Alarm function interaction control method and system for smart watch

Abstract

The present application relates to the field of smart watch interaction control, and particularly relates to a smart watch alarm function interaction control method and system. The smart watch alarm function interaction control method comprises the following steps: S10: setting a pre-shaking time, when it is obtained from the collected wrist physiological signal that the user is in a wake state, wrist movement information is obtained as reference information within the pre-shaking time before the alarm trigger. The present application identifies the closing intention by collecting wrist movement information, extracting the value and frequency change characteristics, and then comparing the characteristic differences before and after the alarm trigger, which not only breaks through the limitation of fixed gesture input instruction, but also adapts to the closing instruction input requirements of different gestures in multiple scenes, effectively solves the problem of insufficient coverage of the user's use demand caused by the way of closing the smart watch alarm through the touch screen, resulting in low user experience, and finally realizes natural, accurate and personalized alarm closing interaction.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of smart watch interaction control, in particular to a smart watch alarm function interaction control method and system. BACKGROUND

[0002] With the popularity of smart wearable devices, smart watches have become the core tool for users to manage their daily time with their portability and multi-functional integration. The alarm function, as a basic and high-frequency application, covers diversified use requirements, including both the fixed morning wake-up scenario to help users smoothly transition from sleep to wakefulness and the fragmented daily timing reminder scenario, such as meeting duration control in work, Pomodoro timing in study, and cooking duration monitoring in life, to help users efficiently plan their time and improve their transaction processing efficiency. Therefore, the practicality and operation convenience of the alarm function directly affect the overall user experience of the smart watch.

[0003] Currently, the alarm control of the smart watch is mostly based on touch screen interaction. For example, the method for gesture control of alarm clock and smart terminal disclosed in Chinese patent application file CN104102444A realizes alarm control by detecting touch screen touch signals. For example, when detecting a touch signal with a long press time ≥ a preset threshold (such as 2 seconds), the alarm is turned off; when detecting a touch signal with a sliding distance ≥ a preset threshold (such as 2 centimeters), the snooze function is started. This method does not require the user to accurately identify the alarm menu on the screen, and can be operated simply by long pressing or sliding, especially suitable for the needs of users who are not fully conscious and whose body movements are not accurate in the morning scenario, effectively avoiding the inconvenience caused by traditional accurate clicking (on the smart watch screen) operation, and improving the use convenience of the morning alarm.

[0004] However, the user's demand for the smart watch alarm is not limited to the morning scenario, and the use frequency of the daily timing alarm is also high, and there is a significant difference between the core demands of the two. When using the timing alarm, the user mostly only needs to remind the user of the timing end of a specific transaction (such as the end of a meeting, coffee being ready, cooking nodes, etc.) at the preset time point, without the need to provide a persistent and strong wake-up prompt like the morning alarm. More importantly, when the timing alarm is triggered, the user is often in a focused state (such as using a computer to handle work, operating kitchen utensils to cook, holding a book to read, etc.), and both hands may be occupied or inconvenient to leave the current operation object. The touch control success rate of the capacitive screen is greatly reduced when the hands are wet, gloves are worn, or tools are held, making it difficult or inconvenient to turn off the alarm by touching the screen, resulting in insufficient coverage of the user's use demand by the existing touch control method. This insufficient demand coverage problem will cause the alarm to continue ringing and interfere with the user's current transaction, significantly reducing the user experience in the timing alarm scenario. SUMMARY

[0005] The application provides an alarm function interaction control method and system for a smart watch, to solve the problem of insufficient coverage of the user's use demand for the way of closing the smart watch alarm by the touch screen, resulting in low user experience.

[0006] To solve the above technical problems, the application provides the following technical solutions:

[0007] The alarm function interaction control method for the smart watch comprises the following steps:

[0008] S10: setting a pre-shaking time, when it is obtained from the collected wrist physiological signal that the user is in a wake state, wrist movement information is obtained as reference information within the pre-shaking time before the alarm trigger, real-time collection of wrist movement information after the alarm trigger is taken as comparison information, and the change characteristics of the reference information and the comparison information are analyzed in combination with the collection time; the change characteristics include numerical change characteristics and frequency change characteristics;

[0009] S20: judging whether the change characteristics of the reference information meet the preset instruction characteristics, if:

[0010] S20a: if not, the alarm is closed when the change characteristics of the comparison information meet the preset instruction characteristics;

[0011] S20b: if yes, the change characteristics of the reference information are extracted as reference characteristics, the change characteristics of the comparison information are extracted as comparison characteristics, the difference between the reference characteristics and the comparison characteristics is comprehensively evaluated according to the numerical change characteristics and the frequency change characteristics of the comparison characteristics and the reference characteristics, and the alarm is closed when the comprehensive evaluation result meets the preset instruction characteristics.

[0012] The basic scheme principle and benefits are as follows: in the user's conscious state, the scheme collects and analyzes the wrist movement information (such as the fluctuation amplitude of the acceleration value collected by the smart watch) and the frequency change characteristics (such as the dynamic law of the vibration period) before and after the alarm clock is triggered; then, by comparing whether there is a difference in the change characteristics of the wrist movement information before and after the alarm clock is triggered, it is determined whether the user has the intention to close the alarm clock. In this process, the scheme does not limit the operation to a specific gesture (such as fixed point finger, sliding, etc.), but captures the user's natural operation through the "change comparison" of the movement characteristics (numerical characteristics and frequency characteristics) obtained at the hand (wrist). No matter how the user expresses the closing intention through point operation, hand shaking or grip adjustment, as long as the numerical and frequency change characteristics of the wrist movement information before and after the alarm clock is triggered meet the "active operation for the alarm clock" (i.e. the pre-set instruction characteristics) rule (i.e. the user behaves differently after the alarm clock is triggered to close the alarm clock than the current focus), it can be recognized. For example, when the user is taking notes in a meeting (conscious state), the smart watch recognizes the regular change (including numerical change and frequency change) formed by holding a pen to write during the pre-shaking time; after the alarm clock is triggered, the user does not want to put down the pen because of handling the meeting record, and changes the movement rhythm by quickly and lightly shaking the wrist (such as changing the frequency from 1 Hz of uniform writing to 3 Hz of rapid shaking, and changing the numerical characteristics from stable to sudden increase), and the smart watch compares the difference between the reference characteristics and the comparison characteristics to recognize that it is an action to deliberately respond to the alarm clock and close the alarm clock. This design breaks through the limitation of traditional touch or fixed gestures, better fits the user's natural behavior habits in different scenarios (such as small movements when both hands are occupied), makes the acquisition of closing instructions more accurate, greatly improves the user's experience without deliberate cooperation; and can also distinguish between daily habitual actions and response actions for the alarm clock, and strengthen the adaptation to complex scenarios.

[0013] Secondly, by analyzing the numerical change characteristics (such as the trend of signal strength increase and decrease) and the frequency change characteristics (such as the mutation law of vibration frequency), in combination with the vibration when the alarm clock is triggered, the scheme can more accurately capture the user's operation intention under the disturbance of the alarm clock vibration. For example, when the user is cooking and both hands are occupied (for example, holding kitchen utensils), the hand has a continuous movement before the alarm clock is triggered (including specific numerical and frequency change characteristics); when the alarm clock vibrates, the user sends a closing instruction by adjusting the grip (such as tightening or loosening the kitchen utensils), and when the alarm clock vibrates, the vibration will amplify the numerical fluctuation and frequency mutation caused by the grip change, so that the smart watch can accurately distinguish between normal stirring action and deliberate grip adjustment by recognizing this amplified change characteristic, avoid misjudging daily action as closing instruction, and at the same time, without the user interrupting cooking to touch the screen or completing a specific gesture, the difficulty of the user entering the closing alarm clock instruction is reduced, further solving the problem of touch inconvenience in focused scenarios, and improving the user's experience.

[0014] Finally, since the scheme extracts the user personalized numerical and frequency change features (such as the reference feature of the pre-shaking time), it can naturally adapt to the operation habits of different users and solve the problem of recognition failure caused by individual motion differences. For example, elderly users may be used to closing the alarm clock with a large amplitude wrist swing (numerical change feature is high amplitude fluctuation, frequency change feature is low frequency), while young users may prefer fast fingertip point operation (numerical change feature is high frequency small amplitude fluctuation, frequency change feature is high frequency); the smart watch compares the reference feature of the pre-shaking time with the comparison feature after triggering to evaluate the difference, without the need for users to learn a unified standard action, the intention can be accurately recognized, which is difficult to achieve by traditional fixed gesture control, further improving the use experience of different users.

[0015] In summary, the scheme collects wrist motion information, extracts numerical and frequency change features, and then compares the feature differences before and after the alarm trigger to recognize the closing intention, which not only breaks through the limitations of fixed gesture input instructions, but also adapts to the closing instruction input needs of different gestures in multiple scenarios, effectively solving the problem of insufficient coverage of user needs caused by the way of closing the smart watch alarm through touch screen, leading to low user experience, and finally realizing natural, accurate and personalized alarm closing interaction.

[0016] Further, in step S20b, the periodic characteristics of the reference information and the comparison information change features are analyzed, and part of the wrist motion information from the tail of the reference information is extracted as tail similar information, and part of the wrist motion information from the head of the comparison information is extracted as head similar information; the difference between the head similar information and the tail similar information change features is comprehensively evaluated, if the comprehensive evaluation result is less than the preset comparison threshold, then part of the wrist motion information from the tail of the comparison information is extracted as comparison similar information according to the periodic characteristics of the comparison feature; the difference between the head similar information and the comparison similar information change features is comprehensively evaluated, and when the comprehensive evaluation result is greater than the preset comparison threshold, the alarm is closed.

[0017] The scheme can accurately distinguish the continuation and closing operation of the user's focused transaction by analyzing the periodic characteristics of the reference and comparison information, extracting similar information in stages and evaluating the differences, thereby improving the recognition accuracy. Before the alarm is triggered, the user handles the transaction with his hand (such as assembling parts), and after the alarm is triggered, the focused state will continue, and at this time, the change feature difference between the tail of the reference information and the head of the comparison information is small. The scheme first compares the similarity of the change features of the tail of the reference information and the head of the comparison information to obtain the transaction continuation of the user, and then accurately identifies the moment when the user turns to the closing gesture according to the difference between the head and the tail of the comparison information, and closes the alarm in time. At the same time, the scheme can also filter the transition action from transaction continuation to closing gesture (such as action rhythm fine-tuning), neither misjudging the transition action, nor accurately obtaining the user's intention to close the alarm, thereby improving the adaptability of the scheme to multiple scenarios, improving the recognition stability and user experience.

[0018] Further, in step S20b, the comprehensive evaluation result is associated with the acquisition time of the comparison information, the change speed of the comprehensive evaluation result with time is analyzed, and the time length of the comparison similar information is adjusted according to the change speed.

[0019] The scheme can shorten the collection time of comparison similar information under high change speed, reduce the invalid duration of alarm ringing, improve the interaction response efficiency, avoid interfering with the user's demand for quickly handling the closing instruction, and at the same time, retain the original time length for low change speed scenarios, which can capture the action features of the user's hesitation process completely, avoid misjudgment caused by incomplete collection, and does not need to increase the cost of sensor hardware, but only through time correlation analysis, the dual effects of efficiency improvement and recognition accuracy guarantee can be realized, which meets the user's demand for different action conversion rhythm.

[0020] Further, when the change speed of the comprehensive evaluation result with time is less than a preset speed threshold, the change speed of the change feature of the comparison similar information with time is obtained, and if the change speed is still less than the preset speed threshold, the difference between the change features of the comparison similar information and the reference information is taken as a suspected instruction feature, and when the suspected instruction feature meets the preset instruction feature, the alarm is closed.

[0021] Before the alarm clock is triggered, if the wrist movement information value or frequency is high, and the user cannot provide high cooperation for closing the alarm clock (such as the scene of holding a pot and stirring), based on the large hand movement value, after the alarm clock is triggered, it is difficult to obtain the alarm clock closing instruction only through the change difference analysis of the wrist movement signal, at this time, from the change characteristics of the similar information (the wrist movement information of the user entering the most possible closing alarm clock gesture), the change characteristics of the reference information (the wrist movement information of the user before the alarm clock is triggered) are eliminated, and the wrist movement information when the user enters the closing alarm clock gesture is obtained, so that the actual intention of the user is more accurately obtained. In this way, the scheme does not need to obtain and calculate other parameters, reduces the consumption of computing power, realizes the triple effect of accurate recognition, scene adaptation and cost control, and guarantees that the user can efficiently close the alarm clock without interrupting cooking.

[0022] Further, when the value of the suspected instruction feature is small, the vibration intensity of the alarm clock is gradually increased.

[0023] In the user focus scene (such as cooking with kitchen utensils, office computer), when the value of the suspected instruction feature is small, the vibration intensity of the alarm clock is gradually increased, which can amplify the movement feature signal of the weak closing action, solve the problem of difficult recognition of instructions when both hands are occupied, and avoid sudden disturbance of the initial strong vibration to the focused state (such as not interrupting the rhythm when reading), which is suitable for the timing needs that do not need strong awakening.

[0024] Further, in step S10, the sound information is collected under the condition of obtaining the permission, the sound information related to the semantic is extracted from the sound information as the voice information, the proportion of the voice information in the sound information is analyzed, and the vibration intensity when the alarm clock is triggered is adjusted according to the proportion.

[0025] When the user is in a focused state (such as the scenarios of operating kitchenware for cooking, using a computer for work, etc. mentioned in the background), the alarm vibration intensity is adjusted by collecting sound information, extracting speech information, and analyzing the proportion thereof, on the one hand, in a scenario where the proportion of speech information is high (such as when the user is in a meeting and communicating with others, or listening to an audio explanation while cooking), the vibration intensity is automatically reduced to avoid strong vibration from interfering with the user's current voice interaction or information acquisition, which meets the requirement that a timing alarm does not need to be strongly awakened. On the other hand, in a scenario where the proportion of speech information is low (such as when the user is focused on reading without speech input, or processing work in a quiet environment), the vibration intensity is appropriately increased to ensure that the user can perceive the alarm reminder, thereby solving the problem that the alarm reminder is easily ignored when touch is inconvenient in the background technology. At the same time, this way of the present solution does not need to rely on user hand operation, but can dynamically adapt to the reminder requirements of different environments through sound information analysis, which not only avoids the limitations of the touch method in scenarios where both hands are occupied, but also realizes the dual effects of interference control and reminder effectiveness guarantee through a single sound collection and analysis function. Especially in scenarios where the user cannot touch the screen and the environmental sound state is variable, the vibration intensity can be automatically matched to improve the user experience in multiple scenarios, which is difficult to achieve by traditional fixed vibration intensity or touch-dependent alarm control methods.

[0026] Further, in step S10, when no physiological electrical signal is collected in the current rocking time, sound information is collected in the pre-rocking time as pre-rocking sound, the frequency variation feature and the amplitude variation feature of the pre-rocking sound are extracted, and the numerical variation feature and the periodic variation feature of the reference information are extracted; the correlation between the pre-rocking sound and the reference information is analyzed by a feature matching algorithm, the closeness of the correlation is calculated, and the influence value of the sound information and the reference information on closing the alarm is set according to the closeness; after the alarm is triggered, the change feature of the sound information before and after the alarm is triggered is taken as the sound change feature; in step S20, after comprehensive evaluation, the comprehensive evaluation result is adjusted according to the relationship between the influence value, the sound change feature, and the comprehensive evaluation result.

[0027] In a scenario where the physiological electrical signal cannot be collected (i.e., when the smart watch is not worn), the pre-rocking sound is collected and the features are extracted, the correlation is analyzed in combination with the reference information features, and the influence value is set, and the comprehensive evaluation result is adjusted according to the sound change feature after the alarm is triggered, which can solve the problem of inconvenient touch, and improve the adaptability of instruction recognition through multiple signals. For example, when the user is cooking, the physiological electrical signal cannot be collected due to the hand-held kitchenware, the steady-state sound of the range hood and the reference information of the wrist pot action form a correlation in the pre-rocking sound, and the smart watch sets a reasonable influence value accordingly. If the sound change (such as the range hood being paused) and the action feature difference meet the evaluation standard after the alarm is triggered, the alarm can be closed without touch.

[0028] Further, in step S10, after collecting the pre-shaking sound, the pre-shaking sound is divided into different types according to the collection time, including a steady type and a transient type, and the timbre features corresponding to different types of sound are extracted; in step S20b, if the comprehensive evaluation result does not meet the preset instruction characteristics, the preset vibration mode of the alarm clock is gradually adjusted, and the comprehensive evaluation result is recalculated after adjusting the vibration mode; when the comprehensive evaluation result meets the preset instruction characteristics, the sound type and timbre feature collected this time are associated with the adjusted vibration mode and stored to form user usage habit data; a deep learning model is constructed based on the user usage habit data, the sound type and timbre feature are taken as input, and the adaptive vibration mode is output, so as to set the vibration mode when the alarm clock is triggered.

[0029] The scheme can not only solve the problem that the user is difficult to touch after the alarm clock is triggered in different scenarios, but also enable the smart watch to adapt to the user's habit automatically; in particular, it can be adapted across industry scenarios, such as a chef working in the catering industry, the steady operation sound of the range hood and the transient collision sound of the spatula. The deep learning model can output an adaptive vibration mode based on the sound type and timbre feature, avoid the problem that the traditional fixed vibration is easily covered by environmental sound or interfered with cooking action, and does not require the user to manually adjust repeatedly, thereby ensuring the accuracy of alarm clock interaction, improving the universality of the smart watch in different industry scenarios, and gradually adapting the vibration mode to the user's usage preference through habit data accumulation, realizing double adaptation of personalization and scenario.

[0030] Further, in step S10, when extracting the timbre feature of the steady type sound, the harmonic distribution is determined synchronously combined with the frequency change feature, and when extracting the transient type sound, the impact energy peak value is determined combined with the amplitude change feature; the association between the sound feature and the reference information is strengthened or weakened through the harmonic distribution and the impact energy peak value.

[0031] The harmonic distribution of the steady sound determined combined with the frequency change feature in the scheme can quantify the matching rule between the harmonic distribution and the frequency change feature of the reference information (wrist movement), and avoid the misassociation caused by a single timbre feature; the impact energy peak value of the transient sound determined combined with the amplitude change feature can accurately anchor the time sequence synchronization with the numerical change feature of the reference information, reduce environmental noise interference, and make the association closeness calculation more reliable. At the same time, by strengthening the effective association (such as the matching of stable harmonic and regular movement) and weakening the invalid association (such as the misplacement of chaotic harmonic and random movement), the sound timbre feature is adaptively adjusted according to the actual association strength, the weight distribution deviation is avoided, and more reasonable basic parameters are provided for the comprehensive evaluation in step S20b. BRIEF DESCRIPTION OF DRAWINGS

[0032] Figure 1 The flowchart of the alarm clock function interaction control method for the smart watch in the embodiment 1 of the scheme. DETAILED DESCRIPTION

[0033] The concept and technical effects of the present application will be described clearly and completely in combination with the embodiments, so as to fully understand the purposes, features and effects of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, but not all the embodiments. Based on the embodiments of the present application, other embodiments obtained by those skilled in the art without creative labor are within the protection scope of the present application:

[0034] Embodiment 1

[0035] As shown in the figure, the alarm function interaction control method for the smart watch comprises the following steps: Figure 1

[0036] S10: setting a pre-shaking time, when it is obtained from the collected wrist physiological signal that the user is in a wakeful state, acquiring wrist movement information within the pre-shaking time before the alarm trigger as reference information, collecting wrist movement information after the alarm trigger as comparison information in real time, and analyzing the change characteristics of the reference information and the comparison information in combination with the collection time; the change characteristics include numerical change characteristics and frequency change characteristics.

[0037] S20: judging whether the change characteristics of the reference information conform to the preset instruction characteristics (determined by the administrator according to the calculation accuracy, including the preset numerical instruction characteristics and the frequency instruction characteristics, wherein the numerical instruction characteristics include that the signal amplitude standard deviation in a unit time exceeds a threshold value, the absolute value of the slope reaches a set value, etc.; the frequency instruction characteristics include that the main frequency offset exceeds a standard, the high-frequency energy proportion reaches a specified proportion, etc.; the preset instruction characteristics are used to identify whether the user has hand use from the wrist physiological signal collected before the alarm trigger through the numerical change characteristics and the frequency change characteristics), if:

[0038] S20a: if not, the alarm is turned off when the change characteristics of the comparison information conform to the preset instruction characteristics. When the change characteristics of the comparison information do not conform to the preset instruction characteristics, the user has no obvious change in hand use before and after the alarm trigger. At this time, the user may have no hand movement before and after the alarm trigger, or the hand movement before and after the alarm trigger may have no obvious change. The numerical instruction characteristics also include time characteristics. If the user has no hand movement before and after the alarm trigger, the change characteristics of the comparison information conform to the preset instruction characteristics; if the user has hand movement before and after the alarm trigger (the user uses the watch to collect whether the wrist moves or not, if the alarm is triggered at this time, the user can feel the wrist vibration or hear the alarm ring), the user knows that the alarm is triggered, and can judge whether the user has actively entered the instruction to turn off the alarm through the duration (time characteristics) of the hand movement conforming to the frequency instruction characteristics.

[0039] ​S20b: If the hand usage condition is met (i.e., the hand usage condition of the user is included), the change feature of the reference information (the change feature of the hand usage of the user before the alarm trigger) is extracted as the reference feature, the change feature of the contrast information is extracted as the contrast feature (the change feature of the hand usage of the user after the alarm trigger), the difference between the reference feature and the contrast feature is comprehensively evaluated according to the numerical change feature and the frequency change feature of the contrast feature and the reference feature (the change of the hand usage condition of the user before and after the alarm trigger), and when the comprehensive evaluation result meets the preset instruction feature, the alarm is turned off.

[0040] The smart watch hardware device is integrated with a gyroscope and an acceleration sensor. The smart watch mainly emits light through the skin through an optical sensor, receives reflected light and converts it into an electrical signal. The signal difference between sleep and wake is that the heart rate is slow and the fluctuation is stable during sleep, and the signal waveform is flat. When awake, the sympathetic nerve is active, the heart rate rises slightly, the fluctuation increases, and the waveform is steeper. At the same time, by analyzing the heart rate variability (HRV), the proportion of high-frequency components is higher when awake. The signal stability is detected, and specific noise is brought by wrist movement when awake. The auxiliary judgment is made by the breath-holding time. Some smart watches also add ECG sensors in the hardware device to enhance the accuracy through the characteristics of the electrocardiogram waveform, and finally accurately confirm whether the user is awake, providing a prerequisite for collecting motion information during the pre-shaking time.

[0041] The numerical change feature focuses on the dynamic fluctuation of the time domain signal, that is, the original signal is first segmented by the sliding window method, low-pass filtered to remove noise, and normalized, and then the fluctuation intensity (standard deviation, root mean square), peak value feature (peak value number and difference), and trend feature (signal slope) are extracted. For example, when the user stirs the pot with a shovel, the numerical standard deviation is stable, and when the user adjusts the grip, the standard deviation rises sharply, so as to quantify the change of wrist action force and amplitude.

[0042] The frequency change feature focuses on the periodicity of the frequency domain. Specifically, it includes: first performing Fourier transform (FFT) on the preprocessed time domain signal, converting it into a frequency domain representation between frequency and energy (for example, the sampling rate is 50-200Hz to avoid frequency spectrum aliasing), and then extracting the main frequency (including the peak frequency with the highest energy ratio, such as 1Hz when writing and 3Hz when shaking the wrist), the frequency band energy ratio (such as the energy ratio of low and high frequency intervals), and the frequency stability (such as spectral entropy). For example, the high-frequency energy ratio increases from 20% to more than 50% when deliberately operating, so as to capture the difference in wrist action rhythm and period. The combination of the two types of features (the combination between the numerical features and the frequency features) can accurately distinguish the motion difference between the transaction continuation and the deliberate operation, and provide a quantitative basis for the alarm closing instruction recognition.

[0043] In step S20b, the periodic characteristics of the reference information and the comparative information change characteristics are analyzed, partial wrist motion information is extracted from the tail of the reference information as tail similar information according to the periodic characteristics, and partial wrist motion information is extracted from the head of the comparative information as head similar information; the difference between the change characteristics of the head similar information and the tail similar information is comprehensively evaluated, if the comprehensive evaluation result is less than a preset comparison threshold (set by an administrator), partial wrist motion information is extracted from the tail of the comparative information as comparative similar information according to the periodic characteristics of the comparison characteristics; the difference between the change characteristics of the head similar information and the comparative similar information is comprehensively evaluated, and when the comprehensive evaluation result is greater than the preset comparison threshold, the alarm clock is turned off.

[0044] When the comprehensive evaluation is performed, the feature difference is quantified, specifically including quantification of the numerical value change feature difference, the frequency change feature difference, and the high-frequency energy proportion difference.

[0045] When the numerical value change feature difference is quantified, the normalized difference rate of the reference feature and the comparison feature is calculated to eliminate the influence of individual action amplitude difference. The numerical value feature normalized difference rate , wherein, is the numerical value feature mean (such as amplitude standard deviation, peak difference, slope, etc.) of the reference information; is the numerical value feature mean of the comparative information. In the above process, by calculating the ratio of the difference absolute value to the numerical value feature mean of the reference information, the differences of different users and different action amplitudes are standardized to ensure horizontal comparability and intuitively reflect the significant changes of the numerical value feature.

[0046] When the frequency change feature difference is quantified, the main frequency offset and the high-frequency energy proportion difference are further calculated to quantify the changes of the action period and the energy distribution. The main frequency offset , wherein is the main frequency (the frequency with the highest energy proportion, unit: Hz) of the reference information; is the main frequency of the comparative information. The frequency offset directly reflects the change of the action rhythm, so that the smart watch can clearly capture the core difference of the frequency feature.

[0047] The high-frequency energy proportion difference , wherein is the energy proportion of the high-frequency band (such as 5-10 Hz) of the reference information; is the energy proportion of the high-frequency band of the comparative information. The high-frequency energy proportion difference can reflect the change of the action energy distribution, and the high-frequency energy proportion of the deliberate closing action is usually higher. The value can be used to assist in distinguishing between “transaction continuation” (i.e., the user continues the operation before the alarm clock is triggered after the alarm clock is triggered) and “closing instruction” (i.e., the user makes a gesture to turn off the alarm clock).

[0048] After quantifying the differences in features, the quantified differences are combined with scenario adaptability to adjust feature weights and generate a comprehensive evaluation value. , The calculation formula is shown in formula (1) below.

[0049] (1),

[0050] in The numerical feature weights are set by the administrator (for example, a base value of 0.4, adjusted to 0.5 for static, focused scenes). The frequency feature weight is set by the administrator (for example, the base value is 0.6, and it is adjusted to 0.7 in the vibration interference scenario). The unnormalized comprehensive evaluation value is (normalized value ranges from 0 to 1). The normalized comprehensive evaluation value S is... ,in, This is the minimum reference value preset for administrators (based on a large number of user sample statistics, for example, fixed at 0.2, corresponding to the typical original value in the "transaction continuation" scenario). The maximum reference value preset for the administrator (e.g., fixed at 1.8, corresponding to the typical original value in the "deliberately turned off" scenario). Normalization of the comprehensive evaluation value can eliminate the influence of differences in units and numerical ranges of the original features, ensuring that the comprehensive evaluation values ​​of different scenarios and different users are comparable.

[0051] Formula (1) highlights the contribution of key features through dynamic weighting. For example, during vibration interference, frequency features exhibit stronger anti-interference capabilities, thus improving... To a minimum of 0.7, thus avoiding the impact of environmental noise on the assessment results.

[0052] Finally, personalized thresholds are calibrated based on fluctuations in users' daily behavior to reduce misjudgments. Personalized thresholds ,in K is the baseline threshold (e.g., a fixed value of 0.6); K is the calibration coefficient (e.g., a fixed value of 0.1). The standard deviation of the reference information features (reflecting the degree of fluctuation in a user's daily actions) is used to personalize the calculation and adapt to individual differences in actions, such as when a user's daily actions fluctuate greatly. Set to 0.3 to avoid misjudgment due to unstable movement amplitude; when When the alarm is turned off, it is determined to be a valid shutdown command.

[0053] In the above, the preset comparison threshold is a fixed judgment standard set in advance based on a large number of user samples (covering different scenarios and action types). The personalized threshold is based on the preset comparison threshold (e.g., Based on 0.6, and combined with the individual user's action fluctuation characteristics (such as the standard deviation of daily actions), ) dynamically calculated exclusive decision threshold, the personalized threshold of each user can be different.

[0054] In step S20b, the comprehensive evaluation result is associated with the acquisition time of the comparison information, the change speed of the comprehensive evaluation result over time is analyzed, and the time length of the comparison similar information is adjusted according to the change speed.

[0055] Specifically, the acquisition time stamp of the comprehensive evaluation result and the corresponding comparison information is recorded synchronously to form data associated with the evaluation result and the time; the difference value of the evaluation result of the continuous time node and the time interval are calculated to obtain the change speed (unit: normalized evaluation value / second). If the change speed is greater than or equal to a preset threshold (such as 0.2, which is set by an administrator), it indicates that the user's intention to switch from transaction continuation to closing operation is clear, the time length of the comparison similar information is shortened (such as from 5 seconds to 2-3 seconds), and the invalid alarm of the alarm clock is reduced; if the change speed is less than the threshold, the original time length (such as 5 seconds) is retained to ensure the completeness of the feature acquisition and avoid misjudgment, and the response efficiency and the recognition accuracy are balanced.

[0056] In a specific implementation, user A uses a smart watch to perform timing reminding during writing. User A sets the pre-rotation time to 30 seconds, and the smart watch collects physiological signals and motion signals of the wrist of user A during writing. The smart watch collects the physiological signals of the wrist through an optical heart rate sensor, and assumes that the smart watch detects that the heart rate of the user is 72 times per minute (within the normal range in a wakeful state) and the high-frequency component of heart rate variability accounts for 35% (above the threshold of a sleep state), and determines that the user is awake. Then, 30 seconds before the alarm is triggered, the watch collects the wrist motion information of the user writing with a pen as reference information through a six-axis sensor (accelerometer + gyroscope), filters noise, and extracts the value change features (including amplitude standard deviation and peak difference) and the frequency change features (including main frequency and high-frequency energy ratio). After the alarm is triggered, the smart watch collects the wrist motion information in real time as comparison information, for example, the user continues to write for the initial 5 seconds, and adjusts the action for the subsequent 5 seconds due to the alarm ringing.

[0057] Then, the smart watch determines whether the reference information feature meets the preset instruction feature (wherein the numerical feature includes the amplitude standard deviation and the peak difference; the frequency feature includes the main frequency and the high frequency energy proportion), and assumes that the reference information feature meets the preset instruction feature, and executes S20b branch. The smart watch extracts the reference feature and the comparison feature (for example, the motion feature within 15 seconds after the alarm clock is triggered, the numerical feature and the frequency feature within the last 5 seconds), comprehensively evaluates the difference between the two, calculates the comprehensive evaluation value according to the numerical change feature and the frequency change feature, and normalizes, and assumes that the result is 0.925. Again, combined with the user's historical action data (assuming that the standard deviation of daily writing action is 0.2), the personalized threshold is assumed to be 0.612. Since the comprehensive evaluation value 0.925 is greater than the personalized threshold 0.612, it is determined that it meets the preset instruction feature, and the alarm clock is triggered to be turned off.

[0058] Embodiment 2

[0059] The difference between this embodiment and embodiment 1 is only that when the change speed of the comprehensive evaluation result with time is less than a preset speed threshold (which is set by the administrator according to the individual situation of the user, and the reaction speed of young people is generally higher than that of old people, so the preset speed threshold of young people is generally higher than that of old people), the change speed of the comparison similar information change feature with time is obtained, and if the change speed is still less than the preset speed threshold, the change feature difference between the comparison similar information and the reference information is taken as the suspected instruction feature. When the suspected instruction feature meets the preset instruction feature, the alarm clock is turned off.

[0060] When the numerical value of the suspected instruction feature is small, the vibration intensity of the alarm clock is gradually increased.

[0061] Specifically, when the numerical value of the suspected instruction feature is less than a preset feature numerical value threshold (which is set based on the fluctuation range of the reference information feature and the lower limit of the preset instruction feature), the vibration intensity of the alarm clock is gradually increased according to a preset gradient, and after each vibration intensity adjustment, the change feature of the comparison similar information is reacquired and the suspected instruction feature is updated; if the updated suspected instruction feature meets the preset instruction feature, the vibration intensity increase is stopped and the alarm clock is turned off; if the suspected instruction feature still does not meet the preset instruction feature after adjusting the vibration intensity for a plurality of times (generally 3 times, which is set by the user), the current vibration intensity is maintained and the collection time length of the comparison similar information is extended.

[0062] In specific implementation, user 2 is cooking in the kitchen, and sets a prompt alarm clock for stewing soup. At this time, the user needs to continuously shake the pot and cannot make obvious gestures, and only tries to turn off the alarm clock by clicking the pot handle.

[0063] Before the alarm clock is triggered, the smart watch collects wrist movement information of the user when the user is stirring the pot as reference information, and the value change characteristics and the frequency change characteristics of the reference information both present strong regularity, which conforms to the current high-frequency use of the user's hand. After the alarm clock is triggered, the change speed of the comprehensive evaluation result with time is less than the preset speed threshold, and the change speed of the feature of the similar comparison information (i.e., the movement information containing the clicking pot handle action) is further obtained, which is still less than the threshold, and then the feature difference between the similar comparison information and the reference information is taken as the suspected instruction feature.

[0064] Because the value of the suspected instruction feature is small, the smart watch gradually increases the vibration intensity, and finally the suspected instruction feature matches the preset instruction feature, the alarm clock is successfully turned off, and the user does not need to interrupt the stirring operation during the whole process, ensuring that the cooking process is not affected.

[0065] Embodiment 3

[0066] The difference between this embodiment and embodiments 1-2 is only that in step S10, the sound information is collected in the case of obtaining the permission, the sound information related to the semantics is extracted from the sound information as the voice information, the proportion of the voice information in the sound information is analyzed, and the vibration intensity when the alarm clock is triggered is adjusted according to the proportion.

[0067] Specifically, after the sound information is collected with the permission, the non-semantic interference noise (such as wind blowing sound and device running sound) in the environment is removed through a noise filtering module, and then the sound information related to the semantics is extracted as the voice information; when analyzing the proportion of the voice information in the filtered sound information, if the proportion is not less than a preset proportion threshold (specifically set by the user), it is determined that the user is in a voice interaction scene (such as a meeting communication or listening to an audio tutorial), and the vibration intensity when the alarm clock is triggered is reduced to a range that does not interfere with voice reception; if the proportion is less than the preset proportion threshold, it is determined that the user is in a low voice scene (such as a noisy environment), and the vibration intensity when the alarm clock is triggered is increased to ensure that the user can perceive it; and after adjusting the vibration intensity each time, the voice information proportion and the vibration intensity matching data of the current scene are recorded synchronously, and the matching data can be directly called to quickly adjust in the same scene in the future.

[0068] In specific implementation, the user C uses the timing alarm clock of the smart watch to remind the meeting interval and the file submission node in the meeting. Before a certain online project meeting, the user C sets a meeting preparation reminder alarm clock triggered after 1 hour, at which time the smart watch has obtained the sound information collection permission.

[0069] During the meeting, the smart watch removes the air conditioner running sound, keyboard tapping sound and other non-semantic interference noises in the environment in real time through the noise filtering module, and extracts the communication voice of user C and colleagues and the meeting explanation audio as voice information. Assuming that it is found through analysis that the proportion of voice information in filtered sound information is not less than a preset proportion threshold, it is intelligently determined that user C is in a voice interaction scene, and the vibration intensity of the alarm trigger is automatically adjusted to a low intensity mode (for example, only a slight wrist perception is produced, and no obvious vibration sound is produced), thereby avoiding interference with the meeting communication when the alarm is triggered.

[0070] After the meeting, user C enters a document writing state, and there is no obvious voice interaction in the environment. The smart watch analyzes that the proportion of voice information is less than the preset proportion threshold, determines that it is a low voice scene, and adjusts the alarm vibration intensity to a medium intensity mode to ensure that the reminder can be perceived. After each adjustment, the watch synchronously records the voice proportion and vibration intensity matching data of the current scene, and when user C is in a similar online meeting or document writing scene again in the future, the historical matching data can be directly called to quickly adjust the vibration intensity without reanalysis, thereby balancing the reminder effectiveness and scene interference control.

[0071] Embodiment 4

[0072] The difference between this embodiment and embodiments 1-3 is that in step S10, when no physiological electrical signal is collected in the current warm-up time, sound information is collected in the pre-warm-up time as pre-warm-up sound, frequency variation features and amplitude variation features of the pre-warm-up sound are extracted, and numerical variation features and periodic variation features of reference information are extracted; the correlation between the pre-warm-up sound and the reference information variation features is analyzed through a feature matching algorithm, the closeness of the correlation is calculated, and the influence values of the sound information and the reference information on closing the alarm are set according to the closeness.

[0073] After the alarm is triggered, the variation features of the sound information before and after the alarm are taken as sound variation features; in step S20, after comprehensive evaluation, the comprehensive evaluation result is adjusted according to the relationship between the influence value, the sound variation features and the comprehensive evaluation result.

[0074] Specifically, when no physiological electrical signal is collected in the pre-warm-up time, the smart watch collects sound information and wrist movement information at the same time. More specifically, the ambient sound is first collected as pre-warm-up sound, and the frequency (fluctuation period, peak timing) and amplitude (rise rate, stable duration) variation features are extracted after filtering out interference; the synchronously collected wrist movement is taken as reference information, and the numerical (amplitude fluctuation, mutation frequency) and periodic (repetition duration, peak distribution) variation features of the reference information and the pre-warm-up sound are extracted.

[0075] The dynamic time warping algorithm is used to align the feature curves of the reference information and the pre-rocking sound. The correlation is determined by frequency overlap and amplitude trend consistency. Then, the tightness is calculated by weighting the three dimensions of frequency coordination (curve overlap rate), amplitude coordination (correlation coefficient), and timing synchronization (peak time difference), and the influence value is set according to the tightness.

[0076] The pre-rock sound feature sequence is set as follows (m is the number of sampling points, (It is the eigenvalue of point i), and the reference information (wrist movement) sequence is: (n is the number of motion sampling points). The difference between individual feature points is calculated using Euclidean distance: Then construct an m×n distance matrix D. , represents the Euclidean distance between the feature of the i-th sampling point of the pre-rocking sound and the feature of the j-th sampling point of the reference information. The smaller the value, the higher the similarity.

[0077] Then, a dynamic programming algorithm is used to find the distance matrix. arrive Optimal path (k is the path length, m is the pre-rocking sound feature sequence) , where n is the feature sequence of reference information (e.g., wrist movement). ,satisfy ). Here It is the t-th point on the path, corresponding to a position in the distance matrix. ,So It is the element value of that position in the distance matrix D, that is, the Euclidean distance value of the t-th point on the path.

[0078] Use dynamic programming to find the path with the minimum cumulative distance. Where k is the path length. The DTW value represents the path point distance; the smaller the DTW value, the higher the alignment between the reference information and the pre-rock sound characteristic curve. Frequency overlap is used... calculate, It represents the frequency overlap, where k is the frequency characteristic sequence of the pre-rocking sound after alignment. and reference information frequency associated feature sequence The number of sampling points (because the two sequences have the same length after dynamic time warping and alignment, k represents the number of common sampling points). t is the index variable for summation, from 1 to k, which sequentially judges and sums the condition of each sampling point. It is an indicator function, when At the time of its establishment, The value is 1; when the condition is not met, the value is 0, which is used to count the number of sampling points that meet the frequency difference within the tolerance range. is the frequency feature value of the pre-roll sound at the tth sampling point, that is, the frequency value at the tth position in the pre-roll sound frequency sequence after alignment. is the frequency correlation feature value of the reference information at the tth sampling point, that is, the frequency correlation value at the tth position in the reference information frequency correlation sequence after alignment. is the preset frequency tolerance threshold, used to define the range within which the frequency difference between the pre-roll sound and the reference information can be considered "coincidence", such as being set to a specific value of 50 Hz according to actual scene requirements.

[0079] Amplitude trend Pearson coefficient is calculated. represents the Pearson correlation coefficient, used to quantify the degree of linear correlation between the pre-roll sound amplitude feature sequence and the reference information amplitude correlation feature sequence. k is the number of sampling points of the pre-roll sound amplitude feature sequence and the reference information amplitude correlation feature sequence (the lengths of the two sequences are consistent after processing, and k represents the common number of sampling points). t is the index variable of summation, from 1 to k, and the correlation data of each sampling point is calculated and summed in turn. is the amplitude feature value of the pre-roll sound at the tth sampling point, that is, the amplitude value at the tth position in the pre-roll sound amplitude sequence. is the amplitude correlation feature value of the reference information at the tth sampling point, that is, the amplitude correlation value at the tth position in the reference information amplitude correlation sequence. is the mean value of the pre-roll sound amplitude feature sequence, calculated by the formula , representing the average level of the pre-roll sound amplitude. is the mean value of the reference information amplitude correlation feature sequence, calculated by the formula , representing the average level of the reference information amplitude correlation.

[0080] Close degree total score , , , are the weights of the frequency coordination degree, the amplitude coordination degree, and the timing synchronization degree, respectively, and the sum of the weights is 1 (which can be adjusted according to actual scene requirements to highlight the importance of different dimensions). represents the amplitude coordination degree, which is obtained by taking the non-negative value of the Pearson correlation coefficient , that is, , which can exclude the interference of negative correlation on the amplitude coordination degree and only focus on the consistency of positive correlation amplitude trend. represents the timing synchronization degree, used to quantify the synchronization tightness of the pre-roll sound and the reference information in the time sequence, and its calculation method is usually , where is the average time difference of the characteristic peaks, is the maximum allowed time difference.

[0081] is the different value sets different association levels. For example, S≥0.8 is the association strong level, the sound influence value , the reference influence value ; 0.5≤S<0.8 is the association medium level, , ; S<0.5 is the association weak level, , , and .

[0082] After the alarm clock is triggered, the sound change characteristics such as the frequency and amplitude change before and after the sound, and the characteristic persistence are extracted. The difference between the reference and the real-time motion characteristics is calculated according to step S20b to obtain an initial comprehensive evaluation result. Then, the correction term is obtained by multiplying the matching degree of the sound change characteristics and the preset instruction characteristics by the sound influence value, the motion term is obtained by multiplying the initial result by the motion influence value, and the adjusted result is obtained by adding the correction term and the motion term. If the result meets the standard, the alarm clock is turned off. If the result does not meet the standard, the influence value can be fine-tuned according to the sound stability to recalculate, thereby ensuring the recognition accuracy.

[0083] In step S10, after the pre-shaking sound is collected, the pre-shaking sound is divided into different types according to the collection time, including a steady-state type and a transient type, and the timbre characteristics corresponding to different types of sound are extracted. In step S20b, if the comprehensive evaluation result does not meet the preset instruction characteristics, the preset vibration mode of the alarm clock is adjusted step by step (set by the user, for example, the vibration mode of the alarm clock is adjusted to a gradually increasing intensity mode), and the comprehensive evaluation result is recalculated after the vibration mode is adjusted. When the comprehensive evaluation result meets the preset instruction characteristics, the sound type and timbre characteristics of this collection are associated and stored with the adjusted vibration mode to form user usage habit data. A deep learning model is constructed based on the user usage habit data, the sound type and timbre characteristics are input, and the adapted vibration mode is output to set the vibration mode when the alarm clock is triggered.

[0084] In step S10, when the timbre characteristics of the steady-state type sound are extracted, the harmonic distribution is determined synchronously combined with the frequency change characteristics, and the impact energy peak value is determined combined with the amplitude change characteristics when the transient type sound is extracted. The association relationship between the sound characteristics and the reference information is strengthened or weakened through the harmonic distribution and the impact energy peak value.

[0085] The harmonic frequency sequence of the steady-state sound is ; wherein is the harmonic frequency of the tth sampling point, which is extracted combined with the frequency change characteristics of the steady-state sound. The period change characteristics of the reference information correspond to the frequency sequence , wherein The periodic correlation frequency of the t-th sampling point is obtained by converting the periodic variation characteristics of the reference information. The correlation formula between the two is: ,in This represents the correlation between the harmonic distribution and the periodic characteristics of the reference information, with a value range of [0,1]. The closer the value is to 1, the stronger the correlation. k is the number of sampling points of the two sequences after alignment (consistent with the "length of the aligned feature sequence"). t is the summation index variable, which iterates through each sampling point. This is the frequency sensitivity coefficient, used to adjust the sensitivity to frequency differences (e.g., it can be increased when there is a lot of ambient noise). (to enhance the identification of subtle differences) As an exponential function, it rapidly increases the correlation between similar frequencies through nonlinear mapping, while significantly reducing the correlation between frequencies with large differences, which aligns with the real-world principle that "closer frequencies have stronger correlations".

[0086] Let the peak energy sequence of transient sound be... ,in The peak impact energy at the t-th sampling point is extracted by combining the amplitude variation characteristics of the transient sound. The numerical variation feature sequence of the reference information is as follows. ,in Let be the numerical feature value of the t-th sampling point, directly obtained from the numerical change characteristics of the reference information. The correlation formula between the two is: ,in This represents the correlation between the peak impact energy and the numerical characteristics of the reference information, with a value range of [0,1]. The closer the value is to 1, the stronger the correlation. k is the number of sampling points in the two sequences after alignment. t is the summation index variable. The reference information is the time corresponding to the peak value of the numerical characteristics (the benchmark for judging the temporal synchronization with the peak value of the transient sound). This is the timing decay coefficient, used to control the impact of timing deviations on correlation (e.g., in fast-motion scenarios). Take the largest value to strictly limit timing deviations). This is a minimum value (e.g., 1e-6) used to avoid calculation errors where the denominator is 0; the first term Used to quantify timing synchronization; the greater the timing deviation, the smaller this value; the second term... Used to quantify amplitude coherence; the closer the amplitudes are, the closer this value is to 1.

[0087] The overall correlation between harmonic distribution, peak impact energy, and reference information is obtained by weighted fusion of the two dimensions mentioned above: ,in To determine the overall correlation, the value range is [0,1], reflecting the overall correlation strength between sound characteristics (harmonic distribution, peak impact energy) and reference information; is the weight coefficient, dynamically adapting to the sound type—when the proportion of steady-state sound is high Take a large value (such as 0.7), focus on the correlation contribution of harmonic distribution; when the proportion of transient sound is high Take a small value (such as 0.3), focus on the correlation contribution of impact energy peak value; , Harmonic distribution correlation degree and impact energy peak value correlation degree respectively, ensure the parameter uniformity with the previous formula.

[0088] In specific implementation, user D is an office worker, and uses the smart watch alarm to wake up in daily life. The default pre-shaking time of the smart watch is 5 minutes before the alarm trigger. In the morning of a working day, the wrist of user D is pressed when he is sleeping on his side, and the watch does not collect physiological electrical signals within the pre-shaking time. At this time, the watch automatically starts the sound and motion collection mode: synchronously collects the steady-state sound (air conditioner running sound) and transient sound (bed sheet rubbing sound when user D turns over) in the room as pre-shaking sound, and collects the wrist motion (turning over and lifting hand action) as reference information. After the collection is completed, the watch filters out the interference of the pre-shaking sound, extracts the frequency variation characteristics (assuming that the fluctuation period is 20 Hz and the peak timing is stable) and amplitude variation characteristics (assuming that the stable duration is 4 minutes and the rise and fall rate is gentle) of the air conditioner sound, extracts the frequency variation characteristics (assuming that the peak timing is concentrated at the turning over moment) and amplitude variation characteristics (assuming that the rise and fall rate is fast and the stable duration is 0.5 seconds) of the bed sheet rubbing sound; synchronously extracts the numerical variation characteristics (assuming that the amplitude fluctuation is ±30° and the mutation frequency is 2 times when turning over) and periodic variation characteristics (assuming that the repeated duration of the lifting hand action is 1.5 seconds and the peak distribution is at the moment when the rubbing sound appears) of the wrist motion.

[0089] Then use the dynamic time warping algorithm to align the characteristic curves: set the sound characteristic sequence X (assuming m=200 sampling points) and the motion characteristic sequence Y (assuming n=180 sampling points), construct a 200×180 distance matrix through the Euclidean distance formula, and then find the optimal path through dynamic programming, assuming that the calculated DTW value is 0.12 (high alignment degree). Then calculate the frequency coincidence degree, assuming that is 0.92 (assuming that the difference between the associated frequencies of the air conditioner sound and the turning over period is ≤50 Hz), the amplitude Pearson coefficient is 0.88 (the amplitude trend of the rubbing sound is consistent with the lifting hand), and the timing synchronization degree is 0.9, and the weight , , is 0.3, 0.3, and 0.4 respectively to calculate the closeness S, assuming that S is 0.9, and is 0.5, is 0.5.

[0090] After the alarm clock is triggered, the user D claps hands (the sound change feature is a sudden increase in frequency and a doubling of amplitude), the watch calculates an initial evaluation result of 0.7, and then the clapping hands and the preset instruction matching degree 0.9 are multiplied Suppose the correction term is 0.45, and the initial result is multiplied Suppose the motion term is 0.35, and the adjusted result is assumed to be 0.8 (up to standard), and the alarm clock is turned off. Subsequently, the user D does not meet the standard at a certain time, and the watch gradually adjusts the vibration to be gradually strong, and stores the transient clapping sound, the corresponding tone, and the gradually strong vibration as habit data when the standard is met; after accumulating 10 data, a deep learning model is constructed, and when the user D claps hands subsequently, the deep learning model directly outputs the gradually strong vibration, without the need for repeated adjustment.

[0091] The above is only an embodiment of the present application, and the application is not limited to this embodiment. The field involved in the embodiment, common knowledge of specific structures and characteristics in the scheme, and the like are not described in detail here. The ordinary skilled person in the art knows all the ordinary technical knowledge in the field to which the present application belongs before the filing date or the priority date, can know all the prior art in the field, and has the ability to apply conventional experimental means before that date. The ordinary skilled person in the art can perfect and implement the present scheme in combination with their own ability under the inspiration given in the present application, and some typical known structures or known methods should not be an obstacle for the ordinary skilled person in the art to implement the present application. It should be noted that, for those skilled in the art, without departing from the structure of the present application, a number of modifications and improvements can be made, which should also be considered as the protection scope of the present application, and these will not affect the implementation effect and practicality of the patent. The protection scope claimed in the present application should be subject to the content of its claims, and the specific implementation modes and the like in the specification can be used to explain the content of the claims.

Claims

1. A method for interactive control of the alarm clock function in a smartwatch, characterized in that, Includes the following steps: S10: Set the pre-animation time. When the user is awake as determined from the collected wrist physiological signals, wrist movement information is acquired during the pre-animation time before the alarm is triggered as reference information. Wrist movement information after the alarm is triggered is acquired in real time as comparison information. The change characteristics of the reference information and comparison information are analyzed in combination with the acquisition time. The change characteristics include numerical change characteristics and frequency change characteristics. S20: Determine whether the change characteristics of the reference information conform to the preset instruction characteristics. If: S20a: If not met, the alarm will be turned off when the change characteristics of the comparison information match the preset instruction characteristics; S20b: If it meets the requirements, extract the change features of the reference information as the reference feature, extract the change features of the comparison information as the comparison feature, and comprehensively evaluate the differences between the reference feature and the comparison feature based on the numerical change features and frequency change features of the comparison feature and the reference feature. When the comprehensive evaluation result meets the preset instruction features, turn off the alarm clock.

2. The interactive control method for the alarm clock function of a smartwatch according to claim 1, characterized in that: In step S20b, the periodic characteristics of the change features of the reference information and the comparison information are analyzed. Based on the periodic characteristics, some wrist movement information is extracted from the tail of the reference information as tail similarity information, and some wrist movement information is extracted from the head of the comparison information as head similarity information. The difference between the change features of the head similarity information and the tail similarity information is comprehensively evaluated. If the comprehensive evaluation result is less than the preset comparison threshold, some wrist movement information is extracted from the tail of the comparison information as comparison similarity information based on the periodic characteristics of the comparison features. The alarm clock is turned off when the difference between the changes in head-similar information and the comparison-similar information is comprehensively evaluated and the comprehensive evaluation result is greater than the preset comparison threshold.

3. The interactive control method for the alarm clock function of a smartwatch according to claim 2, characterized in that: In step S20b, the comprehensive evaluation results are correlated with the acquisition time of the comparison information, the rate of change of the comprehensive evaluation results over time is analyzed, and the time length for comparing similar information is adjusted according to the rate of change.

4. The interactive control method for the alarm clock function of a smartwatch according to claim 3, characterized in that: When the rate of change of the comprehensive evaluation result over time is less than the preset rate threshold, the rate of change of the change characteristics of the comparative similar information over time is obtained. If the rate of change is still less than the preset rate threshold, the difference in the change characteristics between the comparative similar information and the reference information is taken as the suspected instruction characteristics. When the suspected instruction characteristics meet the preset instruction characteristics, the alarm clock is turned off.

5. The interactive control method for the alarm clock function of a smartwatch according to claim 4, characterized in that: When the value of the suspected instruction characteristic is small, gradually increase the vibration intensity of the alarm clock.

6. The interactive control method for the alarm clock function of a smartwatch according to claim 1 or 5, characterized in that: In step S10, with the permission obtained, sound information is collected, semantically related sound information is extracted from the sound information as speech information, the proportion of speech information in the sound information is analyzed, and the vibration intensity when the alarm clock is triggered is adjusted according to the proportion.

7. The interactive control method for the alarm clock function of a smartwatch according to claim 6, characterized in that: In step S10, if no physiological electrical signal is collected during the current rocking time, sound information is collected during the pre-rocking time as the pre-rocking sound, and the frequency change characteristics and amplitude change characteristics of the pre-rocking sound are extracted. At the same time, the numerical change characteristics and periodic change characteristics of the reference information are extracted. The correlation between the pre-earthquake sound and the change features of the reference information is analyzed by feature matching algorithm, the tightness of the correlation is calculated, and the influence values ​​of the sound information and the reference information on turning off the alarm are set according to the tightness. After the alarm is triggered, the changes in sound information before and after the alarm is triggered are used as sound change features. In step S20b, after conducting a comprehensive evaluation, the comprehensive evaluation results are adjusted based on the relationship between the impact value, sound change characteristics, and the comprehensive evaluation results.

8. The interactive control method for the alarm clock function of a smartwatch according to claim 7, characterized in that: In step S10, after the pre-rocking sound is collected, the pre-rocking sound is divided into different types based on the collection time, including steady-state type and transient type, and the timbre features corresponding to different types of sound are extracted at the same time. In step S20b, if the comprehensive evaluation result does not meet the preset instruction characteristics, the preset vibration mode of the alarm clock is gradually adjusted, and the comprehensive evaluation result is recalculated after adjusting the vibration mode. When the comprehensive evaluation result meets the preset instruction characteristics, the sound type and timbre characteristics collected this time are associated and stored with the adjusted vibration mode to form user habit data. A deep learning model is built based on the user habit data, taking the sound type and timbre characteristics as input and outputting the appropriate vibration mode, thereby setting the vibration mode when the alarm clock is triggered.

9. The interactive control method for the alarm clock function of a smartwatch according to claim 8, characterized in that: In step S10, when extracting the timbre features of steady-state sound, the harmonic distribution is determined simultaneously by combining its frequency change features; when extracting transient sound, the peak impact energy is determined by combining the amplitude change features; the correlation between sound features and reference information is strengthened or weakened by harmonic distribution and peak impact energy.

10. An interactive control system for the alarm clock function of a smartwatch, characterized in that, The alarm clock function interactive control method for smartwatches as described in any one of claims 1-9 is used.

Images