n is the total number of sensors. The voice assistance system uses natural language processing technology to allow users to interact with home appliances through voice commands. When the user issues a command, the voice recognition module converts the voice into text and controls the device according to the content of the command. For example, when the user says "turn up the air conditioner temperature", the voice system will recognize the command and map it to the corresponding control command, which will be sent to the home appliance via Bluetooth or infrared signal. The combination of voice and environmental sensors makes the system more intelligent and can respond to user needs and environmental changes in real time. Smart home AI adaptive control can intelligently adjust the operating status of home appliances based on comprehensive environmental factors and user feedback, providing users with a personalized and comfortable living experience.

[0095] Gesture and environmental perception combined control achieves more intelligent and precise home control by combining user gestures and environmental conditions. When a user draws an L-shaped gesture during the day, the system recognizes it as daytime based on the data provided by the environmental sensor. The ambient light intensity is usually high, so the system only adjusts the curtains to avoid direct sunlight and provide appropriate indoor brightness. At night, when the user draws the same L-shaped gesture, the system recognizes it as a low-light environment. The smart home system decides to turn on the lights based on the ambient light perception information to increase the indoor lighting. The ambient light intensity is collected by the light sensor, and the collected value is calculated using the following formula:

[0096]

[0097] Among them, I light is the normalized ambient light intensity, I sensor is the actual light value measured by the sensor, I max is the maximum range of the sensor. light When the set threshold is exceeded, the system determines that it is daytime and performs curtain adjustment; when I light When the light level falls below the threshold, the system determines it's nighttime and turns on the lights. By combining gesture recognition with environmental awareness, the system can automatically adjust home appliances based on time of day and lighting conditions, providing intelligent control services tailored to user needs. This allows the system to not only respond to user intent but also make intelligent decisions based on environmental conditions, achieving a more user-friendly control approach.

[0098] The "draw an L" gesture and the "draw a circle" gesture are designed to trigger the control of multiple home appliances. Through the gesture recognition system, users can achieve complex multi-device linkage through simple gestures. When the user performs the "draw an L" gesture, the system recognizes it as a basic operation, which usually corresponds to the control of a single device, such as curtain adjustment. When the user changes to the "draw a circle" gesture, the system recognizes it as a more complex command, triggering the joint control of multiple home appliances. For example, when the user draws a "circle" gesture, the system can control devices such as lights, curtains, and air conditioners at the same time, and coordinate operations according to user needs and environmental conditions. The gesture recognition system recognizes gestures through a lightweight CNN or LSTM model, and combines preset mapping rules to match different gestures with corresponding device control instructions. For the "draw a circle" gesture, the system generates device control instructions through the following mapping function:

[0099] C=f(gesture)

[0100] Here, C represents a set of device control commands, and f(gesture) is the mapping function from gestures to control commands. The gesture type determines the type of device being controlled and the operational behavior. For a "circle" gesture, f(gesture) is mapped to control commands for multiple devices, triggering multiple device interactions such as turning on lights, adjusting air conditioning temperature, and controlling curtains. This mapping relationship between gestures and device control commands enables users to intelligently control multiple devices with simple and intuitive movements, providing a more convenient smart home experience.

[0101] S5 sends control commands to home appliances via Bluetooth or infrared signals;

[0102] Specifically, the system uses a Bluetooth module to establish a connection with smart home devices and transmits control commands as Bluetooth signals. This system is suitable for smart home appliances with Bluetooth reception capabilities, such as smart speakers and smart air conditioners, enabling short-range wireless control and facilitating user operation from anywhere. Infrared control uses an infrared transmitter module to encode control commands into infrared signals and transmit them to the infrared receivers of traditional home appliances, such as TV and air conditioner remote controls. This allows for compatible control of these appliances and expands the system's application scope.

[0103] The S6 home appliance performs the corresponding operation after receiving the control command.

[0104] Specifically, after receiving the control instruction, the home appliance performs corresponding operations according to the requirements of the instruction.

[0105] See also Figure 4 In a second aspect, a household appliance adaptive environment control system is provided, which is used in the household appliance adaptive environment control method described in the first aspect, and includes a data acquisition module 1, a data processing module 2, a gesture recognition module 3, a decision and mapping module 4, a signal transmission module 5, and a household appliance execution module 6, wherein the data acquisition module 1, the data processing module 2, the gesture recognition module 3, the decision and mapping module 4, the signal transmission module 5, and the household appliance execution module 6 are connected in sequence;

[0106] The data acquisition module 1 is used to collect environmental data;

[0107] The data processing module 2 is used to filter and normalize the collected data and extract eigenvalues;

[0108] The gesture recognition module 3 is used to calculate the pre-processed data and identify different gesture categories;

[0109] The decision and mapping module 4 maps the identified gesture type to a corresponding control instruction;

[0110] The signal transmission module 5 sends instructions to the smart home device via Bluetooth control or infrared control;

[0111] The household appliance execution module 6 is used to receive instructions and respond to them to achieve adaptive control.

[0112] In this embodiment, the data acquisition module 1 uses an MPU6050 sensor to collect acceleration and angular velocity IMU data, as well as air pressure, magnetometer, and temperature and humidity sensor data; the data processing module 2, composed of an STM32 microcontroller, is responsible for filtering, normalizing, and preprocessing the collected data, and extracting eigenvalues; the gesture recognition module 3 uses a lightweight 1D-CNN algorithm to calculate the preprocessed data and identify different gesture categories; the decision and mapping module 4 maps the recognized gesture type to the corresponding control instruction; the signal transmission module 5 sends instructions to smart home devices via Bluetooth control, or sends infrared signals to control traditional home appliances via infrared control; and the home appliance execution module 6 receives instructions and responds to achieve adaptive control.

[0113] Beneficial effects:

[0114] (1) Technical performance improvement

[0115] 1. Precision Control: Multi-sensor fusion technology enables comprehensive and accurate environmental information perception. Combined with an optimized CNN algorithm for data analysis and processing, this technology enables home appliances to make precise control decisions based on subtle environmental changes. For example, traditional air conditioners can have significant deviations in temperature control. However, the system of this invention can precisely control the temperature, wind speed, and operating mode of the air conditioner based on a variety of factors, such as indoor temperature, humidity, and human activity. This improves temperature control accuracy to within ±0.5°C, effectively enhancing indoor comfort.

[0116] Fast Response: The high efficiency of the 1D-CNN algorithm and the powerful processing capabilities of the STM32 microcontroller enable the system to achieve extremely fast data processing and gesture recognition. From the moment a user makes a gesture to the moment the appliance executes the control command, the entire response time can be shortened to less than 500 milliseconds. Compared to traditional appliance control methods, this significantly improves real-time control, allowing users to instantly experience the appliance's response and enhance the user experience.

[0117] 3. Enhanced Environmental Adaptability: The system of the present invention possesses strong environmental adaptability, capable of handling a variety of complex and changing environmental conditions. Whether in different seasons, time periods, or even when the number of people indoors or the intensity of activity changes, the system can automatically adjust the operating status of home appliances by monitoring environmental information and user behavior in real time, providing users with a consistently comfortable home environment. For example, in winter when indoor air is dry, the system can automatically adjust the operating status of the humidifier based on data from the humidity sensor to maintain indoor humidity within a suitable range. During summer gatherings, the system can promptly adjust the air conditioning's cooling capacity and wind speed based on the changes in heat and humidity generated by the activities of the people, ensuring a comfortable indoor environment.

[0118] (2) Improved user experience

[0119] 1. Convenient Operation: Users can control home appliances with simple gestures, eliminating the need to search for a remote control or manually operate buttons on appliances, greatly simplifying the operation process. Gesture control is particularly convenient when both hands are busy or when operating from a distance. For example, when cooking with greasy hands, users can control kitchen appliances, such as turning on the range hood and adjusting the stove power, with simple gestures. While relaxing in the living room, users can easily control devices such as the TV and lights with gestures, without having to get up and search for the remote control.

[0120] 2. Personalized Services: The self-learning system allows users to customize gestures and control commands, meeting the individual needs of different users. The system also automatically adjusts the operating mode of home appliances based on the user's historical usage data and habits, providing users with more intimate and personalized services. For example, the system can automatically adjust the air conditioning temperature and shut down unnecessary electrical devices based on the user's nightly bedtime habits to create a comfortable sleeping environment. For special user groups such as the elderly and people with disabilities, they can customize simple and easy-to-use gesture commands based on their physical condition and operating habits, improving the convenience and usability of home appliances.

[0121] 3. Intelligent Experience: Smart home AI adaptive control enables appliances to automatically make intelligent decisions based on environmental changes and user needs, creating a more intelligent home experience. For example, the system can automatically adjust the opening and closing of curtains and the brightness of lights based on ambient light intensity and time, providing users with a comfortable lighting environment. When users return home, the system can automatically turn on lights and adjust the air conditioning temperature based on user habits and environmental information, creating a warm and comfortable home atmosphere, allowing users to experience the convenience and comfort brought by intelligent living.

[0122] (3) Enhanced market competitiveness

[0123] 1. Technical Advantages: This invention's technology is significantly innovative and advanced in the smart home field. Innovations such as D-CNN algorithm optimization, multi-sensor fusion technology, low-power design, and a self-learning system make this product superior in performance and functionality to traditional smart home products. These technological advantages are expected to attract consumer attention and enhance the product's market acceptance and competitiveness.

[0124] 2. Commercial Value: With the growing demand for smart homes, smart home products featuring advanced technology and a positive user experience have broad market prospects. The technology of this invention can be applied to a variety of home appliances, such as smart air conditioners, smart lighting, and smart curtains, providing new technical solutions for appliance manufacturers, helping to increase product added value, expand market share, and create greater commercial value. Furthermore, this technology can be integrated with other smart home systems to form a more comprehensive smart home ecosystem, further enhancing their commercial competitiveness.

[0125] The above disclosure is merely a preferred embodiment of a method and system for adaptive environmental control of household appliances of the present invention. It is certainly not intended to limit the scope of the present invention. A person skilled in the art will understand that all or part of the processes of the above embodiment and equivalent changes made in accordance with the claims of the present invention still fall within the scope of the invention.

Claims

1. A method for adaptively controlling the environment of household appliances, characterized in that: The following steps are involved: Collect environmental data and IMU data; Process the collected raw data and extract feature values; The pre-processed data is used for gesture recognition based on a lightweight 1D-CNN algorithm. Local features are extracted and processed through activation functions (ReLU) and pooling layers to identify gesture categories. According to the recognized gesture type, it is mapped to the corresponding control instruction; Send control instructions to home appliances via Bluetooth or infrared signals; After receiving the control instruction, the home appliance performs the corresponding operation.

2. The method for controlling the home appliance's adaptive environment according to claim 1, wherein: In "Collecting Environmental Data and IMU Data", environmental data and IMU data are collected through the MPU6050 sensor inside the controller and multiple temperature, humidity, and air pressure sensors distributed throughout the room.

3. The method for controlling the home appliance's adaptive environment according to claim 1, wherein: In "Processing the collected raw data and extracting eigenvalues", the following steps are included: The Kalman filter algorithm is used to filter the collected raw data; Normalize the filtered data; Extract the eigenvalues from the normalized data.

4. The method for controlling the home appliance's adaptive environment according to claim 1, wherein: The following steps are included in the "Performing gesture recognition on preprocessed data based on a lightweight 1D-CNN algorithm, extracting local features, and identifying gesture categories after processing using the ReLU activation function and pooling layers": The convolution kernel slides on the data sequence, calculates the convolution result at each position, and obtains the feature map; The feature map is nonlinearly transformed through the activation function ReLU function; After average pooling in the pooling layer, the feature map is downsampled; The extracted features are globally analyzed and decided through the fully connected layer to identify the category of the gesture.

5. A household appliance adaptive environment control system, used in the household appliance adaptive environment control method according to any one of claims 1 to 4, characterized in that: It includes a data acquisition module, a data processing module, a gesture recognition module, a decision and mapping module, a signal transmission module and a home appliance execution module, wherein the data acquisition module, the data processing module, the gesture recognition module, the decision and mapping module, the signal transmission module and the home appliance execution module are connected in sequence; The data acquisition module is used to collect environmental data; The data processing module is used to filter and normalize the collected data and extract eigenvalues; The gesture recognition module is used to calculate the pre-processed data and identify different gesture categories; The decision and mapping module maps the identified gesture type to a corresponding control instruction; The signal transmission module sends instructions to the smart home device via Bluetooth control or infrared control; The household appliance execution module is used to receive instructions and respond to achieve adaptive control.

Images