Intelligent medicine management and reminding system based on multi-dimensional environment perception

Abstract

The invention provides an intelligent medicine management and reminding system based on multi-dimensional environment perception, and relates to the technical field of intelligent wearable equipment and medical health management. According to the system, a medicine prescription is set in the smart watch, user behaviors and environment are monitored in real time, a three-point positioning technology is applied to manage medicine positions, and dynamic graded reminding is implemented according to medicine importance. The system adaptively adjusts the reminding mode, improves the medication compliance of the user, reduces the risk of missing taking, optimizes the energy management, and prolongs the cruising ability.

Description

Technical Field

[0001] The present invention relates to the field of smart wearable devices and medical health management technology, and in particular to an intelligent drug management and reminder system based on multi-dimensional environmental perception. Background Art

[0002] With the aging population and rising incidence of chronic diseases, long-term medication and polypharmacy management have become major challenges in modern healthcare management. Improving medication adherence is crucial for improving patient outcomes and reducing healthcare costs.

[0003] Currently, there are many medication reminder products and applications on the market, but they generally have the following limitations:

[0004] First, traditional pill boxes and medication reminder apps primarily rely on fixed-time reminders, ignoring the impact of the user's current state and environmental conditions. For example, when a user is driving, in a meeting, or sleeping, simple reminders can be disruptive, leading them to turn off reminders or become resistant to them. This one-size-fits-all reminder strategy fails to meet the actual needs of users in different situations.

[0005] Second, existing solutions lack the ability to effectively manage medication location. Users often experience issues with medication routines, such as forgetting medications, misplacing medications, or losing them. For those who manage multiple medications simultaneously, quickly locating specific medications becomes a major pain point in daily medication management.

[0006] Third, the existing system fails to fully consider the varying therapeutic importance of different medications. Certain life-sustaining medications (such as insulin and anticoagulants) must be taken strictly on schedule, while some auxiliary medications have a certain degree of time flexibility. The lack of a differentiated reminder strategy based on medication characteristics can lead to the risk of missing important medications and can also cause user fatigue due to excessive reminders.

[0007] Fourth, as personal health management tools, smart wearable devices have limited battery capacity, which restricts the continuous operation of complex functions. Minimizing system energy consumption while ensuring reminder accuracy and user experience remains a pressing technical challenge.

[0008] Fifth, most existing medication management systems lack comprehensive recording and analysis capabilities for medication compliance, and are unable to provide users and medical professionals with intuitive compliance data and long-term trend analysis, which limits the development and adjustment of personalized treatment plans.

[0009] To address the above problems, there is an urgent need to develop an intelligent drug management and reminder system that can perceive user behavior status and environmental conditions, realize intelligent drug positioning, implement differentiated reminder strategies based on drug importance, and take energy efficiency into consideration, so as to improve user medication compliance and usage experience. Summary of the Invention

[0010] In order to overcome the shortcomings of the existing technology, the purpose of the present invention is to provide an intelligent drug management and reminder system based on multi-dimensional environmental perception, so as to improve user medication compliance, reduce the risk of missed doses, optimize energy management, and extend battery life.

[0011] To achieve the above object, the present invention provides the following solutions:

[0012] An intelligent drug management and reminder system based on multi-dimensional environmental perception, comprising: a drug management layer, a spatial positioning layer, a multi-dimensional perception layer, an intelligent decision-making layer, and an interactive reminder layer;

[0013] The drug management layer provides a prescription setting interface on the smartwatch terminal, supports naming and classification management of drug location modules, establishes a standardized drug information database, and protects the user's patient privacy data obtained through a secure encryption protocol;

[0014] The spatial positioning layer uses three-point positioning technology to establish a spatial relationship management model between users, drugs and the environment. By tracking the user's location, home location and drug location in real time, it can identify abnormal conditions such as forgotten or lost drugs and trigger an early warning.

[0015] The multi-dimensional perception layer integrates multi-source sensors, collects and processes user behavior data in real time through the fusion activity recognition framework of the multi-source sensors, classifies user behavior patterns based on the user behavior data, and monitors environmental noise, light and motion state parameters;

[0016] The intelligent decision-making layer uses a context-adaptive decision-making engine to comprehensively analyze user behavior, environmental conditions, and drug characteristics, determine the optimal reminder timing based on a dynamic threshold mechanism and reinforcement learning algorithm, and implement a dynamic hierarchical reminder strategy based on the importance of drug treatment;

[0017] The interactive reminder layer is used to automatically select the reminder method according to the user status and environmental conditions, optimize the personalized reminder parameters through user feedback data, record medication compliance data and generate trend analysis reports.

[0018] The present invention discloses the following technical effects:

[0019] First, this invention leverages multi-dimensional environmental sensing technology to identify user behavior and environmental conditions in real time, avoiding the limitations of traditional fixed-time reminders. The system intelligently selects the optimal reminder timing, issuing reminders when the user is in the right state. It also automatically adjusts the reminder method based on varying environmental conditions, significantly improving reminder effectiveness and user experience.

[0020] Secondly, this invention innovatively applies three-point positioning technology to establish a spatial relationship management model between users, medications, and the environment, enabling precise tracking and management of medication locations. The system can promptly detect abnormal situations such as "forgotten medication" and "potentially lost medications" and issue warnings, effectively solving the location management challenges inherent in traditional medication management.

[0021] Third, the present invention establishes a dynamic, hierarchical reminder strategy based on medication importance, implementing differentiated reminder measures for medications of varying importance. This system provides the highest level of reminder protection for life-sustaining medications while avoiding excessive reminders for adjunctive or temporary medications. This ensures the optimal allocation of reminder resources, balancing medication safety and user experience.

[0022] Fourth, the present invention optimizes energy management through a hierarchical processing mechanism and adaptive sampling strategy, combined with ambient energy harvesting technology to significantly extend system endurance. By performing the majority of data processing tasks on a low-power coprocessor and activating the main processor only when necessary, the system minimizes energy consumption while maintaining functional integrity.

[0023] Fifth, the present invention provides comprehensive medication adherence recording and analysis capabilities, automatically recording medication reminders and user response data, and generating intuitive adherence trend reports. This data not only helps users understand their medication habits but also provides valuable reference for medical professionals to develop and adjust personalized treatment plans.

[0024] In summary, the system of the present invention has built a comprehensive intelligent drug management solution by integrating technologies such as multi-dimensional environmental perception, spatial positioning, situational adaptive decision-making and intelligent interactive reminders, which effectively improves users' medication compliance, reduces the risk of missed and wrong doses, and is of great value in improving the treatment effect and quality of life of patients with chronic diseases. BRIEF DESCRIPTION OF THE DRAWINGS

[0025] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0026] Figure 1 A schematic diagram of the overall system architecture of the present invention provided in an embodiment of the present invention;

[0027] Figure 2 A functional module structure diagram of the drug management layer provided in an embodiment of the present invention;

[0028] Figure 3 A schematic diagram of a three-point positioning technology for a spatial positioning layer provided by an embodiment of the present invention;

[0029] Figure 4 A flow chart showing the sensor arrangement and data processing of the multi-dimensional perception layer provided by an embodiment of the present invention;

[0030] Figure 5 This is an architecture diagram of the context-adaptive decision engine for the intelligent decision layer provided by an embodiment of the present invention;

[0031] Figure 6 A schematic diagram of a multimodal reminder strategy for an interactive reminder layer provided by an embodiment of the present invention;

[0032] Figure 7 A diagram of the hierarchical processing architecture of the system energy management mechanism provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0033] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0034] The purpose of this invention is to provide an intelligent drug management and reminder system based on multi-dimensional environmental perception. By integrating technologies such as multi-dimensional environmental perception, spatial positioning, context-adaptive decision-making and intelligent interactive reminders, a comprehensive intelligent drug management solution is constructed, which effectively improves user medication compliance and reduces the risk of missed and wrong doses. It is of great value to improving the treatment effect and quality of life of patients with chronic diseases.

[0035] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the present invention is further described in detail below with reference to the accompanying drawings and specific embodiments.

[0036] Figure 1 The overall system architecture diagram of the present invention is provided in the embodiment of the present invention, such as Figure 1 As shown, the present invention provides an intelligent drug management and reminder system based on multi-dimensional environmental perception, which is characterized by including a drug management layer, a spatial positioning layer, a multi-dimensional perception layer, an intelligent decision-making layer and an interactive reminder layer; wherein:

[0037] The drug management layer provides a prescription setting interface on the smartwatch terminal, supports naming and classification management of drug location modules, establishes a standardized drug information database containing key information such as drug name, dosage, frequency of use, drug characteristics and therapeutic importance, and ensures patient privacy data protection through a secure encryption protocol;

[0038] The spatial positioning layer uses three-point positioning technology to establish a spatial relationship management model between users, medications, and the environment. By deploying positioning devices on smart watches to determine the user's location, by deploying positioning devices on chargers to determine the home location, and by deploying micro-positioning devices near the medications to determine the medication's location, the layer can intelligently identify and provide early warnings for abnormal situations such as forgetting to bring medication or losing medications, based on the relative position of the user and medications and whether the user is at home.

[0039] The multi-dimensional perception layer uses a multi-sensor fusion activity recognition framework, integrating multiple sensors such as accelerometers, gyroscopes, ambient light sensors, and microphones to collect and process user behavior data in real time. It uses a low-power neural network model to convert the collected data into recognizable behavior pattern classifications, including resting, walking, eating, sleeping, and other states. It also monitors environmental parameters such as ambient noise levels, lighting conditions, and user movement status.

[0040] The intelligent decision-making layer uses a context-adaptive decision-making engine to comprehensively analyze the user's current behavior, environmental conditions, and drug characteristics. Based on preset rules and machine learning algorithms, it calculates the probability distribution of the optimal reminder timing. It determines the optimal time to trigger the reminder through a dynamic threshold mechanism and implements a dynamic hierarchical reminder strategy based on the importance of drug treatment, allocating differentiated reminder resources to drugs of different priority levels.

[0041] The interactive reminder layer automatically selects and adjusts reminder methods based on the user status and environmental conditions determined by the intelligent decision-making layer. It increases vibration intensity in high-noise environments, increases screen brightness in low-light environments, and uses a non-intrusive reminder mode when the user is in a meeting or driving. It also continuously optimizes personalized reminder parameters based on user feedback data, records medication compliance data, and generates trend analysis reports.

[0042] The system uses a hierarchical processing mechanism to optimize energy management, completing sensor data collection and primary processing on a low-power coprocessor, activating the main processor only when complex calculations are required, and integrating ambient energy harvesting technology to significantly extend system endurance, thereby minimizing power consumption while ensuring reminder accuracy.

[0043] like Figure 2 As shown, the drug management layer specifically includes:

[0044] The drug locating module is configured with an initial quantity and supports dynamic expansion based on the user's actual medication needs to accommodate the application scenario of managing multiple drugs simultaneously. The locating module is made of biomedical-grade inert materials and has a volume of less than 2 cubic centimeters, ensuring that contact with various drugs will not affect the efficacy of the drugs. It has a built-in low-power LED indicator and a micro vibrator to respond to user search requests.

[0045] The prescription upload area provided by the smartwatch terminal supports users to upload the doctor's prescription information and medical order content to the system database by scanning the prescription QR code through the smartwatch camera or manually entering the prescription information. The system automatically parses the key information in the prescription, such as the drug name, specifications, usage and dosage, and precautions, and converts this information into a standardized structure and stores it in the drug information database. At the same time, it records the special requirements of the doctor's order on the time and conditions of medication.

[0046] The drug location module management area provided by the smartwatch terminal allows users to name each location module, associate drugs with each module, and mark each module by category. Users can use this area to add, delete, and update location modules. When the user selects "Find Drug Module" on the interface and specifies a number (such as "Drug Module 1"), the system triggers the corresponding unnamed module's LED indicator to flash. The user then identifies and removes the module, then names the module on the watch interface and associates it with a specific drug. The system assigns a unique identification code to each module and establishes a one-to-one or one-to-many mapping relationship with a specific drug. At the same time, drugs can be classified and grouped according to therapeutic purposes (such as cardiac medications, antihypertensive drugs, antibiotics, etc.).

[0047] The smartwatch terminal provides a search device function area to realize the positioning function of the user and the named medicine. When the user selects a specific medicine or the medicine positioning module on the watch interface, the system displays the relative distance and direction between the medicine and the user on the screen to help the user quickly find the location of the medicine;

[0048] The standardized drug information database adopts a hierarchical structure design, including four layers: basic drug information layer, medication plan layer, therapeutic importance layer, and medication record layer. The basic drug information layer stores static information such as drug name, ingredients, specifications, appearance, and contraindications; the medication plan layer records dynamic plan information such as the frequency, time point, and dosage of each drug; the therapeutic importance layer automatically calculates and stores the therapeutic importance classification results of each drug based on medical advice and drug characteristics; and the medication record layer continuously records users' actual medication behavior data, providing the system with a historical basis for medication compliance analysis.

[0049] The security encryption protocol uses end-to-end encryption technology to protect user privacy data. All medications and personal health information are stored locally in encrypted form, and limited data sharing with medical institution systems is only possible with user authorization. At the same time, the system supports the setting of emergency contacts. When multiple consecutive medication abnormalities are detected, status reminders can be sent to designated contacts with the user's pre-authorization.

[0050] like Figure 3 As shown, the spatial positioning layer specifically includes:

[0051] The three positioning components in the three-point positioning technology are the built-in positioning device of the smart watch, the positioning base station in the medicine box charging device, and the medicine positioning module of each medicine. The three constitute a complete spatial positioning network, supporting Bluetooth Low Energy (BLE) and Ultra-Wideband (UWB) dual communication modes to ensure high-precision positioning in different scenarios;

[0052] The smartwatch's built-in positioning device uses an integrated GPS chip, accelerometer, and gyroscope to track the user's location in real time. The system automatically records the user's location coordinates at preset time intervals while maintaining a Bluetooth signal connection with the drug positioning module, calculating the relative distance and direction between the two in real time.

[0053] The medicine box charging device acts as a home location positioning base station with a built-in Bluetooth signal transmission module. The user can set a custom home coverage radius through the watch interface. The default value is 10 meters. The system determines whether the user is in the home environment by measuring the signal strength between the watch and the charging device. The user can adjust this home coverage range parameter on the watch interface to meet the actual needs of different living environments.

[0054] The micro positioning module is integrated into each drug container or attached to the outer packaging of the drug, and uses low-power Bluetooth technology to maintain periodic signal communication with the smart watch to report location information and power status in real time;

[0055] The spatial relationship management model sets the following three main scenario judgment rules based on the relative positional relationship between the smartwatch, the medication positioning module, and the home positioning base station: when the smartwatch and the medication positioning module are both within the home coverage area, the system determines that the user is in the "home state"; when the smartwatch is outside the home coverage area and the medication positioning module is within the home coverage area, and the distance between the two exceeds 3 meters, the system determines that the user has forgotten to bring medication and triggers the corresponding early warning mechanism; when the smartwatch and the medication positioning module are both outside the home coverage area and the distance between the two exceeds 3 meters, the system determines that the user has lost medication and determines the optimal reminder method based on the current environmental conditions and user activity status;

[0056] The warning mechanism automatically adjusts its response level based on the severity of the situation. For example, if the user forgets to bring medication while leaving home, the system will first alert the user with a mild vibration and screen prompt. If the user continues to move away from home and does not return to retrieve medication, the system will increase the intensity of the reminder. For the "possibly lost medication" situation, the system will prioritize displaying the medication's last known location and distance, while also recording the geographical coordinates of the environment to facilitate tracing back.

[0057] The spatial positioning layer is designed with an automatic calibration mechanism. Every time the system detects that the drug positioning module has returned to the vicinity of the charging device, it automatically recalibrates the distance parameters and signal strength threshold to compensate for possible environmental interference and improve positioning accuracy. At the same time, the system periodically checks and records the power status of each positioning component, and sends an early warning prompt to the user through the smart watch when the power is low.

[0058] like Figure 4 As shown, the multi-dimensional perception layer specifically includes:

[0059] The multi-source sensors include a three-axis accelerometer, a three-axis gyroscope, an ambient light sensor, an air pressure sensor, a microphone array, and a GPS location module. These sensors operate synchronously at different sampling frequencies. The accelerometer and gyroscope record user body movements at a sampling frequency of 50Hz, the microphone monitors ambient sound characteristics at a sampling frequency of 16kHz, the ambient light sensor detects ambient light intensity at a frequency of 5Hz, and the air pressure sensor monitors altitude changes at a frequency of 1Hz. All sensor data is preliminarily filtered and integrated by a low-power coprocessor;

[0060] The multi-sensor fusion activity recognition framework divides user activities into four categories and a total of 14 specific states: light activity states include sitting still, standing, light walking, and steady strolling; moderate activity states include moderate-speed walking, fast walking, daily housework, and dining activities; high activity states include running and strenuous exercise; and special states include sleeping, driving, meetings, and social chatting. The system achieves real-time recognition of these states through sensor feature extraction and deep learning models, and each state corresponds to a specific reminder strategy.

[0061] The low-power neural network model uses an LSTM-CNN hybrid architecture, applying a one-dimensional convolutional layer to accelerometer and gyroscope data to extract local temporal features, and capturing long-term temporal patterns through an LSTM layer. The model is compressed to a size of no more than 300KB using knowledge distillation technology and can run in real time on a low-power coprocessor in a smartwatch, generating a state classification result every 30 seconds with a classification accuracy of no less than 95%.

[0062] The user status classification is further divided into three levels according to the difficulty of reminder intervention: standard reminder status, including seven states such as sitting, standing, light walking, steady stroll, medium-speed walking, daily housework and social chatting, in which standard vibration and visual reminder methods are used; cautious reminder status, including four states such as fast walking, dining activities, running and strenuous exercise, in which the system will wait for the user's exercise intensity to decrease or the eating activity to be suspended before triggering the reminder; special reminder status, including three states such as sleeping, driving and meeting, in which only progressive vibration wake-up is used for high-priority drugs in the sleeping state, only voice prompts are used to avoid distraction in the driving state, and only slight vibration and screen visual prompts are used in the meeting state;

[0063] The environmental parameter monitoring module analyzes five key features of the user's environment in real time: ambient light illumination is measured by an ambient light sensor and is divided into three levels: dark light environment (<10 lux), indoor standard light (10-500 lux), and bright environment (>500 lux); ambient noise level is obtained through microphone array analysis and is divided into three levels: quiet environment (<40dB), normal environment (40-70dB), and noisy environment (>70dB); movement state is comprehensively analyzed by the accelerometer and gyroscope and is divided into three levels: stillness, light movement, and intense movement; spatial location type is inferred through GPS and WiFi positioning and is divided into four categories: home, workplace, public place, and outdoor; altitude is estimated by a pressure sensor to detect whether the user is in a high-altitude environment such as an airplane;

[0064] The multi-dimensional perception layer adopts a multimodal sensor data fusion algorithm, combined with short-term window feature extraction and long-term window activity understanding, and improves the robustness of state recognition through a hierarchical reasoning structure. The system performs adaptive calibration based on the behavioral habits of different users and regularly updates the user behavior pattern library to improve the personalized accuracy of recognition. At the same time, an activation threshold control mechanism is introduced to automatically adjust the sampling frequency and activation state of each sensor according to the power status and environmental complexity, balancing recognition accuracy and energy consumption requirements.

[0065] like Figure 5 As shown, the intelligent decision-making layer specifically includes:

[0066] The context-adaptive decision engine adopts a three-layer fusion architecture, including a context perception module, a drug property analysis module, and a decision optimization module. By integrating user behavior status data and environmental parameter data provided by the multi-dimensional perception layer and drug property information provided by the drug management layer, it can accurately determine the optimal reminder timing.

[0067] The context awareness module uses a weighting matrix to comprehensively score the user's state and environmental parameters identified by the multi-dimensional perception layer, categorizing the user's situation into five levels: optimal alert level (user sitting or standing in a quiet environment), appropriate alert level (user engaging in light activity in a moderate environment), general alert level (user engaging in moderate activity or slightly disturbed environment), cautious alert level (user engaging in high activity or significantly disturbed environment), and delayed alert level (user in special situations such as sleeping, driving, or in an important meeting). The system continuously optimizes this weighting matrix based on historical medication records and user feedback to improve the accuracy of context assessment.

[0068] The drug characteristic analysis module divides drugs into four levels of therapeutic importance based on the medical properties of the drugs, frequency of use, potential risks of missed use, and importance of the doctor's order: life-sustaining drugs (such as insulin, heart disease drugs, etc., which must be taken strictly on time), with a score of 90-100 points; symptom control drugs (such as antihypertensive drugs, hypoglycemic drugs, etc., which are used to control the symptoms of chronic diseases in the long term), with a score of 70-89 points; treatment auxiliary drugs (such as vitamins, probiotics, etc., which are auxiliary treatment or health care drugs), with a score of 40-69 points; temporary drugs (such as cold medicines, painkillers, etc., which are used in the short term), with a score of 10-39 points; the system will automatically make a preliminary assessment of the importance of the drugs based on the urgency keywords and frequency of use in the doctor's order, and then confirm or adjust the assessment with the user;

[0069] The decision optimization module uses a reinforcement learning algorithm to establish a reminder decision model with maximizing long-term medication compliance as the objective function. Input parameters include the situational suitability score, the medication importance score, the deviation value from the preset medication time, and user historical feedback data, and outputs a reminder trigger probability distribution curve. The system is designed with a dynamic time window mechanism. For life-sustaining drugs, the allowed time deviation window is the smallest, usually ±15 minutes; for symptom control drugs, the time window is ±30 minutes; for treatment-assisted drugs, the time window is ±60 minutes; and for temporary drugs, the time window is the largest, up to ±120 minutes.

[0070] The dynamic threshold mechanism automatically adjusts the probability threshold for triggering reminders based on the combined relationship between the importance of the drug and the appropriateness of the situation: for life-sustaining drugs, the trigger threshold is as low as 0.3 even in the delayed reminder level scenario, ensuring that reminders are issued in almost all situations; for symptom control drugs, the trigger threshold is 0.4 in the most appropriate and appropriate reminder level scenarios, and is increased to 0.6 in the cautious and delayed reminder level scenarios; for therapeutic adjuvant drugs, the trigger threshold is 0.5 in the most appropriate reminder level scenario, and is increased to 0.8 in the delayed reminder level scenario; for temporary medications, reminders are triggered only in the most appropriate and appropriate reminder level scenarios, with thresholds of 0.6 and 0.7, respectively;

[0071] The dynamic hierarchical reminder strategy allocates differentiated reminder resources according to the importance of drugs with different priorities: for life-sustaining drugs, multiple reminders are allowed until the user confirms, with a maximum of 5 reminders, and the interval time gradually shortened from 5 minutes to 1 minute; for symptom control drugs, a maximum of 3 reminders are issued, with an interval of 10 minutes; for treatment-adjuvant drugs, a maximum of 2 reminders are issued, with an interval of 15 minutes; for temporary medications, only 1 reminder is issued; at the same time, the system will reasonably arrange the reminder order according to the reminder priority of multiple drugs in the same time period to avoid triggering multiple reminders at the same time and causing trouble to the user;

[0072] The intelligent decision-making layer also includes a user feedback learning module, which records the user's response to each reminder (taking medicine immediately, delaying taking medicine, ignoring reminders, etc.) and feedback satisfaction, and continuously optimizes the decision model parameters through a supervised learning algorithm to improve the personalization of the reminder strategy; this module adopts an incremental learning method, enabling the system to gradually adapt to changes in user habits and maximize user experience while ensuring medication safety.

[0073] like Figure 6 As shown, the interactive reminder layer specifically includes:

[0074] The interactive reminder layer supports five basic reminder modes: visual reminder, tactile reminder, auditory reminder, multimodal reminder and non-intrusive reminder. It automatically selects the best reminder mode or a combination of multiple modes based on the user status classification, environmental condition parameters and drug importance level provided by the intelligent decision-making layer;

[0075] The visual reminder mode includes screen lighting, customized reminder interface display, and intelligent brightness adjustment mechanism. In low-light environments (<10 lux), the screen brightness is automatically increased to the highest comfortable level, in standard lighting environments (10-500 lux), the brightness is maintained at a medium level, and in bright light environments (>500 lux), a high-contrast display mode is activated. The system displays differentiated interface designs based on the importance of the medication, with eye-catching red borders and drug icons for life-sustaining drugs, orange borders for symptom-control drugs, and blue and green borders for treatment-aid and temporary medications, respectively.

[0076] The tactile reminder mode uses a precisely controlled vibration motor and supports a combination of four different intensities and six vibration modes: mild vibration intensity (amplitude 20%) and brief vibration mode in silent environments (<40dB); moderate vibration intensity (amplitude 50%) and double pulse vibration mode in normal environments (40-70dB); high-intensity vibration (amplitude 80%) and continuous wave mode in high-noise environments (>70dB); for life-sustaining and symptom-control drugs, the system will gradually increase the vibration intensity and duration after the user does not respond to the initial reminder; at the same time, the system automatically calibrates the optimal vibration parameters for different wristband materials and wearing tightness through accelerometer feedback data;

[0077] The auditory reminder mode uses the smartwatch's built-in speaker to provide audio prompts, with an intelligent volume control function. In a quiet environment, the volume is set to 30%, in a normal environment, the volume is set to 60%, and in a noisy environment, the volume is set to 90%. The system has designed unique reminder sound effects for different categories of medications. Life-sustaining drugs use three short, rapid tones, symptom-control drugs use two medium-paced tones, and treatment-aid and temporary medications use a single, gentle tone. When the ambient noise exceeds 80dB, the system automatically abandons the auditory reminder and switches to an enhanced tactile reminder.

[0078] The multimodal reminder mode comprehensively utilizes three reminder methods: visual, tactile, and auditory. According to the most appropriate reminder level situation determined by the intelligent decision-making layer, screen display, moderate vibration, and appropriate sound prompts are activated simultaneously. According to different reminder levels determined by the situational awareness module, the system adopts different combination strategies: in the most appropriate reminder level situation, the three reminder methods are activated simultaneously; in the appropriate reminder level situation, visual and tactile reminders are combined; in the general reminder level situation, tactile reminders are preferred in addition to visual reminders; in the cautious reminder level situation, visual reminders are preferred or the tactile reminder timing is delayed;

[0079] The non-intrusive reminder mode is designed for special conditions, including: sleep state reminders use gradual mild vibrations, starting from the lowest intensity and slowly increasing to the awakening intensity, and such reminders are only triggered for life-sustaining drugs; driving state reminders use only brief voice prompts and limited visual prompts to avoid distracting driving attention; meeting state reminders use only a single mild vibration and a low-brightness screen prompt, and the screen content automatically switches to a simple mode, displaying only the drug name and dosage, without detailed medical instructions; for social occasions, the system uses mild intermittent vibration prompts to avoid causing social interference;

[0080] The personalized reminder parameter optimization mechanism records the user's response behavior to each reminder, including four types: immediate response, delayed response, ignoring reminders, and closing reminders. The system uses machine learning algorithms to analyze the user's preferences and response speed for different reminder methods, and continuously adjusts the selection probability and parameter settings of the reminder mode. At the same time, it builds a user activity pattern map to identify the user's daily regular activities and prioritizes reminders at natural interruptions such as activity transition points (such as the end of a meal or a meeting).

[0081] The medication compliance recording and analysis module automatically records the issuance time of each medication reminder, the user response time and the actual medication confirmation time, calculates key indicators such as on-time medication rate, missed medication rate and delayed medication rate, and generates daily, weekly and monthly medication compliance reports; the system evaluates the effectiveness of the reminder strategy based on compliance data, and automatically strengthens the reminder strategy for drugs with low compliance; users can view graphical trend analysis reports through smart watches, including comparisons of medication compliance in different time periods, comparisons of compliance rates of various drugs, and long-term changes in compliance trends.

[0082] like Figure 7 As shown, the energy management mechanism of the system specifically includes:

[0083] The hierarchical processing mechanism adopts a three-level computing architecture, including a low-power resident monitoring unit, a medium-power coprocessor and a high-performance main processor, and automatically distributes the load among the three-level processing units according to the complexity of the computing task;

[0084] The low-power resident monitoring unit consumes less than 1mW and is responsible for collecting raw sensor data and performing simple threshold judgment. When a potential state change is detected, it triggers the coprocessor to wake up.

[0085] The medium-power coprocessor consumes 10-50mW of power and is responsible for performing feature extraction, activity state classification, and environmental parameter analysis. It has a built-in lightweight neural network model to identify user activity states. The system dynamically adjusts the coprocessor's operating frequency and wake-up interval based on battery power.

[0086] The high-performance main processor is activated only when it is needed to perform complex decision analysis, user interaction, high-priority medication reminders or data analysis, and the average daily activation time is controlled within a cumulative total of 30 minutes;

[0087] The system adopts an adaptive sampling strategy to dynamically adjust the sensor sampling frequency according to the stability of the user state: reduce the frequency when the state is stable, increase the frequency when the state changes, and reduce the number of wake-up times for batch processing of data;

[0088] The environmental energy harvesting technology integrates three energy recovery mechanisms: light energy conversion module, thermal energy conversion module and micro-kinetic energy conversion module. Under ideal conditions, it can extend the system life by more than 40%;

[0089] The system automatically adjusts its functional configuration based on the battery level: full functionality in high battery mode (>70%); reduced sampling frequency of non-critical sensors in power conservation mode (30%-70%); emergency mode (<30%) retains only high-priority medication reminders; and critical mode (<10%) retains only basic reminders for life-sustaining medications.

[0090] The system transfers non-real-time medication compliance historical data and user behavior pattern analysis tasks to the cloud server for processing, and the optimized model parameters are updated locally during charging, achieving optimal allocation of computing tasks and energy consumption.

[0091] As an optional implementation, Figure 1 As shown in the figure, the intelligent drug management and reminder system based on multi-dimensional environmental perception of the present invention includes five functional layers: drug management layer, spatial positioning layer, multi-dimensional perception layer, intelligent decision-making layer and interactive reminder layer. These functional layers are interconnected through information interaction interfaces to jointly complete the intelligent management and reminder of user medication behavior. System deployment and drug management based on smart watches are as follows: Figure 2 As shown, the drug management layer includes five main components: drug positioning module, prescription upload area, drug positioning module management area, search device function area and standardized drug information database.

[0092] In this example, for a 65-year-old heart patient who takes long-term antihypertensive, antidiabetic, and anticoagulant medications, the system is configured with three medication location modules: "Antihypertensive," "Antidiabetic," and "Warfarin." Each location module is made of biomedical-grade inert material, has a volume of 1.5 cubic centimeters, and houses a built-in low-power Bluetooth module, an LED indicator, and a microvibrator.

[0093] For the first time, users scan the QR code on their hospital prescription through the prescription upload area on the smartwatch. The system automatically parses and extracts information such as the drug name, specifications, and usage and dosage. For warfarin, the system identifies it as an anticoagulant, with a recommended dosage of one 3mg tablet once daily, recommended after breakfast. After parsing this information, the system guides the user to associate the drug with a specific location module (labeled "Warfarin").

[0094] The Medication Location Module Management area allows users to name and group location modules. In this example, three medications are grouped into "Cardiac Medications" and "Metabolic Medications" for centralized management. To locate a specific medication, users can select "Find Warfarin" in the Find Device area. This triggers the corresponding location module's LED to flash and briefly vibrate. Directions and distance information are also displayed on the watch screen, helping users quickly locate the medication.

[0095] The standardized drug information database utilizes a hierarchical design, storing warfarin information in four layers: the basic information layer records warfarin's generic name, trade name, strength, and appearance; the medication plan layer sets a daily dosing time of 8:00 AM; the therapeutic importance layer classifies it as a "life-sustaining" drug (scored 95 points); and the medication record layer continuously records the user's actual daily dosing time and dosage compliance. All data is protected by end-to-end encryption to ensure user privacy.

[0096] As another optional implementation, spatial positioning and position relationship management such as Figure 3 As shown in the figure, the spatial positioning layer uses three-point positioning technology to establish a spatial relationship management model between users, drugs, and the environment. The three positioning components include the positioning device built into the smartwatch, the positioning base station in the drug box charging device, and the drug positioning module.

[0097] In this example, the user places the pill box charger in a fixed location at home and sets a home coverage radius of 12 meters, covering the entire residential area. The smartwatch uses its built-in GPS chip, accelerometer, and gyroscope to track the user's location in real time, while maintaining a Bluetooth connection with the medication location module to calculate relative distance and direction.

[0098] One morning, a user was preparing to go out for medical treatment. The system detected that the user was moving away from home (the distance between the watch and the charger increased), but the "Warfarin" positioning module remained at home (the distance from the charger remained the same). When the user was about 5 meters from the door, the system determined that they had forgotten to bring their medication and immediately triggered a reminder mechanism. The watch vibrated and the screen prompted the user: "You may have forgotten your anticoagulant warfarin. I recommend returning to pick it up. You have a follow-up appointment at the hospital this afternoon." The user then returned to pick up the medication, avoiding the risk of missing an anticoagulant.

[0099] In another scenario, while the user was out and about, the system detected that the distance between the "antihypertensive medication" location module and the smartwatch had suddenly changed from close range (carried on the person) to long distance, triggering a "possible medication loss" alert. The system immediately displayed the geographic coordinates and relative distance of the medication's last known location on the watch screen, with a prompt: "Your antihypertensive medication may have been left in a café (approximately 120 meters from your current location). We recommend retrieving it immediately." The user followed the prompt and returned to the café to retrieve the medication, avoiding the risk of missing a dose due to lost medication.

[0100] The system incorporates an automatic calibration mechanism. Whenever the drug locating module returns to the charging station, it automatically recalibrates distance parameters and signal strength thresholds to ensure accurate positioning. The system also regularly checks the battery status of each locating component. If the battery level of a drug locating module falls below 20%, the user is automatically notified via a smartwatch to recharge immediately.

[0101] As another optional implementation method, multi-dimensional perception and user status recognition are as follows: Figure 4 As shown in Figure 3, the multi-dimensional perception layer collects and processes user behavior data in real time by integrating multi-source sensors, converts it into recognizable behavior pattern classifications, and monitors environmental parameters at the same time.

[0102] In this embodiment, the system uses the smartwatch's built-in sensors, including a three-axis accelerometer, a three-axis gyroscope, an ambient light sensor, an air pressure sensor, and a microphone array, to collect data at different sampling frequencies. The accelerometer and gyroscope sample at 50Hz to record the user's body movement characteristics; the microphone monitors ambient noise at 16kHz; and the ambient light sensor detects ambient light intensity at 5Hz.

[0103] Using a multi-sensor fusion activity recognition framework, the system categorizes user activities into four categories and fourteen specific states. A low-power neural network model, employing a hybrid LSTM-CNN architecture, runs on the watch's coprocessor, generating state classification results every 30 seconds with 96% accuracy. The model is compressed to 280KB using knowledge distillation technology, ensuring efficient operation on resource-constrained smartwatches.

[0104] User status is further divided into three levels based on the difficulty of reminder intervention. For example, when the system identifies the user as "sitting still" (standard reminder state), it uses normal vibration and screen prompts for reminders; when it identifies the user as "eating" (cautious reminder state), the system waits for the user to pause in their meal before triggering the reminder; and when it identifies the user as "driving" (special reminder state), it only uses a brief voice prompt to avoid distraction.

[0105] The environmental parameter monitoring module analyzes the characteristics of the user's environment in real time. For example, if the system detects that the user has entered a noisy environment (noise > 70dB), it automatically adjusts the reminder strategy, increasing the vibration intensity and disabling the sound reminder. If it detects a dim environment (<10lux), it increases the screen brightness to ensure the effectiveness of the visual reminder.

[0106] The system adaptively calibrates itself based on the behavioral habits of different users, and its state recognition accuracy continues to improve with usage time. At the same time, an activation threshold control mechanism is introduced to automatically reduce the sampling frequency of non-critical sensors when the battery level drops below 30%, balancing recognition accuracy and energy consumption requirements.

[0107] As another optional implementation, the situational adaptive decision and the optimal reminder timing calculation are as follows: Figure 5 As shown in the figure, the intelligent decision-making layer uses a situational adaptive decision engine to comprehensively analyze the user status, environmental conditions, and drug characteristics to calculate the optimal reminder time.

[0108] In this embodiment, the context awareness module scores and grades the user's context based on data provided by the multi-dimensional perception layer. For example, if sensor data indicates that the user is sitting quietly in a quiet environment with moderate lighting, the system will assess the current context as "Optimal Alert Level" (score of 90 points); if the user is in a meeting with multiple people talking nearby, the system will assess the context as "Delayed Alert Level" (score of 30 points).

[0109] The medication property analysis module assesses the patient's three medications based on the doctor's orders and their properties: the anticoagulant warfarin is classified as "life-sustaining" (95 points), the antihypertensive drug is classified as "symptom control" (85 points), and vitamin D is classified as "therapeutic aid" (50 points). These scores directly influence the system's alert strategy and priority allocation.

[0110] The decision optimization module uses a reinforcement learning algorithm to build a reminder decision model, with maximizing long-term medication adherence as the objective function. The system sets differentiated time windows for different medication categories: the allowable time deviation window for warfarin is ±15 minutes, for antihypertensive drugs it is ±30 minutes, and for vitamin D it is ±60 minutes.

[0111] A dynamic threshold mechanism automatically adjusts the probability threshold for triggering reminders based on medication importance and contextual appropriateness. For example, for life-sustaining medications like warfarin, the trigger threshold is as low as 0.3, even in delayed reminder-level scenarios (such as meetings), ensuring reminders are issued in almost all situations. Meanwhile, for adjunctive medications like vitamin D, reminders are triggered only in optimal and appropriate reminder-level scenarios, with thresholds of 0.5 and 0.7, respectively.

[0112] A dynamic, hierarchical reminder strategy allocates differentiated reminder resources to medications of different priorities. For example, if a user doesn't respond to a warfarin dose after the due date, the system will issue up to five repeated reminders, with the intervals gradually decreasing from five minutes to one minute. Vitamin D, on the other hand, will only issue a single reminder. When multiple medications are due close together, the system prioritizes reminders for more important medications, avoiding the inconvenience of multiple simultaneous reminders.

[0113] The user feedback learning module records how users respond to each reminder. For example, the system observed that users typically delayed responding to the 9 PM reminder for their blood pressure medication, but responded almost immediately to the 8 AM reminder for their warfarin. Based on this data, the system automatically adjusted the default evening reminder time for blood pressure medication to 8:45 AM, 15 minutes earlier, to accommodate the user's actual medication habits.

[0114] As another optional implementation, multimodal interactive reminder and personalized optimization Figure 6 As shown in the figure, the interactive reminder layer automatically selects the best reminder method based on the judgment of the intelligent decision-making layer, and continuously optimizes the reminder parameters through user feedback.

[0115] In this embodiment, the system supports five basic reminder modes, automatically selecting the optimal combination based on the situation. For example, when the 8:00 AM warfarin reminder arrives, the system detects that the user is reading a newspaper in a quiet environment (the optimal reminder-level scenario) and automatically selects a multimodal reminder mode, activating screen display, moderate vibration, and appropriate audio. The screen displays a prominent red-bordered medication icon, clearly labeling the medication information: "Anticoagulant Warfarin 1 Tablet (3mg)."

[0116] When the system detects that the user is in a noisy environment (noise > 75dB), it automatically increases the vibration intensity to 80%, adopts a continuous fluctuation mode, and disables sound reminders; when it detects that the user is in a dim environment, it automatically increases the screen brightness and enables high-contrast display mode to ensure the effectiveness of visual reminders.

[0117] For special situations, the system uses a non-intrusive reminder mode. For example, if it detects that the user is driving, it only uses a brief voice prompt: "It's time to take warfarin," to avoid distracting visual reminders. If it detects that the user is in a meeting, it only uses a slight vibration and a simple on-screen prompt to avoid causing social interference.

[0118] The personalized reminder parameter optimization mechanism records user preferences and response speeds for different reminder methods. For example, the system observes that users typically respond faster to vibration reminders than audio reminders and automatically increases the probability of selecting vibration reminders accordingly. Simultaneously, the system maps user activity patterns and finds that users are more receptive to medication reminders after breakfast or after watching TV in the evening, optimizing reminder timing accordingly.

[0119] The Medication Adherence Recording and Analysis module automatically records the time each medication reminder is issued, the user's response time, and the actual time it confirms medication intake, calculating various adherence metrics. For example, the system generates a monthly report showing that the on-time take rate for warfarin is 92%, with an average delay of 8 minutes; the on-time take rate for antihypertensive medication is 85%, with an average delay of 22 minutes; and the missed dose rate for vitamin D is 15%. Based on this data, the system automatically strengthens reminder strategies for medications with low adherence rates, such as increasing reminder frequency or adjusting reminder timing.

[0120] As another optional implementation method, energy management and system optimization such as Figure 7 As shown in the figure, the system adopts a hierarchical processing mechanism to optimize energy management, completing sensor data acquisition and primary processing on a low-power coprocessor, and activating the main processor only when complex calculations are required.

[0121] In this embodiment, the system employs a three-tiered computing architecture, consisting of a low-power resident monitoring unit (LRMU), a medium-power coprocessor, and a high-performance main processor. The LRMU consumes only 0.8mW and continuously operates to collect raw sensor data and perform simple threshold checks, such as monitoring for sudden changes in motion or significant changes in ambient noise. Upon detecting a potential state change, the coprocessor automatically wakes up.

[0122] The medium-power coprocessor consumes 30mW and is responsible for performing moderately complex tasks such as feature extraction, activity state classification, and environmental parameter analysis. For example, it wakes up every 30 seconds to run a compressed neural network model to analyze the user's current activity state, such as sitting still, walking, or eating. The system automatically adjusts the coprocessor's operating frequency based on the battery level. When the battery level drops below 30%, the state recognition interval is extended to once every 60 seconds.

[0123] The high-performance main processor activates only when needed for complex decision analysis, user interaction, high-priority medication reminders, or data analysis. For example, when a user clicks to view an adherence report or when the system needs to retrain a user behavior model. Through optimized design, the system limits the average daily main processor activation time to less than 25 minutes, significantly reducing energy consumption.

[0124] The system uses an adaptive sampling strategy to dynamically adjust the sensor sampling frequency. For example, when the user's state remains stable (such as sitting still for more than 10 minutes), the accelerometer and gyroscope sampling frequency is reduced from 50Hz to 10Hz. When a change in the user's state is detected, the high sampling frequency is immediately restored to capture detailed motion characteristics. At the same time, the system batches data to reduce the number of processor wake-ups, such as accumulating 30 seconds of sensor data and then processing it all at once, rather than processing each data point in real time.

[0125] Ambient energy harvesting technology integrates three energy recovery mechanisms: light, heat, and micro-kinetic energy. The smartwatch dial incorporates a high-efficiency photoelectric conversion film to convert ambient light into electricity; the back of the watch incorporates a thermoelectric conversion element that generates electricity from the temperature difference between the human body and the surrounding environment; and the strap incorporates a micro-kinetic energy converter that converts the user's daily arm movement into electricity. Under ideal conditions (such as during outdoor activities), these energy harvesting technologies can provide approximately 25% of the system's energy needs, extending battery life by over 40%.

[0126] The system automatically adjusts its functional configuration based on the battery charge level. When the battery level is above 70% in "Battery Full Mode," the system operates with full functionality, including high-frequency sensor sampling and complete activity status recognition. In "Battery Conservation Mode," between 30% and 70%, the sampling frequency of non-critical sensors is reduced, such as reducing the ambient light sensor sampling frequency from 5Hz to 1Hz. In "Battery Emergency Mode," below 30%, only high-priority medication reminders are retained, disabling complex status recognition algorithms. In "Battery Critical Mode," below 10%, only basic reminders for "life-sustaining" medications are retained, employing a simple time-based reminder strategy and disabling all other non-essential features.

[0127] The system offloads non-real-time data analysis tasks to cloud servers. For example, computationally intensive tasks such as long-term trend analysis of medication adherence, user behavior pattern mining, and reminder parameter optimization are automatically uploaded to the cloud for processing when the user charges their device and connects to a Wi-Fi network. Optimized model parameters and analysis results are updated locally after charging is complete, optimizing the allocation of computing tasks and energy consumption. This strategy enables the system to significantly reduce local energy consumption while maintaining intelligent and personalized services.

[0128] As can be seen from the above examples, the system of the present invention utilizes innovative features such as multi-dimensional environmental perception technology, three-point positioning technology, a context-adaptive decision engine, and multimodal interactive reminders to build a comprehensive intelligent medication management and reminder system. The system intelligently senses user behavior and environmental conditions, accurately locates medications, implements differentiated reminder strategies based on medication importance, and extends system battery life through optimized energy management, effectively improving user medication compliance and enhancing the medication management experience for patients with chronic diseases.

[0129] The various embodiments in this specification are described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same or similar parts between the various embodiments can be referenced to each other.

[0130] This document uses specific examples to illustrate the principles and implementation methods of the present invention. The above examples are only intended to help understand the method and core concept of the present invention. At the same time, those skilled in the art will find that the specific implementation methods and application scopes may vary based on the concept of the present invention. In summary, the contents of this specification should not be construed as limiting the present invention.

Claims

1. An intelligent drug management and reminder system based on multi-dimensional environmental perception, characterized in that: include: It includes drug management layer, spatial positioning layer, multi-dimensional perception layer, intelligent decision-making layer and interactive reminder layer; The drug management layer provides a prescription setting interface on the smartwatch terminal, supports naming and classification management of drug location modules, establishes a standardized drug information database, and protects the user's patient privacy data obtained through a secure encryption protocol; The spatial positioning layer uses three-point positioning technology to establish a spatial relationship management model between users, drugs and the environment. By tracking the user's location, home location and drug location in real time, it can identify abnormal conditions such as forgotten or lost drugs and trigger an early warning. The multi-dimensional perception layer integrates multi-source sensors, collects and processes user behavior data in real time through the fusion activity recognition framework of the multi-source sensors, classifies user behavior patterns based on the user behavior data, and monitors environmental noise, light and motion state parameters; The intelligent decision-making layer uses a context-adaptive decision-making engine to comprehensively analyze user behavior, environmental conditions, and drug characteristics, determine the optimal reminder timing based on a dynamic threshold mechanism and reinforcement learning algorithm, and implement a dynamic hierarchical reminder strategy based on the importance of drug treatment; The interactive reminder layer is used to automatically select the reminder method according to the user status and environmental conditions, optimize the personalized reminder parameters through user feedback data, record medication compliance data and generate trend analysis reports.

2. The intelligent drug management and reminder system based on multi-dimensional environmental perception according to claim 1 is characterized in that: The number of drug locating modules is dynamically increased and expanded based on actual user medication needs to accommodate scenarios where multiple medications are managed simultaneously. Made of biomedical-grade inert materials, the modules are smaller than 2 cubic centimeters in size and feature built-in low-power LED indicators and micro-vibrators to respond to user search requests. The prescription upload area provided by the smartwatch terminal allows users to upload doctor-issued prescription information and medical instructions to the system database by scanning the prescription QR code with the smartwatch camera or manually entering the information. The system automatically parses key information in the prescription, such as the drug name, specifications, usage, dosage, and precautions, and converts this information into a standardized structure and stores it in the drug information database. It also records special requirements for medication time and conditions in the medical instructions. The drug location module management area provided by the smartwatch terminal allows users to name, associate and pair drugs, and classify each drug location module. Users can add, delete, and update drug location modules through the management area. When the user selects "Find Drug Location Module" on the interface and specifies a number, the LED indicator of the corresponding unnamed module flashes to remind the user to remove the drug location module. The drug location module is then named and associated with a specific drug on the watch interface. A unique identification code is assigned to each drug location module and a one-to-one or one-to-many mapping relationship is established with a specific drug. Drugs are then classified and grouped according to therapeutic purposes. The search device function area provided by the smartwatch terminal enables the user to locate the named medication. When the user selects a specific medication or medication location module on the watch interface, the relative distance and direction between the specific medication or medication location module and the user are displayed on the screen to help the user find the corresponding location. The standardized medication information database adopts a hierarchical structure design; the hierarchical structure includes a basic medication information layer, a medication plan layer, a treatment importance layer, and a medication record layer. The basic drug information layer stores static information such as drug name, ingredients, specifications, appearance, and contraindications; the medication plan layer records dynamic plan information such as the frequency, time point, and dosage of each drug; the therapeutic importance layer is used to automatically calculate and store the therapeutic importance grading results of each drug based on medical advice and drug characteristics; the medication record layer is used to continuously record the user's actual medication behavior data to provide a historical basis for medication compliance analysis; the security encryption protocol uses end-to-end encryption technology to protect user privacy data, and all drugs and personal health information are encrypted and stored locally, and limited data sharing is only performed with the medical institution system with the user's authorization; it also supports the setting of emergency contacts. When multiple consecutive medication abnormalities are detected, status reminders will be sent to designated contacts with the user's pre-authorization.

3. The intelligent drug management and reminder system based on multi-dimensional environmental perception according to claim 1 is characterized in that: The three positioning components of the three-point positioning technology are the built-in positioning device of the smart watch, the positioning base station in the medicine box charging device, and the medication positioning module for each medication. The three together form a complete spatial positioning network. The three-point positioning technology supports Bluetooth low energy and ultra-wideband dual communication modes to ensure high-precision positioning in different scenarios. The built-in positioning device of the smart watch uses an internally integrated GPS chip, accelerometer and gyroscope to achieve real-time tracking of the user's location. The built-in positioning device of the smart watch automatically records the user's location coordinates at preset time intervals while maintaining a Bluetooth signal connection with the medication positioning module to calculate the relative distance and relative direction between the user and the medication positioning module in real time. The medicine box charging device acts as a home location positioning base station and has a built-in Bluetooth signal transmission module. The medicine box charging device can set a custom home coverage radius through the watch interface and determine whether the user is in the home environment by measuring the signal strength between the watch and the charging device. The user can adjust the parameters of the home coverage range on the watch interface to meet the actual needs of different living environments. The miniature drug positioning module is integrated into each drug container or attached to the drug packaging, using low-power Bluetooth technology to maintain periodic signal communication and report location information and battery status in real time; The spatial relationship management model sets the following three main scenario judgment rules based on the relative positional relationship between the smartwatch, the medication positioning module, and the home positioning base station: when the smartwatch and the medication positioning module are both within the home coverage area, it is determined to be "home state"; when the smartwatch is outside the home coverage area and the medication positioning module is within the home coverage area, and the distance between the smartwatch and the medication positioning module exceeds 3 meters, it is determined to be "forgotten to bring medication when going out", and the corresponding early warning mechanism is triggered; when the smartwatch and the medication positioning module are both outside the home coverage area, and the distance between the smartwatch and the medication positioning module exceeds 3 meters, it is determined to be "possibly lost medication" state, and the optimal reminder method is determined based on the current environmental conditions and user activity status; The warning mechanism automatically adjusts the response level based on the severity of the situation. For example, if the user forgets to bring medication, the system will first alert the user with a slight vibration and screen prompt. If the user continues to move away from home and does not return to retrieve the medication, the reminder intensity will increase. For the "possibly lost medication" situation, the system will prioritize displaying the medication's last known location and distance, while also recording the geographical coordinates of the environment to facilitate tracing back. The spatial positioning layer is designed with an automatic calibration mechanism. Every time it detects that the drug positioning module has returned to the vicinity of the charging device, it automatically recalibrates the distance parameters and signal strength threshold to compensate for any environmental interference. At the same time, it periodically checks and records the power status of each positioning component and sends an early warning to the user through the smart watch when the power is low.

4. The intelligent drug management and reminder system based on multi-dimensional environmental perception according to claim 1 is characterized in that: The multi-source sensor includes a three-axis accelerometer, a three-axis gyroscope, an ambient light sensor, an air pressure sensor, a microphone array, and a GPS location module. Each sub-sensor in the multi-source sensor operates synchronously at different sampling frequencies. The accelerometer and gyroscope record user body movements at a sampling frequency of 50Hz, the microphone monitors ambient sound characteristics at a sampling frequency of 16kHz, the ambient light sensor detects ambient light intensity at a frequency of 5Hz, and the air pressure sensor monitors altitude changes at a frequency of 1Hz. The data from each sub-sensor is preliminarily filtered and integrated by a low-power coprocessor; The multi-source sensor fusion activity recognition framework is used to classify user activities into light activity state, moderate activity state, high activity state and special state class: the light activity state includes four types: sitting still, standing, light walking and steady stroll; the moderate activity state includes four types: medium-speed walking, fast walking, daily housework and dining activities; the high activity state includes two types: running and strenuous exercise; the special state class includes four types: sleeping, driving, meeting and social chatting; the light activity state, the moderate activity state, the high activity state and the special state class are recognized in real time through sensor feature extraction and deep learning models, and each state corresponds to a specific reminder strategy; The multi-dimensional perception layer adopts an LSTM-CNN hybrid architecture, applies a one-dimensional convolutional layer to the accelerometer and gyroscope data to extract local time features, and captures long-term time series patterns through the LSTM layer; the multi-dimensional perception layer is used to analyze five key features of the user's environment in real time; the ambient light illumination is measured by an ambient light sensor and is divided into three levels: dark light environment, indoor standard light and bright environment; the ambient noise level is obtained by microphone array analysis and is divided into three levels: quiet environment, normal environment and noisy environment; the motion state is comprehensively analyzed by a three-axis accelerometer and a three-axis gyroscope and is divided into three levels: stillness, slight movement and intense movement; the spatial position type is inferred by the GPS location module and is divided into four categories: home, workplace, public place and outdoor; the altitude is estimated by a pressure sensor to detect whether the user is in a high-altitude environment such as an airplane; The multi-dimensional perception layer uses a multimodal sensor data fusion algorithm, combining short-term window feature extraction and long-term window activity understanding. It improves the robustness of state recognition through a hierarchical reasoning structure, performs adaptive calibration based on the behavioral habits of different users, regularly updates the user behavior pattern library, and introduces an activation threshold control mechanism to automatically adjust the sampling frequency and activation state of each sub-sensor based on the power status and environmental complexity to balance recognition accuracy and energy consumption requirements. The multi-dimensional perception layer is divided into three levels according to the difficulty of reminder intervention: standard reminder state, cautious reminder state and special reminder state; the standard reminder state includes sitting quietly, standing, light walking, steady stroll, medium-speed walking, daily housework and social chatting, and standard vibration and visual reminder methods are used in the standard reminder state; the cautious reminder state includes fast walking, dining activities, running and strenuous exercise, and in the cautious reminder state, the reminder is triggered only after the user's exercise intensity is reduced or the eating activity is suspended; the special reminder state includes sleeping, driving and meeting, and in the sleeping state, only progressive vibration wake-up is used for high-priority drugs, in the driving state, only voice prompts are used to avoid distraction, and in the meeting state, only slight vibration and screen visual prompts are used.

5. The intelligent drug management and reminder system based on multi-dimensional environmental perception according to claim 1 is characterized in that: The context-adaptive decision engine adopts a three-layer fusion architecture, including a context-awareness module, a drug property analysis module, and a decision optimization module. The context-adaptive decision engine determines the optimal reminder timing by integrating user behavior status data and environmental parameter data provided by the multi-dimensional perception layer, as well as drug property information provided by the drug management layer. The context awareness module uses a weighting matrix to comprehensively score the user status and environmental parameters identified by the multi-dimensional perception layer, categorizing the user's context into five levels: optimal reminder level, appropriate reminder level, cautious reminder level, and delayed reminder level. This weighting matrix is continuously optimized based on historical medication records and user feedback to improve the accuracy of context assessment. The drug property analysis module is used to classify drugs into four therapeutic importance levels: life-sustaining, symptom-control, treatment-assisting, and temporary medication based on the drug's medical properties, frequency of use, potential risk of missed medication, and doctor's order importance. It also automatically and preliminarily assesses drug importance based on the urgency keyword and frequency of use in the doctor's order, which is then confirmed or adjusted by the user. The life-sustaining category is scored on a scale of 90-100 points; the symptom-control category is scored on a scale of 70-89 points; the treatment-assisting category is scored on a scale of 40-69 points; and the temporary medication category is scored on a scale of 10-39 points. The decision optimization module is used to establish a reminder decision model using a reinforcement learning algorithm, with maximizing long-term medication compliance as the objective function. A dynamic time window mechanism is designed. For life-sustaining drugs, the allowed time deviation window is the smallest, typically ±15 minutes; for symptom control drugs, the time window is ±30 minutes; for treatment-assisted drugs, the time window is ±60 minutes; and for temporary drugs, the time window is the largest, up to ±120 minutes. The input parameters of the reminder decision model include the situational suitability score, the drug importance score, the deviation value from the preset medication time, and user historical feedback data; The dynamic threshold mechanism is used to automatically adjust the probability threshold for triggering reminders based on the combined relationship between the importance of the drug and the appropriateness of the situation: for life-sustaining drugs, the trigger threshold is as low as 0.3 in the delayed reminder level scenario; for symptom control drugs, the trigger threshold is 0.4 in the optimal and appropriate reminder level scenarios, and is increased to 0.6 in the cautious and delayed reminder level scenarios; for therapeutic adjuvant drugs, the trigger threshold is 0.5 in the optimal reminder level scenario, and is increased to 0.8 in the delayed reminder level scenario; for temporary medications, reminders are triggered only in the optimal and appropriate reminder level scenarios, with thresholds of 0.6 and 0.7, respectively; The dynamic hierarchical reminder strategy is used to allocate differentiated reminder resources according to the importance of drugs with different priorities: for life-sustaining drugs, multiple reminders are allowed until the user confirms, with a maximum of 5 reminders, and the interval time is shortened from 5 minutes to 1 minute; for symptom control drugs, a maximum of 3 reminders are issued, with an interval of 10 minutes; for treatment-adjuvant drugs, a maximum of 2 reminders are issued, with an interval of 15 minutes; for temporary medications, only 1 reminder is issued; at the same time, the reminder order is arranged according to the reminder priority of multiple drugs in the same time period to avoid triggering multiple reminders at the same time and causing trouble to the user; The intelligent decision-making layer also includes a user feedback learning module, which is used to record the user's response to each reminder and feedback satisfaction, and continuously optimize the decision model parameters through a supervised learning algorithm to improve the personalization of the reminder strategy; The user feedback learning module adopts an incremental learning approach to adapt to changes in user habits and maximize user experience while ensuring medication safety.

6. The intelligent drug management and reminder system based on multi-dimensional environmental perception according to claim 5 is characterized in that: The interactive reminder layer supports five basic reminder modes; The basic reminder modes include visual reminders, tactile reminders, auditory reminders, multimodal reminders, and non-intrusive reminders. The interactive reminder layer automatically selects the best reminder mode or a combination of multiple modes based on the user status classification, environmental condition parameters, and medication importance level provided by the intelligent decision-making layer. The visual reminder includes screen lighting, customized reminder interface display, and intelligent brightness adjustment mechanism. It automatically increases the screen brightness to the highest comfortable level in low-light environments, maintains a medium brightness level in standard lighting environments, and activates a high-contrast display mode in bright light environments. Under the visual reminder, a differentiated interface design is displayed according to the importance of the medication. For life-sustaining drugs, a striking red border and drug icon are used. For symptom control drugs, an orange border is used. For treatment support drugs, a blue border and a green border are used for temporary medications. The tactile reminder uses a precisely controlled vibration motor that supports a combination of four different intensities and six vibration modes: mild vibration intensity and brief vibration mode in silent environments; moderate vibration intensity and double-pulse vibration mode in general environments; high-intensity vibration and continuous wave mode in high-noise environments; for life-sustaining and symptom-control drugs, the vibration intensity and duration are gradually increased after the user does not respond to the initial reminder; and the optimal vibration parameters are automatically calibrated for different wristband materials and wearing tightness using feedback data from a three-axis accelerometer. The auditory reminder uses the smartwatch's built-in speaker to provide sound prompts, with an intelligent volume control function. In a quiet environment, the volume is used at a minimum of 30%, in a normal environment, the volume is used at 60%, and in a noisy environment, the volume is used at 90%. The auditory reminder is designed with unique reminder sound effects for different categories of medications. Life-sustaining drugs use three short, rapid tones, symptom-control drugs use two medium-rhythm tones, and treatment-aid and temporary medications use a single, gentle tone. When the ambient noise exceeds 80dB, the auditory reminder is automatically abandoned and switched to an enhanced tactile reminder. The basic reminder mode uses a combination of visual, tactile and auditory reminders, and activates screen display, moderate vibration and appropriate sound prompts according to the most appropriate reminder level situation determined by the intelligent decision-making layer. Different combination strategies are adopted for different reminder levels determined by the situational awareness module: in the most appropriate reminder level situation, all three reminder methods are activated simultaneously; in the appropriate reminder level situation, visual and tactile reminders are combined; in the general reminder level situation, tactile reminders are given priority in addition to visual reminders; in the cautious reminder level situation, visual reminders are given priority or the tactile reminder timing is delayed; The non-intrusive reminder mode is designed for special conditions, including: sleep state reminders use gradual mild vibrations, starting from the lowest intensity and slowly increasing to the awakening intensity, and such reminders are only triggered for life-sustaining drugs; driving state reminders use only brief voice prompts and limited visual prompts to avoid distracting driving attention; meeting state reminders use only a single mild vibration and a low-brightness screen prompt, and the screen content automatically switches to a simple mode, showing only the drug name and dosage, without detailed doctor's instructions; for social occasions, the system uses mild intermittent vibration prompts; The personalized reminder parameter optimization mechanism records the user's response behavior to each reminder, including four types: immediate response, delayed response, ignoring reminders, and closing reminders. Through machine learning algorithms, the user's preferences and response speed for different reminder methods are analyzed, and the selection probability and parameter settings of the reminder mode are continuously adjusted. At the same time, a user activity pattern map is established to identify the user's daily regular activities and prioritize reminders at natural interruptions in activity transition points. The interactive reminder layer has a built-in medication compliance recording and analysis module; the medication compliance recording and analysis module automatically records the issuance time of each medication reminder, the user response time and the actual medication confirmation time, calculates the key indicators of on-time medication rate, missed medication rate and delayed medication rate, and generates daily, weekly and monthly medication compliance reports; and evaluates the effectiveness of the reminder strategy based on the compliance data, and automatically strengthens the reminder strategy for drugs with low compliance; users view the graphical trend analysis report through the smart watch; the graphical trend analysis report includes a comparison of medication compliance in different time periods, a comparison of compliance rates of various drugs and the long-term change trend of compliance.

7. The intelligent drug management and reminder system based on multi-dimensional environmental perception according to claim 1 is characterized in that: It also includes a hierarchical processing mechanism; The hierarchical processing mechanism adopts a three-level computing architecture, including a low-power resident monitoring unit, a medium-power coprocessor and a high-performance main processor; the three-level computing architecture automatically distributes the load among the three-level processing units according to the complexity of the computing task; the low-power resident monitoring unit consumes less than 1mW of power, is responsible for collecting raw sensor data and performing simple threshold judgment, and triggers the coprocessor to wake up when a potential state change is detected; the medium-power coprocessor has a power consumption range of 10-50mW, is responsible for performing feature extraction, activity state classification and environmental parameter analysis, has a built-in lightweight neural network model to identify user activity status, and dynamically adjusts the coprocessor's operating frequency and wake-up interval according to the battery power; the high-performance main processor is only activated when it is necessary to perform complex decision analysis, user interaction, high-priority drug reminders or data analysis, and the average daily activation time is controlled within a cumulative 30 minutes.

8. The intelligent drug management and reminder system based on multi-dimensional environmental perception according to claim 1 is characterized in that: Also includes: Adaptive sampling strategy; The adaptive sampling strategy dynamically adjusts the sensor sampling frequency according to the user state stability, reducing the frequency when the state is stable and increasing the frequency when the state changes, and reduces the number of wake-up times for data batch processing.

9. The intelligent drug management and reminder system based on multi-dimensional environmental perception according to claim 1 is characterized in that: Also included: ambient energy harvesting technology; The environmental energy harvesting technology integrates three energy recovery mechanisms: a light energy conversion module, a thermal energy conversion module, and a micro-kinetic energy conversion module to extend battery life by more than 40%, and automatically adjusts functional configuration according to the battery power level: full function operation when the power level is >70%, reducing the sampling frequency of non-critical sensors when the power level is 30%-70%, retaining only the high-priority drug reminder function when the power level is <30%, and retaining only the basic reminder function for life-sustaining drugs when the power level is <10%.

10. The intelligent drug management and reminder system based on multi-dimensional environmental perception according to claim 1 is characterized in that: It also includes: a cloud server; the cloud server is used to process the transferred non-real-time medication compliance historical data and user behavior pattern analysis tasks, and update the optimized model parameters to the local during charging to achieve optimal allocation of computing tasks and energy consumption.

Images