Equipment awakening method and device based on millimeter wave sensor, equipment and medium

Abstract

The embodiment of the invention provides an equipment awakening method and device based on a millimeter wave sensor, equipment and a medium. The method comprises the following steps: acquiring millimeter wave data acquired by the millimeter wave sensor; constructing a feature vector corresponding to the millimeter wave data; inputting the feature vector into a preset user approaching judgment model to obtain a judgment result; according to the feature vector and the judgment result, determining whether the user is close to the equipment or not; under the condition that the user is close to the equipment, switching the equipment from a standby mode to a working mode; the power consumption of the standby mode is lower than that of the working mode. The user approaching behavior can be accurately and efficiently recognized, the device is awakened when the user approaches, the device is prevented from being in a high-power-consumption standby state for a long time, and energy consumption is remarkably reduced.

Description

Technical Field

[0001] The present invention relates to the technical field of device control, and particularly to a device wake-up method based on a millimeter-wave sensor, a device wake-up device based on a millimeter-wave sensor, an electronic device, and a computer-readable storage medium. Background Art

[0002] With the rapid development of smart home, voice interaction technology has become a highly concerned human-computer interaction method. Users can control smart home devices through voice commands. Especially for users with high demands and high requirements and extreme sensitivity to privacy protection, there is no need for real-time voice detection, and the device relies on always staying in the standby state to listen for the voice wake-up word. When using the voice wake-up method, the device in the waiting-to-wake state continuously listens to the surrounding environment sounds and analyzes whether the fixed wake-up word is matched. If the match is successful, the device will be immediately woken up and activated.

[0003] The voice wake-up system usually can only rely on audio signals, and the device needs to keep a resident microphone listening to capture the wake-up word, which causes energy consumption waste when the user is not near the device or the device has not been used for a long time. Summary of the Invention

[0004] In view of the above problems, embodiments of the present invention are proposed to provide a device wake-up method, device, device, and medium based on a millimeter-wave sensor that overcome the above problems or at least partially solve the above problems.

[0005] According to a first aspect of an embodiment of the present invention, there is provided a device wake-up method based on a millimeter-wave sensor, the method including:

[0006] Obtain millimeter-wave data collected by the millimeter-wave sensor;

[0007] Construct a feature vector corresponding to the millimeter-wave data;

[0008] Input the feature vector into a preset user proximity determination model to obtain a determination result;

[0009] Determine whether a user is close to the device according to the feature vector and the determination result;

[0010] In the case where the user is close to the device, switch the device from the standby mode to the working mode; the power consumption of the standby mode is lower than that of the working mode.

[0011] Optionally, the obtaining the feature vector corresponding to the millimeter-wave data includes:

[0012] Combine the millimeter-wave data into a structure; the structure includes: a timestamp, amplitude information, phase information, and a flag bit;

[0013] Preprocess the said structure body;

[0014] Obtain the preprocessed structure body, and perform feature extraction on the preprocessed structure body; the feature extraction includes time-domain feature extraction and frequency-domain feature extraction;

[0015] Obtain the feature vector according to the extracted time-domain features and frequency-domain features.

[0016] Optionally, the obtaining the feature vector according to the extracted time-domain features and frequency-domain features includes:

[0017] Perform feature fusion on the time-domain features and the frequency-domain features to obtain an initial feature vector;

[0018] Reduce the dimension of the initial feature vector through feature selection and dimensionality reduction processing to obtain the feature vector.

[0019] Optionally, the feature vector includes multiple feature indicators;

[0020] The determining whether the user is close to the device according to the feature vector and the judgment result includes:

[0021] Determine whether the device meets the first preset user proximity condition according to the feature indicators in the feature vector;

[0022] Determine whether the device meets the second preset user proximity condition according to the judgment result; the judgment result is the probability that the user is close to the device;

[0023] If the first preset user proximity condition and the second preset user proximity condition are met, it is determined that the user is close to the device.

[0024] Optionally, the determining whether the device meets the first preset user proximity condition according to the feature indicators in the feature vector includes:

[0025] If each feature indicator in the feature vector is greater than or equal to the corresponding first preset threshold, it is determined that the first preset user proximity condition is met;

[0026] If at least one of the feature indicators in the feature vector is less than the corresponding first preset threshold, it is determined that the first preset user proximity condition is not met.

[0027] Optionally, the determining whether the device meets the second preset user proximity condition according to the judgment result includes;

[0028] If the judgment result is greater than or equal to the second preset threshold, it is determined that the second preset user proximity condition is met;

[0029] If the determination result is less than the second preset threshold, it is determined that the second preset user proximity condition is satisfied.

[0030] Optionally, the user proximity determination model is generated in the following manner:

[0031] Obtain a sample data set;

[0032] Generate a feature vector corresponding to the sample data set;

[0033] Input the feature vector into the user proximity determination model for training;

[0034] Obtain the determination result of the user proximity determination model, and calculate the accuracy rate of the determination result;

[0035] Adjust the user proximity determination model according to the accuracy rate of the determination result.

[0036] Optionally, the method further includes:

[0037] After the device switches from the standby mode to the working mode, if it is determined that the user is not close to the device, switch the device from the working mode to the standby mode.

[0038] Optionally, the method further includes:

[0039] When at least two users are close to the device, determine the priorities of the at least two users;

[0040] According to the priorities of the at least two users, respond to the voice commands of the at least two users.

[0041] According to the second aspect of the embodiments of the present invention, there is provided a device wake-up device based on a millimeter-wave sensor, and the device includes:

[0042] A data acquisition module, configured to acquire millimeter-wave data collected by the millimeter-wave sensor;

[0043] A vector construction module, configured to construct a feature vector corresponding to the millimeter-wave data;

[0044] A result acquisition module, configured to input the feature vector into a preset user proximity determination model to obtain a determination result;

[0045] A result determination module, configured to determine whether a user is close to the device according to the feature vector and the determination result;

[0046] A first mode switching module, configured to switch the device from the standby mode to the working mode when the user approaches the device; the power consumption of the standby mode is lower than that of the working mode.

[0047] Optionally, the vector acquisition module includes:

[0048] A structure combination sub-module, configured to combine the millimeter-wave data into a structure; the structure includes: a timestamp, amplitude information, phase information, and a flag bit;

[0049] A preprocessing sub-module, configured to preprocess the structure;

[0050] A feature extraction sub-module, configured to obtain the preprocessed structure and perform feature extraction on the preprocessed structure; the feature extraction includes time-domain feature extraction and frequency-domain feature extraction;

[0051] A vector generation sub-module, configured to obtain the feature vector according to the extracted time-domain features and frequency-domain features.

[0052] Optionally, the vector generation sub-module includes:

[0053] A vector fusion unit, configured to perform feature fusion on the time-domain features and the frequency-domain features to obtain an initial feature vector;

[0054] A vector dimensionality reduction unit, configured to reduce the dimension of the initial feature vector through feature selection and dimensionality reduction processing to obtain the feature vector.

[0055] Optionally, the feature vector includes a plurality of feature indicators;

[0056] The result determination module includes:

[0057] A first condition determination sub-module, configured to determine whether the device meets a first preset user approach condition according to the feature indicators in the feature vector;

[0058] A second condition determination sub-module, configured to determine whether the device meets a second preset user approach condition according to the judgment result; the judgment result is the probability that the user approaches the device;

[0059] A user approach determination sub-module, configured to determine that the user approaches the device if the first preset user approach condition and the second preset user approach condition are met.

[0060] Optionally, the first condition determination sub-module includes:

[0061] A first condition judgment unit, configured to determine that the first preset user proximity condition is satisfied if each feature index in the feature vector is greater than or equal to the corresponding first preset threshold; and determine that the first preset user proximity condition is not satisfied if at least one of the feature indexes in the feature vector is less than the corresponding first preset threshold.

[0062] Optionally, the second condition determination sub-module includes:

[0063] A second condition judgment unit, configured to determine that the second preset user proximity condition is satisfied if the judgment result is greater than or equal to a second preset threshold; and determine that the second preset user proximity condition is satisfied if the judgment result is less than the second preset threshold.

[0064] Optionally, the user proximity judgment model is generated in the following manner:

[0065] A sample acquisition module, configured to acquire a sample data set;

[0066] A feature vector generation module, configured to generate a feature vector corresponding to the sample data set;

[0067] A model training module, configured to input the feature vector into the user proximity judgment model for training;

[0068] An accurate judgment module, configured to obtain the judgment result of the user proximity judgment model and calculate the accuracy rate of the judgment result;

[0069] A model adjustment module, configured to adjust the user proximity judgment model according to the accuracy rate of the judgment result.

[0070] Optionally, the device further includes:

[0071] A second mode switching module, configured to switch the device from the working mode to the standby mode if it is determined that the user is not close to the device after the device is switched from the standby mode to the working mode.

[0072] Optionally, the device further includes:

[0073] A priority determination module, configured to determine the priorities of at least two users when the at least two users are close to the device;

[0074] An instruction response module, configured to respond to voice instructions of the at least two users according to the priorities of the at least two users.

[0075] According to a third aspect of the present invention, there is provided an electronic device, comprising: a processor, a memory, and a computer program stored on the memory and capable of running on the processor, wherein when the computer program is executed by the processor, the steps of the device wake-up method based on a millimeter-wave sensor as described above are implemented.

[0076] According to a fourth aspect of the present invention, there is provided a computer-readable storage medium, on which a computer program is stored, and when the computer program is executed by a processor, the steps of the device wake-up method based on a millimeter-wave sensor as described above are implemented.

[0077] The technical solutions provided by the embodiments of the present invention may include the following beneficial effects:

[0078] The embodiments of the present invention provide a device wake-up method, device, equipment, and medium based on a millimeter-wave sensor. The method includes: collecting millimeter-wave data of a user through a millimeter-wave sensor, extracting a feature vector from the collected millimeter-wave data, inputting the extracted feature vector into a preset user proximity determination model, and the model will output a result according to the input data, so as to determine whether the user is close to the device based on the feature vector and the result. If the user is close to the device, switch the device from the low-power standby mode to the high-power working mode. By detecting the user's approach behavior through the millimeter-wave sensor, the problem of false wake-up in a noisy environment is reduced, and the device is only woken up when the user is close, avoiding unnecessary energy consumption waste. BRIEF DESCRIPTION OF THE DRAWINGS

[0079] Figure 1 is a flowchart of the steps of a device wake-up method based on a millimeter-wave sensor provided by an embodiment of the present invention;

[0080] Figure 2 is a flowchart of the steps of another device wake-up method based on a millimeter-wave sensor provided by an embodiment of the present invention;

[0081] Figure 3 is a schematic flowchart of a device wake-up method based on a millimeter-wave sensor provided by an embodiment of the present invention;

[0082] Figure 4 is a structural block diagram of a device wake-up device based on a millimeter-wave sensor provided by an embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0083] To make the above objects, features, and advantages of the present invention more obvious and understandable, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0084] One of the core concepts of the embodiments of the present invention is to generate corresponding feature vectors from the millimeter-wave data collected by the millimeter-wave sensor, determine whether the user is approaching the device according to the judgment result of the preset user approach judgment model and the feature vectors, and wake up the device when the user approaches, avoiding the device being in a high-power standby state for a long time and significantly reducing energy consumption.

[0085] Referring to Figure 1 , a flowchart of steps of a device wake-up method based on a millimeter-wave sensor provided by an embodiment of the present invention is shown. The method may specifically include the following steps:

[0086] Step 101, obtain the millimeter-wave data collected by the millimeter-wave sensor;

[0087] The embodiments of the present invention are applicable to the wake-up process of devices in a home scenario, and the device may be a smart home device. For example, a smart speaker, a central control screen, etc. A millimeter-wave sensor may be provided in the device to collect millimeter-wave data, and the device may obtain the millimeter-wave data collected by the millimeter-wave sensor.

[0088] The millimeter-wave sensor is based on millimeter-wave radar technology and uses electromagnetic waves in the millimeter-wave band for target detection and measurement. Its core principle is to transmit millimeter-wave signals and receive reflected signals, and analyze the changes in the signals to obtain information such as the distance, speed, angle, and motion state of the target object.

[0089] The millimeter-wave sensor frequency band determines the detection range of the millimeter-wave signal of the millimeter-wave sensor. The larger the frequency band, the larger the detection range and the higher its cost. In the wake-up scenario of smart home devices, a millimeter-wave sensor with a 24GHz frequency band may be used. The millimeter-wave sensor with a 24GHz frequency band is suitable for short-distance detection and is applicable in a small space (such as a room, a corridor, etc.). Its detection range is usually about 5 meters, which can not only meet the needs of most home environments but also save costs.

[0090] The sampling frequency of the millimeter-wave sensor is the number of times the millimeter-wave sensor samples the signal per unit time, which is used to determine the sampling ability of the millimeter-wave sensor signal (such as the reflected wave). For example, a sampling frequency of 1kHz means that 1000 data points are collected per second. The sampling frequency of the millimeter-wave sensor depends on the signal change rate, detection accuracy requirements, and computing resources in the application scenario. A high sampling frequency will generate more data, require a more powerful processor and more complex algorithms to process these data, and thus increase costs. Since in the actual smart home scenario, the user moves slowly, a sampling frequency of 100Hz is sufficient to capture subtle changes and can also save costs.

[0091] Specifically, a millimeter-wave sensor on the device emits a signal, and the millimeter-wave sensor receives the echo signal formed by the reflection of the emitted millimeter wave on the user. The millimeter-wave sensor transmits the received echo signal to the controller for subsequent determination of whether the user is approaching the device.

[0092] Step 102, construct a feature vector corresponding to the millimeter-wave data;

[0093] A feature vector is a core concept in machine learning and pattern recognition. It is a mathematical representation of the original data and is used to describe the key features of the data. A feature vector is usually a numerical vector, where each element represents a certain feature of the data. By extracting the feature vector, complex data can be transformed into a form that can be processed by a computer, and thus can be used for tasks such as classification, clustering, and regression.

[0094] Specifically, to construct a feature vector from the raw data collected by the millimeter-wave sensor, operations such as data preprocessing, feature extraction, and dimensionality reduction are usually required. Before extracting the feature vector, the raw millimeter-wave data needs to be preprocessed, including denoising, normalization, and outlier removal, etc., to ensure data quality and reduce noise. Key information that can characterize the target characteristics is extracted from the preprocessed data. If the extracted feature dimension is high, dimensionality reduction techniques can be used to reduce the number of features while retaining the main information, and finally, a feature vector corresponding to the millimeter-wave data is obtained.

[0095] Step 103, input the feature vector into a preset user approach determination model to obtain a determination result;

[0096] The user approach determination model is a machine learning or rule-based model based on sensor data (such as millimeter-wave data) and is used to detect whether a user is approaching the device. By analyzing the feature vector corresponding to the millimeter-wave data, it can be detected in real time whether a user is approaching the device. It can be obtained through training of a machine learning model.

[0097] The purpose of the user approach determination model is to determine whether a user is approaching the device. By inputting the feature vector corresponding to the millimeter-wave data, and then outputting the probability value (possibility of approaching) of the user approaching the device.

[0098] Specifically, the user approach determination model is pre-trained and deployed on the device. The extracted feature vector is input into the preset user approach determination model. The dimension of the feature vector needs to be consistent with the input dimension of the model. The model is used to infer the input feature vector to obtain a determination result. The probability value output by the model represents the probability of the user approaching.

[0099] Step 104, determine whether the user is approaching the device according to the feature vector and the determination result;

[0100] Specifically, the device determines whether the user is close to the device through the feature vector (i.e., the threshold method), and sets reasonable thresholds for each key index in the feature vector. When the feature index in the feature vector exceeds the threshold, it can be determined that the user is close. At the same time, it also determines based on the output result of the user proximity judgment model (i.e., the probability of the user being close is output). The probability value output by the model is compared with the threshold. If the probability value is greater than the threshold, it is determined that the user is close to the device. When both the threshold method and the judgment result of the model indicate that the user is close to the device, it can be determined that the user is close to the device; otherwise, it is determined that the user is not close to the device.

[0101] Step 105, when the user is close to the device, switch the device from the standby mode to the working mode; the power consumption of the standby mode is lower than that of the working mode.

[0102] When the user is not close to the device or the device is not in use, the device is in the standby mode, avoiding the device being in a high-power working state for a long time.

[0103] Specifically, data is collected through a millimeter-wave sensor, the feature vector is extracted and input into a pre-trained model. Based on the feature vector and the model output result, it is determined whether the user is close. If the user is close, the device is switched from the standby mode to the working mode, thus achieving low-power operation. If the user is not close, the device remains in the standby mode, and unnecessary functions (such as voice interaction functions, microphone listening, etc.) are turned off in the standby mode; in the working mode, all working modules are enabled.

[0104] In an embodiment of the present invention, millimeter-wave data of the user is collected through a millimeter-wave sensor, and a feature vector is extracted from the collected millimeter-wave data to describe the state or behavior of the user. The extracted feature vector is input into a preset user proximity judgment model, and the model will output a result based on the input data. Thus, it is determined whether the user is close to the device according to the feature vector and the result. If the user is close to the device, the device is switched from the low-power standby mode to the high-power working mode. The device has a lower power consumption in the standby mode and only switches to the working mode when the user is close, effectively saving energy and reducing the running time of the device in the high-power mode, which helps to extend the service life of the device.

[0105] Refer to Figure 2 , which shows a flowchart of steps of another device wake-up method based on a millimeter-wave sensor provided by an embodiment of the present invention. The method may specifically include the following steps:

[0106] Step 201, obtain the millimeter-wave data collected by the millimeter-wave sensor;

[0107] Specifically, a signal is emitted by a millimeter-wave sensor on the device, and an echo signal formed by the reflection of the emitted millimeter wave on the user is received by the millimeter-wave sensor. The millimeter-wave sensor transmits the received echo signal to the controller for subsequent determination of whether the user is approaching the device.

[0108] Step 202, constructing a feature vector corresponding to the millimeter-wave data;

[0109] To construct a feature vector from the raw data collected by the millimeter-wave sensor, operations such as data preprocessing, feature extraction, and dimensionality reduction are usually required. Before extracting the feature vector, the raw millimeter-wave data needs to be preprocessed, including denoising, normalization, outlier removal, etc., to ensure data quality and reduce noise. Key information that can characterize the target characteristics is extracted from the preprocessed data. If the extracted feature dimension is high, dimensionality reduction techniques can be used to reduce the number of features while retaining the main information, and finally a feature vector corresponding to the millimeter-wave data is obtained.

[0110] In one embodiment, step 202 may include the following sub-steps:

[0111] Sub-step S11, combining the millimeter-wave data into a structure; the structure includes: timestamp, amplitude information, phase information, and flag bit;

[0112] A structure (Struct) is a user-defined composite data type used to combine multiple different types of data. It allows developers to organize related data fields together to form a logical unit, thereby improving the readability and maintainability of the code. Each field of the variables contained in the structure has its own data type and name. The structure can improve the efficiency of data access and transmission, enabling different types of data (such as timestamps, amplitudes, phases, flag bits) to be stored compactly and processed uniformly, facilitating subsequent data analysis and processing; it also helps to ensure data consistency and reduce the additional computational overhead caused by multiple reads of different variables.

[0113] A timestamp is a way of representing time, usually a number or a string, used to record the specific time when an event occurs. It is recorded using a timestamp (uint64_t), representing the number of milliseconds since the device was started.

[0114] Amplitude Information generally refers to the intensity or magnitude of a signal at a certain moment. In fields such as signal processing, communication, sensor data analysis, and audio processing, amplitude information is an important feature used to describe the physical characteristics of a signal. The amplitude information is stored using a 32-bit floating-point number (float), representing the intensity value of the signal returned by the sensor and reflecting the signal's variation;

[0115] Phase Information is an important concept in signal processing, used to describe the phase characteristics of a signal in the time domain or frequency domain. Phase Information is usually used in combination with amplitude information to jointly describe the complete characteristics of a signal. Phase Information uses a 32-bit floating-point number (float) to represent the phase angle for subsequent signal processing;

[0116] A flag is a binary digit (usually 1 or 0) used to represent a certain state or condition, used to indicate a certain state or condition. The flag (uint8_t) is used to indicate the validity of the data and whether the state is normal.

[0117] Specifically, the data collected by the millimeter-wave sensor is stored in a structured manner, and each collection point of the data is combined into a structure. The purpose of structured data storage is to store complete sensor data, improve access efficiency and data management convenience, rather than just serving the sampling frequency. It is used to store all key information collected by the millimeter-wave sensor, including timestamp, amplitude, phase, and flag. The timestamp records the time of data collection, the amplitude information stores the amplitude data of the millimeter-wave signal, the phase information stores the phase data of the millimeter-wave signal, and the flag is used to mark the special state of the data (such as whether it is valid, whether it has been processed, etc.).

[0118] The data collected by the millimeter-wave sensor is transmitted to the device controller. For data transmission, a high-speed and low-latency communication interface needs to be selected. SPI (Serial Peripheral Interface) or I2C (Inter-Integrated Circuit) is used as the communication protocol between the millimeter-wave sensor and the microcontroller (MCU) or digital signal processor (DSP).

[0119] Sub-step S12, preprocess the structure;

[0120] Specifically, preprocessing the millimeter-wave data structure is to improve data quality, reduce noise, and prepare for subsequent feature extraction or model inference. When preprocessing the structure, it usually includes denoising, time-domain and frequency-domain analysis, and normalization processing.

[0121] First, a low-pass filter is used to filter the noise of the original signal to remove high-frequency noise and ensure that the useful signal is retained. The cut-off frequency of the filter should be adjusted according to the actual application scenario. Through time-domain and frequency-domain analysis, in the time domain, statistical features such as the mean and standard deviation of the signal are calculated to identify the overall shape of the signal. The statistical features can help establish a baseline and determine whether the signal change conforms to the pattern close to the user. The mean measures the overall energy level of the signal and reflects the trend of the target approaching or moving away; the standard deviation describes the degree of signal fluctuation and distinguishes stationary background noise from moving targets; kurtosis and skewness are used to identify abnormal changes and help judge non-normal targets (such as pets, environmental interference). In the frequency domain, the fast Fourier transform is used to analyze the signal and extract the main frequency components.

[0122] Finally, signal normalization is performed. Here, normalization methods are adopted, including min-max normalization and Z-score standardization, to ensure the consistency of signal intensity under different environments, so that subsequent feature extraction and user detection will be more accurate. Different application scenarios require different normalization methods, and the specific selection depends on the variation characteristics of sensor data and application requirements. Min-max normalization is more suitable when the signal amplitude range of millimeter-wave sensors is relatively stable and mainly based on thresholds to judge proximity to the device. Z-score standardization is applicable when the signal is greatly affected by the environment (such as different signal ranges in different rooms and different devices), because it can eliminate the influence of device or environmental factors and improve detection accuracy. It is also possible to first use Z-score to handle environmental differences and then use Min-Max normalization to scale to [0,1], which can not only ensure stability but also adapt to different environments.

[0123] Sub-step S13: Obtain the preprocessed structure and perform feature extraction on the preprocessed structure; the feature extraction includes time-domain feature extraction and frequency-domain feature extraction;

[0124] Based on the preprocessed millimeter-wave data structure, feature extraction on the preprocessed structure includes time-domain feature extraction and frequency-domain feature extraction, thereby obtaining time-domain features and frequency-domain features

[0125] Feature extraction is to extract meaningful information from the processed signal. In time-domain feature extraction, statistical features such as the average value, maximum value, minimum value, variance, kurtosis, and skewness of the signal are calculated as the overall characteristics reflecting the signal. In frequency-domain feature extraction, the results of Fourier transform are used to extract frequency-domain features, including the amplitude of the main frequency components, energy spectral density, and frequency center, etc.

[0126] Sub-step S14: Obtain the feature vector based on the extracted time-domain features and frequency-domain features.

[0127] Based on the extracted time-domain features and frequency-domain features, these features can be combined into a feature vector. The feature vector is a one-dimensional array, where each element corresponds to a feature value. Extract time-domain features (such as mean, variance, peak value, etc.) from the preprocessed data, extract frequency-domain features (such as spectral centroid, total energy, etc.) from the preprocessed data, and combine the time-domain features and frequency-domain features in sequence to form a feature vector.

[0128] Structurally store the millimeter-wave data as a structure, then preprocess the structure, and perform feature extraction on the preprocessed structure, including time-domain feature extraction and frequency-domain feature extraction, and obtain the feature vector according to the extracted time-domain features and frequency-domain features. Storing data in the form of a structure facilitates management and access. Preprocessing the structure improves the data quality and lays a foundation for subsequent analysis. Time-domain and frequency-domain feature extraction provides a comprehensive data description and enhances the accuracy of analysis.

[0129] In one embodiment, the step S14 may include the following sub-steps:

[0130] Sub-step S141, perform feature fusion on the time-domain feature and the frequency-domain feature to obtain an initial feature vector;

[0131] Feature fusion is the process of combining time-domain features and frequency-domain features into a unified feature vector. Through feature fusion, the information in the time domain and frequency domain can be fully utilized to improve the performance of subsequent machine learning or data analysis tasks. Extract time-domain features from the preprocessed data, extract frequency-domain features from the preprocessed data, and fuse the time-domain features and frequency-domain features into an initial feature vector, which should be able to fully represent the user's approaching behavior and serve as the input for subsequent user detection algorithms.

[0132] Sub-step S142, reduce the dimension of the initial feature vector through feature selection and dimensionality reduction processing to obtain the feature vector.

[0133] At the same time, adopt feature selection and dimensionality reduction techniques, including principal component analysis and redundant feature elimination, to reduce the dimension of the initial feature vector to make the feature vector more concise.

[0134] Feature selection is used to select the most representative features, reduce redundancy, and improve computational efficiency. It includes correlation analysis to remove redundant features with high similarity; recursive feature elimination (RFE), which iteratively deletes unimportant features according to the importance of the machine learning model. Dimensionality reduction techniques are used to reduce the data dimension and improve computational efficiency. It includes principal component analysis (PCA), which extracts the main components through linear transformation and retains the maximum amount of information; linear discriminant analysis (LDA), which improves the classification effect by optimizing the ratio of between-class variance and within-class variance.

[0135] Fuse time-domain features and frequency-domain features to generate an initial feature vector. The goal of feature selection is to screen out the most useful features for the task from the initial feature vector, reduce redundant information, and the purpose of dimensionality reduction is to further reduce the dimension of the feature vector while retaining the main information. After feature selection and dimensionality reduction, the final feature vector is obtained. Combining time-domain and frequency-domain features provides a more comprehensive data description, removes redundant features, reduces computational complexity, improves the generalization ability of the model, reduces the dimension of the feature vector, improves computational efficiency, and retains the main information at the same time.

[0136] Step 203: Input the feature vector into a preset user proximity determination model to obtain a determination result.

[0137] The user proximity determination model is pre-trained and deployed on the device. Input the extracted feature vector into the preset user proximity determination model. The dimension of the feature vector needs to be consistent with the input dimension of the model, and use the model to perform inference on the input feature vector to obtain a determination result. The probability value output by the model represents the probability of user proximity.

[0138] Step 204: Determine whether the user is close to the device according to the feature vector and the determination result.

[0139] The device determines whether the user is close to the device through the feature vector (i.e., the threshold method), and sets reasonable thresholds through each key index in the feature vector. Set the thresholds for each key index. When the feature index in the feature vector exceeds this threshold, it can be determined that the user is close. At the same time, it also passes the output result of the user proximity determination model (i.e., outputs the probability of user proximity), compares the probability value output by the model with the threshold. If the probability value is greater than the threshold, it is determined that the user is close to the device. When both the threshold method and the determination result of the model are that the user is close to the device, it can be determined that the user is close to the device, otherwise it is determined that the user is not close to the device.

[0140] In one embodiment, the feature vector includes multiple feature indexes; step 204 may include the following sub-steps:

[0141] Sub-step S21: Determine whether the device meets the first preset user proximity condition according to the feature indexes in the feature vector.

[0142] The feature indexes are each time-domain feature and frequency-domain feature in the feature vector, including mean value, variance, frequency amplitude, etc.

[0143] The first preset user proximity condition is used to determine whether the user is close to the device, and is analyzed through the millimeter-wave data collected by the millimeter-wave sensor, and is used to determine the distance between the user and the device, so as to determine whether the user is close to the device.

[0144] By means of a threshold detection method, a reasonable threshold is set according to the key indicators in the feature vector. In other words, a threshold for the signal mean is set. When the feature indicators in the feature vector exceed this threshold, it can be determined that the user is approaching. That is, when the mean > threshold A and the frequency amplitude > threshold B, the next step of processing is triggered.

[0145] In one embodiment, the step S21 may include the following sub-steps:

[0146] Sub-step S211, if each feature indicator in the feature vector is greater than or equal to the corresponding first preset threshold, it is determined that the first preset user approach condition is satisfied; if at least one of the feature indicators in the feature vector is less than the corresponding first preset threshold, it is determined that the first preset user approach condition is not satisfied.

[0147] By comparing the comparison results of each indicator in the feature vector with the preset threshold, it is judged whether a specific condition is satisfied. A threshold is set for each feature indicator. If the condition is satisfied, each feature indicator in the feature vector is greater than or equal to the corresponding first preset threshold, and it is determined that the "first preset user approach condition" is satisfied; if the condition is not satisfied, at least one feature indicator in the feature vector is less than the corresponding first preset threshold, and it is determined that the "first preset user approach condition" is not satisfied.

[0148] According to whether each feature indicator in the feature vector is greater than or equal to the corresponding first preset threshold, it is judged whether the first preset user approach condition is satisfied. Only when all feature indicators meet the threshold is it determined that the condition is satisfied. This strict standard can avoid misjudgment and improve the reliability of judgment. The threshold of each feature indicator can be adjusted according to the specific application scenario to adapt to the needs of different devices and environments. The judgment logic is based on simple comparison operations, with low computational complexity and suitable for real-time processing.

[0149] Sub-step S22, according to the judgment result, determine whether the device meets the second preset user approach condition; the judgment result is the probability that the user approaches the device.

[0150] The feature vector is input into a preset "user approach judgment model", and whether the "second preset user approach condition" is satisfied is determined through the judgment result output by the model (i.e., the probability that the user approaches the device). The time-domain and frequency-domain feature vectors are input into the machine learning model to obtain the output result. For example: the model output probability > 0.8 (i.e., the probability that the user approaches exceeds 80%), and it is considered that the user does approach.

[0151] In one embodiment, the step S22 may include the following sub-steps:

[0152] Sub-step S221: If the judgment result is greater than or equal to the second preset threshold, it is determined that the second preset user approach condition is met; if the judgment result is less than the second preset threshold, it is determined that the second preset user approach condition is met.

[0153] Input the feature vector into a preset "user approach judgment model". The model outputs a judgment result, which is the probability of the user approaching the device (usually a value between 0 and 1). According to the probability value output by the model, it is judged whether the second preset user approach condition is met. If the probability value is greater than or equal to a preset probability threshold, it is determined that the second preset user approach condition is met; if the probability value is less than the probability threshold, it is determined that the second preset user approach condition is not met.

[0154] According to whether the judgment result (the probability of the user approaching the device) is greater than or equal to the second preset threshold, it is determined whether the second preset user approach condition is met. The judgment result is usually a probability value calculated by a machine learning model or statistical method, which can more accurately reflect the possibility of the user approaching. The probability value can be dynamically adjusted to adapt to different environments and user behaviors.

[0155] Sub-step S23: If the first preset user approach condition and the second preset user approach condition are met, it is determined that the user is approaching the device.

[0156] For the first preset user approach condition, each feature index in the feature vector is greater than or equal to the corresponding first preset threshold; for the second preset user approach condition, the user approach probability output by the user approach judgment model is greater than or equal to the preset probability threshold, then it is finally determined that the user is approaching the device. If both the first preset user approach condition and the second preset user approach condition are met, it is determined that the user is approaching the device. If any one of the conditions is not met, it is determined that the user is not approaching the device. This dual judgment mechanism combines rule judgment (the first condition) and model prediction (the second condition), and can more accurately determine whether the user is approaching the device. The first condition is used to quickly screen out obvious non-conformities, and the second condition is further verified by the model, which is applicable to scenarios with high accuracy requirements.

[0157] The first preset user approach condition sets thresholds based on some key indicators (such as amplitude, phase, time-domain features, etc.) in the feature vector. The second preset user approach condition is based on the probability of the user approaching the device (through a machine learning model). If both the first preset condition and the second preset condition are met, it is determined that the user is approaching the device. Through the dual verification of the first condition and the second condition, the accuracy of the judgment is improved, and the thresholds and conditions can be adjusted according to specific application scenarios, with wide applications.

[0158] Step 205, when the user is close to the device, switch the device from the standby mode to the working mode; the power consumption of the standby mode is lower than that of the working mode.

[0159] Collect data through a millimeter-wave sensor, extract feature vectors and input them into a pre-trained model. According to the feature vectors and the model output results, determine whether the user is close. If the user is close, switch the device from the standby mode to the working mode, so as to achieve low-power operation. If the user is not close, keep the device in the standby mode and turn off unnecessary functions (such as voice interaction functions, microphone listening, etc.) in the standby mode; in the working mode, enable all working modules.

[0160] Step 206, after the device is switched from the standby mode to the working mode, if it is determined that the user is not close to the device, switch the device from the working mode to the standby mode.

[0161] After the device is switched from the standby mode to the working mode, the millimeter-wave sensor will also continuously emit signals to detect whether the user is near the device. The system can re-detect whether the user exists within a short period of time (such as 1 second). If a misjudgment is found or the user leaves the device, quickly restore the low-power state to avoid power consumption waste. Through regular detection and mode switching, ensure that the device can quickly respond to user behavior.

[0162] The device is switched from the standby mode to the working mode (for example, by detecting possible user activities through a sensor), and uses feature vectors and a preset user proximity judgment model to determine whether the user is close to the device. If the user is not close to the device, switch the device back from the working mode to the standby mode to save energy or avoid unnecessary operation. If the user is close to the device, the device remains in the working mode and continues to provide services.

[0163] Step 207, when at least two users are close to the device, determine the priorities of the at least two users;

[0164] Step 208, respond to the voice commands of the at least two users according to the priorities of the at least two users.

[0165] Once the user is detected successfully, the system should immediately trigger real-time wake-up processing, including activating the voice recognition module of the device and preparing to receive the user's voice commands. After waking up, the device gives the user obvious feedback, such as through sound or visual cues, indicating that the device is ready to receive commands. After the user is close to the device and enters the specified range, the millimeter-wave sensor triggers the wake-up mechanism and at the same time activates the microphone to enter the listening state. At this time, the user only needs to say the preset voice wake-up word to wake up the device and interact with it. By combining millimeter-wave sensing and voice recognition, the false wake-up rate in a noisy environment can be reduced and the accuracy of wake-up can be improved.

[0166] When multiple users approach the device simultaneously, through the multi-target tracking ability of the millimeter-wave radar, combined with user identification, according to voice features or Bluetooth signals, prioritize responding to specific users. At the same time, set a priority mechanism, first come first served or specific users have priority. It can process voice commands of multiple users simultaneously and sort them according to priority to ensure that the needs of high-priority users are responded to in a timely manner while taking into account other users. The priority can be dynamically adjusted according to factors such as user identity and scenario to enhance the usage experience in multi-user scenarios.

[0167] The user approach judgment model is generated in the following way:

[0168] Obtain a sample data set; generate feature vectors corresponding to the sample data set; input the feature vectors into the user approach judgment model for training; obtain the judgment results of the user approach judgment model and calculate the accuracy of the judgment results; adjust the user approach judgment model according to the accuracy of the judgment results.

[0169] Combined with a machine learning model, use supervised learning algorithms (such as support vector machines, decision trees, or random forests, etc.) to train the feature vectors. Use a labeled data set for model training to improve the accuracy and reliability of the model.

[0170] Collect data from the actual scenario to construct a sample data set. The sample data should include various feature indicators when the user approaches and does not approach the device (such as user distance, signal strength, user movement speed, etc.). Each sample needs to be labeled with its true category (such as "approach" or "not approach"). Preprocess the sample data set, extract features and generate feature vectors, and use the generated feature vectors to train the "user approach judgment model". The model can be a machine learning model (such as logistic regression, support vector machine, random forest, etc.) or a deep learning model (such as a neural network). The model outputs a judgment result (such as the probability of "approach" or "not approach") for each sample in the test set, compare the judgment result of the model with the true category, calculate the accuracy, and adjust and optimize the model according to the accuracy until the accuracy of the model reaches the expectation.

[0171] In an embodiment of the present invention, millimeter-wave data of a user is collected by a millimeter-wave sensor, and the millimeter-wave data is structurally stored as a structure including a timestamp, amplitude information, phase information, and a flag bit, and preprocessed. Time-domain features and frequency-domain features are extracted from the preprocessed structure, the time-domain features and the frequency-domain features are fused into an initial feature vector, and the dimension is reduced through feature selection and dimensionality reduction techniques to obtain a final feature vector. The feature vector is input into a preset user approach judgment model to obtain the probability (judgment result) of the user approaching. According to the feature indicators and the judgment result in the feature vector, it is respectively judged whether the first preset user approach condition and the second preset user approach condition are satisfied. If both conditions are satisfied at the same time, it is determined that the user is approaching the device. If the user is approaching the device, the device is switched from the low-power standby mode to the high-power working mode. If the user is not approaching the device, the device is switched back from the working mode to the standby mode. When multiple users approach the device, the priority of the users is determined, and voice commands are responded to according to the priority. The device has low power consumption in the standby mode and only switches to the working mode when the user approaches, significantly reducing energy consumption. The millimeter-wave sensor has a fast response speed. Combined with an efficient judgment model, it can quickly switch the device mode, improve the user experience, reduce the running time of the device in the high-power mode, and significantly reduce energy consumption.

[0172] It should be noted that for the method embodiments, for the sake of simple description, they are all expressed as a series of action combinations. However, those skilled in the art should know that the embodiments of the present invention are not limited by the described action sequence, because according to the embodiments of the present invention, certain steps can be performed in other sequences or simultaneously. Secondly, those skilled in the art should also know that the embodiments described in the specification are all preferred embodiments, and the actions involved are not necessarily essential for the embodiments of the present invention.

[0173] Refer to Figure 3 , which shows a schematic flowchart of a device wake-up method based on a millimeter-wave sensor provided by an embodiment of the present invention, specifically including the following processes:

[0174] Data is collected by a millimeter-wave sensor on the device, and the collected millimeter-wave data is sent to a controller. The device controller preprocesses the millimeter-wave data including time-domain and frequency-domain analysis. After the data preprocessing is completed, feature extraction is performed to obtain time-domain features and frequency-domain features, and the time-domain features and the frequency-domain features are fused to obtain a feature vector. Whether the user is approaching the device is detected through the feature vector and the judgment result of the user approach judgment model. If the user is approaching the device, the device is woken up and the microphone and voice recognition module of the device are activated. If the user is not approaching the device, the user is continuously detected.

[0175] Refer to Figure 4, showing the structural block diagram of a device wake-up device based on a millimeter-wave sensor provided by an embodiment of the present invention, which may specifically include the following modules:

[0176] A data acquisition module 401, configured to acquire millimeter-wave data collected by the millimeter-wave sensor;

[0177] A vector construction module 402, configured to construct a feature vector corresponding to the millimeter-wave data;

[0178] A result acquisition module 403, configured to input the feature vector into a preset user proximity judgment model to obtain a judgment result;

[0179] A result determination module 404, configured to determine whether a user is close to the device according to the feature vector and the judgment result;

[0180] A first mode switching module 405, configured to switch the device from a standby mode to a working mode when the user is close to the device; the power consumption of the standby mode is lower than that of the working mode.

[0181] Optionally, the vector acquisition module 402 includes:

[0182] A structure combination sub-module, configured to combine the millimeter-wave data into a structure; the structure includes: a timestamp, amplitude information, phase information, and a flag bit;

[0183] A preprocessing sub-module, configured to preprocess the structure;

[0184] A feature extraction sub-module, configured to obtain the preprocessed structure and perform feature extraction on the preprocessed structure; the feature extraction includes time-domain feature extraction and frequency-domain feature extraction;

[0185] A vector generation sub-module, configured to obtain the feature vector according to the extracted time-domain features and frequency-domain features.

[0186] Optionally, the vector generation sub-module includes:

[0187] A vector fusion unit, configured to perform feature fusion on the time-domain feature and the frequency-domain feature to obtain an initial feature vector;

[0188] A vector dimensionality reduction unit, configured to reduce the dimension of the initial feature vector through feature selection and dimensionality reduction processing to obtain the feature vector.

[0189] Optionally, the feature vector includes multiple feature indicators; the result determination module 404 includes:

[0190] A first condition determination sub-module, configured to determine whether the device meets a first preset user approach condition according to the feature indicators in the feature vector;

[0191] A second condition determination sub-module, configured to determine whether the device meets a second preset user approach condition according to the judgment result; the judgment result is the probability that the user approaches the device;

[0192] A user approach determination sub-module, configured to determine that the user approaches the device if the first preset user approach condition and the second preset user approach condition are met.

[0193] Optionally, the first condition determination sub-module includes:

[0194] A first condition judgment unit, configured to determine that the first preset user approach condition is met if each feature indicator in the feature vector is greater than or equal to the corresponding first preset threshold; and determine that the first preset user approach condition is not met if at least one of the feature indicators in the feature vector is less than the corresponding first preset threshold.

[0195] Optionally, the second condition determination sub-module includes;

[0196] A second condition judgment unit, configured to determine that the second preset user approach condition is met if the judgment result is greater than or equal to a second preset threshold; and determine that the second preset user approach condition is met if the judgment result is less than the second preset threshold.

[0197] Optionally, the user approach judgment model is generated in the following manner:

[0198] A sample acquisition module, configured to acquire a sample data set;

[0199] A feature vector generation module, configured to generate a feature vector corresponding to the sample data set;

[0200] A model training module, configured to input the feature vector into the user approach judgment model for training;

[0201] An accurate judgment module, configured to obtain the judgment result of the user approach judgment model and calculate the accuracy rate of the judgment result;

[0202] A model adjustment module, configured to adjust the user approach judgment model according to the accuracy rate of the judgment result.

[0203] Optionally, the device further includes:

[0204] A second mode switching module, configured to switch the device from the working mode to the standby mode if it is determined that the user is not close to the device after the device is switched from the standby mode to the working mode.

[0205] Optionally, the device further includes:

[0206] A priority determination module, configured to determine the priorities of at least two users when the at least two users are close to the device;

[0207] An instruction response module, configured to respond to voice instructions of the at least two users according to the priorities of the at least two users.

[0208] In an embodiment of the present invention, a data acquisition module acquires millimeter-wave data of a user through a millimeter-wave sensor. A structure combination sub-module in a vector acquisition module combines and stores the millimeter-wave data as a structure including a timestamp, amplitude information, phase information, and a flag bit. A preprocessing sub-module preprocesses the structure to prepare for feature extraction. A feature extraction sub-module in the vector acquisition module extracts time-domain features and frequency-domain features from the preprocessed structure. A vector generation sub-module fuses the time-domain features and the frequency-domain features into an initial feature vector, and reduces the dimension through feature selection and dimensionality reduction techniques to obtain a final feature vector. A result acquisition module inputs the feature vector into a preset user proximity determination model to obtain the probability (judgment result) of user proximity. A result determination module determines whether a first preset user proximity condition is satisfied according to feature indicators in the feature vector; and determines whether a second preset user proximity condition is satisfied according to the judgment result (probability of user proximity). If the user is close to the device, the device is switched from the low-power standby mode to the high-power working mode. If the user is not close to the device, the device is switched back from the working mode to the standby mode. The device has low power consumption in the standby mode and is only switched to the working mode when the user is close, significantly reducing energy consumption. Through the feature vector and dual-condition judgment (feature indicators and model probability), misjudgment is reduced and the judgment accuracy is improved. It supports multi-user scenarios and can process voice instructions according to priorities, enhancing the multi-user applicability of the device.

[0209] For the device embodiment, since it is basically similar to the method embodiment, the description is relatively simple. For related parts, refer to the partial description of the method embodiment.

[0210] An embodiment of the present invention further provides an electronic device, including:

[0211] It includes a processor, a memory, and a computer program stored on the memory and capable of running on the processor. When the computer program is executed by the processor, it implements each process of the above-described method embodiment of the device wake-up method based on a millimeter-wave sensor, and can achieve the same technical effects. To avoid repetition, it will not be elaborated here.

[0212] An embodiment of the present invention also provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, it implements each process of the above-described embodiment of the device wake-up method based on a millimeter-wave sensor and can achieve the same technical effects. To avoid repetition, it will not be elaborated here.

[0213] Each embodiment in this specification is described in a progressive manner. The key point of each embodiment is to illustrate the differences from other embodiments. For the same or similar parts among the embodiments, reference can be made to each other.

[0214] Those skilled in the art should understand that the embodiments of the present invention can be provided as a method, a device, or a computer program product. Therefore, the embodiments of the present invention can take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware aspects. Moreover, the embodiments of the present invention can take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk memories, CD-ROMs, optical memories, etc.) containing computer-usable program codes.

[0215] The embodiments of the present invention are described with reference to the flowcharts and / or block diagrams of methods, terminal devices (systems), and computer program products according to the embodiments of the present invention. It should be understood that each process and / or block in the flowcharts and / or block diagrams can be implemented by computer program instructions, and the combination of the processes and / or blocks in the flowcharts and / or block diagrams can also be implemented by computer program instructions. These computer program instructions can be provided to the processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing terminal devices to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing terminal devices generate a device for implementing the functions specified in Figure 1 one process or multiple processes and / or blocks Figure 1 one block or multiple blocks.

[0216] These computer program instructions can also be stored in a computer-readable memory that can direct a computer or other programmable data processing terminal device to work in a specific manner, so that the instructions stored in the computer-readable memory generate a manufactured product including an instruction device, and the instruction device implements the functions specified in Figure 1 one process or multiple processes and / or blocks Figure 1 one block or multiple blocks.

[0217] These computer program instructions can also be loaded onto a computer or other programmable data processing terminal device, so that a series of operation steps are executed on the computer or other programmable terminal device to generate a computer-implemented process. Thus, the instructions executed on the computer or other programmable terminal device provide steps for implementing the functions specified in one process or multiple processes and / or blocks Figure 1 one process or multiple processes and / or blocks Figure 1 steps for implementing the functions specified in one block or multiple blocks.

[0218] Although the preferred embodiments of the embodiments of the present invention have been described, those skilled in the art can make additional changes and modifications once they learn the basic creative concepts. Therefore, the appended claims are intended to be construed to include the preferred embodiments as well as all changes and modifications falling within the scope of the embodiments of the present invention.

[0219] Finally, it should also be noted that in this article, relational terms such as first and second are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the term "comprising", "including" or any other variant thereof is intended to cover non-exclusive inclusion, so that a process, method, article or terminal device comprising a series of elements not only includes those elements, but also includes other elements not expressly listed, or also includes elements inherent to such process, method, article or terminal device. Without further limitation, an element defined by the statement "comprising an..." does not exclude the presence of additional identical elements in the process, method, article or terminal device comprising the element.

[0220] The above has introduced in detail a device wake-up method based on a millimeter-wave sensor and a device wake-up device based on a millimeter-wave sensor provided by the present invention. Specific examples are used in this article to elaborate on the principle and implementation manner of the present invention. The description of the above embodiments is only used to help understand the method and its core idea of the present invention; at the same time, for those of ordinary skill in the art, according to the idea of the present invention, there will be changes in the specific implementation manner and application scope. In summary, the content of this specification should not be construed as a limitation to the present invention.

Claims

1. A device wake-up method based on a millimeter-wave sensor, characterized in that, The method includes: Obtaining millimeter-wave data collected by the millimeter-wave sensor; Constructing a feature vector corresponding to the millimeter-wave data; Inputting the feature vector into a preset user proximity determination model to obtain a determination result; Determining whether a user is approaching the device according to the feature vector and the determination result; When the user approaches the device, switching the device from the standby mode to the working mode; the power consumption of the standby mode is lower than that of the working mode.

2. The device wake-up method based on a millimeter-wave sensor according to claim 1, wherein, The obtaining of the feature vector corresponding to the millimeter-wave data includes: Combining the millimeter-wave data into a structure; the structure includes: a timestamp, amplitude information, phase information, and a flag bit; Preprocessing the structure; Obtaining the preprocessed structure and performing feature extraction on the preprocessed structure; the feature extraction includes time-domain feature extraction and frequency-domain feature extraction; Obtaining the feature vector according to the extracted time-domain features and frequency-domain features.

3. The device wake-up method based on a millimeter-wave sensor according to claim 2, characterized in that The obtaining of the feature vector according to the extracted time-domain features and frequency-domain features includes: Performing feature fusion on the time-domain features and the frequency-domain features to obtain an initial feature vector; Reducing the dimension of the initial feature vector through feature selection and dimensionality reduction processing to obtain the feature vector.

4. The method for waking up a device based on a millimeter-wave sensor according to claim 1, characterized in that The feature vector includes multiple feature indicators; The determining whether a user is approaching the device according to the feature vector and the determination result includes: Determining whether the device meets a first preset user proximity condition according to the feature indicators in the feature vector; Determining whether the device meets a second preset user proximity condition according to the determination result; the determination result is the probability that the user is approaching the device; If the first preset user proximity condition and the second preset user proximity condition are met, it is determined that the user is approaching the device.

5. The device wake-up method based on a millimeter-wave sensor according to claim 4, characterized in that The determining whether the device meets the first preset user proximity condition according to the feature indicators in the feature vector includes: If each feature indicator in the feature vector is greater than or equal to the corresponding first preset threshold, it is determined that the first preset user proximity condition is met; If at least one of the feature indicators in the feature vector is less than the corresponding first preset threshold, it is determined that the first preset user proximity condition is not met.

6. The device wake-up method based on a millimeter-wave sensor according to claim 4, wherein The determining whether the device meets the second preset user proximity condition according to the determination result includes; If the determination result is greater than or equal to a second preset threshold, it is determined that the second preset user proximity condition is met; If the determination result is less than the second preset threshold, it is determined that the second preset user proximity condition is met.

7. The method for waking up a millimeter-wave sensor-based device according to any one of claims 1-6, characterized in that, The user proximity determination model is generated in the following manner: Obtaining a sample data set; Generating a feature vector corresponding to the sample data set; Inputting the feature vector into the user proximity determination model for training; Obtaining the determination result of the user proximity determination model and calculating the accuracy rate of the determination result; Adjusting the user proximity determination model according to the accuracy rate of the determination result.

8. The device wake-up method based on a millimeter-wave sensor according to claim 1, characterized in that, The method further includes: After the device switches from the standby mode to the working mode, if it is determined that the user is not close to the device, the device is switched from the working mode to the standby mode.

9. The device wake-up method based on a millimeter-wave sensor according to claim 1, wherein The method further includes: When at least two users are close to the device, determining the priorities of the at least two users; Responding to the voice commands of the at least two users according to the priorities of the at least two users.

10. A device wake-up device based on a millimeter-wave sensor, characterized in that The device includes: A data acquisition module, configured to acquire millimeter-wave data collected by the millimeter-wave sensor; A vector construction module, configured to construct a feature vector corresponding to the millimeter-wave data; A result acquisition module, configured to input the feature vector into a preset user proximity determination model to obtain a determination result; A result determination module, configured to determine whether a user is close to the device according to the feature vector and the determination result; A first mode switching module, configured to switch the device from the standby mode to the working mode when the user is close to the device; the power consumption of the standby mode is lower than that of the working mode.

11. An electronic device, characterized in that, Including: A processor, a memory, and a computer program stored on the memory and capable of running on the processor, where when the computer program is executed by the processor, the steps of the millimeter-wave sensor-based device wake-up method according to any one of claims 1-9 are implemented.

12. A computer-readable storage medium, characterized in that, A computer program is stored on the computer-readable storage medium, and when the computer program is executed by a processor, the steps of the millimeter-wave sensor-based device wake-up method according to any one of claims 1-9 are implemented.

Images