CN-122005302-A - Intelligent closed-loop massage system based on brain-computer interface

Abstract

The invention discloses an intelligent closed-loop massage system based on a brain-computer interface, which comprises a diagnosis module, a physiotherapy module, a real-time optimization module and a curative effect evaluation module, wherein the diagnosis module is used for automatically identifying discomfort of a user and generating an initial physiotherapy scheme, the physiotherapy module is used for executing accurate automatic massage, the real-time optimization module is used for realizing self-adaptive adjustment in a physiotherapy process and is an intelligent control core of the system, and the curative effect evaluation module is used for quantitatively evaluating curative effects and driving the system to evolve for a long time. The invention creates a massage system which is truly intelligent, personalized and continuously optimized in curative effect through the multi-mode fusion of BCI nerve signal leading, the accurate targeting of acupoint level and the reproduction of expert methods, the real-time optimization of online reinforcement learning and the long-term closed-loop learning evolution, and the massage system is substantially improved in curative effect, comfort, safety and long-term health management value.

Inventors

• ZHOU QI
• LIU ZHILONG
• YAN JUNTAO
• WANG PEICHUANG

Assignees

• 深圳市灵手科技有限公司

Dates

Publication Date

20260512

Application Date

20260211

Claims (6)

1. 1. An intelligent closed-loop massage system based on a brain-computer interface is characterized by comprising a diagnosis module, a physiotherapy module, a real-time optimization module and a curative effect evaluation module; The diagnosis module is used for automatically identifying discomfort of a user and generating an initial physiotherapy scheme, receiving three types of multi-mode input data, namely a human body diagram collected by a high-density pressure sensor array, an EEG/EMG/GSR baseline signal collected by brain-computer interface equipment and a human body posture and body surface image collected by a depth camera, firstly completing high-precision automatic positioning of acupuncture points of the whole body through an acupuncture point positioning module, then carrying out depth fusion on pressure abnormal data and pain and relaxation nerve signals identified by BCI, and finally outputting an uncomfortable thermodynamic diagram and an acupuncture point priority ordering list which are centered on the acupuncture points, and the initial physiotherapy scheme comprising a target acupuncture point sequence, a recommended method and an intensity curve; The physiotherapy module is used for carrying out accurate automatic massage, loading an initial physiotherapy scheme output by the diagnosis module by the physiotherapy module, controlling a massage component to complete massage actions, carrying out accurate motion-mechanical parameterization modeling on a plurality of clinically verified massage manipulations by the physiotherapy module, planning an optimal massage track along channels and collaterals by the physiotherapy module according to the initial scheme through a reinforcement learning algorithm, automatically generating a manipulation sequence matched with the type of the acupoint, and calling a corresponding mathematical model to drive an actuator to reproduce expert manipulation effects; The real-time optimization module is used for realizing self-adaptive adjustment in the physical therapy process, acquiring three types of multi-modal feedback signals in real time in a period, wherein the three types of multi-modal feedback signals comprise BCI nerve signals, mechanical signals of an end effector, EMG (electro-mechanical transducer) and HRV (human-magnetic transducer) physiological signals, preprocessing the signals and inputting the signals into the real-time optimization algorithm, wherein the real-time optimization algorithm takes multi-modal characteristics as states, takes massage strength, speed, frequency, contact area, acupoint residence time, manipulation switching and meridian direction parameters as actions, takes pain signal descent, force feedback comfort and physiological relaxation as rewarding targets, continuously carries out strategy iteration, and outputs dynamic adjustment instructions to the physical therapy module to realize real-time optimization of massage parameters; The therapeutic effect evaluation module is used for quantitatively evaluating therapeutic effects and driving the system to evolve for a long time, comparing multi-mode data before and after massage by the therapeutic effect evaluation module after the treatment course is finished, calculating pain index reduction amplitude, muscle tension reduction rate and physiological release degree improvement quantization indexes by taking BCI signal change as core indexes, weighting to obtain therapeutic effect total and grade, generating a personalized report containing data comparison graph and natural language total by the therapeutic effect evaluation module based on the evaluation result, updating a user long-term archive by utilizing a transducer model, predicting and generating a future personalized physiotherapy plan, and loading the updated model to the diagnosis module when the user uses next time to realize cross-session long-term closed-loop optimization.
2. 2. The intelligent closed loop massage system based on brain-computer interface according to claim 1, wherein the massage component is a multi-degree of freedom robotic arm or massage mechanism.
3. 3. An intelligent closed loop massage system based on brain-computer interface according to claim 1 or 2, wherein the real-time optimization algorithm is a reinforcement learning algorithm PPO.
4. 4. An intelligent closed loop massage system based on brain-computer interface according to claim 1 or 2, wherein the real-time optimization algorithm employs model-based predictive control MPC or adaptive fuzzy PID control.
5. 5. The intelligent closed-loop massage system based on the brain-computer interface according to claim 1 or 2, wherein the acupoint positioning module is a depth camera and a non-rigid registration, and the human body posture estimation is acquired through the depth camera and then is in non-rigid registration with a standard acupoint database.
6. 6. The intelligent closed loop massage system based on brain-computer interface according to claim 1 or 2, wherein the diagnosis module performs deep fusion of the pressure anomaly data with the pain and relaxation nerve signals identified by BCI through a multimodal fusion neural network fusing CNN and transducer.

Description

Intelligent closed-loop massage system based on brain-computer interface Technical Field The invention relates to a massage physiotherapy robot, in particular to an intelligent closed-loop massage system based on a brain-computer interface. Background With the rapid development of artificial intelligence, robotics and biosignal sensing technologies, intelligent massage devices are becoming a research hotspot in the field of health management. In the prior art, the massaging robot mostly adopts a pre-programming or simple force feedback control mode, and can execute basic massaging actions such as pressing, kneading and the like. Meanwhile, an adaptive massage system based on physiological signals such as myoelectricity, heart rate, skin conductance and the like has been initially explored, and the massage strength and rhythm can be adjusted by monitoring the muscle tension or heart rate variation of a user. Brain-computer interface (BCI) technology is also beginning to be applied in the fields of pain management and nerve feedback treatment, and can directly capture nerve signals related to pain and relaxation in the brain. However, the prior art multi-focus single technology path has not yet achieved a complete closed loop for multi-modal sensing, real-time dynamic optimization and long-term personalized learning. Meanwhile, in the aspect of fusion of meridian theory and accurate acupoint conditioning of traditional Chinese medicine, the prior art lacks a systematic and intelligent solution. The data-driven intelligent acupuncture system is provided in the academic paper "Intelligent acupuncture: data-driven revolution of traditional Chinese medicine"（Acupuncture and Herbal Medicine, vol.3, no.4, pp.271-284, 2023）, is a typical scheme for combining traditional Chinese medicine acupuncture and artificial intelligence technology, and has the following core technical contents: 1. based on the acupoint sensitization theory, the electrical impedance sensor is used for detecting the electrophysiological characteristics of the acupoints, and the Convolutional Neural Network (CNN) is combined for analyzing the body surface image, so that the automatic identification and positioning of the acupoints are realized, and the positioning error is controlled within +/-1.2 mm. 2. The data-driven acupoint prescription generation adopts the technologies of association rule mining (Apriori algorithm), complex network analysis and the like to extract acupoint compatibility rules from massive clinical data and ancient books to generate an acupoint combination scheme aiming at specific diseases. 3. And (3) evaluating the acupuncture curative effect, namely monitoring cerebral cortex nerve activity in the treatment process by utilizing nerve imaging technologies such as a functional near infrared spectrum (fNIRS), an electroencephalogram (EEG) and the like, quantitatively evaluating the curative effect by combining a Support Vector Machine (SVM) algorithm, and establishing a stimulus-response association model. 4. The method is digitalized, namely, the force, frequency and track data of the acupuncture and moxibustion technique of a doctor are collected through the PVDF film touch sensor, a mechanical-kinematic mathematical model is established, and standardized reproduction of the acupuncture and moxibustion technique is realized. The prior art 1 has the defects that 1. The BCI real-time nerve feedback closed loop is lacking, namely, although the system adopts EEG and fNIRS to monitor the nerve activity, the system is only used for postoperative efficacy evaluation, nerve signals are not fed back to a treatment execution module in real time, and manipulation parameters cannot be dynamically adjusted in the conditioning process, and the system belongs to an open loop mode of diagnosis-treatment-evaluation. 2. The technical scene is limited by acupuncture and moxibustion, and the massage requirement cannot be met, namely the system focuses on the digitization of needling actions, the massage core techniques such as massage, kneading and rolling are not covered, and the correlation model of the techniques and the meridian stimulation depth is not established, so that the 'feeling of the feeling' requirement of the massage is difficult to meet. 3. The individuation and long-term learning capability is insufficient, the acupoint prescription generation depends on historical data mining, the real-time physiological state and neural response characteristic dynamic optimization of a user are not combined, a cross-course user model updating mechanism is lacked, and more-used and more-accurate individuation conditioning cannot be realized. 4. The multi-mode data fusion degree is low, the system does not integrate physical signals such as pressure sensing, human body posture and the like, only depends on electrophysiological and neuro image data, is easy to be interfered by environment, and has insufficient robustness in diagnosis and cu