A non-verbal / letter emotional interaction system, method, device and software

Abstract

The application discloses a system, method, device and software for non-verbal / textual emotional two-way interaction. The system comprises a wearable device in a screen-free form, a remote server and a care end application. The wearable device collects preset interaction operation to generate an active intention signal and collects physiological data. The server adopts a double-layer decision funnel model, infers a passive situation state by using the physiological data, performs semantic disambiguation in combination with the active intention, and generates high-situation information. The information is visually presented on the care end and feedback is given, which is pushed to the wearable device by the server to perform vibration or flicker response. The application solves the problem of semantic ambiguity by fusing the active intention and the passive situation, and realizes remote emotional communication with low use threshold and high situation understanding ability for communication-impaired people.

Description

Technical Field

[0001] This invention belongs to the fields of wearable smart devices, emotion computing and remote health monitoring. Specifically, it relates to a non-verbal / textual emotion interaction system, method, device and software based on the fusion of multimodal physiological signals and active intentions. Background Technology

[0002] With the accelerating aging of society and changes in family structure, separated and solitary elderly care has become a common phenomenon. For elderly people with aphasia, limited mobility, or difficulty using smart devices, effective emotional communication with distant relatives faces significant obstacles. They not only face physiological limitations in their language expression abilities but also often experience psychological barriers of "unwillingness / fear" to communicate due to concerns about disturbing their children's work or causing generational conflicts, which can easily lead to emotional isolation in the long run.

[0003] Existing technologies, such as patent CN115153545A, focus on emotional fluctuation alarms and lack an entry point for users to actively express themselves. Patent US20180103901A1 only serves as general monitoring data and does not address the semantic mapping gap between simple non-verbal / textual signals (such as a single click) and complex emotions (such as "I feel very sad" and "I miss you").

[0004] The current state of commercial solutions in the market is as follows: First, mainstream wearable devices, represented by smartwatches, have complex functions, small interfaces, and multiple interaction layers, creating a significant "digital divide" for elderly users with declining vision and fine motor skills, resulting in huge learning costs and cognitive burdens. Meanwhile, professional assistive communication devices or applications are often cumbersome to operate and are more suitable for medical rehabilitation scenarios than daily life. Second, products such as health bracelets or smart rings mainly serve as individual data recorders, collecting physiological data such as heart rate and blood oxygen to provide users with personal health analysis. Their function is introverted, lacking extroverted functions for emotional communication with family members. Third, there are also products for remote heart rate and blood sugar monitoring, but these are primarily functional. The caregiver's end only provides cold, impersonal physiological data, lacking two-way interaction and neglecting the daily emotional connection and companionship needs of the elderly. Furthermore, their strong medical attributes can easily bring psychological "stigma" to the wearer. Fourth, there are wearable devices that use vibration for remote emotional interaction. These devices can only transmit single, low-context tactile signals and cannot distinguish between distinct intentions such as "daily greetings" and "emergency help requests," resulting in ambiguous communication semantics and limited information value.

[0005] Therefore, how to provide a non-verbal / textual emotional interaction solution with low usage threshold, high contextual understanding ability, and de-medicalization, and establish an effective and burden-free emotional connection bridge between people with communication barriers and their families while ensuring remote security monitoring, is a technical problem that urgently needs to be solved in this field. Summary of the Invention

[0006] The main objective of this invention is to provide a non-verbal / text-based emotional interaction system, method, device, and software that solves problems such as high barriers to entry for smart devices, neglect of daily emotional needs, and semantic ambiguity in non-verbal / text-based interaction in the prior art.

[0007] To achieve the above objectives, the first aspect of the present invention provides a non-verbal / text-based two-way emotional interaction system, comprising: a wearable device worn by a first user, the wearable device comprising: an active intention input module for collecting preset interactive operations of the first user and generating an active intention signal; a multi-modal physiological sensor for collecting physiological data characterizing the physiological state of the first user in real time; a motion sensor for collecting motion data characterizing the motion state of the first user in real time; and a feedback module for expressing feedback information sent by a second user to the first user, via vibration or LED display.

[0008] The processor, electrically connected to the active intention input module and the multimodal physiological sensor, sends this data, along with motion sensor data, to a remote server. The remote server is configured to: infer the passive context of the first user based on the physiological data collected by the physiological sensor using an emotion computing algorithm; upon receiving the active intention signal, fuse the active intention signal with the passive context to generate high-context information, and push it to the caregiver application; the caregiver application, used by a second user, receives and presents the high-context information in a preset visualization manner, and receives feedback from the second user, which is then forwarded by the remote server to the wearable device.

[0009] To achieve the above objectives, a second aspect of the present invention provides a non-verbal / text-based two-way emotional interaction method, comprising: acquiring a preset interactive operation performed by a first user through an active intention input module of a wearable device, and generating an active intention signal; acquiring physiological data characterizing the physiological state of the first user in real time through multiple physiological sensors of the wearable device; and acquiring motion state data of the first user in real time through motion sensors. A processor sends all data to a remote server. Emotional computing software on the remote server, based on the physiological data, infers the passive context of the first user using an emotional computing algorithm; upon receiving the active intention signal, it fuses the active intention signal with the passive context to generate high-context information, which is then sent to a caregiver application for a second user; the high-context information is presented in a preset visualization manner on the caregiver application, and feedback from the second user is received through a preset interactive interface, which is then forwarded to the wearable device by the remote server.

[0010] The present invention also provides a non-verbal / textual emotion interaction device, including the aforementioned wearable device.

[0011] The present invention also provides embedded software for the aforementioned wearable device, emotion computing software deployed on a remote server, and caregiver software used by a second user, for executing the aforementioned method.

[0012] The present invention, through the above technical solution, has at least the following beneficial effects:

[0013] 1. Semantic Disambiguation Mechanism: This invention proposes for the first time an architecture of "passive context verifying active intent." By using physiological data as "context," simple interactive actions can convey rich emotional information. For example, the system can distinguish between "touch in a calm state" (interpreted as a daily greeting) and "touch in an anxious state" (interpreted as a request for attention), solving the pain point of semantic ambiguity in traditional non-verbal interaction devices. This greatly reduces the cognitive load on aphasic users and enables caregivers to more accurately understand the user's true state and needs.

[0014] 2. Motion entropy correction: Compared with the distributed multimodal detection of CN114795209A, this invention solves the interference of motion artifacts on the determination of psychological arousal by using motion entropy, which is more suitable for real home care scenarios.

[0015] 3. Enhance emotional connection: Provides a communication channel for separated family members with zero psychological burden and zero conflict. Users can convey emotions without worrying about disturbing each other or language barriers, while caregivers can also gain a sense of security from being online at all times without invading the privacy of their loved ones (video monitoring), effectively alleviating users' feelings of social isolation and strengthening intergenerational emotional bonds.

[0016] 4. HCI Design Philosophy: Employing a screenless design, users do not need to learn complex interface operations; they can express their intentions simply through touch and tapping. This reduces the learning and usage difficulty for older adults, especially those with fine motor skills impairments such as aphasia or Parkinson's disease, effectively bridging the digital divide.

[0017] 5. Increase acceptance of wearability: Wearable devices adopt an ornamental and de-medical appearance design, which reduces the stigma and sense of being monitored that wearers may feel due to wearing monitoring devices, making them more easily accepted and worn as everyday accessories in the long term.

[0018] 6. Integrated Comprehensive Care: Building upon core emotional interaction functions, it integrates reliable safety warning functions such as fall detection. With active long press or squeeze, it effectively eliminates missed and false alarms, providing a comprehensive remote care solution that ranges from daily emotional support to emergency response, offering more holistic care. Attached Figure Description

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

[0020] Figure 1 This is a structural block diagram of a non-verbal / text-based emotional interaction system provided in an embodiment of the present invention.

[0021] Figure 2 This is a flowchart of a non-verbal / text-based emotional interaction method provided in an embodiment of the present invention.

[0022] Figure 3 This is a detailed flowchart of the passive context reasoning algorithm in this embodiment of the invention.

[0023] In the diagram: 100-wearable device, 101-processor, 102-active intention input module, 103-multiple physiological sensors, 103a-PPG sensor, 103b-conductivity skin sensor, 103c-motion sensing unit, 104-feedback module, 105-power module, 200-remote server, 300-caregiver application. Detailed Implementation

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

[0025] Example 1

[0026] This embodiment provides a non-verbal / text-based emotional interaction system and method. The description of this embodiment will be combined with... Figures 1 to 3 conduct.

[0027] like Figure 1 As shown, the nonverbal / textual emotional interaction system of this embodiment includes a wearable device 100, a remote server 200, and a caregiver application 300. This system is used to establish a low-threshold, high-context nonverbal / textual emotional communication link between a first user (such as an elderly person wearing the wearable device 100) and a second user (such as a child using the caregiver application 300).

[0028] Wearable device 100 is the core data acquisition and preliminary processing terminal of the system. In this embodiment, its physical form can be designed as an ornamental item such as a bracelet, pendant, or brooch. For example, it can adopt a hinged bracelet design, which utilizes a spring hinge mechanism to facilitate easy one-handed wear by users with diminished fine motor skills. Simultaneously, its rigid structure and inward-curving shape provide constant centripetal pressure, ensuring a tight and stable fit between the sensors inside the device and the user's skin. This effectively suppresses motion artifacts caused by loosening, providing a high-quality data foundation for subsequent physiological signal analysis.

[0029] Specifically, the wearable device 100 integrates multiple functional modules:

[0030] Processor / Communication 101: As the control center of the wearable device 100, it is responsible for coordinating the work of various modules and performing data processing and algorithm calculations. In this embodiment, the processor 101 is preferably a microcontroller unit with low power consumption and built-in wireless communication capabilities, such as Espressif's ESP32-C3SuperMini. It provides sufficient computing power to acquire and preprocess multi-mode sensor data and motion data.

[0031] The processor 101 transmits all data to the remote server 200 via its built-in 2.4GHz Wi-Fi. In this embodiment, to achieve low-power, low-latency, and highly reliable real-time data transmission, the system uses the message-oriented lightweight Internet of Things (MQTT) protocol for communication. The wearable device 100, acting as an MQTT client, actively pushes data to the server, avoiding the latency and power consumption issues associated with traditional HTTP polling. The MQTT protocol also incorporates heartbeat keep-alive and will message mechanisms, ensuring that the server can immediately detect and notify the caregiver when the device unexpectedly goes offline, thus guaranteeing the resilience of the communication link.

[0032] Active intent input module 102: Used to collect the explicit interaction intent of the first user. To lower the operational threshold, this module adopts a screenless "single-byte interaction" method. In this embodiment, the active intent input module 102 is specifically a capacitive touch sensor embedded in the surface of the device casing. The first user can trigger the module to generate corresponding active intent signals with simple semantics through preset interactive operations, such as single click, double click, long press, squeeze, etc.

[0033] Multimode sensor 103: Used to collect objective data characterizing the physiological state of the first user in real time and without physical contact. This data forms the basis for subsequent reasoning about the "passive situation". In this embodiment, the physiological sensors include at least a photoplethysmography (PPG) sensor 103a, a nerve conductance sensor 103b, and a motion sensing measurement unit 103c. The PPG sensor 103a, such as the MAX30102 module, measures the blood flow volume pulsation of the wrist by emitting light of a specific wavelength and detecting changes in the intensity of the reflected light, thereby calculating indicators such as heart rate, heart rate variability, and blood oxygen saturation. Heart rate variability is a key indicator for assessing the balance of the autonomic nervous system and is closely related to emotional stress levels. The nerve conductance sensor 103b measures changes in skin conductivity caused by sweat gland activity (controlled by the sympathetic nervous system) by applying a weak, constant voltage to the skin surface. Skin conductance activity is a standard for measuring emotional arousal and can effectively reflect the user's state of tension, anxiety, or excitement. The motion measurement unit 103c, such as the MPU6050 module, integrates a three-axis accelerometer and a three-axis gyroscope. This sensor is used to monitor the body posture and motion state of the first user in real time. Its data is used to correct other physiological signals (such as removing motion artifacts in photoplethysmography signals) and to identify specific activities (such as stillness, walking) and special events (such as falls).

[0034] Power module 105: Provides power to the entire wearable device 100. This module includes a rechargeable lithium battery and a corresponding power management unit, which is responsible for battery charge and discharge management, voltage conversion, and low-power mode control.

[0035] Feedback module 104: Used to provide non-verbal tactile or visual feedback to the first user. In this embodiment, feedback module 104 includes multiple linear resonant motors and an addressable RGB LED. Compared with traditional eccentric rotor motors, linear resonant motors have a faster response speed and a crisper vibration, and can simulate tactile patterns with rich social emotional meanings such as "light grip," "stroking," and "heartbeat" through specific timing and intensity combinations. The RGB LED can be used to indicate device status or as a supplement to information feedback, and can also express different emotional meanings through display methods such as running lights and breathing lights.

[0036] Remote server 200: As the cloud hub of the system, it deploys the system's core intelligent algorithms. During the development phase of this embodiment, edge computing devices such as NVIDIA Jetson Orin can be used as server prototypes. Remote server 200 includes at least one MQTT broker responsible for receiving, filtering, and distributing all messages from wearable device 100. The core function of server 200 is to execute complex emotion computing algorithms, perform deep analysis and state reasoning on the received high-dimensional physiological data, and push the final "high-context information" to the caregiver application 300. In addition, server 200 also includes a database system for storing users' historical physiological data, personalized baseline models, and interaction event records. To protect user privacy, communication between the device and the server is encrypted using TLS / SSL throughout.

[0037] The caregiver application 300 serves as the second user interface to the system, typically an application installed on a smartphone or tablet. This application subscribes to messages from a remote server 200, receiving and presenting high-context information in real time. Its interface design follows the principle of "atmospheric comfort," replacing complex raw data charts with emotional and intuitive visualizations. For example, based on the content of the received high-context information, different colored or dynamic graphic icons (such as a bird icon, displaying different postures depending on the state, such as flying, resting, or agitated) are displayed on the interface, accompanied by brief natural language explanations (such as "Status stable, currently resting").

[0038] Based on the above system architecture, the non-verbal / text-based emotion interaction method in this embodiment is as follows: Figure 2 As shown, the specific steps include:

[0039] Step S201: Acquire active intent signal. When acquiring active intent signal, the active intent input module 102 detects a preset interactive operation (such as a click) performed on the wearable device 100, generates a corresponding active intent signal, and transmits it to the processor 101.

[0040] Step S202: Real-time acquisition of physiological / motor data. The multimodal physiological sensor 103 operates continuously: the photoplethysmography (PPG) sensor 103a acquires pulse wave data, the electrodermal transfer sensor 103b acquires electrodermal transfer data, and the motion sensing unit 103c acquires triaxial acceleration and angular velocity data. This raw physiological / motor data is sent to the processor 101 in real time. After preprocessing such as noise reduction, the processor sends it to the remote server 200 in MQTT protocol format.

[0041] In this embodiment, the processor 101 first preprocesses the acquired physiological data. For photoplethysmography (PPG) signal processing, to ensure the accuracy of heart rate and heart rate variability calculations, the system performs multi-layer optimization on the PPG signal. First, motion artifact correction is performed on the PPG signal using motion data acquired by the motion measurement unit 103c. This correction logic can be expressed by the following formula: ,in, It is the corrected signal. It is the filtered signal. It is a correction factor. It is the current total acceleration. This refers to gravitational acceleration, a step used to remove signal noise caused by body motion. Next, an adaptive moving average filter is used to further smooth the signal, the logic of which can be expressed as follows:

[0042] ,

[0043] in, It is the filtered value at the current moment. It is the original sampled value at the current moment. It uses adaptive weights that can be dynamically adjusted based on motion status. Furthermore, to balance power consumption and accuracy, the system employs a dynamic sampling rate method; for example, the sampling rate is reduced to 25Hz when the user is stationary and increased to 100Hz when motion is detected.

[0044] Step S203: Reasoning about Passive Contexts. This is one of the core steps of the method of this invention, executed by the emotion computing software on the remote server. The purpose of this step is to transform the raw, multi-dimensional physiological / motor data stream into an understandable, discrete "passive context" through emotion computing algorithms. The system employs a hybrid intelligent algorithm of "data-driven + physiological rule verification" to reason about passive contexts; the specific process can be found in [reference needed]. Figure 3 First, step S301, multidimensional feature extraction, is performed. The system extracts key features from the processed signal, including: calculating heart rate variability using photoplethysmography (PPG) data. Indicators (such as time-domain indicators) , ), using electroskin response data to calculate the electroskin response Indicators (such as the number of peak responses) are used to identify motion states and calculate motion entropy using motion sensing unit data. Then, step S302, emotion arousal determination, is performed, inputting the extracted multidimensional features into a pre-trained machine learning model. In this embodiment, a support vector machine model is used, which outputs a preliminary emotion arousal state. Next, step S303, combining motion context segmentation, is performed. The preliminary emotion arousal state determination result is combined with the motion entropy judgment to form a "two-layer decision funnel" to solve the representational ambiguity problem of difficulty in distinguishing between "psychological stress" and "physiological movement."

[0045] In the emotion arousal judgment, the WESAD dataset, short for Wearable Error Stress and Affect Detection, was used. It is one of the most authoritative and widely used public datasets in the field of stress monitoring and emotion recognition. The SVM model was pre-trained and the accuracy of high arousal state judgment was AUC > 96%.

[0046] In the segmentation of motion context, under the high arousal state, the motion sensor measurement unit data is further combined. If the user's body is basically still or the motion entropy is large, that is, the motion is chaotic, it is judged as "anxious state". If the motion entropy is small, it is judged as "active state". If the user is in the low arousal state, it is further distinguished as "resting state" or "stable state" according to the motion measurement unit data.

[0047] Finally, step S304, steady-state control, is executed. To prevent frequent state transitions, the output is smoothed. For example, a state requires `k_confirm` consecutive confirmations before being finally output, and there is a `cooldown_s` second cooldown period after each state transition. Through the above process, the system can map the user's physiological state to a preset set of passive scenarios, which includes at least: resting state, stable state, active state, and agitated state.

[0048] Step S204: Generating High-Context Information. When the remote server receives the active intent signal generated in step S201, it performs a fusion operation with the current passive context. This fusion is based on a preset fusion rule matrix, which defines the high-context information with specific semantics corresponding to different combinations of active intent signals and passive contexts. For example, the fusion rule matrix may include at least: mapping the combination of the "click" active intent signal and the "stable state" passive context to "daily greeting" high-context information; mapping the combination of the "click" active intent signal and the "anxious state" passive context to "request for attention" high-context information; and mapping the combination of the "long press" active intent signal and any passive context to "emergency help" high-context information. In this way, a simple "click" action is given completely different emotional connotations, thereby generating information-rich high-context information. The remote server pushes the current passive context and the fused high-context information to the corresponding caregiver application 300.

[0049] In a preferred embodiment, the system also supports personalization. A remote server is configured to build a personal physiological model for the first user, containing baseline values ​​for various physiological data, within a preset calibration period after the device is first activated. During subsequent inference of passive scenarios, real-time collected physiological data is compared with the baseline values ​​in this model, the deviation is calculated, and the mapping relationships in the fusion rule matrix are dynamically adjusted based on the deviation to achieve more accurate personalized scenario interpretation.

[0050] Step S205: Presenting high-context information. After receiving the high-context information, the caregiver application 300 presents it to the second user in a preset visualization. As mentioned earlier, this method is intuitive and emotional. For example, the icon representing the first user on the interface turns red and flashes, while a text prompt pops up: "He / She may be a little anxious right now and needs your attention."

[0051] In addition, this system integrates an important safety function—fall detection. The processor 101 continuously analyzes the acceleration and angular velocity data collected by the motion measurement unit 103c. Specifically, the system calculates the resultant acceleration and resultant angular velocity: , ,in, It is a triaxial acceleration. This refers to the three-axis angular velocity. When an impact characteristic satisfying a preset fall model is detected, for example, when... and At this time, the processor 101 will suspend or ignore other routine physiological data collection and active intention signals, directly generate "fall warning data" as high-context information, and send it to the remote server 200 for emergency alarm via the communication module.

[0052] In summary, this embodiment integrates the active intention of the first user with the passive context objectively reflected by physiological / motor signals to generate and transmit high-context information. This not only realizes the basic remote monitoring function, but more importantly, it builds a bridge between the two parties that can accurately convey emotional context, effectively solving many problems of the prior art.

[0053] Those skilled in the art will understand that the above embodiments are for illustrating the present invention and not for limiting the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A non-verbal / textual emotional two-way interactive system, characterized in that, include: 1) Wearable device 100, in the form of a bracelet, wristband, hat, etc., worn by a first user, wherein the wearable device includes: The active intent input module 102 is used to collect preset interactive operations such as non-language / text, screenless pressing, and stroking, and generate active intent signals. The multimodal sensing module (103) is used to collect physiological and behavioral characteristic data of the first user in real time, including heart rate variability (HRV), skin conductance (EDA), and motion state (IMU). The processor / communication module 101 performs local multimodal signal acquisition and sends it to the remote server 200 via the MQTT protocol; after receiving feedback information from the care application 300 forwarded by the remote server 200, it sends a corresponding command to the feedback module 104. Feedback module 104, upon receiving the command from 101, executes predefined non-verbal / textual feedback, such as various forms of vibration or LED flashing. 2) The remote server 200, as the AI ​​reasoning hub, completes the passive context reasoning and active intention semantic fusion: based on physiological and behavioral feature data, it identifies the current context state of the first user. When the active intention signal is received, it uses the context state to perform semantic disambiguation on the signal, generates high context information containing emotional context, and sends the current context state and high context information of the first user to the caregiver application 300 through the MQTT protocol. 3) The caregiver application 300 is used by a second user to receive and present the high-context information in a preset visual manner, and to provide a corresponding non-verbal / text feedback human-machine interface. The non-verbal / text feedback information is sent to the remote server 200 via the MQTT protocol.

2. The system according to claim 1, characterized in that, The remote server 200, acting as the AI ​​inference hub, employs a "two-layer decision funnel" model: 1) First layer: Using AI models such as Support Vector Machine (SVM) to make a preliminary judgment on whether physiological characteristics have entered a high arousal state; 2) Second layer: Introducing motion entropy features based on IMU signals to distinguish between arousal caused by physiological movement and arousal caused by psychological stress: If the EDA is highly awake and the IMU motion entropy shows that it is stationary or struggling chaotically, it is judged as agitation. If the EDA is highly awake but the IMU shows regular motion, it is considered to be active (Vitality). If the EDA does not have high wake-up, it is determined to be either resting or stable based on the IMU motion.

3. The system according to claim 1, characterized in that, When the remote server performs the fusion, it uses a preset fusion rule matrix to define different combinations of the active intent signal and the passive context, including at least: 1) Integrate the "single touch signal" with the "anxiety" context to generate a "request for attention" message; 2) Integrate "single touch signal" with other passive contexts to generate "daily greeting" messages; 3) Regardless of the passive situation, when a "long press / grip signal" or "fall impact signal" is detected, an "emergency help / warning" message will be generated first.

4. The system according to claim 1, characterized in that, The remote server is also configured to: establish a personal physiological model for the first user, including baseline values ​​of various physiological data, within a preset calibration period after the wearable device is first activated; compare the real-time collected physiological data with the baseline values ​​in the personal physiological model and calculate the deviation when reasoning about the passive situation; and dynamically adjust the mapping relationship in the fusion rule matrix based on the deviation to achieve personalized situation interpretation.

5. The system according to claim 1, characterized in that, The processor is also configured to: continuously analyze the acceleration data and angular velocity data collected by the motion sensing unit; when an impact characteristic that meets the preset fall model is detected, it will suspend or ignore other current routine physiological data collection and active intention signals, directly generate fall warning data as the high-situation information, and send it to the remote server 200 for emergency alarm via the communication module.

6. The wearable device according to claim 1, characterized in that, The bracelet / wristband is designed to be worn in a hinged bracelet shape, using a spring hinge mechanism to provide constant centripetal pressure. This centripetal pressure is used to ensure that the inner sensor array fits tightly against the skin, physically suppressing motion artifacts and improving the signal-to-noise ratio of physiological signals.

7. The system according to claim 1, characterized in that, The active intent input module includes a capacitive touch sensor and a pressure sensor, and the preset interactive operations include, but are not limited to, single click, double click, long press, and squeeze.

8. The system according to claim 1, characterized in that, The wearable device also includes a feedback module (104) that uses a linear resonant motor (LRA) array to simulate social touch through haptic programming. Light grip mode: Multi-motor synchronous enhancement simulates deep pressure tactile (DPT) to deliver a sense of security; Touch mode: The motors are activated sequentially to simulate a flowing tactile sensation that delivers comfort and care; Using LED arrays, emotional expression is achieved through visual choreography: Marquee mode: The LED array flashes rapidly in sequence to convey a signal that someone is confirming a request for help, that they are on their way, or that appropriate measures have been arranged. Breathing light mode: The LED array flashes slowly in sync, simulating comfort and care.

9. The system according to claim 1, characterized in that, The communication module uses the message-oriented lightweight IoT message transmission protocol MQTT for data transmission and utilizes its "will message" mechanism to monitor the online status of the device in real time, ensuring the resilience of the communication link for remote care.

10. The system according to claim 1, characterized in that, The interface of the caregiver application includes: inferred passive contextual information, high-contextual information after semantic fusion of active intent signals, and graphical icons with different colors or dynamic effects displayed on the interface according to the information category, as well as corresponding text prompts. The interface also includes predefined feedback buttons. For "daily greetings", the feedback button is "heart" / "daily care", and the second user can also take the initiative to initiate daily greetings; For the "Request Attention" request, the feedback button is a wave / I'm coming, allowing the second user to delay taking their next action. In response to "emergency help," in addition to pressing the button to wave / come (I'm coming) and receiving confirmation from the first user, preset emergency measures will be invoked, such as dialing the corresponding caregiver or ambulance number.

11. The system according to claim 1, characterized in that, The wearable device adopts a bracelet-like, jewelry-like design.

12. The system according to claim 1, characterized in that, The sentiment computing algorithm is a pre-trained machine learning model, which includes a support vector machine model, a random forest model, or a deep neural network model.

13. A non-verbal / textual two-way emotional interaction method, characterized in that, include: The processor of the wearable device collects preset interactive operations performed by the first user through the active intent input module of the wearable device and generates an active intent signal; Physiological data characterizing the physiological state of the first user are collected in real time through the multi-mode physiological sensors of the wearable device. And by collecting acceleration and angular velocity data through motion sensing units; The above data is sent to the remote server through the communication module of the wearable device; After receiving feedback messages from the second user forwarded by the remote server, the processor of the wearable device drives the feedback module to vibrate the motor and flash the LED. The remote server, based on the physiological data, infers the passive context of the first user through an emotion computing algorithm; upon receiving the active intention signal, it fuses the active intention signal with the passive context to generate high-context information. The high-context information is then sent to the second user's care application. The high-context information is presented in a preset visualization format on the caregiving application. The caregiver application presents the high-context information in a preset visual manner while providing a non-verbal / text feedback interface. After the second user clicks, a predefined feedback signal is sent to the remote server, which then forwards it to the processor of the wearable device.

14. The method according to claim 13, characterized in that, The steps for collecting physiological data include: collecting pulse wave data through a photoplethysmography (PPG) sensor, collecting skin conductance data through a skin conductance sensor, and collecting acceleration and angular velocity data through a telekinetic sensing unit.

15. The method according to claim 13, wherein the step of reasoning out the passive situation specifically includes: Calculate the heart rate variability of the pulse wave data Indicators; calculating the skin conductance response of the stated skin conductance data. Indicators; identifying the motion state corresponding to the acceleration and angular velocity data; and the heart rate variability. Indicators, the aforementioned skin electrical response The indicators and the motion state are used as multi-dimensional inputs and mapped to a specific situation in a preset set of passive situations through the emotion computing algorithm. The set of passive situations includes at least resting state, stable state, active state, and anxious state.

16. The method according to claim 13, characterized in that, In the step of reasoning out the passive situation, the following logic is executed: First, AI models such as Support Vector Machine (SVM) are used to make a preliminary judgment on whether physiological characteristics have entered a high arousal state; based on IMU signals, motion entropy is calculated to further distinguish between arousal caused by physiological movement and arousal caused by psychological stress. If the EDA is highly awake and the IMU motion entropy shows that it is stationary or struggling chaotically, it is judged as agitation. If the EDA is highly awake but the IMU shows regular motion, it is considered to be active (Vitality). If the EDA does not have high wake-up, it is determined to be either resting or stable based on the IMU motion.

17. The method according to claim 13, characterized in that, The fusion step is based on a preset fusion rule matrix, which defines the high-context information of specific semantics corresponding to different combinations of the active intent signal and the passive context. The fusion steps include at least: when the active intent signal is a click operation and the passive context is a stable state, generating "daily greeting" high-context information; When the active intent signal is a click operation and the passive situation is an anxious state, a "request attention" high-context information is generated; when the active intent signal is a long press operation, regardless of the state of the passive situation, an "emergency help" high-context information is generated.

18. The method according to claim 15, characterized in that, The method further includes: establishing a personal physiological model containing the baseline values ​​of various physiological data of the first user within a preset calibration period after the wearable device is first activated; when performing the step of inferring the passive situation in the subsequent execution, first calculating the deviation of the physiological data collected in real time relative to the baseline values ​​in the personal physiological model; and dynamically adjusting the mapping relationship in the fusion rule matrix based on the deviation to achieve personalized situation interpretation.

19. The method according to claim 16, characterized in that, The method further includes: continuously analyzing the acceleration data and angular velocity data collected by the motion sensing unit; when an impact characteristic that meets the preset fall model is detected, the method will suspend or ignore other current routine physiological data collection and active intention signals, directly generate fall warning data as the high-situation information, and prioritize the step of sending the high-situation information to the remote server for emergency alarm.

20. The method according to claim 13, characterized in that, In the step of collecting preset interactive operations, the preset interactive operations include single click, double click, long press and squeeze.

21. The method according to claim 13, characterized in that, The system employs the message-oriented, lightweight IoT message transmission protocol MQTT.

22. The method according to claim 13, characterized in that, The interface of the caregiver application includes: inferred passive contextual information, high-contextual information after semantic fusion of active intent signals, and graphical icons with different colors or dynamic effects displayed on the interface according to the information category, as well as corresponding text prompts. The interface also includes predefined feedback buttons. For "daily greetings", the feedback button is "heart" / "daily care", and the second user can also take the initiative to initiate daily greetings; For the "Request Attention" request, the feedback button is a wave / I'm coming, allowing the second user to delay taking their next action. In response to "emergency help," in addition to pressing the button to wave / come (I'm coming) and receiving confirmation from the first user, preset emergency measures will be invoked, such as dialing the corresponding caregiver or ambulance number.

23. The method according to claim 13, characterized in that, The sentiment computing algorithm is a pre-trained machine learning model, which includes a support vector machine model, a random forest model, or a deep neural network model.

24. A non-verbal / text-based emotional interaction device, for first-time user use, characterized in that... Includes the wearable device as described in claim 1.

25. A software product comprising embedded software in a non-verbal / text-based interactive device used by a first user, emotion computing software on a remote server, and a second user care application software, the three forming a system that performs the method according to any one of claims 12 to 21.

Images