[go: up one dir, main page]

CN120508819A - Biological feedback equipment data processing method based on artificial intelligence - Google Patents

Biological feedback equipment data processing method based on artificial intelligence

Info

Publication number
CN120508819A
CN120508819A CN202510620298.4A CN202510620298A CN120508819A CN 120508819 A CN120508819 A CN 120508819A CN 202510620298 A CN202510620298 A CN 202510620298A CN 120508819 A CN120508819 A CN 120508819A
Authority
CN
China
Prior art keywords
vector
respiratory
data processing
subspace
component
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Granted
Application number
CN202510620298.4A
Other languages
Chinese (zh)
Other versions
CN120508819B (en
Inventor
朱凤勤
林音
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Hunan Ruimida Life Technology Co ltd
Original Assignee
Beijing Ruimida International Biotechnology Co ltd
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Beijing Ruimida International Biotechnology Co ltd filed Critical Beijing Ruimida International Biotechnology Co ltd
Priority to CN202510620298.4A priority Critical patent/CN120508819B/en
Publication of CN120508819A publication Critical patent/CN120508819A/en
Application granted granted Critical
Publication of CN120508819B publication Critical patent/CN120508819B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Classifications

    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B5/00—Measuring for diagnostic purposes; Identification of persons
    • A61B5/08—Measuring devices for evaluating the respiratory organs
    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B5/00—Measuring for diagnostic purposes; Identification of persons
    • A61B5/24—Detecting, measuring or recording bioelectric or biomagnetic signals of the body or parts thereof
    • A61B5/316—Modalities, i.e. specific diagnostic methods
    • A61B5/318—Heart-related electrical modalities, e.g. electrocardiography [ECG]
    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B5/00—Measuring for diagnostic purposes; Identification of persons
    • A61B5/24—Detecting, measuring or recording bioelectric or biomagnetic signals of the body or parts thereof
    • A61B5/316—Modalities, i.e. specific diagnostic methods
    • A61B5/318—Heart-related electrical modalities, e.g. electrocardiography [ECG]
    • A61B5/346—Analysis of electrocardiograms
    • A61B5/349—Detecting specific parameters of the electrocardiograph cycle
    • A61B5/352—Detecting R peaks, e.g. for synchronising diagnostic apparatus; Estimating R-R interval
    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B5/00—Measuring for diagnostic purposes; Identification of persons
    • A61B5/72—Signal processing specially adapted for physiological signals or for diagnostic purposes
    • A61B5/7225—Details of analogue processing, e.g. isolation amplifier, gain or sensitivity adjustment, filtering, baseline or drift compensation
    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B5/00—Measuring for diagnostic purposes; Identification of persons
    • A61B5/72—Signal processing specially adapted for physiological signals or for diagnostic purposes
    • A61B5/7235—Details of waveform analysis
    • A61B5/7253—Details of waveform analysis characterised by using transforms
    • A61B5/7257—Details of waveform analysis characterised by using transforms using Fourier transforms
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
    • G06F18/213—Feature extraction, e.g. by transforming the feature space; Summarisation; Mappings, e.g. subspace methods
    • G06F18/2135—Feature extraction, e.g. by transforming the feature space; Summarisation; Mappings, e.g. subspace methods based on approximation criteria, e.g. principal component analysis
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/22—Matching criteria, e.g. proximity measures
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/08—Learning methods
    • G06N3/09—Supervised learning
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/08—Learning methods
    • G06N3/094—Adversarial learning

Landscapes

  • Health & Medical Sciences (AREA)
  • Engineering & Computer Science (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • Biomedical Technology (AREA)
  • Biophysics (AREA)
  • Molecular Biology (AREA)
  • General Health & Medical Sciences (AREA)
  • Data Mining & Analysis (AREA)
  • Artificial Intelligence (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Veterinary Medicine (AREA)
  • Public Health (AREA)
  • Animal Behavior & Ethology (AREA)
  • Surgery (AREA)
  • Medical Informatics (AREA)
  • Heart & Thoracic Surgery (AREA)
  • Pathology (AREA)
  • Evolutionary Computation (AREA)
  • General Physics & Mathematics (AREA)
  • Cardiology (AREA)
  • General Engineering & Computer Science (AREA)
  • Mathematical Physics (AREA)
  • Signal Processing (AREA)
  • Physiology (AREA)
  • Evolutionary Biology (AREA)
  • Software Systems (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Computing Systems (AREA)
  • Computational Linguistics (AREA)
  • Psychiatry (AREA)
  • Power Engineering (AREA)
  • Pulmonology (AREA)
  • Measurement And Recording Of Electrical Phenomena And Electrical Characteristics Of The Living Body (AREA)
  • Measurement Of The Respiration, Hearing Ability, Form, And Blood Characteristics Of Living Organisms (AREA)

Abstract

本发明公开了一种基于人工智能的生物反馈设备数据处理方法，涉及生物反馈数据处理技术领域，本发明通过潜空间正交分解机制，在非线性特征空间内实现呼吸节律与自主神经调节信号的精准解耦，避免了传统频域滤波导致的HRV低频段信息丢失问题；基于变分自编码器的呼吸分量估计无需依赖外部传感器，降低硬件成本与用户负担，同时通过动态正交约束与在线微调机制，自适应个体呼吸模式差异及频率漂移，保障了长期分析的稳定性。

The present invention discloses a biofeedback device data processing method based on artificial intelligence, which relates to the technical field of biofeedback data processing. The present invention realizes the precise decoupling of respiratory rhythm and autonomic nervous system regulation signal in nonlinear feature space through the latent space orthogonal decomposition mechanism, thereby avoiding the problem of HRV low-frequency band information loss caused by traditional frequency domain filtering; the respiratory component estimation based on variational autoencoder does not need to rely on external sensors, reducing hardware costs and user burden; at the same time, through dynamic orthogonal constraints and online fine-tuning mechanisms, it adapts to individual respiratory pattern differences and frequency drift, thereby ensuring the stability of long-term analysis.

Description

Biological feedback equipment data processing method based on artificial intelligence
Technical Field
The invention relates to the technical field of biofeedback data processing, in particular to a biofeedback equipment data processing method based on artificial intelligence.
Background
The traditional HRV analysis relies on frequency domain analysis of RR interval signals, extracts low-frequency LF,0.04-0.15Hz and high-frequency HF,0.15-0.4Hz components through fast Fourier transform FFT or wavelet transform, and quantifies the balance state of sympathetic nerves and parasympathetic nerves.
The complexity of human physiological signals can lead to HRV analysis to be easily interfered by respiratory rhythms, respiratory motion is coupled with central nerves through mechanical stretch reflex to cause periodic fluctuation of RR intervals, the frequency range (0.1-0.3 Hz) of the HRV is overlapped with the LF frequency band of the HRV, the spectrum confusion makes the traditional method difficult to distinguish autonomous nerve regulation signals and respiratory rhythms to couple noise, the clinical interpretation efficacy of HRV characteristics is reduced, part of early biological feedback equipment adopts two strategies for inhibiting respiratory interference, namely frequency domain band-pass filtering, fixed cut-off frequency is set for filtering low-frequency respiratory components, the LF frequency critical information of the HRV is lost, autonomous nerve function assessment distortion is caused, multi-sensor fusion is realized, respiratory signals are synchronously collected through independent respiratory sensors, respiratory related components are removed from heart rate signals by utilizing linear regression or blind source separation algorithm, the former is difficult to preset universal filtering parameters due to individual differences of respiratory frequency, the later is required to increase hardware cost and user wearing burden, and the linear model cannot analyze nonlinear dynamic coupling relation between respiratory and the HRV.
However, the influence of respiration on HRV is not simple spectrum superposition or linear modulation, but a non-stable dynamic process formed by a nerve-mechanical feedback loop, for example, in a rapid shallow respiratory state of an anxiety patient, respiratory frequency may briefly invade an LF frequency band, and the traditional FFT analysis can wrongly attribute respiratory rhythm to sympathetic hyperactivity and mislead treatment decisions, although partial research attempts to introduce adaptive filtering or time-frequency analysis (such as Wigner-Ville distribution), the computational complexity is high and the sensitivity to noise is difficult to meet the real-time requirement of biological feedback equipment, so that an accurate decoupling model of respiratory-HRV coupling relation is needed to be constructed from an algorithm level, and the sensorless and low-complexity respiratory interference suppression is realized on the premise of retaining the physiological significance of HRV, thereby providing a technical basis for high-precision autonomic nerve function assessment.
Disclosure of Invention
The present invention has been made in view of the above-described problems occurring in the prior art.
The invention provides a biological feedback equipment data processing method based on artificial intelligence, which solves the problems that the traditional HRV analysis is interfered by breathing rhythm coupling, signal distortion is caused by frequency domain filtering, the multi-sensor scheme increases the hardware burden, and a linear model cannot analyze a dynamic nonlinear coupling relation.
In order to solve the technical problems, the invention provides the following technical scheme:
the embodiment of the invention provides a biological feedback equipment data processing method based on artificial intelligence, which comprises the following steps of,
Step S1, acquiring an electrocardiosignal and a respiratory signal of a user, wherein the respiratory signal is generated by a chest and abdomen motion sensor or from the electrocardiosignal based on periodic fluctuation characteristic estimation;
Step S2, converting the electrocardiosignal into an RR interval sequence, inputting the RR interval sequence into a variation self-encoder for latent space decomposition to obtain a latent space representation comprising a respiratory rhythm component and a nerve regulation component;
step S3, orthogonalizing the latent space representation to enable the respiratory rhythm component and the nerve regulation component to be in an orthogonalization relationship in the latent space;
step S4, reconstructing RR interval sequences without breathing coupling based on orthogonalized latent space representation;
And S5, calculating frequency domain characteristic parameters according to the reconstructed RR interval sequence, and outputting the frequency domain characteristic parameters to a visual interface of the biofeedback equipment.
As a preferable scheme of the artificial intelligence-based data processing method of the biofeedback device, the respiratory signal estimation and generation step further comprises the following steps:
performing adaptive sliding window Fourier transform on the RR interval sequence, and extracting amplitude variation characteristics of a frequency band of 0.1-0.3 Hz;
An analog respiratory waveform is generated based on the amplitude variation characteristic as an alternative input to the respiratory signal.
As a preferable scheme of the artificial intelligence-based data processing method of the biofeedback equipment, the invention comprises the following steps of:
constructing a latent variable subspace of the respiratory rhythm component and a latent variable subspace of the neuromodulation component;
The distribution of the respiratory rhythm component subspaces is constrained by an countermeasure training to match the time-dependent characteristics of the respiratory signal.
As a preferred scheme of the artificial intelligence-based data processing method of the biofeedback equipment, in the step S2, the RR interval sequence and the potential space decomposition are generated in the following way:
Detecting R wave peak on original electrocardiosignal x (t) by using Pan-Tompkins algorithm to obtain peak time sequence And calculates RR intervals: Wherein, the Represents the (n+1) th R peak time,Indicating the nth R-peak time, RR n indicating the nth RR interval;
The { RR n } is normalized to construct a column vector RR= [ RR 1,RR2,…,RRN]T, wherein RR represents the input RR interval sequence column vector, N represents the sequence length, and the sequence length is sent to a variable self-encoder (VAE) encoder to obtain a latent space representation: Wherein z ε R d represents the latent space ensemble vector, The representation of the encoder map is made,Is a parameter thereof;
Dividing z into respiratory rhythm component subspace and neuromodulation component subspace in dimension:
Wherein z r represents a respiratory rhythm subspace vector, the dimension d r,an represents a neuromodulation subspace vector, the dimension d n, d represents a total latent space dimension, and R represents a real set;
introducing countermeasure training, defining the countermeasure loss as:
Wherein L adv denotes the contrast training loss, Representing a desire for a sample x r of real breathing characteristics,Representing the expectation of the encoder output respiratory vector z r, a φ (·) representing the challenge-discriminator map, Φ being the discriminator parameter, x r representing the representation of the real respiratory signal in the feature domain, z r representing the encoder generated respiratory subspace vector;
The encoder and the discriminator pass through:
The distribution of z r is aligned stepwise with the distribution of x r.
As a preferable scheme of the artificial intelligence-based data processing method of the biofeedback equipment, the orthogonalization processing step comprises the following steps:
calculating the principal component direction of the respiratory rhythm component in the latent space;
and applying orthogonal projection constraint to the nerve modulation component so that an included angle between the nerve modulation component and the principal component direction is not smaller than 80 degrees.
As a preferable scheme of the artificial intelligence-based data processing method of the biofeedback equipment, in the step S3, the step of the latent space orthogonalization processing is as follows:
Calculating a covariance matrix of the breathing rhythm subspace vector set, wherein the formula is C r＝Cov(zr), A covariance matrix representing a respiratory subspace vector z r, cov (·) being a covariance operator;
On the basis, principal component analysis is solved to obtain a principal direction vector u of the breathing subspace:
Wherein, the Representing principal component direction vectors, |v| representing the euclidean norm of vector v;
orthographic projection is carried out on the nerve regulation component vector to obtain an orthogonalized vector:
z' n＝zn-(uTzn) u, wherein, Representing the normalized neuromodulation vector, u Tzn representing the vector inner product;
Defining an orthogonal penalty term:
Lorth＝max(0,|uTzn|-|zn|cos(80°)),
Wherein, L orth represents the loss of orthogonality penalty, max (·) is the maximum operator, z n represents the Euclidean norm of the neuromodulation vector, cos (80 °) represents the cosine value of 80 degrees;
This incorporates an overall training penalty to suppress inter-subspace overlap.
As a preferable scheme of the artificial intelligence-based data processing method of the biofeedback device, in the step S3, orthogonal projection constraint is applied to the nerve modulation component, specifically:
The included angle between the defined vectors is:
Wherein, the A principal component direction vector representing a breathing subspace,Representing a neuromodulation component vector, |·| representing an euclidean norm, arccos (·) being an inverse cosine function, θ being the angle between the vectors;
The geometric constraint condition is written as θ being greater than or equal to 80 °, which is equivalent to:
Wherein cos (80 °) represents a cosine value of 80 °;
in practical training, a soft constraint penalty term is employed:
wherein, L orth is the loss of orthogonality penalty, max (·) is the maximum operator;
the threshold setting logic comprises that in HRV analysis, 80 degrees correspond to cosine values of about 0.1736, subspace high separation is guaranteed, nerve characteristics are not weakened too much, too small cosine threshold can cause unstable training, too large cosine threshold is insufficient in separation, 80 degrees are balance points of performance and robustness in experience, and in different individuals or application scenes, 80 degrees can be used as initial values, and fine adjustment can be performed according to verification set performance.
As a preferable scheme of the artificial intelligence-based data processing method of the biofeedback equipment, the method comprises the following steps of:
and carrying out frequency domain correction on the reconstructed RR interval sequence, specifically dynamically restraining a power density value of an overlapping region of the reconstructed RR interval sequence and a respiratory estimated frequency band within 0.04-0.15Hz of an LF frequency band.
As a preferable scheme of the artificial intelligence-based data processing method of the biofeedback equipment, the method further comprises the step of dynamically adjusting:
Monitoring the frequency drift amount of a respiratory signal, and triggering the parameter of the variation self-encoder to be finely adjusted on line when the frequency drift amount exceeds a preset threshold;
The online fine tuning employs an unsupervised loss function that updates model weights based only on the latent spatial orthogonality constraints of the current input signal.
As a preferred scheme of the artificial intelligence-based data processing method of the biofeedback equipment, the method is deployed on edge computing equipment and specifically comprises the following steps:
compiling the variation self-encoder and the orthogonalization processing step into a lightweight inference engine, wherein the lightweight inference engine performs model compression through structured pruning and 8-bit integer quantization;
And carrying out streaming processing on the RR interval sequence through a memory mapping technology, wherein the memory mapping technology is based on a DMA controller of target edge equipment to carry out zero copy data transmission.
The method has the advantages that accurate decoupling of the breathing rhythm and the autonomic nerve regulation signals is achieved in a nonlinear characteristic space through a latent space orthogonal decomposition mechanism, the problem of HRV low-frequency-band information loss caused by traditional frequency domain filtering is avoided, the breathing component estimation based on a variation self-encoder does not need to depend on an external sensor, hardware cost and user load are reduced, meanwhile, stability of long-term analysis is guaranteed through dynamic orthogonal constraint and an on-line fine tuning mechanism, the self-adaptive individual breathing mode difference and frequency drift are guaranteed, a lightweight compiling and streaming processing technology in edge deployment enables a complex model to run on low-power equipment in real time, the defect of redundancy of traditional self-adaptive algorithm calculation is overcome, and robustness and system applicability of breathing interference suppression are remarkably improved on the premise that the physiological significance of the HRV is maintained.
Drawings
In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings required for the description of the embodiments will be briefly described below, it being obvious that the drawings in the following description are only some embodiments of the present invention, and that other drawings may be obtained according to these drawings without inventive effort for a person skilled in the art.
Fig. 1 is a schematic flow chart of the biofeedback data processing of embodiment 1.
Detailed Description
In order that the above-recited objects, features and advantages of the present invention will become more readily apparent, a more particular description of the invention will be rendered by reference to specific embodiments thereof which are illustrated in the appended drawings.
In the following description, numerous specific details are set forth in order to provide a thorough understanding of the present invention, but the present invention may be practiced in other ways other than those described herein, and persons skilled in the art will readily appreciate that the present invention is not limited to the specific embodiments disclosed below.
Further, reference herein to "one embodiment" or "an embodiment" means that a particular feature, structure, or characteristic can be included in at least one implementation of the invention. The appearances of the phrase "in one embodiment" in various places in the specification are not necessarily all referring to the same embodiment, nor are separate or alternative embodiments mutually exclusive of other embodiments.
Embodiment 1, referring to fig. 1, this embodiment provides a biofeedback data processing method based on artificial intelligence, including:
Step S1, acquiring an electrocardiosignal and a respiratory signal of a user, wherein the respiratory signal is generated by a chest and abdomen motion sensor or from the electrocardiosignal based on periodic fluctuation characteristic estimation;
the step of generating an estimate of the respiratory signal further comprises:
performing adaptive sliding window Fourier transform on the RR interval sequence, and extracting amplitude variation characteristics of a frequency band of 0.1-0.3 Hz;
Generating an analog respiratory waveform based on the amplitude variation characteristics as a surrogate input for the respiratory signal;
Step S2, converting electrocardiosignals into RR interval sequences, inputting the RR interval sequences into a variation self-encoder for latent space decomposition to obtain a latent space representation containing respiratory rhythm components and nerve regulation components;
The step of latent spatial decomposition of the variant self-encoder comprises:
constructing a latent variable subspace of the respiratory rhythm component and a latent variable subspace of the neuromodulation component;
Constraining the distribution of the respiratory rhythm component subspace to match the time-dependent features of the respiratory signal by countermeasure training;
in step S2, the manner of generating the RR interval sequence and the latent space decomposition is as follows:
Detecting R wave peak on original electrocardiosignal x (t) by using Pan-Tompkins algorithm to obtain peak time sequence And calculates RR intervals: Wherein, the Represents the (n+1) th R peak time,Indicating the nth R-peak time, RR n indicating the nth RR interval;
The { RR n } is normalized to construct a column vector RR= [ RR 1,RR2,…,RRN]T, wherein RR represents the input RR interval sequence column vector, N represents the sequence length, and the sequence length is sent to a variable self-encoder (VAE) encoder to obtain a latent space representation: Wherein z ε R d represents the latent space ensemble vector, The representation of the encoder map is made,Is a parameter thereof;
Dividing z into respiratory rhythm component subspace and neuromodulation component subspace in dimension:
Wherein z r represents a respiratory rhythm subspace vector, the dimension d r,zn represents a neuromodulation subspace vector, the dimension d n, d represents a total latent space dimension, and R herein represents a real set, different from RR;
introducing countermeasure training, defining the countermeasure loss as:
Wherein L adv denotes the contrast training loss, Representing a desire for a sample x r of real breathing characteristics,Representing the expectation of the encoder output respiratory vector z r, a φ (·) representing the challenge-discriminator map, Φ being the discriminator parameter, x r representing the representation of the real respiratory signal in the feature domain, z r representing the encoder generated respiratory subspace vector;
The encoder and the discriminator pass through:
Gradually aligning the z r distribution with the x r distribution;
step S3, orthogonalizing the latent space representation to enable the respiratory rhythm component and the nerve regulation component to be in an orthogonalization relationship in the latent space;
The orthogonalization processing step includes:
calculating the principal component direction of the respiratory rhythm component in the latent space;
applying orthogonal projection constraint to the nerve regulation component to ensure that the included angle between the nerve regulation component and the direction of the main component is not less than 80 degrees;
In step S3, the step of the latent space orthogonalization processing is as follows:
Calculating a covariance matrix of the breathing rhythm subspace vector set, wherein the formula is C r＝Cov(zr), A covariance matrix representing a respiratory subspace vector z r, cov (·) being a covariance operator;
On the basis, principal component analysis is solved to obtain a principal direction vector u of the breathing subspace:
Wherein, the Representing principal component direction vectors, |v| representing the euclidean norm of vector v;
orthographic projection is carried out on the nerve regulation component vector to obtain an orthogonalized vector:
z' n＝zn-(uTzn) u, wherein, Representing the normalized neuromodulation vector, u Tzn representing the vector inner product;
Defining an orthogonal penalty term:
Lorth＝max(0,|uTzn|-|zn|cos(80°)),
Wherein, L orth represents the loss of orthogonality penalty, max (·) is the maximum operator, z n represents the Euclidean norm of the neuromodulation vector, cos (80 °) represents the cosine value of 80 degrees;
this incorporates an overall training penalty to suppress inter-subspace overlap;
in step S3, orthographic projection constraints are applied to the neuromodulation components, specifically:
The included angle between the defined vectors is:
Wherein, the A principal component direction vector representing a breathing subspace,Representing a neuromodulation component vector, |·| representing an euclidean norm, arccos (·) being an inverse cosine function, θ being the angle between the vectors;
The geometric constraint condition is written as θ being greater than or equal to 80 °, which is equivalent to:
Wherein cos (80 °) represents a cosine value of 80 °;
in practical training, a soft constraint penalty term is employed:
wherein, L orth is the loss of orthogonality penalty, max (·) is the maximum operator;
The threshold setting logic comprises that in HRV analysis, 80 degrees correspond to cosine values of about 0.1736, so that subspace high separation is ensured, nerve characteristics are not weakened too much, the fact that the training is unstable is caused by too small cosine threshold, separation is insufficient if too large cosine threshold, and 80 degrees are balance points of performance and robustness in experience;
Specifically, the geometric constraint directly limits the minimum included angle between the nerve regulation component and the main breathing direction in the latent space, ensures that the nerve regulation component and the main breathing direction have obvious separation degree in the high-dimensional space through the upper bound of cosine values, and compared with a method of punishing after pure projection, the soft constraint introduces an angle threshold, so that the filtering effect between subspaces is more interpretable, and due to the adoption of a continuous punishment function, the loss is smooth in the training process, the convergence stability of a model is improved, the 80-degree threshold not only meets the decoupling requirement of HRV signals, but also reserves the fine fluctuation information of nerve regulation, and provides purer RR interval reconstruction results which are rich in physiological significance for the subsequent frequency domain feature extraction;
step S4, reconstructing RR interval sequences without breathing coupling based on orthogonalized latent space representation;
s5, calculating frequency domain characteristic parameters according to the reconstructed RR interval sequence, and outputting the frequency domain characteristic parameters to a visual interface of biofeedback equipment;
the calculating step of the frequency domain characteristic parameter comprises the following steps:
Carrying out frequency domain correction on the reconstructed RR interval sequence, specifically dynamically restraining a power density value of an overlapping region of an estimated frequency band and respiration in an LF frequency band of 0.04-0.15 Hz;
in step S5, the mode of performing frequency domain correction on the reconstructed RR interval sequence is as follows:
Performing power spectrum estimation, wherein the formula is P (f) =welch (RR ') (f), wherein RR' represents a reconstructed RR interval column vector, P (f) represents a power spectrum density function of the RR interval column vector, welch (·) represents a Welch spectrum estimation operator adopting an overlapping window and averaging method, and f represents frequency;
Defining a low frequency band and a respiratory overlap region [ f LF,low,fLF,high ]:
[fLF,low,fLF,high]=[0.04,0.15]Hz,
Fovl＝[fr-δ,fr+δ]∩[fLF,low,fLF,high],
Wherein f LF,low and f LF,high represent the lower and upper frequency bands, respectively, f r represents the current estimated respiratory frequency, and δ represents the respiratory frequency drift half-window width;
Constructing a suppression mask function expressed as:
If F e F ovl, H (F) =0, otherwise H (F) =1,
Wherein H (F) is a frequency rejection mask function, which is completely rejected when the frequency falls within the breath and low frequency band overlap region F ovl, otherwise the original power is preserved;
Performing power spectrum correction, wherein the correction formula is P corr (f) =H (f) P (f), and P corr (f) is a corrected power spectrum density function;
Calculating low band correction power P LF:
wherein P LF represents the low-band power corrected for biofeedback;
The suppression rules and band selection logic include:
the low frequency band 0.04,0.15 Hz reflects the sympatho-parasympathetic balance, and is therefore the target power calculation interval,
The respiratory frequency band is usually 0.1-0.3Hz and fluctuates with the respiratory depth and the respiratory rate, and the respiratory frequency f r and the drift half-window width delta thereof can be estimated in real time and can be set as the standard deviation of the respiratory frequency sliding windowOr an empirical value of 0.01Hz, dynamically constructing an overlapping region F ovl, ensuring that the inhibition rule can adapt to individual respiratory variation;
The hard threshold mask is adopted to finish power zero setting, so that the influence of residual respiration peaks on LF power is avoided, and the method ensures the simplicity of calculation and the real-time requirement while retaining non-respiration components;
specifically, the mechanism ensures that all power peaks possibly generated by respiratory coupling are dynamically removed in a low-frequency band range, so that purer neuromodulation indexes are provided for subsequent biofeedback;
the data processing method of the biofeedback device further comprises a dynamic adjustment step:
Monitoring the frequency drift amount of the respiratory signal, and triggering the parameter of the variation self-encoder to be finely adjusted on line when the frequency drift amount exceeds a preset threshold;
the online fine tuning adopts an unsupervised loss function, and model weights are updated only based on the potential space orthogonality constraint of the current input signal;
the data processing method of the biofeedback device is deployed on the edge computing device and specifically comprises the following steps:
Compiling the variation self-encoder and the orthogonalization processing step into a lightweight inference engine, wherein the lightweight inference engine performs model compression through structured pruning and 8-bit integer quantization;
And carrying out streaming processing on the RR interval sequence by a memory mapping technology, wherein the memory mapping technology is based on the DMA controller of the target edge device to carry out zero copy data transmission.
It should be noted that the above embodiments are only for illustrating the technical solution of the present invention and not for limiting the same, and although the present invention has been described in detail with reference to the preferred embodiments, it should be understood by those skilled in the art that the technical solution of the present invention may be modified or substituted without departing from the spirit and scope of the technical solution of the present invention, which is intended to be covered in the scope of the claims of the present invention.

Claims (10)

1.一种基于人工智能的生物反馈设备数据处理方法，其特征在于：包括，1. A biofeedback device data processing method based on artificial intelligence, characterized by: comprising: 步骤S1，获取用户的心电信号及呼吸信号，其中所述呼吸信号通过胸腹运动传感器或从心电信号中基于周期波动特征估计生成；Step S1, obtaining the user's electrocardiogram (ECG) signal and respiratory signal, wherein the respiratory signal is generated by a chest and abdomen motion sensor or estimated based on periodic fluctuation characteristics of the ECG signal; 步骤S2，将所述心电信号转换为RR间期序列，并输入至变分自编码器进行潜空间分解，获得包含呼吸节律分量与神经调节分量的潜空间表示；Step S2, converting the ECG signal into a RR interval sequence and inputting it into a variational autoencoder for latent space decomposition to obtain a latent space representation including a respiratory rhythm component and a neural regulation component; 步骤S3，对所述潜空间表示进行正交化处理，使呼吸节律分量与神经调节分量在潜空间内呈正交关系；Step S3, performing orthogonal processing on the latent space representation so that the respiratory rhythm component and the neural regulation component are orthogonal in the latent space; 步骤S4，基于正交化后的潜空间表示，重建去除非呼吸耦合的RR间期序列；Step S4, reconstructing the RR interval sequence without non-respiratory coupling based on the orthogonalized latent space representation; 步骤S5，根据重构后的RR间期序列计算频域特征参数，并输出至生物反馈设备的可视化界面。Step S5: Calculate frequency domain characteristic parameters based on the reconstructed RR interval sequence and output them to the visualization interface of the biofeedback device. 2.如权利要求1所述的一种基于人工智能的生物反馈设备数据处理方法，其特征在于：所述呼吸信号的估计生成步骤进一步包括：2. The artificial intelligence-based biofeedback device data processing method according to claim 1, wherein the step of estimating and generating the respiratory signal further comprises: 对RR间期序列进行自适应滑动窗口傅里叶变换，提取0.1-0.3Hz频段的幅值变化特征；Adaptive sliding window Fourier transform was performed on the RR interval sequence to extract the amplitude change characteristics in the 0.1-0.3 Hz frequency band; 基于所述幅值变化特征生成模拟呼吸波形，作为所述呼吸信号的替代输入。A simulated respiratory waveform is generated based on the amplitude variation characteristic as an input substitute for the respiratory signal. 3.如权利要求1所述的一种基于人工智能的生物反馈设备数据处理方法，其特征在于：所述变分自编码器的潜空间分解步骤包括：3. The artificial intelligence-based biofeedback device data processing method according to claim 1, wherein the step of decomposing the latent space of the variational autoencoder comprises: 构建呼吸节律分量的潜变量子空间与神经调节分量的潜变量子空间；Constructing the latent variable subspace of respiratory rhythm component and the latent variable subspace of neural regulation component; 通过对抗训练约束所述呼吸节律分量子空间的分布与呼吸信号的时间相关性特征匹配。The distribution of the respiratory rhythm component subspace is constrained to match the time correlation characteristics of the respiratory signal through adversarial training. 4.如权利要求3所述的一种基于人工智能的生物反馈设备数据处理方法，其特征在于，步骤S2中，生成所述RR间期序列及潜空间分解的方式为：4. The artificial intelligence-based biofeedback device data processing method according to claim 3, wherein in step S2, the RR interval sequence and latent space decomposition are generated by: 利用Pan-Tompkins算法在原始心电信号x(t)上检测R波峰，得到峰值时刻序列并计算RR间期：其中，表示第(n+1)次R波峰时刻，表示第n次R波峰时刻，RRn表示第n个RR间期；The Pan-Tompkins algorithm is used to detect the R wave peak on the original ECG signal x(t) to obtain the peak time sequence And calculate the RR interval: in, represents the (n+1)th R peak moment, represents the moment of the nth R wave peak, RR n represents the nth RR interval; 将{RRn}标准化后构造列向量RR＝[RR1,RR2,…,RRN]T，其中，RR表示输入的RR间期序列列向量，N表示序列长度；将其送入变分自编码器VAE编码器，得到潜空间表示：其中，z∈Rd表示潜空间总体向量，表示编码器映射，为其参数；After normalizing {RR n }, we construct a column vector RR = [RR 1 ,RR 2 ,…,RR N ] T , where RR represents the input RR interval sequence column vector and N represents the sequence length. This vector is fed into the variational autoencoder (VAE) encoder to obtain the latent space representation: Among them, z∈R d represents the overall vector of the latent space, represents the encoder mapping, For its parameters; 将z按维度划分为呼吸节律分量子空间与神经调节分量子空间：Divide z into respiratory rhythm component subspace and neural regulation component subspace according to dimension: 其中，zr表示呼吸节律子空间向量，维度为dr，zn表示神经调节子空间向量，维度为dn，d表示潜空间总维度，此处的R表示实数集；Where z r represents the respiratory rhythm subspace vector with dimension d r , z n represents the neural modulation subspace vector with dimension d n , d represents the total dimension of the latent space, and R here represents the set of real numbers; 引入对抗训练，定义对抗损失为：Introduce adversarial training and define adversarial loss as: 其中，Ladv表示对抗训练损失，表示对真实呼吸特征样本xr的期望，表示对编码器输出呼吸向量zr的期望，Aφ(·)表示对抗判别器映射，φ为判别器参数，xr表示真实呼吸信号在特征域的表示，zr表示编码器生成的呼吸子空间向量；Among them, L adv represents the adversarial training loss, represents the expectation of the real breathing feature sample xr , represents the expectation of the encoder output breathing vector z r , A φ (·) represents the adversarial discriminator mapping, φ is the discriminator parameter, x r represents the representation of the real breathing signal in the feature domain, and z r represents the breathing subspace vector generated by the encoder; 编码器和判别器通过：The encoder and discriminator pass: 的对抗过程，使zr的分布逐步对齐xr的分布。 The adversarial process makes the distribution of z r gradually align with the distribution of x r . 5.如权利要求1所述的一种基于人工智能的生物反馈设备数据处理方法，其特征在于，所述正交化处理步骤包括：5. The artificial intelligence-based biofeedback device data processing method according to claim 1, wherein the orthogonalization step comprises: 计算呼吸节律分量在潜空间内的主成分方向；Calculate the principal component direction of the respiratory rhythm component in the latent space; 对神经调节分量施加正交投影约束，使其与所述主成分方向的夹角不小于80度。An orthogonal projection constraint is imposed on the neuromodulation component so that the angle between it and the principal component direction is not less than 80 degrees. 6.如权利要求5所述的一种基于人工智能的生物反馈设备数据处理方法，其特征在于，步骤S3中，所述潜空间正交化处理的步骤为：6. The artificial intelligence-based biofeedback device data processing method according to claim 5, wherein in step S3, the step of orthogonalizing the latent space is: 计算呼吸节律子空间向量集合的协方差矩阵，公式为：Cr＝Cov(zr)，其中，表示呼吸子空间向量zr的协方差矩阵，Cov(·)为协方差算子；The covariance matrix of the respiratory rhythm subspace vector set is calculated using the formula: C r = Cov(z r ), where: represents the covariance matrix of the respiratory subspace vector z r , Cov(·) is the covariance operator; 在此基础上通过求解主成分分析，得到呼吸子空间的主方向向量u：On this basis, by solving the principal component analysis, the main direction vector u of the breathing subspace is obtained: 其中，表示主成分方向向量，|v|表示向量v的欧氏范数；in, represents the principal component direction vector, |v| represents the Euclidean norm of vector v; 对神经调节分量向量进行正交投影，得到正交化后的向量：Perform orthogonal projection on the neuromodulation component vector to obtain the orthogonalized vector: z'n＝zn-(uTzn)u，其中，表示正交化后的神经调节向量，uTzn表示向量内积；z' n =z n -(u T z n )u, where, represents the orthogonalized neural modulation vector, u T z n represents the vector inner product; 定义正交惩罚项：Define the orthogonality penalty term: Lorth＝max(0,|uTzn|-|zn|cos(80°))，L orth =max(0,|u T z n |-|z n |cos(80°)), 其中，Lorth表示正交性惩罚损失，max(·)为取最大值算子，|zn|表示神经调节向量的欧氏范数，cos(80°)表示80度的余弦值。Where L orth represents the orthogonality penalty loss, max(·) is the maximum operator, |z n | represents the Euclidean norm of the neural modulation vector, and cos(80°) represents the cosine value of 80 degrees. 7.如权利要求6所述的一种基于人工智能的生物反馈设备数据处理方法，其特征在于，步骤S3中，对神经调节分量施加正交投影约束，具体为：7. The artificial intelligence-based biofeedback device data processing method according to claim 6, wherein in step S3, an orthogonal projection constraint is applied to the neuromodulation component, specifically: 定义向量间夹角为：The angle between vectors is defined as: 其中，表示呼吸子空间的主成分方向向量，表示神经调节分量向量，|·|表示欧氏范数，arccos(·)为反余弦函数，θ为向量间夹角；in, represents the principal component direction vector of the breathing subspace, represents the neuromodulatory component vector, |·| represents the Euclidean norm, arccos(·) is the arccosine function, and θ is the angle between vectors; 几何约束条件写为θ≥80°，等价于：The geometric constraint is written as θ≥80°, which is equivalent to: 其中，cos(80°)表示80°的余弦值； Where cos(80°) represents the cosine value of 80°; 在实际训练中，采用软约束惩罚项：In actual training, a soft constraint penalty term is used: 其中，Lorth为正交性惩罚损失，max(·)取最大值算子。Among them, L orth is the orthogonality penalty loss, and max(·) takes the maximum value operator. 8.如权利要求1所述的一种基于人工智能的生物反馈设备数据处理方法，其特征在于，所述频域特征参数的计算步骤包括：8. The artificial intelligence-based biofeedback device data processing method according to claim 1, wherein the step of calculating the frequency domain characteristic parameters comprises: 对重构后的RR间期序列进行频域校正，具体为：在LF频段0.04-0.15Hz内动态抑制与呼吸估计频段重叠区域的功率密度值。The reconstructed RR interval sequence was corrected in the frequency domain, specifically by dynamically suppressing the power density value in the region overlapping with the respiratory estimation frequency band within the LF frequency band of 0.04-0.15 Hz. 9.如权利要求8所述的一种基于人工智能的生物反馈设备数据处理方法，其特征在于，进一步包括动态调整步骤：9. The artificial intelligence-based biofeedback device data processing method according to claim 8, further comprising a dynamic adjustment step: 监测呼吸信号的频率漂移量，当所述频率漂移量超过预设阈值时，触发所述变分自编码器的参数在线微调；Monitoring the frequency drift of the respiratory signal, and triggering online fine-tuning of the parameters of the variational autoencoder when the frequency drift exceeds a preset threshold; 所述在线微调采用无监督损失函数，仅基于当前输入信号的潜空间正交性约束更新模型权重。The online fine-tuning adopts an unsupervised loss function and updates the model weights only based on the orthogonality constraint of the latent space of the current input signal. 10.如权利要求9所述的一种基于人工智能的生物反馈设备数据处理方法，其特征在于，该方法部署于边缘计算设备，具体包括：10. The artificial intelligence-based biofeedback device data processing method according to claim 9, wherein the method is deployed on an edge computing device and specifically comprises: 将所述变分自编码器与正交化处理步骤编译为轻量化推理引擎；所述轻量化推理引擎通过结构化剪枝与8位整数量化进行模型压缩；Compiling the variational autoencoder and the orthogonalization processing step into a lightweight inference engine; the lightweight inference engine performs model compression through structured pruning and 8-bit integer quantization; 通过内存映射技术对RR间期序列进行流式处理；所述内存映射技术基于目标边缘设备的DMA控制器进行零拷贝数据传输。The RR interval sequence is streamed using a memory mapping technique that performs zero-copy data transfer based on a DMA controller of a target edge device.
CN202510620298.4A 2025-05-14 2025-05-14 A Data Processing Method for Biofeedback Devices Based on Artificial Intelligence Active CN120508819B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202510620298.4A CN120508819B (en) 2025-05-14 2025-05-14 A Data Processing Method for Biofeedback Devices Based on Artificial Intelligence

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202510620298.4A CN120508819B (en) 2025-05-14 2025-05-14 A Data Processing Method for Biofeedback Devices Based on Artificial Intelligence

Publications (2)

Publication Number Publication Date
CN120508819A true CN120508819A (en) 2025-08-19
CN120508819B CN120508819B (en) 2025-11-21

Family

ID=96703990

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202510620298.4A Active CN120508819B (en) 2025-05-14 2025-05-14 A Data Processing Method for Biofeedback Devices Based on Artificial Intelligence

Country Status (1)

Country Link
CN (1) CN120508819B (en)

Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
KR20040091337A (en) * 2003-04-21 2004-10-28 삼성전자주식회사 Sequence estimating method and device in mimo ofdm communication system
US20210081804A1 (en) * 2017-05-30 2021-03-18 GTN Ltd. Tensor network machine learning system
US20210353224A1 (en) * 2018-10-15 2021-11-18 The Board Of Trustees Of The Leland Stanford Junior University Treatment of depression using machine learning
US20230079744A1 (en) * 2021-08-27 2023-03-16 Huawei Technologies Co., Ltd. Methods and systems for source coding using a neural network
CN118094157A (en) * 2024-03-12 2024-05-28 成都晟泽科技有限公司 High-dimensional uncertainty analysis method based on adaptive deep convolutional model
CN118521526A (en) * 2023-02-20 2024-08-20 Fei 公司 Deep learning techniques for analyzing experimental data generated by two or more detectors

Patent Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
KR20040091337A (en) * 2003-04-21 2004-10-28 삼성전자주식회사 Sequence estimating method and device in mimo ofdm communication system
US20210081804A1 (en) * 2017-05-30 2021-03-18 GTN Ltd. Tensor network machine learning system
US20210353224A1 (en) * 2018-10-15 2021-11-18 The Board Of Trustees Of The Leland Stanford Junior University Treatment of depression using machine learning
US20230079744A1 (en) * 2021-08-27 2023-03-16 Huawei Technologies Co., Ltd. Methods and systems for source coding using a neural network
CN118521526A (en) * 2023-02-20 2024-08-20 Fei 公司 Deep learning techniques for analyzing experimental data generated by two or more detectors
CN118094157A (en) * 2024-03-12 2024-05-28 成都晟泽科技有限公司 High-dimensional uncertainty analysis method based on adaptive deep convolutional model

Non-Patent Citations (2)

* Cited by examiner, † Cited by third party
Title
ZHAO QIBIN: "Higher order partial least squares (HOPLS): a generalized multilinear regression method.", IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE, 24 July 2013 (2013-07-24) *
赵春晖等: "从分解视角出发:基于多元统计方法的工业时序建模与过程监测综述", 控制与决策, 20 August 2024 (2024-08-20) *

Also Published As

Publication number Publication date
CN120508819B (en) 2025-11-21

Similar Documents

Publication Publication Date Title
US7225013B2 (en) Adaptive prediction of changes of physiological/pathological states using processing of biomedical signals
Shi et al. Feature extraction and classification of lower limb motion based on sEMG signals
Mendez et al. Sleep apnea screening by autoregressive models from a single ECG lead
Wu et al. Assess sleep stage by modern signal processing techniques
Quiceno-Manrique et al. Detection of obstructive sleep apnea in ECG recordings using time-frequency distributions and dynamic features
US20150164413A1 (en) Method of creating anesthetic consciousness index with artificial neural network
Gutta et al. Cardiorespiratory model-based data-driven approach for sleep apnea detection
Hersek et al. A globalized model for mapping wearable seismocardiogram signals to whole-body ballistocardiogram signals based on deep learning
CN111353972B (en) Condition assessment system, diagnosis and treatment system and their operation methods
CN120470527A (en) Method and system for spatiotemporal alignment and fusion processing of multi-source physiological signals
CN108814579A (en) A method of electrocardio, breathing combined calculation heart rate variability based on EMD decomposition
CN118675693A (en) Pulmonary rehabilitation training method based on respiratory electromyographic signal feedback
Nguyen-Ky et al. Measuring and reflecting depth of anesthesia using wavelet and power spectral density
Nazmi et al. Electromyography (EMG) based signal analysis for physiological device application in lower limb rehabilitation
Alotaiby et al. ECG‐Based Subject Identification Using Common Spatial Pattern and SVM
CN117786470A (en) A method for separation and reconstruction of ECG signals and respiratory signals based on radar signals
Bali et al. Detection of sleep apnea in ECG signal using Pan-Tompkins algorithm and ANN classifiers
KR20220158462A (en) EMG signal-based recognition information extraction system and EMG signal-based recognition information extraction method using the same
CN120508819B (en) A Data Processing Method for Biofeedback Devices Based on Artificial Intelligence
CN120241095A (en) A screening method for paroxysmal atrial fibrillation based on heart rate variability characteristics
CN106510716B (en) A method and device for calculating respiration rate based on electrocardiographic signals
Johnson et al. Improved respiration rate estimation using a Kalman filter and wavelet cross-coherence
Toulni et al. Heart problems diagnosis using ECG and PCG signals and a k-nearest neighbor classifier
Xiong et al. A biofeedback system for mobile healthcare based on heart rate variability
Yan et al. An EMG Denoising Method Based on Flexible Wearable Sensors

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
GR01 Patent grant
GR01 Patent grant
TR01 Transfer of patent right
TR01 Transfer of patent right

Effective date of registration: 20251225

Address after: 415000 Hunan Province Changde City Economic and Technological Development Zone Deshan Street Qingshan Community Ganming Road (Second Phase of Smart Electronic Industrial Park, 8th Building, 5th Floor)

Patentee after: Hunan Ruimida Life Technology Co.,Ltd.

Country or region after: China

Address before: 100100 No. 9 Building, 1st to 13th Floor, 10th Yard, West Auto Museum Road, Fengtai District, Beijing

Patentee before: BEIJING RUIMIDA INTERNATIONAL BIOTECHNOLOGY Co.,Ltd.

Country or region before: China