Wearing detection method and device, wearable device and computer readable storage medium

Abstract

Embodiments of the present application disclose a wearing detection method and device, a wearable device and a computer readable storage medium, and relate to the technical field of wearing detection; the method comprises: scanning sensors currently available for the wearable device, and obtaining configuration information of the sensors; inputting the configuration information of the sensors into a pre-trained algorithm selection model to obtain a target algorithm output by the algorithm selection model; the target algorithm has an accuracy rate and power consumption for wearing detection based on data collected by part or all of the sensors satisfying a preset condition; a target sensor corresponding to the target algorithm is determined, and target data collected by the target sensor is obtained; and the target data is processed based on the target algorithm to obtain a wearing detection result of the wearable device. Thus, the present application can accurately and low-power detect wearing of various wearable devices at low research and development cost.

Description

Technical Field

[0001] This application relates to the field of wear detection technology, specifically to a wear detection method, apparatus, wearable device, and computer-readable storage medium. Background Technology

[0002] Wear detection in smart wearable devices determines the accuracy of monitoring core data such as heart rate and activity, and can also adjust power consumption by recognizing non-wearing states. Wear detection in related technologies relies on multiple sensors working continuously to ensure accuracy. However, the small size and limited power supply of smart wearable devices make power consumption a persistent problem. Furthermore, the sensor configurations of different brands and models vary significantly, requiring the development of separate detection algorithms, model training, and parameter tuning for each type of device, resulting in a cumbersome development process and high costs. Summary of the Invention

[0003] This application provides a wear detection method, apparatus, wearable device, and computer-readable storage medium, which can perform accurate and low-power wear detection for various wearable devices with low development costs.

[0004] In a first aspect, embodiments of this application provide a wear detection method applied to wearable devices, the method comprising:

[0005] Scan the currently available sensors of the wearable device and obtain the configuration information of the sensors;

[0006] The configuration information of the sensor is input into the pre-trained algorithm selection model to obtain the target algorithm output by the algorithm selection model; the target algorithm meets preset conditions in terms of accuracy and power consumption for wear detection based on some or all of the data collected by the sensor.

[0007] Determine the target sensor corresponding to the target algorithm, and acquire the target data collected by the target sensor;

[0008] The target data is processed based on the target algorithm to obtain the wearing detection result of the wearable device.

[0009] Secondly, embodiments of this application provide a wear detection device for use in wearable devices, the device comprising:

[0010] The information acquisition module is used to scan the currently available sensors of the wearable device and acquire the configuration information of the sensors;

[0011] The algorithm output module is used to input the configuration information of the sensor into the pre-trained algorithm selection model to obtain the target algorithm output by the algorithm selection model; the target algorithm meets preset conditions in terms of accuracy and power consumption for wear detection based on some or all of the data collected by the sensor.

[0012] The data acquisition module is used to determine the target sensor corresponding to the target algorithm and to acquire the target data collected by the target sensor.

[0013] The data processing module is used to process the target data based on the target algorithm to obtain the wearing detection result of the wearable device.

[0014] Thirdly, embodiments of this application also provide a wearable device, which includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When the computer program is executed by the processor, it implements the steps in the above-described wear detection method.

[0015] Fourthly, embodiments of this application also provide a computer-readable storage medium storing a computer program, which, when executed by a processor, implements the steps in the above-described wearing detection method.

[0016] Fifthly, embodiments of this application also provide a computer program product or computer program, which includes computer instructions stored in a computer-readable storage medium. A processor of a computer device reads the computer instructions from the computer-readable storage medium and executes the computer instructions, causing the computer device to perform the methods provided in the various optional implementations described in embodiments of this application.

[0017] The embodiments of this application have the following beneficial effects:

[0018] When different wearable devices have different available sensors, the pre-trained algorithm selection model can select the target algorithm for each wearable device. In this way, there is no need to develop detection algorithms, train models and debug parameters separately for various wearable devices with different sensor configurations, which can save resources and reduce R&D costs. Moreover, the accuracy and power consumption of the target algorithm based on the data collected by the sensor can meet the preset conditions. Therefore, by processing the target data collected by the target sensor according to the target algorithm, the accuracy of the wear detection result of the wearable device is high and the power consumption of wear detection is low. Attached Figure Description

[0019] To more clearly illustrate the technical solutions in this application, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0020] Figure 1 This is a schematic diagram of the steps of a wear detection method provided in an embodiment of this application;

[0021] Figure 2 This is a schematic flowchart of a wear detection method provided in an embodiment of this application;

[0022] Figure 3 This is a schematic diagram of the structure of a wear detection device provided in one embodiment of this application;

[0023] Figure 4 This is a schematic diagram of the structure of a wearable device provided in an embodiment of this application. Detailed Implementation

[0024] The technical solutions of this application will now be clearly and completely described with reference to the accompanying drawings. Obviously, the described embodiments are merely some embodiments of the present invention, and not all embodiments. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.

[0025] In one embodiment, such as Figure 1As shown, a wear detection method is provided. Although the logical order is illustrated in the step diagram, in some cases, the steps shown or described may be performed in a different order than that shown in the diagram. Specifically, this wear detection method can be applied to wearable devices, which may include, but are not limited to, one or more of smartwatches, smart bracelets, smart headphones, smart glasses, medical wearable devices (such as ambulatory blood pressure monitors and blood glucose monitors), children's watches, and health bracelets for the elderly. The smart glasses may be extended reality (XR) glasses, which may include, but are not limited to, augmented reality (AR) glasses, virtual reality (VR) glasses, and mixed reality (MR) glasses. The smart glasses may be wearable optical see-through smart glasses. Specifically, the wearable optical see-through smart glasses may at least include an optical display module consisting mainly of a microdisplay, optical lenses, and waveguide sheets to display virtual content to the user; an audio module consisting mainly of microphones and speakers to collect and play sound; sensors consisting mainly of cameras, gyroscopes, barometers, and infrared light emitters and receivers to collect information data related to the human body, the glasses themselves, and the external environment; an integrated processor with a microcontroller unit (MCU) or central processing unit (CPU) as its core to perform data processing and calculation; circuit boards, flexible or rigid, to connect other electronic components to form an electronic circuit system, which is generally placed inside the cavities of the glasses frame and temples; and a battery power supply.

[0026] The following sections provide detailed descriptions of each example. It should be noted that the order in which the embodiments are described is not intended to limit the priority of the embodiments.

[0027] according to Figure 1 The wear detection method shown includes at least steps S110 to S140, which are described in detail below:

[0028] In step S110, the currently available sensors of the wearable device are scanned, and the configuration information of the sensors is obtained.

[0029] In step S120, the configuration information of the sensor is input into a pre-trained algorithm selection model to obtain the target algorithm output by the algorithm selection model. The target algorithm achieves a wear detection accuracy and power consumption that meet preset conditions based on some or all of the data collected by the sensor.

[0030] In step S130, the target sensor corresponding to the target algorithm is determined, and the target data collected by the target sensor is acquired.

[0031] In step S140, the target data is processed based on the target algorithm to obtain the wearing detection result of the wearable device.

[0032] Wearable devices can scan for currently available sensors. Specifically, a wearable device can access multiple sensors integrated within itself and detect whether each sensor is currently online or in a wake-up state, thereby determining which sensors are currently available.

[0033] In one embodiment, the wearable device includes multiple different types of sensors, including sensors that are completely unrelated to wear detection. Multiple sensors related to wear detection can be predetermined, and the wearable device only needs to scan the currently available sensors among these.

[0034] In one embodiment, the sensor associated with wear detection may include, but is not limited to, one or more of the following: an accelerometer, a heart rate sensor, a gyroscope, a capacitive sensor, an infrared proximity sensor, a temperature sensor, a pressure sensor, and a switch sensor.

[0035] Among these technologies, accelerometers can detect minute vibrations or posture changes related to human activity to determine whether a wearable device is being worn. Heart rate sensors utilize the characteristic of detecting blood flow signals from the skin; they can only capture heart rate data when the device is in contact with the skin, directly distinguishing between wearing and not wearing the device, and also helping to determine the tightness of the fit. Gyroscopes can sense the angular motion and spatial posture of the wearable device to identify whether it is in a stable state in contact with the body; for example, if the device's posture is erratic or it remains stationary for a long time at an angle not in contact with the body, it is likely not being worn. Capacitive sensors rely on the fact that the human body is a conductor; when the electrodes of the wearable device come into contact with the skin, the capacitance value changes specifically, and monitoring this change allows for quick and accurate determination of whether the device is being worn. Infrared proximity sensors emit infrared light and receive reflected signals; when the device is close to the skin, the intensity of the reflected signal is significantly different from that in a suspended state, thus determining whether the device is being worn. Temperature sensors detect the temperature of objects in contact with the wearable device. If the temperature is close to the normal human body temperature range, it is likely that the device is being worn; if the detected temperature is ambient temperature, it is likely that the device is not being worn. When the wearable device is worn on the wrist, fingers, or other parts of the body, the strap or the device itself exerts pressure on the sensor. The pressure sensor identifies this pressure signal to determine whether the device is being worn and can also help detect the tightness of the fit. Switch sensors can be integrated into watch strap buckles or the connection between the temples and frames of eyeglasses. Opening or closing the wearable device triggers a switch signal, and the wearing status of the device can be inferred from this switch information.

[0036] Based on data collected by sensors related to wear detection, the wearing status of wearable devices can be accurately determined, thereby avoiding false data measurement when not worn and ensuring the accuracy of data such as heart rate and exercise; it can also trigger a low-power mode when not worn to extend battery life; at the same time, it reduces false triggering of functions and improves the user experience.

[0037] Wearable devices can obtain sensor configuration information by reading preset parameters and monitoring sensor parameters in real time. Sensor configuration information may include, but is not limited to, one or more of the following: sensor type, quantity of each type of sensor, sensor performance parameters, and sensor power consumption parameters. Sensor performance parameters may include, but are not limited to, one or more of the following: sampling frequency, detection accuracy, and response speed. Sensor power consumption parameters may include, but are not limited to, one or more of the following: current consumption in different operating modes and standby power consumption. Based on the number of sensors, it can be determined whether the wearable device uses a single accelerometer sensor or dual heart rate sensors, etc.

[0038] After acquiring the sensor configuration information, the wearable device can organize the collected configuration information into a standardized sensor configuration dataset and store it in a temporary database on the wearable device's local machine.

[0039] The algorithm selection model can be deployed locally on the wearable device or in the cloud. It can acquire execution parameters and information for various algorithms. Algorithm information may include data from the sensors the algorithm will use, the algorithm's accuracy, and the power consumption required to execute it. This algorithm is used for wear detection.

[0040] The configuration information of the currently available sensors of the wearable device is input into a pre-trained algorithm selection model. This model outputs the most suitable target algorithm (or the execution parameters of the target algorithm) for the currently available sensors of the wearable device. The accuracy and power consumption of wear detection based on the target algorithm both meet preset conditions. The power consumption of wear detection refers to the power consumption caused by executing the target algorithm, which may include the power consumption of the sensor in collecting and transmitting data for executing the target algorithm. The preset conditions can be set according to actual needs. For example, the preset conditions could be that the accuracy of wear detection is greater than an accuracy threshold, and the power consumption of wear detection is lower than a power consumption threshold, where the accuracy threshold and power consumption threshold can be set according to actual needs.

[0041] The target algorithm can be an algorithm that detects wearability based on data collected by all currently available sensors, or it can be an algorithm that detects wearability based on data collected by some currently available sensors. For example, wearable devices currently have four types of sensors available: A, B, C, and D. The first algorithm detects wearability using data collected by all four sensors (A, B, C, and D), while the second algorithm detects wearability using data collected by three sensors (A, B, and C). Compared to the second algorithm, although the first algorithm uses data collected by one more sensor (sensor D), its accuracy may not be as high as that of the second algorithm (for example, data collected by sensor D is prone to errors). Therefore, the first algorithm can be selected as the target algorithm.

[0042] After determining the target algorithm, the corresponding target sensor can be identified. The target sensor is the sensor that corresponds to the data used by the target algorithm. The target sensor is some or all of the sensors currently available in the wearable device.

[0043] The system acquires target data collected by the target sensor and performs preprocessing steps such as missing value imputation, outlier removal, standardization, normalization, and deduplication. Based on the target algorithm, the preprocessed target data is further processed to obtain the wear detection result of the wearable device. The wear detection result is used to characterize whether the wearable device is currently being worn, not being worn, or being worn but not properly.

[0044] When the wearable device is in use, it can accurately activate heart rate, exercise and other monitoring functions to ensure the validity of health and exercise data and provide users with reliable references. When the wearable device is not in use, it can automatically switch to a low-power mode to reduce power waste and extend battery life. If the wearable device is in a state where it is worn but not properly, it can issue a prompt to remind the user to adjust it in time, avoid data deviation caused by insufficient fit, prevent accidental function triggering, and improve the usability of the wearable device.

[0045] By adopting the technical solution of this application embodiment, when different wearable devices have different currently available sensors, the pre-trained algorithm selection model can select the target algorithm for the wearable device. In this way, there is no need to develop detection algorithms, train models and debug parameters separately for various wearable devices with different sensor configurations, which can save resources and reduce R&D costs. Moreover, the accuracy and power consumption of the target algorithm for wear detection based on the data collected by the sensor can meet the preset conditions. Therefore, by processing the target data collected by the target sensor according to the target algorithm, the accuracy of the wear detection result of the wearable device is high and the power consumption for wear detection is low.

[0046] Based on the above technical solution, as an embodiment, the step of inputting the configuration information of the sensor into a pre-trained algorithm selection model to obtain the target algorithm output by the algorithm selection model may include: inputting the configuration information of the sensor into the algorithm selection model to determine multiple candidate algorithms that use some or all of the data collected by the sensor; determining the accuracy and power consumption of the multiple candidate algorithms; obtaining the accuracy threshold and power consumption threshold in the preset conditions; determining multiple preliminary algorithms from the multiple candidate algorithms whose accuracy is greater than the accuracy threshold and whose power consumption is lower than the power consumption threshold; obtaining user preferences and the current remaining battery power of the wearable device; and determining the target algorithm from the multiple preliminary algorithms based on the user preferences, the remaining battery power, and the accuracy and power consumption of the multiple preliminary algorithms.

[0047] The algorithm selection model can obtain execution parameters and information for various algorithms. Algorithm information may include the data from the sensors the algorithm will use, the algorithm's accuracy, and the power consumption required to execute it. This algorithm is used for wear detection. Based on the input sensor configuration information, the algorithm selection model can identify multiple candidate algorithms from a pool of algorithms whose data matches the sensor configuration information. For example, if the first algorithm uses data collected by four sensors (A, B, C, and D) for wear detection, and the second algorithm uses three sensors (A, B, and C), and based on the sensor configuration information, it can be determined that the wearable device currently uses four sensors (A, B, C, and D), then both algorithms use data that matches the sensor configuration information.

[0048] The algorithm selection model can obtain the accuracy and power consumption of multiple candidate algorithms, as well as the accuracy threshold and power consumption threshold in the preset conditions, so as to determine multiple preliminary algorithms from multiple candidate algorithms with an accuracy greater than the accuracy threshold and a power consumption lower than the power consumption threshold.

[0049] It can acquire user preferences; user preferences indicate whether the user prefers to improve accuracy or reduce power consumption; user preferences can be user-defined. It can also acquire the wearable device's remaining battery power in real time. By combining user preferences, remaining battery power, the accuracy and power consumption of the current target algorithm and multiple preliminary algorithms, the target algorithm is determined from among the preliminary algorithms.

[0050] By employing the technical solution of this application embodiment, a dynamic balance between the performance and battery life of wearable devices can be achieved by combining user preferences and remaining battery power, making the decision-making target algorithm more aligned with current practical application scenarios. By considering user preferences, user needs can be met, avoiding an excessive pursuit of accuracy or power consumption reduction; combined with the current remaining battery power of the wearable device, accuracy can be fully guaranteed when the battery is sufficient, while low-power algorithms are automatically favored when the battery is low, extending usage time.

[0051] Based on the above technical solution, as an embodiment, determining the target algorithm from multiple preliminary algorithms according to the user preference, the remaining battery power, and the accuracy and power consumption of multiple preliminary algorithms may include: determining the current target based on the user preference and the remaining battery power; when the current target is to reduce power consumption, determining the algorithm with the lowest power consumption among the multiple preliminary algorithms as the target algorithm; when the current target is to improve accuracy, determining the algorithm with the highest accuracy among the multiple preliminary algorithms as the target algorithm.

[0052] The current objectives can include reducing power consumption and improving accuracy. If the user's preference leans more towards improving accuracy, it can be determined whether the remaining battery level is greater than a battery threshold. If the remaining battery level is greater than the threshold, and the user's preference leans more towards improving accuracy, then the current objective is to improve accuracy. If the user's preference leans more towards improving accuracy, but the remaining battery level is not greater than the threshold, then the current objective is to reduce power consumption. If the user's preference leans more towards reducing power consumption, then the current objective can be directly set as reducing power consumption.

[0053] When the current goal is to reduce power consumption, the algorithm with the lowest power consumption among several preliminary algorithms can be identified as the target algorithm. This is because the accuracy of the preliminary algorithms is greater than the accuracy threshold and the power consumption is lower than the power consumption threshold; therefore, the target algorithm with the lowest power consumption can still guarantee an accuracy greater than the accuracy threshold. When the current goal is to improve accuracy, the algorithm with the highest accuracy among several preliminary algorithms can be identified as the target algorithm. The preliminary algorithm with the highest accuracy can still guarantee a power consumption lower than the power consumption threshold.

[0054] By adopting the technical solution of this application embodiment, when the accuracy of the algorithm is greater than the accuracy threshold and the power consumption is lower than the power consumption threshold, the target algorithm that fits the current actual application scenario can be selected according to user preferences and remaining power.

[0055] Based on the above technical solution, as an example, determining the current target according to the user preference and the remaining battery power may include:

[0056] An initial target is set based on user preferences, and this initial target aligns with the user preferences. The current remaining battery level is then determined, and the initial target is adjusted accordingly. The battery level range includes high, medium, and low ranges. The high range can be a range not less than a first battery threshold (e.g., 50%), the medium range can be a range less than the first battery threshold but not less than a second battery threshold (e.g., 20%), and the low range can be a range less than the second battery threshold.

[0057] When the remaining battery power is in the high range, the battery is sufficient, and there is no need to adjust the initial target; the initial target can be directly used as the final target. For example, if the user prefers accuracy, the final target is to ensure accuracy to the fullest extent.

[0058] When the remaining battery power is in the medium range, the initial target can be fine-tuned. For example, if the user prefers high accuracy, the initial target can be revised to prioritize algorithms with lower power consumption while ensuring high accuracy. Alternatively, the algorithm with the lowest power consumption among several preliminary algorithms with accuracy greater than a first accuracy threshold can be selected as the target algorithm, where the first accuracy threshold is greater than a preset accuracy threshold. Conversely, if the user prefers low power consumption, the initial target can be revised to control power consumption without sacrificing basic accuracy. Alternatively, the algorithm with the highest accuracy among several preliminary algorithms with power consumption lower than a first power consumption threshold can be selected as the target algorithm, where the first power consumption threshold is lower than a preset power consumption threshold.

[0059] When the remaining battery power is low and power is scarce, the initial goal can be strongly constrained to reduce power consumption. For example, even if the user prefers accuracy, the goal should be adjusted to prioritize low power consumption while maintaining basic accuracy (accuracy greater than the accuracy threshold in the preset conditions), ensuring that the wearable device can continue to operate.

[0060] Based on the established current objective, the target algorithm can be selected from several preliminary algorithms.

[0061] Based on the above technical solution, as an embodiment, a preliminary algorithm can be determined from multiple candidate algorithms by combining the response time. The response time of each algorithm can be obtained. Preset conditions may include a response time threshold. From the multiple candidate algorithms, multiple preliminary algorithms are determined that have an accuracy greater than the accuracy threshold, a power consumption lower than the power consumption threshold, and a response time lower than the response time threshold, and then the target algorithm is determined from the multiple preliminary algorithms.

[0062] Based on the above technical solution, as an example, after determining multiple candidate algorithms, algorithms with an accuracy greater than the accuracy threshold can be selected from the multiple candidate algorithms. Under the premise that the accuracy meets the standard, the algorithm with the lowest power consumption or the power consumption lower than the power consumption threshold can be selected from the candidate algorithms with an accuracy greater than the accuracy threshold. When there are multiple algorithms with the lowest power consumption or the power consumption lower than the power consumption threshold among the candidate algorithms with an accuracy greater than the accuracy threshold, the algorithm with the shortest response time can be selected.

[0063] Based on the above technical solution, as an embodiment, the target algorithm may include a general detection algorithm; the general detection algorithm may include a random forest constructed based on multiple support vector machines; the multiple support vector machines are connected based on an objective function; each support vector machine is used to process data collected by different sensors.

[0064] A general detection algorithm can serve as a fallback, adaptable to any combination of sensors. This algorithm includes multiple Support Vector Machines (SVMs), with a separate SVM designed for each sensor type. Based on the hardware characteristics of the corresponding sensor (e.g., sampling frequency, accuracy range) and the characteristics of the collected data (e.g., signal fluctuation patterns, noise distribution), preprocessed effective features (e.g., time-domain statistical features, frequency-domain peak features, etc.) are extracted. These features are then used to train the corresponding SVM for that sensor using labeled data samples of the wear status, enabling each SVM to independently detect data from its corresponding sensor. Each trained SVM can be viewed as a decision tree in a random forest; this decision tree can output a clear detection result based on its own input features and possesses strong classification and generalization capabilities.

[0065] A random forest framework is built based on each Support Vector Machine (SVM) step, integrating all SVMs into the basic unit of the forest, and a custom objective function is designed. This objective function breaks down the isolation of individual SVMs by establishing feature mappings between them through feature association mechanisms (e.g., associating trend features of a temperature sensor with abrupt change features of a humidity sensor). Furthermore, the objective function can set weight allocations for each SVM through result fusion rules (e.g., assigning higher weights to sensor models with high detection accuracy), enabling collaborative interaction between decision trees rather than simple result aggregation.

[0066] The random forest is trained using a large number of different sensor combinations. The sensor data includes feature samples from different sensor combinations and environmental scenarios, allowing the objective function to continuously optimize feature association and weight allocation logic, ultimately resulting in an integrated general detection algorithm.

[0067] The technical solution adopted in this application embodiment allows the universal detection algorithm to adapt to any sensor combination. Regardless of whether the input is a conventional or unconventional sensor combination, the universal detection algorithm can output detection results with high accuracy through the collaborative calculation of its internal decision tree. Thus, even if some sensors in the wearable device malfunction or the sensor type is temporarily changed, the universal detection algorithm can still be quickly invoked to ensure uninterrupted wear detection function, avoiding wear detection failure due to changes in sensor configuration.

[0068] Based on the above technical solution, as an embodiment, the training steps of the algorithm selection model may include at least: acquiring configuration information of each sensor sample in multiple sensor combinations, and acquiring the actual wearing state corresponding to the sensor combination; the sensor combination is a combination constructed from multiple available sensor samples on a test wearable device; acquiring multiple algorithms; inputting the configuration information of each sensor sample in multiple sensor combinations into an initial algorithm selection model to obtain target algorithm samples selected from multiple algorithms for each sensor combination sample in the initial algorithm selection model; processing the data samples collected by each sensor sample in the sensor combination sample based on the target algorithm samples to obtain the wearing detection results corresponding to each sensor combination sample; training the initial algorithm selection model according to the difference between the wearing detection results and the actual wearing state corresponding to each sensor combination sample to obtain the trained algorithm selection model.

[0069] To train the model, training samples are needed. In this embodiment, sensor combinations can be determined as training samples for the algorithm selection model. To ensure the accuracy of the trained algorithm selection model, multiple sensor combinations can be obtained from a large number of different test wearable devices. Each sensor combination can be a set of multiple sensor samples currently available on each test wearable device. It is understood that multiple sensor combinations corresponding to a test wearable device can be obtained by changing the availability status of multiple sensor samples on a test wearable device.

[0070] For each sensor combination, obtain the configuration information of each sensor sample in that combination. The method for obtaining the configuration information of each sensor sample can refer to the method for obtaining the configuration information of each sensor described above, and will not be repeated here.

[0071] While acquiring configuration information, it is necessary to simultaneously collect the actual wearing status corresponding to each sensor combination. This actual wearing status can include wearing status, not wearing status, and wearing status but not properly. To avoid model bias due to imbalanced samples, a sufficient number of sensor combinations are needed for each actual wearing status.

[0072] Multiple algorithms can be acquired. These algorithms can be stored in an algorithm library, which can be set up locally or in the cloud. These algorithms can be collected or developed based on the needs of wearable device wear detection, and possess different characteristics. The algorithms can be tailored to different sensor combinations and can include high-accuracy algorithms (such as deep learning-based classification algorithms, suitable for accurate identification of wearing status, but with higher power consumption), low-power algorithms (such as traditional threshold-based algorithms, with low computational load, suitable for battery-sensitive scenarios), and balanced algorithms (such as lightweight machine learning algorithms, balancing accuracy and power consumption). Each algorithm can undergo basic testing in advance to ensure it can handle data collected from different sensors and has callable interfaces for easy subsequent model selection and data processing.

[0073] The configuration information of multiple sensor samples in each sensor combination is input into the initial algorithm selection model. The initial algorithm selection model will automatically determine the target algorithm sample based on the configuration information of the sensor samples and multiple algorithms in the algorithm library.

[0074] For example, for a combination of motion sensors with a high sampling frequency, the initial algorithm selection model may prioritize a high-accuracy algorithm; for a combination of physiological sensors with low accuracy, a balanced algorithm may be selected.

[0075] Based on the target algorithm samples corresponding to each sensor combination, the data samples collected by each sensor sample in the sensor combination are processed to obtain the wear detection results. Information such as power consumption, accuracy, and processing time for executing each target algorithm sample can be recorded to provide a reference for subsequent model optimization.

[0076] By comparing the wear detection results with the actual wear status, the model's decision logic is adjusted. This involves calculating a detection difference index for each sensor combination sample, which can include accuracy (the percentage of samples correctly identifying the wear status), false positive rate (the percentage of samples misclassified as not wearing the sensor), and false negative rate (the percentage of samples that fail to detect a half-wearing state), quantifying the detection performance of the target algorithm samples. If the detection results of a certain combination sample differ significantly from the actual state (e.g., accuracy below 85%), the selection logic of the initial algorithm selection model is traced back to analyze whether key parameters in the sensor configuration information were ignored (e.g., a high-accuracy algorithm fails due to an excessively low sampling frequency) or whether the algorithm matching rules are unreasonable (e.g., a high-power algorithm is selected for a low-power sensor combination). This leads to adjustments in the decision weights of the initial algorithm selection model (e.g., increasing the priority of sampling frequency in algorithm selection) and optimization of algorithm matching rules (e.g., increasing the probability of selecting a balanced algorithm for unconventional combinations).

[0077] The initial algorithm selection model is iteratively trained until all detection difference indicators of the initial algorithm selection model under multiple sensor combination samples meet preset thresholds (e.g., overall accuracy higher than 95% and false positive rate lower than 3%), resulting in a well-trained algorithm selection model. The well-trained algorithm selection model can automatically select the appropriate target algorithm based on the configuration information of any sensor combination, ensuring the reliability of wear detection.

[0078] The technical solution adopted in this application embodiment can provide a comprehensive and realistic training basis for the algorithm selection model with a large amount of sample data, avoiding decision bias caused by one-sided data; the trained algorithm selection model has strong adaptability and can adapt to any combination of sensors, solving the detection problem under unconventional configurations; through multiple rounds of iterative optimization, the trained algorithm selection model can be guaranteed to have high accuracy and low false positive and false negative rates, providing a stable guarantee for the wearing detection of wearable devices.

[0079] Based on the above technical solution, as an example, the wear detection method may further include: acquiring weather information;

[0080] The step of inputting the configuration information of the sensor into a pre-trained algorithm selection model to obtain the target algorithm output by the algorithm selection model includes: inputting the weather information and the configuration information of the sensor into the algorithm selection model to obtain the target algorithm output by the algorithm selection model.

[0081] Weather information may include, but is not limited to, one or more of temperature and humidity. Weather information can be obtained from weather forecast websites, etc. This application study found that the data collected by the sensor is affected by the weather, and therefore different weather conditions have a significant impact on the accuracy of wear detection.

[0082] When the weather temperature is too low (<10℃) or too high (>35℃), it will affect the measurement of the temperature of the user's exposed skin, thus reducing the accuracy of wear detection based on body temperature data collected by the temperature sensor. Weather humidity will affect the measurement of capacitive sensors; in high humidity environments, moisture may be located at or near the capacitive sensor and be collected, thus significantly affecting the value and performance of the capacitive sensor, leading to incorrect wear status detection results.

[0083] The accuracy of each algorithm under different weather conditions can be determined in advance through experiments. Weather information and sensor configuration information are input into the algorithm selection model, which automatically outputs the target algorithm for that weather condition. The target algorithm, based on data collected by some or all of the sensors, achieves accuracy and power consumption for wear detection under that weather condition, meeting preset conditions. The target sensor corresponding to the target algorithm is determined, and the target data collected by the target sensor is acquired. The target data is processed based on the target algorithm to obtain the wear detection results of the wearable device.

[0084] In this embodiment, the algorithm selection model is obtained by supervised training of the initial algorithm selection model based on weather information samples and sensor combinations.

[0085] The technical solution adopted in this application takes into account the impact of weather information on the wear detection results. The weather information and sensor configuration information are input together into the algorithm selection model. The target algorithm obtained is an algorithm that can still meet the preset conditions of wear detection accuracy and power consumption under the current weather information, thereby ensuring the accuracy of wear detection.

[0086] Figure 2 This is a schematic flowchart of a wear detection method provided in an embodiment of this application; see reference. Figure 2 The wear detection method may include: the wearable device scanning available sensors at startup and acquiring their configuration information, then integrating this information into a sensor configuration dataset. It can detect whether the wearable device has a local adaptive algorithm library, which includes multiple algorithms. If the wearable device has a local adaptive algorithm library, it retrieves it directly; otherwise, it retrieves it from the cloud. An algorithm selection model can be used to select a sensor-compatible target algorithm from the adaptive algorithm library. If the algorithm selection model selects a sensor-compatible target algorithm from the adaptive algorithm library, the most suitable target algorithm can be selected based on multi-objective optimization. This multi-objective optimization may include selection based on accuracy, power consumption, and / or response time. If the algorithm selection model does not select a sensor-compatible target algorithm from the adaptive algorithm library, a fallback general detection algorithm can be selected. Wear detection of the wearable device is then performed based on the determined most suitable target algorithm or general detection algorithm.

[0087] In one embodiment, the wear detection method may further include building an adaptive algorithm library. The adaptive algorithm library is used to provide accurately matched wear detection algorithms for wearable devices with different configurations and in different scenarios. Building the adaptive algorithm library may include building wear detection algorithms for different sensor combinations.

[0088] For example, one algorithm could be a wear detection algorithm for accelerometers and heart rate sensors. This algorithm can determine the basic wear status by extracting the motion amplitude characteristics of the accelerometer (such as the motion pattern when wearing the device) and the signal stability characteristics of the heart rate sensor (such as large signal fluctuations when not wearing the device). This algorithm does not require complex computation and can run efficiently on devices with limited hardware resources, solving the detection needs of basic sensor combinations.

[0089] One algorithm is a wear detection algorithm for accelerometers, gyroscopes, and capacitive sensors. This algorithm can use multi-dimensional feature collaboration, using accelerometers to determine the overall motion state, gyroscopes to capture subtle posture changes (such as whether the device fits the skin), and capacitive sensors to directly detect skin contact signals, thereby performing wear detection. This significantly improves the detection accuracy in complex scenarios (such as when the device is loose during exercise or when sweating causes signal interference), and meets the high requirements of high-end devices for detection reliability.

[0090] One algorithm is a wear detection algorithm for a single accelerometer sensor. This algorithm can extract only the static gravity component (e.g., when the device is worn at a fixed angle, the gravity component is stable) and dynamic activity frequency (e.g., when the device is not worn, it shakes irregularly) from the accelerometer sensor, and quickly determine the wearing status by using a preset threshold. This algorithm has extremely low computational complexity, which can reduce device power consumption and solve the detection pain points in single-sensor, low-power scenarios.

[0091] The adaptive algorithm library can reserve extension interfaces to support the addition of new algorithms in the future, ensuring the flexibility and iterability of the adaptive algorithm library.

[0092] Each algorithm in the adaptive algorithm library can be assigned quantifiable and verifiable evaluation metrics, forming a table showing the correspondence between algorithm, performance, and power consumption, providing a clear basis for subsequent algorithm selection. Evaluation metrics can include accuracy, response time, and power consumption. The evaluation metrics for each algorithm can be determined through testing with a large number of labeled samples.

[0093] Considering the storage space limitations of wearable devices, the adaptive algorithm library can be stored on the device itself or in the cloud, and the selection can be made by querying the device's preset information. If the device's preset information indicates that the adaptive algorithm library is stored locally, the wearable device's algorithm selection model can call the sensor configuration dataset in a temporary database and match it with the algorithms in the adaptive algorithm library to filter out target algorithms compatible with the current device's sensor configuration. If the device's preset information indicates that the adaptive algorithm library needs to be obtained from the cloud, the wearable device will first check the network connection. If it is normal, it will actively retrieve the adaptive algorithm library for adaptation and filter out algorithms compatible with the current device's available sensor configurations. If there is no network connection, the user will be prompted to connect to the network.

[0094] It can monitor the availability status of sensors in real time (such as sensor failure, offline status, etc.). If a change in the availability status of the sensor is detected, the algorithm matching and selection process is immediately re-triggered to ensure that the wear detection function remains effective.

[0095] In one embodiment, the two wearable devices are a smartwatch A and a smart bracelet B. Smartwatch A is equipped with an accelerometer (sampling frequency 50Hz), a heart rate sensor (detection accuracy ±2bpm), a gyroscope (sampling frequency 100Hz), and a skin capacitive sensor (response time 50ms). Smart bracelet B is equipped only with an accelerometer (sampling frequency 30Hz) and a basic heart rate sensor (detection accuracy ±5bpm). After smartwatch A is powered on, the sensor information acquisition module scans and obtains its sensor configuration: "accelerometer + heart rate sensor + gyroscope + capacitive sensor," and records the performance and power consumption parameters of each sensor. After smart bracelet B is powered on, the sensor information acquisition module scans and obtains its sensor configuration: "accelerometer + basic heart rate sensor," and records the corresponding parameters.

[0096] The trained algorithm selection model was deployed to smartwatch A and smart bracelet B respectively. Smartwatch A called the algorithm selection model, input the sensor configuration information, and the model automatically output the execution parameters of algorithm B, initiating the wear detection function. Smart bracelet B called the algorithm selection model, input the sensor configuration information, and the model automatically output the execution parameters of algorithm A, initiating the wear detection function. Specifically, the algorithm selection model of smartwatch A called the sensor configuration data, matched it with the adaptive algorithm library, and selected algorithm B ("accelerometer + gyroscope + capacitance" combination) as a candidate algorithm. Evaluation showed that algorithm B had an accuracy of 98%, a response time of 80ms, and an average daily power consumption of 4mAh, meeting the preset conditions, so algorithm B was selected. The algorithm selection model of smart bracelet B selected algorithm A ("accelerometer + heart rate" combination) as a candidate algorithm after matching. Evaluation showed that algorithm A had an accuracy of 96%, a response time of 90ms, and an average daily power consumption of 2mAh, meeting the low power consumption requirement, so algorithm A was selected.

[0097] When the capacitive sensor of smartwatch A is disconnected, the sensor configuration change can be detected, the matching process is retried, and the algorithm D with the combination of "accelerometer + heart rate + gyroscope" (accuracy 97%, response time 85ms, daily power consumption 4.5mAh) is selected. The device automatically switches to algorithm D, and the detection function continues to work normally.

[0098] Among them, smartwatch A and smart bracelet B share the same algorithm selection model, eliminating the need for separate training and shortening the development time by 70%. The daily power consumption of smartwatch A has been reduced from 6 mAh to 4 mAh using the traditional algorithm, extending battery life by 33%. The daily power consumption of smart bracelet B has been reduced from 3 mAh to 2 mAh using the traditional algorithm, extending battery life by 33%. Both devices have a wear detection accuracy of ≥96% and a response time of ≤90ms, meeting user needs.

[0099] The wear detection method of this application can be used in smart wristbands or smart bracelets as a pre-wear judgment for health monitoring (such as heart rate and blood oxygen detection), avoiding accidental sensor activation when not worn, thereby reducing power consumption; it can be used in smart headphones to realize the function of pausing when removed and playing when worn through wear detection, and is compatible with the sensor configurations of different headphones (such as in-ear detection sensors and capacitive sensors); it can be used in smart glasses as wear status recognition in AR / VR scenarios, ensuring that the displayed content is only activated when worn, and is compatible with the lightweight sensor configurations of different glasses; it can be used in medical-grade wearable devices to ensure the validity of monitoring data through accurate wear detection, and is compatible with the dedicated sensors of medical devices (such as pressure sensors); it can be used in children's / elderly wearable devices, for children's watches and elderly health bracelets with low power consumption and simple sensor configurations, reducing power consumption and extending battery life through simplified algorithms, while ensuring the accuracy of wear detection.

[0100] The technical solution adopted in this application breaks through the traditional "one-to-one" algorithm adaptation mode. Through automatic sensor configuration matching and multi-target optimization selection, dynamic adaptation between the wear detection algorithm and device sensors can be achieved without manual intervention. The algorithm selection model covers a variety of sensor combinations and has the ability to identify sensor configurations and adaptively adjust algorithm parameters, realizing "one-time model training, multi-device deployment," greatly simplifying the R&D process. During the algorithm selection process, detection accuracy, response time, and power consumption are considered simultaneously, avoiding unnecessary sensor calls and minimizing device power consumption while ensuring functional performance. The sensor's working status can be monitored in real time. If the sensor configuration changes, the algorithm adaptation process is automatically re-triggered, ensuring the stability and continuity of the wear detection function.

[0101] Based on actual experimental results, the "one-time model training, multi-device deployment" approach can eliminate the need for repetitive algorithm design, training, and debugging for different devices, shortening the R&D cycle by more than 60% and improving production efficiency by more than 50%. The adaptive algorithm selection mechanism can reduce unnecessary sensor calls by 20%-40%, and combined with a low-power algorithm priority selection strategy, the average daily power consumption of the device is reduced by about 30%, and the battery life is extended by 25%-35%. The algorithm selection model, after being trained with multi-sensor combination samples, maintains a detection accuracy of over 95% when adapted to different devices, with a response time of ≤100ms, meeting users' requirements for real-time and accurate wearable detection. It can be adapted to smart wearable devices with various sensor combinations such as accelerometers, heart rate sensors, gyroscopes, and capacitive sensors, and can support the access of new types of sensors without modifying the core logic, reducing product iteration costs.

[0102] To facilitate better implementation of the wearing detection method of this application, this application also provides a wearing detection device based on the above-described wearing detection method. The meanings of the terms used are the same as in the wearing detection method described above, and specific implementation details can be found in the description of the method embodiments.

[0103] Please see Figure 3 , Figure 3 This is a schematic diagram of the wear detection device provided in an embodiment of this application, wherein the wear detection device includes:

[0104] The information acquisition module 301 is used to scan the currently available sensors of the wearable device and acquire the configuration information of the sensors;

[0105] The algorithm output module 302 is used to input the configuration information of the sensor into the pre-trained algorithm selection model to obtain the target algorithm output by the algorithm selection model; the target algorithm meets preset conditions in terms of accuracy and power consumption for wear detection based on some or all of the data collected by the sensor.

[0106] The data acquisition module 303 is used to determine the target sensor corresponding to the target algorithm and to acquire the target data collected by the target sensor.

[0107] The data processing module 304 is used to process the target data based on the target algorithm to obtain the wearing detection result of the wearable device.

[0108] In one embodiment, the algorithm output module 302 is specifically used to perform:

[0109] The configuration information of the sensor is input into the algorithm selection model to determine multiple candidate algorithms that use some or all of the data collected by the sensor.

[0110] Determine the accuracy and power consumption of multiple candidate algorithms;

[0111] Obtain the accuracy threshold and power consumption threshold from the preset conditions;

[0112] From the candidate algorithms, determine a plurality of preliminary algorithms in which the accuracy is greater than the accuracy threshold and the power consumption is lower than the power consumption threshold;

[0113] Obtain user preferences and the current remaining battery power of the wearable device;

[0114] The target algorithm is determined from among the preliminary algorithms based on the user preferences, the remaining battery power, and the accuracy and power consumption of the preliminary algorithms.

[0115] In one embodiment, determining the target algorithm from a plurality of preliminary algorithms based on the user preference, the remaining battery power, and the accuracy and power consumption of the plurality of preliminary algorithms includes:

[0116] Determine the current target based on the user preferences and the remaining battery power;

[0117] When the current goal is to reduce power consumption, the algorithm with the lowest power consumption among the multiple preliminary algorithms is determined as the target algorithm;

[0118] When the current goal is to improve accuracy, the algorithm with the highest accuracy among the multiple preliminary algorithms is determined as the target algorithm.

[0119] In one embodiment, the training steps of the algorithm for selecting the model include at least:

[0120] The configuration information of each sensor sample in a multi-sensor combination is obtained, as well as the actual wearing state corresponding to the sensor combination; the sensor combination is a combination constructed from multiple available sensor samples on the test wearable device.

[0121] Acquire multiple algorithms;

[0122] The configuration information of each sensor sample in the multiple sensor combinations is input into the initial algorithm selection model to obtain the target algorithm sample selected by each sensor combination sample from multiple algorithms.

[0123] Based on the target algorithm sample, the data samples collected by each sensor sample in the sensor combination sample are processed to obtain the wear detection result corresponding to each sensor combination sample;

[0124] Based on the difference between the wearing detection results and the actual wearing state corresponding to each of the sensor combination samples, the initial algorithm selection model is trained to obtain the trained algorithm selection model.

[0125] In one embodiment, the target algorithm includes a general detection algorithm; the general detection algorithm includes a random forest constructed based on multiple support vector machines; the multiple support vector machines are connected based on an objective function; each support vector machine is used to process data collected by different sensors.

[0126] In one embodiment, the device further includes:

[0127] The weather acquisition module is used to obtain weather information;

[0128] The algorithm output module 302 is specifically used to perform the following: inputting the weather information and the configuration information of the sensor into the algorithm selection model to obtain the target algorithm output by the algorithm selection model.

[0129] In one embodiment, the sensor includes one or more of the following: an accelerometer, a heart rate sensor, a gyroscope, a capacitive sensor, an infrared proximity sensor, a temperature sensor, a pressure sensor, and a switch sensor.

[0130] By adopting the technical solution of this application embodiment, when different wearable devices have different currently available sensors, the pre-trained algorithm selection model can select the target algorithm for the wearable device. In this way, there is no need to develop detection algorithms, train models and debug parameters separately for various wearable devices with different sensor configurations, which can save resources and reduce R&D costs. Moreover, the accuracy and power consumption of the target algorithm for wear detection based on the data collected by the sensor can meet the preset conditions. Therefore, by processing the target data collected by the target sensor according to the target algorithm, the accuracy of the wear detection result of the wearable device is high and the power consumption for wear detection is low.

[0131] Specific limitations regarding the wear detection device can be found in the limitations of the wear detection method described above, and will not be repeated here. Each module in the aforementioned wear detection device can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in or independent of the processor in a computer device in hardware form, or stored in the memory of a computer device in software form, so that the processor can call and execute the operations corresponding to each module.

[0132] Furthermore, this application also provides a wearable device, which may include, but is not limited to, one or more of the following: smartwatches, smart bracelets, smart headphones, smart glasses, medical wearable devices (such as ambulatory blood pressure monitors and blood glucose monitors), children's watches, and health bracelets for the elderly. Figure 4 As shown, it illustrates the structural diagram of the wearable device involved in this application, specifically:

[0133] The wearable device may include components such as a processor 401 with one or more processing cores and a memory 402 of one or more computer-readable storage media. Those skilled in the art will understand that... Figure 4 The wearable device structure shown does not constitute a limitation on the wearable device and may include more or fewer components than illustrated, or combine certain components, or have different component arrangements. Wherein:

[0134] The processor 401 is the control center of the wearable device. It connects various parts of the wearable device via various interfaces and lines. By running or executing software programs and / or modules stored in the memory 402, and by calling data stored in the memory 402, it performs various functions and processes data, thereby providing overall monitoring of the wearable device. Optionally, the processor 401 may include one or more processing cores; preferably, the processor 401 may integrate an application processor and a modem processor, wherein the application processor mainly handles the operating system, user interface, and applications, while the modem processor mainly handles wireless communication. It is understood that the modem processor may not be integrated into the processor 401.

[0135] The memory 402 can be used to store software programs and modules. The processor 401 executes various functional applications and data processing by running the software programs and modules stored in the memory 402. The memory 402 may mainly include a program storage area and a data storage area. The program storage area may store the operating system, at least one application program required for a function (such as sound playback function, image playback function, etc.), etc.; the data storage area may store data created based on the use of the wearable device, etc. In addition, the memory 402 may include high-speed random access memory, and may also include non-volatile memory, such as at least one disk storage device, flash memory device, or other volatile solid-state storage device. Accordingly, the memory 402 may also include a memory controller to provide the processor 401 with access to the memory 402.

[0136] In one embodiment, the wearable device further includes a power supply 403 that supplies power to the various components. Preferably, the power supply 403 can be logically connected to the processor 401 through a power management system, thereby enabling functions such as charging, discharging, and power consumption management through the power management system. The power supply 403 may also include one or more DC or AC power supplies, recharging systems, power equipment debugging circuits, power converters or inverters, power status indicators, and other arbitrary components.

[0137] In one embodiment, the wearable device may further include an input unit 404, which can be used to receive input digital or character information and generate keyboard, mouse, joystick, optical or trackball signal inputs related to user settings and function control.

[0138] Although not shown, wearable devices may also include display units, etc., which will not be described in detail here. Specifically, in this embodiment, the processor 401 in the wearable device loads the executable files corresponding to the processes of one or more applications into the memory 402 according to the following instructions, and the processor 401 runs the applications stored in the memory 402, thereby implementing the steps in any of the wear detection methods provided in the embodiments of this application.

[0139] Those skilled in the art will understand that Figure 4 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the wearable device to which the present application is applied. A specific wearable device may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0140] In one embodiment, a wearable device is provided, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the methods described in any embodiment of this application.

[0141] In one embodiment, a computer-readable storage medium is provided having a computer program stored thereon, which, when executed by a processor, implements the method described in any embodiment of this application.

[0142] In some embodiments, a computer program product is also provided, including a computer program or instructions that, when executed by a processor, implement the methods described in any embodiment of this application.

[0143] For details on the implementation of each of the above operations, please refer to the previous examples, which will not be repeated here.

[0144] Those skilled in the art will understand that all or part of the steps in the various methods of the above embodiments can be performed by instructions, or by instructions controlling related hardware. These instructions can be stored in a computer-readable storage medium and loaded and executed by a processor.

[0145] Therefore, this application provides a computer-readable storage medium storing a computer program that can be loaded by a processor to execute the steps of any of the wear detection methods provided in this application.

[0146] For details on the implementation of each of the above operations, please refer to the previous examples, which will not be repeated here.

[0147] The computer-readable storage medium may include: read-only memory (ROM), random access memory (RAM), disk or optical disk, etc.

[0148] Since the instructions stored in the computer-readable storage medium can execute the steps of any of the wearing detection methods provided in this application, the beneficial effects that any of the wearing detection methods provided in this application can achieve can be realized, as detailed in the preceding embodiments, and will not be repeated here.

[0149] Finally, it should be noted that in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or terminal device that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or terminal device. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or terminal device that includes said element.

[0150] The above provides a detailed description of a wear detection method, apparatus, wearable device, and computer-readable storage medium provided in this application. Specific examples have been used to illustrate the principles and implementation methods of the present invention. The description of the above embodiments is only for the purpose of helping to understand the method and core ideas of the present invention. At the same time, those skilled in the art will recognize that there will be changes in the specific implementation methods and application scope based on the ideas of the present invention. Therefore, the content of this specification should not be construed as a limitation of the present invention.

Claims

1. A method for detecting wearability, characterized in that, The wear detection method, applied to wearable devices, includes: Scan the currently available sensors of the wearable device and obtain the configuration information of the sensors; The configuration information of the sensor is input into a pre-trained algorithm selection model to obtain a target algorithm selected by the algorithm selection model from multiple candidate algorithms. The target algorithm, based on some or all of the data collected by the sensor, achieves a wear detection accuracy and power consumption that meet preset conditions. The process of inputting the sensor configuration information into the pre-trained algorithm selection model to obtain a target algorithm selected by the algorithm selection model from multiple candidate algorithms includes: inputting the sensor configuration information into the algorithm selection model to determine multiple candidate algorithms using some or all of the data collected by the sensor; determining the accuracy and power consumption of the multiple candidate algorithms; obtaining an accuracy threshold and a power consumption threshold from the preset conditions; determining multiple preliminary algorithms from the multiple candidate algorithms whose accuracy is greater than the accuracy threshold and whose power consumption is lower than the power consumption threshold; obtaining user preferences and the current remaining battery power of the wearable device; determining the current target based on the user preferences and the remaining battery power; when the current target is to reduce power consumption, determining the algorithm with the lowest power consumption among the multiple preliminary algorithms as the target algorithm; when the current target is to improve accuracy, determining the algorithm with the highest accuracy among the multiple preliminary algorithms as the target algorithm. Determine the target sensor corresponding to the target algorithm, and acquire the target data collected by the target sensor; The target data is processed based on the target algorithm to obtain the wearing detection result of the wearable device.

2. The wearing detection method according to claim 1, characterized in that, The training steps for selecting the model in the algorithm include at least the following: The configuration information of each sensor sample in a multi-sensor combination sample is obtained, as well as the actual wearing state corresponding to the sensor combination sample; the sensor combination sample is a combination constructed from multiple available sensor samples on the test wearable device; Acquire multiple algorithms; The configuration information of each sensor sample in the multiple sensor combination samples is input into the initial algorithm selection model to obtain the target algorithm sample selected by each sensor combination sample from multiple algorithms. Based on the target algorithm sample, the data samples collected by each sensor sample in the sensor combination sample are processed to obtain the wear detection result corresponding to each sensor combination sample; Based on the difference between the wearing detection results and the actual wearing state corresponding to each of the sensor combination samples, the initial algorithm selection model is trained to obtain the trained algorithm selection model.

3. The wearing detection method according to claim 1, characterized in that, The target algorithm includes a general detection algorithm; the general detection algorithm includes a random forest built based on multiple support vector machines; the multiple support vector machines are connected based on an objective function; each support vector machine is used to process data collected by different sensors.

4. The wearing detection method according to claim 1, characterized in that, The wear detection method also includes: Get weather information; The step of inputting the configuration information of the sensor into a pre-trained algorithm selection model to obtain the target algorithm selected by the algorithm selection model from multiple candidate algorithms further includes: The weather information and the configuration information of the sensors are input into the algorithm selection model, and the target algorithm is selected and output by the algorithm selection model from multiple candidate algorithms.

5. The wearing detection method according to claim 1, characterized in that, The sensors include one or more of the following: accelerometer, heart rate sensor, gyroscope, capacitive sensor, infrared proximity sensor, temperature sensor, pressure sensor, and switch sensor.

6. A wear detection device, characterized in that, The wear detection device, applied to wearable devices, includes: The information acquisition module is used to scan the currently available sensors of the wearable device and acquire the configuration information of the sensors; An algorithm output module is used to input the configuration information of the sensor into a pre-trained algorithm selection model to obtain a target algorithm selected by the algorithm selection model from multiple candidate algorithms. The target algorithm satisfies preset conditions in terms of accuracy and power consumption for wear detection based on some or all of the data collected by the sensor. The process of inputting the configuration information of the sensor into the pre-trained algorithm selection model to obtain the target algorithm selected by the algorithm selection model from multiple candidate algorithms includes: inputting the configuration information of the sensor into the algorithm selection model to determine multiple candidate algorithms using some or all of the data collected by the sensor; determining the accuracy and power consumption of the multiple candidate algorithms; obtaining the accuracy threshold and power consumption threshold in the preset conditions; determining multiple preliminary algorithms from the multiple candidate algorithms whose accuracy is greater than the accuracy threshold and whose power consumption is lower than the power consumption threshold; obtaining user preferences and the current remaining battery power of the wearable device; determining the current target based on the user preferences and the remaining battery power; when the current target is to reduce power consumption, determining the algorithm with the lowest power consumption among the multiple preliminary algorithms as the target algorithm; when the current target is to improve accuracy, determining the algorithm with the highest accuracy among the multiple preliminary algorithms as the target algorithm. The data acquisition module is used to determine the target sensor corresponding to the target algorithm and to acquire the target data collected by the target sensor. The data processing module is used to process the target data based on the target algorithm to obtain the wearing detection result of the wearable device.

7. A wearable device, characterized in that, It includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps of the wear detection method as described in any one of claims 1 to 5.

8. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program that, when executed by a processor, implements the steps of the wear detection method as described in any one of claims 1 to 5.

Images