Multimodal data-driven hanging neck type negative ion health care and AI translation adaptive regulation method

Abstract

The present application relates to the technical field of air ions, in particular to a multi-modal data-driven hanging neck type negative ion health care and AI translation adaptive regulation method, comprising the following steps: S1, driving the carbon brush head negative ion generation module to generate a negative ion flow, and adjusting the discharge frequency and intensity according to the control instruction; S2, collecting user voice, performing voice recognition and semantic processing, and generating corresponding target language translation results; S3, obtaining the user's heart rate, skin electricity or other physiological parameters through the physiological signal detection module to represent the user's state; S4, scene recognition is performed based on user voice features, physiological parameters, motion information or environmental parameters, and the negative ion output mode and the translation output mode are adjusted according to the recognition results, the present application can automatically select a local or cloud translation model according to different scenes, thereby significantly outperforming the prior art in stability, functional integration, comfort and intelligence, and having cross-border communication and health care application value.

Description

Technical Field

[0001] This invention relates to the field of air ion technology, specifically to a multimodal data-driven neckband-style negative ion therapy and AI translation adaptive control method. Background Technology

[0002] With the development of wearable devices, health monitoring, air purification, and intelligent voice interaction functions are gradually evolving towards lightweight and portable designs. Among these, negative ion health and wellness devices, which generate high concentrations of negative ions through air ionization to improve the respiratory environment, regulate user emotions, and alleviate fatigue, are widely used in homes, offices, and personal settings. For example, existing patents CN203609688U, CN203399273U, and CN110986338B all employ high-voltage discharge structures to generate negative ions for air purification or health and wellness effects. However, these devices are often single-function and typically lack dynamic linkage capabilities with user behavior, environmental changes, and physiological states. Furthermore, the intensity of negative ion release is difficult to adaptively adjust according to the scenario, easily leading to increased energy consumption or unstable performance.

[0003] On the other hand, the demand for cross-language communication has rapidly increased with the rise of cross-border business, tourism, and international cooperation, spurring the development of various AI-powered simultaneous translation devices. For example, the intelligent translation terminal described in patent number CN109036451A primarily achieves language conversion through speech recognition, machine translation, and speech synthesis, meeting the needs of real-time multilingual interaction. However, traditional translation devices are usually independent of health and wellness functions, unable to simultaneously perceive the user's psychological and physiological state while worn, nor can they provide emotional relief or health assistance during communication. Summary of the Invention

[0004] The purpose of this invention is to solve the problems in the existing technology that lack the ability to dynamically link with user behavior, environmental changes and physiological state, and that the intensity of negative ion release is difficult to adaptively adjust according to the scenario, which easily leads to increased energy consumption or unstable use effect. It is also impossible to sense the user's psychological and physiological state while wearing the device, and it is impossible to provide emotional relief or health care assistance during communication.

[0005] To achieve the above objectives, the present invention provides the following technical solution: This invention provides a multimodal data-driven adaptive control method for neck-mounted negative ion therapy and AI translation, comprising the following steps: S1. Drive the carbon brush head negative ion generating module to generate a negative ion flow, and adjust the discharge frequency and intensity according to the control command; S2. Collect user voice, perform speech recognition and semantic processing, and generate corresponding target language translation results; S3. Obtain the user's heart rate, skin conductance, or other physiological parameters through the physiological signal detection module to characterize the user's state; S4. Based on user voice features, physiological parameters, motion information or environmental parameters, perform scene recognition, and adjust the negative ion output mode and translation output mode simultaneously according to the recognition results.

[0006] Furthermore, in step S1, the negative ion generating module includes a carbon brush head composed of multiple bundles of carbon fiber microfilaments, a flexible conductive base, a micro high-voltage boost circuit, and a directional airflow mechanism; the micro high-voltage boost circuit takes DC5V as input and outputs a pulsed high voltage of 2.5–4.5 kV, and the discharge frequency and duty cycle of the micro high-voltage boost circuit are dynamically adjusted by the main control unit according to the scene tags generated by the scene recognition engine, ambient air quality parameters, user physiological state, and audio acquisition status.

[0007] The beneficial effects of the above-mentioned further solutions are as follows: by adopting a structural combination of carbon brush head, flexible base and micro boost circuit, a stable negative ion flow can be generated under low power consumption conditions; and by dynamically adjusting the discharge frequency and duty cycle, the negative ion output can respond in real time to changes in scene, physiological state and audio acquisition, thereby improving the accuracy and safety of negative ion release and reducing the risk of ozone generation.

[0008] Furthermore, in step S1, the carbon brush head achieves ±15° airflow direction adjustment via an adjustable rotating shaft, and enters a discharge pause mode or a low duty cycle mode when the audio acquisition state is in the acquisition phase.

[0009] The beneficial effects of the above-mentioned further solutions are as follows: by setting an adjustable air guide shaft of ±15°, the negative ion flow can be made to fit the user's breathing zone more closely, thereby improving the efficiency of negative ion inhalation; automatically entering pause or low duty cycle mode during audio acquisition can reduce electric field interference, ensure the quality of voice acquisition, and at the same time avoid discomfort caused by noise, thereby improving the overall stability of use.

[0010] Furthermore, in step S2, the collected user speech is subjected to noise reduction processing and speech endpoint detection; speech recognition is performed using an acoustic model and a language model based on a deep semantic recognition model.

[0011] The beneficial effects of the above-mentioned further solutions are as follows: by performing noise reduction and endpoint detection on the speech, the accuracy and stability of speech recognition can be significantly improved; the acoustic model and language model using deep semantic recognition model can maintain a high recognition rate in noisy environments, providing more reliable input for subsequent translation generation and enhancing the translation and interaction experience.

[0012] Furthermore, in step S2, a multilingual machine translation model is invoked to perform cross-language conversion on the identified text; and the corresponding translation tone and output mode are selected based on the user's voice emotion characteristics, which are inferred from acoustic prosodic parameters.

[0013] The beneficial effects of the above-mentioned further solutions are as follows: cross-language conversion can be achieved through multilingual machine translation models, enabling users to conduct real-time cross-language communication; inferring user emotional characteristics based on acoustic prosody and adjusting the translation tone can improve the naturalness and emotional matching of the translation, making the translation results more in line with the user's expressive intent and enhancing the friendliness and authenticity of human-computer interaction.

[0014] Furthermore, in step S3, the user's heart rate signal, skin conductance signal and other physiological parameters are collected by the physiological signal detection module that is in close contact with the user's neck skin; the collected signals are filtered and feature extracted and then input into the physiological state inference model to infer and output the user's physiological state characteristics.

[0015] The beneficial effects of the above-mentioned further solutions are as follows: by acquiring physiological parameters such as heart rate and skin conductance through a sensing structure that is in close contact with the neck skin, signal stability can be guaranteed; after filtering and feature extraction, the input into the physiological state inference model can obtain a more accurate user physiological state, providing a reliable data foundation for negative ion output and scene recognition, and improving the targeting of health and wellness regulation.

[0016] Furthermore, in step S3, the user's physiological state characteristics include stress index, mood level, and fatigue level, and these physiological state characteristics are transmitted in real time to the scene recognition engine and negative ion control module for coordinated adjustment of negative ion output and translation tone.

[0017] The beneficial effects of the above-mentioned further solutions are as follows: by reasoning, the stress index, mood level and fatigue level can be obtained, which can comprehensively present the user's current physiological state; by transmitting this state to the scene recognition engine and negative ion control module in real time, the system can adjust the negative ions and translation output in real time according to the user's stress and mood changes, making health care and interaction more personalized and adaptive.

[0018] Furthermore, in step S4, based on user voice features, motion information, environmental parameters, and physiological state features, a scene recognition model is used to classify the current usage scenario and generate scene labels.

[0019] The beneficial effects of the above-mentioned further solutions are as follows: by multimodal fusion of user voice features, motion information, environmental parameters and physiological state, the accuracy of scene recognition can be effectively improved; by using scene recognition models to generate scene labels, the device can more quickly identify the current usage context, thereby providing a basis for subsequent negative ion and translation scheduling, and improving the overall system intelligence level.

[0020] Furthermore, in step S4, the output mode of the negative ion module, the air guide angle, and the output mode of the voice translation module are synchronously adjusted according to the scene labels. The adjustment of the negative ion module achieves adaptive and coordinated control of negative ion health care and voice translation, and automatically selects the local or cloud translation model in different scenarios. The beneficial effects of the above-mentioned further solutions are as follows: by automatically selecting local or cloud translation models in different scenarios, basic translation capabilities can be guaranteed when the network is limited, and higher quality translations can be obtained when the network is good, thereby enhancing the robustness of the system; at the same time, the synergistic regulation of negative ion health care and voice translation is realized, enabling the system to achieve dual improvement in health management and cross-language communication, and giving full play to the overall advantages of multimodal adaptive regulation.

[0021] Beneficial effects Compared with known public technologies, the technical solution provided by this invention has the following beneficial effects: By deeply integrating carbon brush head negative ion generation technology, AI voice recognition and multilingual translation, physiological signal detection, and multimodal scene recognition and control, a neck-worn, adaptive control system integrating negative ion health care and translation has been constructed. Compared with existing stand-alone negative ion devices and stand-alone translation devices, this invention can simultaneously acquire user voice features, physiological parameters, environmental data, and motion information, and generate scene tags based on a scene recognition model to achieve synchronous dynamic control of negative ion output mode, airflow direction, voice translation tone, and output method. The carbon brush head, composed of multiple carbon fiber bundles, and the pulsed high-voltage boost circuit can generate a stable, low-ozone negative ion flow under low power consumption conditions, and can switch to pause or low duty cycle mode according to the audio acquisition status, effectively reducing electromagnetic and acoustic interference. Through noise reduction, endpoint detection, and deep semantic recognition of user voice, high-precision voice recognition results can be obtained; combined with multilingual translation models and emotional prosody analysis, more natural and emotionally resonant translation output can be generated. The physiological signal detection module, closely attached to the neck skin, acquires physiological data such as heart rate and skin conductance. After processing by a state inference model, this data reflects the user's stress index, mood level, and fatigue level in real time, providing a reliable basis for scene recognition and negative ion regulation. This invention achieves synergistic enhancement of translation and health-care functions, enabling the device to simultaneously soothe emotions and regulate health during cross-language communication. It can automatically select local or cloud-based translation models according to different scenarios, thus significantly outperforming existing technologies in terms of stability, functional integration, comfort, and intelligent performance. It has broad application value in cross-border communication, fatigue relief, and health care. Attached Figure Description

[0022] Figure 1 This is a system control logic block diagram of the present invention; Figure 2This is a flowchart of the AI ​​semantic health regulation algorithm of the present invention; Figure 3 This is a diagram of the negative ion release air guide structure of the carbon brush head of the present invention; Figure 4 This is a flowchart of the method of the present invention. Detailed Implementation

[0023] To enable those skilled in the art to better understand the present invention, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. 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 should fall within the scope of protection of the present invention.

[0024] It should be noted that the terms "first," "second," etc., in the specification, claims, and accompanying drawings of this invention are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of the invention described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover a non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but includes other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0025] The present invention will now be described in further detail with reference to the accompanying drawings: Example: like Figure 1-4 As shown, S1. Drive the carbon brush head negative ion generating module to generate a negative ion flow, and adjust the discharge frequency and intensity according to the control command; Furthermore, in step S1, the negative ion generating module includes a carbon brush head composed of multiple bundles of carbon fiber microfilaments, a flexible conductive base, a micro high-voltage boost circuit, and a directional airflow mechanism; the micro high-voltage boost circuit takes DC5V as input and outputs a pulsed high voltage of 2.5–4.5 kV. The discharge frequency and duty cycle of the micro high-voltage boost circuit are dynamically adjusted by the main control unit according to the scene tags generated by the scene recognition engine, ambient air quality parameters, user physiological state, and audio acquisition status. The carbon brush head achieves ±15° airflow direction adjustment through an adjustable rotating shaft, and enters a discharge pause mode or a low duty cycle mode when the audio acquisition status is in the process of acquisition. Specifically: S1.1: Composed of several bundles of carbon fiber microfilaments (each bundle contains approximately 100–200 microfilaments, with a total number of microfilaments preferably 500–1200), with a multi-point micro-corona discharge array formed at the ends.

[0026] Flexible conductive base: provides electrical connection and mechanical buffer for carbon fiber.

[0027] Miniature high-voltage boost circuit: Input DC5V, output programmable pulse high voltage (preferred range 2.5–4.5 kV), using PWM pulse drive to control the discharge duty cycle.

[0028] Directional airflow mechanism: air guide vanes and adjustable rotating shaft, with a rotation angle range of ±15°, used to direct the ionized airflow to the user's breathing zone.

[0029] Safety and auxiliary sensing: ozone sensor, temperature sensor, IMU (head angle) and audio acquisition status signals.

[0030] S1.2: Scene recognition and provided input: Output: Scene tags (e.g., during a call, resting quietly, in noisy outdoor environments, user anxiety); Confidence score (0–1); Emotional level and physiological stress index (numerical features used to quantify whether a user is in a state of stress or relaxation).

[0031] S1.3: The master's reasoning and decision-making: The main controller receives scene tags and confidence levels; if the confidence level is lower than the threshold, a conservative strategy is adopted (maintaining a low duty cycle / low output voltage). Mapping to baseline discharge levels based on scenario category (e.g., noisy outdoor → high baseline; quiet rest → low baseline; user anxiety → soothing baseline). At baseline, adjustments are made based on physiological stress index (when physiological indicators show high stress, ion density is moderately increased; when they show low or normal, it is maintained or decreased), while being subject to ozone and temperature rise safety constraints. If the voice module reports "Acquiring audio", the main controller will prioritize the time-division multiplexing strategy, pausing or reducing the duty cycle to the lowest priority value. The main controller adjusts the airflow angle based on the head / neck angle information provided by the IMU to ensure that the ion airflow covers the mouth and nose area; when the user is moving vigorously, safety and stable output are prioritized while reducing the target coverage accuracy.

[0032] S1.4: Scene label → Baseline output level (low / medium / high), the baseline level corresponds to the preferred voltage and duty cycle range (e.g., low: 2.5 kV, duty cycle 10–30%; medium: 3.5 kV, 30–60%; high: 4.2 kV, 60–80%). S1.5: When the physiological stress index inferred by the model is higher than the preset threshold and the emotion label shows anxiety / tension, switch from the baseline to "relaxation mode", which is manifested by: increasing ion density (within the safety limit) and increasing the discharge frequency or moderately increasing the duty cycle, while selecting higher frequency pulses to suppress ozone generation.

[0033] S1.6: If audio acquisition is true, immediately adopt the "audio priority" strategy: reduce the duty cycle to the minimum or pause the discharge; after a single translation session ends, the scene engine determines whether to resume and selects the resumption parameters; S1.7: Safety Constraints: Under any circumstances, if the temperature of the ozone sensor or boost module exceeds the safety threshold, the output will be immediately degraded or the power will be cut off, and a warning will be issued to the user. S1.8: Implementation Process: a. After the device is powered on, the carbon brush head and boost circuit enter a self-test. The ozone and temperature sensors confirm that the levels are within safe ranges, and the carbon brush head enters a low duty cycle standby output. b. S2, S3, environmental sensors, and IMU continuously collect data and send it to the scene recognition engine in real time. c. The scene recognition engine infers from audio features, physiological features, motion, and environmental data, and outputs scene labels, emotion levels, physiological stress indices, and corresponding confidence scores. d. Based on the inference results and a preset mapping table, the main controller determines the discharge baseline and correction amount, and assesses whether an audio priority or safety degradation strategy needs to be implemented. E. The main controller sends control commands (adjusting voltage, frequency, and duty cycle) to the boost module and drives the air guide mechanism to adjust its angle. Discharge parameters are implemented in the circuit using PWM pulses. F. Continuously monitor ozone, temperature, and output stability; if an anomaly is detected, immediately downgrade or shut down the system and send the data back to the scene engine to adjust subsequent strategies. G. If the scene label changes or the audio acquisition state switches, return to steps C–E and loop to execute new reasoning and decisions; S1.9: Measure the ion output curve and ozone level for each carbon brush head, record the optimal operating range, and write it into the device parameter storage area. The device can perform short-term adaptive calibration during initial use or regular maintenance. By testing different duty cycle / voltage combinations in a safe environment and selecting the optimal parameters based on feedback from the ozone sensor and ion meter, the circuit has overcurrent, short circuit, overtemperature, and abnormal detection capabilities. Automatic degradation or shutdown is triggered by ozone threshold.

[0034] S1.10: Co-occurrence with S2–S4: During translation sessions: S1 automatically reduces or pauses discharge to eliminate noise interference and ensures accurate speech recognition.

[0035] When the user experiences emotional or physiological abnormalities: S1 executes a soothing mode, increases the negative ion density, and prompts the user with a gentle voice (S2 output).

[0036] When driven by scene recognition: S1 adjusts the discharge parameters and airflow direction within seconds based on scene tags to achieve "scene-adaptive" collaborative control of health care and translation.

[0037] In this embodiment, the main control unit serves as the adjustment center, receiving input signals from the scene recognition engine (S4), the voice recognition module (S2), the physiological signal detection module (S3), and the environmental sensor. Through reasoning and rule-based decision-making, it outputs control commands for the carbon brush head discharge parameters (voltage, pulse frequency, duty cycle) and the air guide angle, thereby achieving uniform, stable, and safe real-time adjustment of the negative ion flow.

[0038] S2. Collect user voice, perform speech recognition and semantic processing, and generate corresponding target language translation results; Further, in step S2, the collected user speech is subjected to noise reduction processing and speech endpoint detection; speech recognition is performed using an acoustic model and a language model based on a deep semantic recognition model; a multilingual machine translation model is called to perform cross-language conversion on the recognized text; and the corresponding translation tone and output mode are selected according to the user's speech emotion characteristics, which are inferred from the acoustic prosodic parameters. Specifically: S2.1: The acquired voice signal will undergo the following preprocessing steps: Endpoint detection (VAD): Used to determine whether a user is speaking; Noise suppression: Spectral subtraction or lightweight neural noise reduction models are used to adapt to wind noise and environmental noise in the wearing scenario; Feature extraction: Generate acoustic features of 40 dimensions or more (such as MFCC or Mel features); The preprocessed speech features are then fed into the backend recognition model. S2.2: A lightweight edge-side deep semantic recognition model is adopted, and the following process is inferred based on the Transformer structure: Perform temporal convolution, attention encoding, or gating unit operations on the input features to obtain the acoustic hidden vector; Using an end-to-end sequence model, the text sequence corresponding to the speech is output based on the acoustic hidden vector.

[0039] S2.3: During the speech emotion recognition process, prosodic features of the speech are extracted simultaneously, including: fundamental frequency variation, speech rate, energy curve, and intonation fluctuation. Based on the prosodic parameters, the emotion inference model determines whether the user is in a state such as "calm, tense, tired, or anxious." When the device is under high computational pressure or when inputting complex sentences, it automatically switches to cloud mode (in conjunction with S4). The translation model can perform bidirectional conversion for the following languages: Chinese. English, Chinese Korean and Chinese Japanese, English Korean; during the translation process, the semantic structure, context, and linguistic elements are reorganized to ensure a natural and fluent output.

[0040] S2.3: After the translated text is generated, different translation tones are selected based on the user's voice emotion characteristics, physiological signal parameters, and scene tags, such as: Soothing mode: Soft tone and slowed speech rate, used for anxious or high-pressure emotional situations; Standard mode: Normal speaking speed and tone; Emphasis mode: Suitable for noisy environments or situations where dialogue needs to be emphasized; Finally, the speech output module generates audio in the target language.

[0041] S2.4: It should be noted that if voice input is detected, the negative ion generation module will pause discharging to avoid interfering with the sound acquisition; the emotion analysis results will be used as input for S4 along with physiological signals (heart rate, skin conductance); different translation models or different output tones can be switched according to scene labels.

[0042] In this embodiment, the AI ​​speech recognition and translation module of the neckband terminal is used to receive user voice input, perform speech recognition, emotion analysis, and multilingual translation, and generate corresponding target language output. The module's operation process includes four stages: speech signal acquisition, acoustic preprocessing, semantic recognition, and model inference.

[0043] S3. Obtain the user's heart rate, skin conductance, or other physiological parameters through the physiological signal detection module to characterize the user's state.

[0044] Furthermore, in step S3, the user's heart rate signal, skin conductance signal, and other physiological parameters are collected through a physiological signal detection module that is in close contact with the user's neck skin. The collected signals are filtered and feature extracted and then input into the physiological state inference model to infer and output the user's physiological state characteristics, including stress index, mood level, and fatigue level. The physiological state characteristics are then transmitted in real time to the scene recognition engine and negative ion control module for coordinated adjustment of negative ion output and translation tone. Specifically: S3.1: The structure of physiological signal acquisition is as follows: Wearing method: The physiological module is integrated into the neck strap, which is in close contact with the skin of the neck to ensure the stability of signal acquisition.

[0045] Sensor type: Heart rate sensor (PPG): Detects changes in blood flow using photoplethysmography to obtain heart rate and heart rate variability (HRV).

[0046] Gestational skin conductance (GSR) sensor: detects changes in skin conductance to reflect autonomic nervous activity and indirectly reflect stress and mood levels.

[0047] Optional additional sensors: temperature sensor, accelerometer (for motion state recognition and signal denoising).

[0048] Hardware features: High-sensitivity miniature optical sensor with low power consumption design (<10mW). Signal sampling frequency: heart rate 100–200 Hz, skin conductance 10–50 Hz; Data interface: Bluetooth / Wi-Fi connection to the main control MCU to achieve real-time data transmission.

[0049] S3.2: Signal Preprocessing and Feature Extraction Filtering: A bandpass filter (0.5–5 Hz) is used to remove motion artifacts and ambient light interference from the heart rate signal; a low-pass filter is used to remove high-frequency noise from the skin conductance signal. Feature extraction: Heart rate: instantaneous heart rate, heart rate variability (RMSSD, SDNN); Skin conductance: mean skin conductance, peak count, skin conductance activity amplitude; Optional motion features: neck acceleration changes, used to eliminate motion artifacts.

[0050] S3.3: Physiological State Inference Model: Input: Heart rate characteristics, skin conductance characteristics, exercise status; Output: User's physiological characteristics, including: stress index (0-100 points, the higher the value, the greater the stress), mood level (calm, mild tension, anxiety), and fatigue level (low, medium, high).

[0051] Reasoning logic: A comprehensive score is given for heart rate variability and skin conductance peak count; motion artifacts are removed by combining motion state → a continuous stress index and a discrete emotion level are output → the results are sent to the scene recognition engine (S4) and the negative ion control module (S1) for adaptive adjustment.

[0052] In this embodiment, the step-by-step implementation process is as follows: 1. Initialization: After the device is powered on, the physiological module performs a self-test to confirm that the sensor functions are normal.

[0053] 2. Signal Acquisition: Heart rate and skin conductance signals are continuously acquired at sampling frequencies of 100–200 Hz and 10–50 Hz, respectively.

[0054] 3. Data preprocessing: Bandpass filtering is applied to the heart rate signal to eliminate optical interference; low-pass filtering is applied to the skin conductance signal to remove high-frequency noise; motion artifacts are removed based on acceleration information.

[0055] 4. Feature Calculation: Calculate heart rate variability (HRV) index, calculate the number and mean of skin conductance peaks. State reasoning: Input features into the physiological state reasoning model; output stress index, mood level, and fatigue level; 6. Data linkage: The output is sent to the scene recognition engine (S4) to generate scene tags; the output is sent to S1 to adjust the discharge intensity and soothing mode; the output is sent to S2 to adjust the tone of voice output.

[0056] It should be noted that the signal acquisition module is isolated from the main control circuit to reduce electromagnetic interference; The optical heart rate sensor and skin conductance sensor are designed for low power consumption and interference resistance. The continuity and outliers of the acquired signals are detected to prevent abnormal output from affecting the negative ion regulation or translation tone; The heart rate and skin conductance signal range are calibrated at the factory to ensure signal reliability under different user wearing conditions. S4. Based on user voice features, physiological parameters, motion information or environmental parameters, perform scene recognition, and adjust the negative ion output mode and translation output mode simultaneously according to the recognition results; Furthermore, in step S4, based on user voice features, motion information, environmental parameters and physiological state features, a scene recognition model is used to classify the current usage scenario and generate scene tags; the output mode of the negative ion module, the air guide angle and the output mode of the voice translation module are adjusted synchronously according to the scene tags to achieve adaptive and coordinated control of negative ion health care and voice translation, and to automatically select the local or cloud translation model in different scenarios. Specifically: S4.1: Source of input signal User voice characteristics: voice signal, emotion level, volume; User physiological characteristics: stress index, mood level, fatigue level; Environmental information: air quality, temperature and humidity, noise level; Motion information: Accelerometers / gyroscopes collect neck / head movement data; Processing module: Scene recognition model: Lightweight deep neural network or multimodal fusion model, which maps input features to scene labels; Control Decision Unit: Outputs control parameters for S1 (negative ion module) and S2 (voice translation module) based on scene labels; S4.2: Multimodal fusion: Voice features, motion information, environmental parameters, and user physiological state are normalized and integrated into a unified feature vector to input the scene recognition model; Different modalities are assigned different weights based on confidence level, for example: During audio acquisition → High audio weight; High-motion state → Increased weight of motion features; Severe air pollution → Increased weight of environmental characteristics; S4.3: Multimodal neural network, Transformer, or lightweight fusion network; input: normalized multimodal feature vector. Output: Scene tags (e.g., indoor quiet, noisy outdoors, user on a phone call, anxious state); Confidence score: Auxiliary parameter (used for control strategies, such as the probability of triggering soothing mode); If the confidence level is below the threshold, a conservative control strategy (low duty cycle / low translation volume) is adopted. If the confidence level is high, select the negative ion output mode and translation model (local / cloud) according to the scene label.

[0057] S4.4: Control Strategy and Linkage S4.4.1: Negative ion module control: Output modes: Baseline, Soothing, Low Duty Cycle, Pause Airflow angle adjustment ±15° prioritizes coverage of the user's breathing zone. The ion current intensity is dynamically adjusted based on the stress index and mood level. S4.4.2: Voice translation module control: Select the output tone (soothing / standard / emphasis) based on the scene tag. Choose between local or cloud-based translation models based on ambient noise and scene. S4.4.3: Priority and Security Mechanisms: Audio acquisition priority → If the voice translation module is acquiring audio, the negative ion module will automatically downgrade or pause. Ozone / temperature monitoring → Immediately reduce negative ion output if the threshold is exceeded; Scene label change → Immediately update the control parameters of the negative ion module and voice translation module.

[0058] Implementation process: (1) Collect multimodal inputs: speech features, motion data, environmental parameters, and user physiological state; (2) Normalize and adjust the weights of the features; (3) Input the scene recognition model and infer to generate scene labels and confidence scores; (4) Based on the scene label mapping table, generate: negative ion module discharge parameters and air guide angle, speech translation tone mode and model selection; (5) The control unit sends parameters to the negative ion module and voice translation, while monitoring the safety threshold. (6) If the scene label or user status changes, return to steps 3-5 and repeat.

[0059] In another embodiment, the device shell is made of medical-grade antibacterial ABS material, weighing approximately 120g, and its neckband design facilitates prolonged wear. The carbon brush head discharge module consists of multiple bundles of carbon fibers, achieving uniform discharge through a miniature high-voltage power supply (DC 5V input, approximately 3.5kV output). The AI ​​translation module is based on a deep semantic recognition model (supporting mainstream languages ​​such as Chinese, English, Korean, and Japanese), and simultaneously utilizes a physiological parameter acquisition module (PPG heart rate sensor and skin conductance sensor) to achieve fusion calculation of semantic and health data. The main control MCU dynamically adjusts the negative ion release intensity and voice prompts based on the AI ​​analysis results. When the AI ​​module detects that the user is in a high-pressure or anxious tone, it determines the emotional level through the speech recognition model, triggers the negative ion module to enter the "soothing mode", outputs a low-noise, high-density negative ion flow, and provides feedback on the translation results in a gentle tone, thereby achieving two-way intervention of language communication and health conditioning. There is a time-division multiplexing control logic between the audio acquisition unit and the negative ion discharge unit. When the device is detected to be in audio input state, the negative ion discharge will be paused to reduce electromagnetic and acoustic interference. The AI ​​module in the device dynamically adjusts the direction of the negative ion air vents based on information provided by the motion sensor, ensuring that negative ions can be directed towards the user's breathing area.

[0060] The above embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions will not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A multimodal data-driven adaptive control method for neck-mounted negative ion therapy and AI translation, characterized in that: Includes the following steps: S1. Drive the carbon brush head negative ion generating module to generate a negative ion flow, and adjust the discharge frequency and intensity according to the control command; S2. Collect user voice, perform speech recognition and semantic processing, and generate corresponding target language translation results; S3. Obtain the user's heart rate, skin conductance, or other physiological parameters through the physiological signal detection module to characterize the user's state; S4. Based on user voice features, physiological parameters, motion information or environmental parameters, scene recognition is performed, and the negative ion output mode and translation output mode are adjusted simultaneously according to the recognition results.

2. The multimodal data-driven adaptive control method for neck-mounted negative ion therapy and AI translation according to claim 1, characterized in that, In step S1, the negative ion generating module includes a carbon brush head composed of multiple bundles of carbon fiber microfilaments, a flexible conductive base, a miniature high-voltage boost circuit, and a directional airflow mechanism; the miniature high-voltage boost circuit takes DC5V as input and outputs a pulsed high voltage of 2.5–4.5 kV, and the discharge frequency and duty cycle of the miniature high-voltage boost circuit are dynamically adjusted by the main control unit according to the scene tags generated by the scene recognition engine, ambient air quality parameters, user physiological state, and audio acquisition status.

3. The multimodal data-driven adaptive control method for neck-mounted negative ion therapy and AI translation according to claim 2, characterized in that, In step S1, the carbon brush head achieves ±15° airflow direction adjustment via an adjustable shaft, and enters discharge pause mode or low duty cycle mode when the audio acquisition state is in the acquisition state.

4. The multimodal data-driven adaptive control method for neck-mounted negative ion therapy and AI translation according to claim 3, characterized in that, In step S2, noise reduction and voice endpoint detection are performed on the collected user voice; speech recognition is performed using an acoustic model and a language model based on a deep semantic recognition model.

5. The multimodal data-driven adaptive control method for neck-mounted negative ion therapy and AI translation according to claim 4, characterized in that, In step S2, a multilingual machine translation model is invoked to perform cross-language conversion on the identified text; and the corresponding translation tone and output mode are selected based on the user's voice emotion characteristics, which are inferred from acoustic prosodic parameters.

6. The multimodal data-driven adaptive control method for neck-mounted negative ion therapy and AI translation according to claim 5, characterized in that, In step S3, the physiological signal detection module, which is in close contact with the user's neck skin, collects the user's heart rate signal, skin conductance signal and other physiological parameters; after filtering and feature extraction, the collected signals are input into the physiological state inference model, and the user's physiological state characteristics are inferred and output.

7. The multimodal data-driven adaptive control method for neck-mounted negative ion therapy and AI translation according to claim 6, characterized in that, In step S3, the user's physiological state characteristics include stress index, mood level and fatigue level, and the physiological state characteristics are transmitted in real time to the scene recognition engine and negative ion control module for coordinated adjustment of negative ion output and translation tone.

8. The multimodal data-driven adaptive control method for neck-mounted negative ion therapy and AI translation according to claim 7, characterized in that, In step S4, based on user voice features, motion information, environmental parameters and physiological state features, a scene recognition model is used to classify the current usage scenario and generate scene labels.

9. The multimodal data-driven adaptive control method for neck-mounted negative ion therapy and AI translation according to claim 8, characterized in that, In step S4, the output mode of the negative ion module, the air guide angle, and the output mode of the voice translation module are adjusted synchronously according to the scene label.

10. The multimodal data-driven adaptive control method for neck-mounted negative ion therapy and AI translation according to claim 9, characterized in that, In step S4, the adjustment of the negative ion module is used to achieve adaptive and coordinated control of negative ion health care and voice translation, and to automatically select the local or cloud translation model in different scenarios.

Images