Search

CN-122004768-A - Massage device control method and system for assisting sleep

CN122004768ACN 122004768 ACN122004768 ACN 122004768ACN-122004768-A

Abstract

The invention discloses a massage device control method for assisting sleep, which comprises the steps of S1, collecting multi-mode physiological signals of a user in a sleep process in real time through a non-contact or micro-contact sensor, preprocessing the multi-mode physiological signals, S2, inputting the collected multi-mode physiological signals into a pre-trained sleep analysis model, calculating and outputting the sleep stage probability distribution and sleep stability index of the current user in real time, S3, dynamically generating an optimal massage control instruction according to the sleep stage probability distribution and the sleep stability index at the current moment and a preset intervention target library, S4, recording physiological signal change and sleep stage evolution data before and after massage intervention, and continuously optimizing the sleep analysis model and a massage decision strategy by combining the subjective sleep quality scores of the user.

Inventors

  • LUO JINGHUA

Assignees

  • 广州扶元保健器材有限公司

Dates

Publication Date
20260512
Application Date
20260207

Claims (10)

  1. 1. A massage apparatus control method for assisting sleep, comprising: s1, acquiring multi-mode physiological signals of a user in a sleeping process in real time through a non-contact or micro-contact sensor, and preprocessing the multi-mode physiological signals; S2, inputting the acquired multi-mode physiological signals into a pre-trained sleep analysis model, and calculating and outputting the sleep stage probability distribution and sleep stability index of the current user in real time; S3, dynamically generating an optimal massage control instruction according to the sleep stage probability distribution and the sleep stability index at the current moment and a preset intervention target library; And S4, recording physiological signal changes and sleep stage evolution data before and after massage intervention, and continuously optimizing a sleep analysis model and a massage decision strategy by combining the subjective sleep quality scores of the user in the next morning.
  2. 2. The method according to claim 1, wherein the multi-modal physiological signals in step S1 include at least a body movement signal, a heart rate variability signal, and a respiratory rhythm signal.
  3. 3. The method according to claim 1, wherein the sleep stability index in step S2 integrates the synergy of body movement frequency, heart rate and respiration, and the smoothness of the phase transition.
  4. 4. The method according to claim 1, wherein the step S1 of preprocessing the multi-modal physiological signal comprises: Identifying and eliminating interference data segments generated by large-scale body movements by comparing and analyzing the phase change of the millimeter wave echo signals and the charge change amplitude of the piezoelectric signals to obtain an anti-interference millimeter wave phase sequence and a piezoelectric loading sequence; And filtering and blind source separation processing are carried out on the anti-interference sequence, and signal components of the physiological information are primarily separated.
  5. 5. The method according to claim 1, wherein the step S2 specifically includes: s21, inputting the preprocessed multi-mode physiological signal sequence into a pre-trained deep neural network to generate a context-aware feature vector; S22, inputting the feature vector into a classification layer, and outputting the real-time probability distribution of the user in each sleep stage; s23, calculating the dynamic time warping distance between the current feature and the recent historical feature and the transition entropy of the stage probability sequence based on the feature vector and the stage probability, merging the indexes and the stage prior value, and normalizing the indexes and the stage prior value into a sleep stability index.
  6. 6. The method according to claim 5, wherein the deep neural network adopts a hybrid architecture of a bi-directional recurrent neural network and an attention mechanism.
  7. 7. The method according to claim 1, wherein the step S3 of generating the optimal massage control command comprises: S31, generating a multidimensional intervention demand vector through a demand calculation function based on a current dominant sleep stage and a stability index; S32, inquiring an intervention strategy knowledge base by using the current state tuple, and dynamically synthesizing a preliminary massage strategy; S33, inputting the preliminary strategy into an effect prediction model, predicting the physiological state change after the preliminary strategy is executed, and under the safety constraint, taking the maximum prediction stability and the deep sleep probability as targets, performing fine adjustment on the strategy parameters to obtain and execute the optimized massage instruction.
  8. 8. The method according to claim 7, wherein the dynamic synthetic preliminary massage strategy in step S32 employs a nearest neighbor matching or neural network strategy generator.
  9. 9. The method according to claim 1, wherein the process of continuously optimizing the sleep analysis model and the massage decision strategy in step S4 specifically comprises: s41, recording a complete experience data packet of each intervention, wherein the complete experience data packet comprises a pre-intervention state, an execution action, a post-intervention state transition and an instant reward calculated by physiological change; S42, combining the subjective sleep quality scores of the user in the next morning, periodically using the accumulated experience data packets, updating the strategy network and the effect prediction model through a reinforcement learning algorithm, and optionally fine-tuning the sleep analysis model; s43, updating the learned new strategy knowledge to an intervention strategy knowledge base, and adjusting user personalized parameters.
  10. 10. A massage device control system for assisting sleep, for realizing a massage device control method for assisting sleep as claimed in any one of claims 1-9, characterized by comprising: the signal acquisition and preprocessing module comprises a millimeter wave radar sensor and a piezoelectric film sensor and is used for acquiring original physiological signals of a user and preprocessing the original physiological signals to eliminate body movement interference; The sleep state analysis module is internally provided with a deep learning model and is used for receiving the preprocessed physiological signals, calculating and outputting the sleep stage probability distribution and the sleep stability index of the user in real time; The self-adaptive decision module is integrated with an intervention strategy knowledge base and an effect prediction model and is used for generating and outputting an optimized massage control instruction according to the sleep stage probability distribution and the sleep stability index; a massage execution module comprising a plurality of independently controllable miniature massage executors for executing the massage control instructions; the user interaction and learning module is used for providing a user interface, collecting subjective sleep quality scores, and continuously optimizing the sleep state analysis module and the self-adaptive decision module by using physiological signal changes before and after intervention, sleep stage evolution data and the subjective scores through a reinforcement learning algorithm; And the central controller is used for coordinating and controlling the operation of the modules.

Description

Massage device control method and system for assisting sleep Technical Field The invention relates to the technical field of physiotherapy equipment control, in particular to a massage device control method and a massage device control system for assisting sleep. Background Sleep disorders have become a common problem affecting global public health, and traditional solutions include pharmaceutical intervention, cognitive behavioral therapy, and various types of sleep aids. Among them, physical sleep-aiding devices having massage function are receiving attention because of their non-invasive and soothing properties. The current sleep assisting massage devices on the market, such as massage pillows, massage mattresses and the like, mostly rely on simple timers or preset fixed program cycles for control. These programs are typically based on a general relaxation logic design, such as a massage that is performed for a fixed period of time, a fixed maneuver, before the user sleeps. However, this "one-shot" control mode has significant technical limitations, namely, one, perceived absence. The device cannot sense and identify in real time the actual sleep state of the user (e.g., whether to fall asleep, what sleep stage is in). A user who is striving to fall asleep receives the same mechanical stimulus as a user who has already fallen into deep sleep. And secondly, the decision is stiff. The intervention is open-loop and blind, and cannot be dynamically adjusted according to the real-time physiological feedback of the user. This not only results in a greatly compromised sleep aiding effect, but also may result in fragmented sleep due to the application of an inappropriate intensity or frequency of stimuli at inappropriate times (e.g., deep sleep periods), disrupting sleep continuity. Third, there is a lack of personalization and evolutionary capability. The device cannot learn and optimize strategies from long-term interactions with a particular user, its function is cured at the time of delivery, and it cannot adapt to individual differences and time-varying sleep needs. To increase the level of intelligence, some high-end products begin to integrate a single biosensor (e.g., heart rate sensor) in an attempt to acquire the user's state. However, it is difficult to distinguish between complex sleep stages (such as distinguishing between N1 phase shallow sleep and REM phase) with high accuracy and high robustness only by using a single mode signal such as heart rate, and it is not possible to quantify the "stability" of sleep. The decision logic is still a finite state machine based on simple threshold judgment, and the true understanding and decision based on the multidimensional physiological context cannot be realized. Thus, prior art solutions have faults on the core chain of "accurate perception-intelligent decision-personalized adaptation". Disclosure of Invention The invention aims to solve the technical problems and provides a control method and a control system for a massage device for assisting sleep. The technical scheme of the invention is realized as follows: A massage device control method for assisting sleep, comprising: s1, acquiring multi-mode physiological signals of a user in a sleeping process in real time through a non-contact or micro-contact sensor, and preprocessing the multi-mode physiological signals; S2, inputting the acquired multi-mode physiological signals into a pre-trained sleep analysis model, and calculating and outputting the sleep stage probability distribution and sleep stability index of the current user in real time; S3, dynamically generating an optimal massage control instruction according to the sleep stage probability distribution and the sleep stability index at the current moment and a preset intervention target library; And S4, recording physiological signal changes and sleep stage evolution data before and after massage intervention, and continuously optimizing a sleep analysis model and a massage decision strategy by combining the subjective sleep quality scores of the user in the next morning. A massage device control system for assisting sleep, comprising: the signal acquisition and preprocessing module comprises a millimeter wave radar sensor and a piezoelectric film sensor and is used for acquiring original physiological signals of a user and preprocessing the original physiological signals to eliminate body movement interference; The sleep state analysis module is internally provided with a deep learning model and is used for receiving the preprocessed physiological signals, calculating and outputting the sleep stage probability distribution and the sleep stability index of the user in real time; The self-adaptive decision module is integrated with an intervention strategy knowledge base and an effect prediction model and is used for generating and outputting an optimized massage control instruction according to the sleep stage probability distribution and the s