Search

CN-122004789-A - Non-contact sleep respiration monitoring method, device and system

CN122004789ACN 122004789 ACN122004789 ACN 122004789ACN-122004789-A

Abstract

The application is suitable for the field of sleep monitoring and provides a non-contact sleep respiration monitoring method, equipment and a system. The method comprises the steps of respectively obtaining chest vibration signals and abdomen vibration signals through a first optical fiber sensor arranged below the chest of a lying and supine object and a second optical fiber sensor arranged below the abdomen of the lying and supine object, synchronously collecting pulse wave signals and blood oxygen saturation values from fingers of the lying and supine object through a PPG sensor, respectively extracting respiratory source signals, heart beat source signals and body movement noise source signals after blind source separation of the chest vibration signals and the abdomen vibration signals, carrying out templated enhancement on the heart beat source signals to obtain a heartbeat event time sequence, constructing a multi-feature fuzzy inference system, judging event types, and calculating sleep apnea hypopnea indexes according to the total number of effective respiratory events and total sleep time. The application realizes the judgment of sleep apnea and hypopnea events under the home-free scene, and realizes the automatic calculation of the sleep apnea hypopnea index.

Inventors

  • ZHAO DONGXING
  • Zhang Nuofu

Assignees

  • 广州医科大学附属第一医院（广州呼吸中心）

Dates

Publication Date
20260512
Application Date
20260408

Claims (10)

  1. 1. A method of non-contact sleep respiration monitoring, the method comprising: s101, respectively acquiring a chest vibration signal and an abdomen vibration signal through a first optical fiber sensor arranged below the chest of a lying and supine subject and a second optical fiber sensor arranged below the abdomen of the lying and supine subject, and synchronously acquiring pulse wave signals and blood oxygen saturation values from fingers of the lying and supine subject through a PPG sensor; s102, blind source separation is carried out on chest vibration signals and abdomen vibration signals, then respiratory source signals, heart beat source signals and body movement noise source signals are respectively extracted, and the heart beat source signals are subjected to templated enhancement to obtain a heart beat event time sequence; s103, calculating a respiration amplitude reduction ratio, a blood oxygen saturation reduction value, a body energy index, a respiration effort index and a chest-abdomen contradiction index based on a respiration source signal, a pulse wave signal, a body movement noise source signal, a blood oxygen saturation value, a chest vibration signal and an abdomen vibration signal; S104, constructing a multi-feature fuzzy inference system, and judging event types based on five features of a respiratory amplitude reduction ratio, a blood oxygen saturation reduction value, a body energy index, a respiratory effort index and a chest-abdomen contradiction index; S105, obtaining a sleep period label of each period according to the heartbeat event time sequence, the respiration source signal and the physical energy index, calculating total sleep time according to the sleep period label, counting the total number of effective respiration events occurring in the total sleep time interval according to the event label, and calculating the sleep apnea low-ventilation index according to the total number of the effective respiration events and the total sleep time.
  2. 2. The method of claim 1, further comprising the step of: the chest vibration signal, the abdomen vibration signal and the pulse wave signal are marked with uniform microsecond-level hardware time stamps.
  3. 3. The method of claim 1, wherein S102 specifically comprises: S1021, preprocessing and filtering the chest vibration signal and the abdomen vibration signal to remove power frequency interference and baseline drift, and then respectively performing Ensemble Empirical Mode Decomposition (EEMD) to obtain a group of Intrinsic Mode Functions (IMF); S1022, respectively calculating the average instantaneous frequency and the normalized energy of each IMF component in each group of eigen mode functions according to the chest vibration signal and the abdomen vibration signal, selecting components which simultaneously meet the average instantaneous frequency within the range of 0.1 Hz-10 Hz and have the normalized energy of more than 1%, and forming a chest candidate component set And abdomen candidate component set ; S1023, constructing variable step size recursive least square VSS-RLS adaptive filter to chest candidate component set And abdomen candidate component set Filtering; S1024, obtaining a low-frequency respiration reference signal through band-pass filtering the chest vibration signal, and mixing the chest signals Abdomen mixed signal The system comprises a low-frequency respiration reference signal, an observation matrix X (t), a plurality of independent components, a power calculation module, a main component identification module, a motion noise source signal S_motion signal S_t, a motion noise source signal S_motion signal S_t, a low-frequency respiration reference signal S_resp (t), a power calculation module, a motion noise source signal S_motion signal S_t, a motion noise source signal S_resp (t) and a motion noise source signal S_motion signal S_t, wherein the low-frequency respiration reference signal jointly forms the observation matrix X (t), the X (t) adopts a FastICA algorithm based on negative entropy maximization, outputs a plurality of independent components with the dimension of the observation matrix, the independent components with the low-frequency respiration reference signal and the highest cross-correlation coefficient is identified as the respiration source signal S_resp (t), power of each component with the maximum power at the frequency band of 0.8-3 Hz; S1025, carrying out templated enhancement on a heart beat source signal, namely carrying out band-pass filtering on the heart beat source signal S_heart (t), detecting candidate heart beat peaks by using a self-adaptive threshold method, intercepting signal fragments with fixed duration by taking each peak point as the center, generating a typical heart beat waveform template through clustering analysis, carrying out cross-correlation calculation on all the signal fragments and the template, reserving fragments with high correlation as qualified fragments, carrying out sub-sampling accurate positioning according to the cross-correlation peaks to obtain a high-precision heart beat event time sequence, aligning all the qualified fragments according to the heart beat event time sequence, carrying out coherent averaging, and finally outputting the enhanced heart beat source signal S_heart_enhanced (t), wherein the heart beat event time sequence is used for calculating heart rate variability characteristics.
  4. 4. A method according to claim 3, wherein S1023 comprises: Set chest candidate components And abdomen candidate component set The sum of the absolute values of the amplitudes of all the components at each instant is used as a reference noise signal For comprehensively characterizing global high-frequency vibration and instantaneous large-amplitude motion by referring to noise signals As reference noise input, chest candidate component set IMF_C and abdomen candidate component set Takes the synthesized signal of (2) as main input to carry out parallel self-adaptive filtering, and the step factor mu is based on Instantaneous power adaptive adjustment of (a): , wherein, For the initial step size to be a step, Is the attenuation coefficient; finally, outputting the chest mixed signal with preliminary denoising And abdomen mixed signal 。
  5. 5. A method according to claim 3, wherein S103 specifically comprises: S1031, performing Hilbert transformation on a respiration source signal, obtaining a respiration signal envelope env_resp (t) representing amplitude variation, calculating a preset percentile of env_resp (t) in a sliding time window, dynamically updating to serve as a current respiration baseline, calculating a moving average value of a blood oxygen saturation value according to the same sliding time window, and dynamically updating to serve as a blood oxygen saturation baseline; Calculating a respiratory amplitude reduction ratio R_drop= [ B_resp (t) -env_resp (t) ]/B_resp (t), namely the reduction ratio of the current respiratory envelope value relative to the dynamic baseline thereof, wherein B_resp (t) is the respiratory baseline and env_resp (t) is the respiratory signal envelope; Blood oxygen saturation decrease value s_drop s_drop=b_spo2 (t) -SpO 2 (t), i.e. the absolute value of the decrease of the current blood oxygen value from its dynamic baseline, where b_spo2 (t) is the blood oxygen saturation baseline and SpO 2 (t) is the blood oxygen saturation value; S1032, performing high-pass filtering on the body motion noise source signal S_motion (t) and the chest vibration signal V_chest (t), respectively calculating root mean square values in a short time window of the filtered signals to obtain body motion noise source signal energy E_motion (t) and original high-frequency vibration energy E_hf (t), and fusing the body motion noise source signal energy E_motion (t) and the original high-frequency vibration energy E_hf (t): Original fusion energy e_raw (t) =sqrt (α×e_motion (t) 2 +β* E_hf(t) 2 ), where α and β are weight coefficients and satisfy α+β=1, calculating 99% fractional number of e_raw (t) in the past time window as a normalized reference value, dividing the current e_raw (t) by the normalized reference value to obtain a body energy index a (t), where a (t) > 1 indicates that there is significant body movement; s1033, respectively carrying out band-pass filtering on the chest vibration signal V_chest (t) and the abdomen vibration signal V_abd (t), and solving Hilbert envelopes to obtain chest breathing envelope env_c (t) and abdomen breathing envelope env_a (t); Calculate instantaneous respiratory effort amplitude, off_inst (t) =sqrt (env_c (t) 2 + Env_a(t) 2 ); Calculating the median value of the Eff_inst (t) in the sliding time window as a dynamic baseline Eff_base (t); the resulting respiratory effort index is eff_idx (t) =eff_inst (t)/eff_base (t), where eff_idx (t) > 1.2 represents respiratory effort enhancement; s1034, performing Hilbert transform on band-pass filtering signals of the chest vibration signal V_chest (t) and the abdomen vibration signal V_abd (t) to obtain instantaneous phases phi_c (t) and phi_a (t); Calculating an instantaneous phase difference delta phi (t) =phi_c (t) -phi_a (t), and regulating to a [ -pi, pi ] interval; Within a judging event window, counting the percentage of sampling points of |delta phi (t) | > (2 pi/3) radians, wherein the percentage is the chest-abdomen contradiction index P (t).
  6. 6. The method of claim 3, wherein S104 specifically comprises: S1041, defining event comprehensive confidence coefficient, wherein the value range of the event comprehensive confidence coefficient is [0,1], and defining fuzzy sets { extremely low, medium, high, extremely high } and corresponding membership functions for the event comprehensive confidence coefficient; S1042, respectively inputting a respiratory amplitude reduction ratio, a blood oxygen saturation reduction value, a body energy index, a respiratory effort index and a chest-abdomen contradiction index into a predefined membership function to obtain membership of the fuzzy linguistic variable belonging to 'extremely low', 'medium', 'high' or 'extremely high'; S1043, constructing a fuzzy rule base containing a plurality of if-then rules; S1044, calculating the minimum value of membership degree of all the former conditions of each rule in the fuzzy rule base as the triggering strength of the rule, and activating the output fuzzy set corresponding to the latter by each rule according to the triggering strength of the rule; S1045, calculating the mass center of the total output fuzzy set by adopting a gravity center method, mapping the mass center into a scalar between 0 and 1, namely, the comprehensive confidence score C of the event, and formally adopting the event by the system only when the duration of a potential event is more than or equal to 10 seconds and the comprehensive confidence score C of the potential event is more than or equal to a threshold value, wherein the specific type of the adopted event is determined by the event type indicated by the rule with highest triggering intensity.
  7. 7. The method of claim 6, wherein S105 specifically comprises: S1051, dividing signals in overnight monitoring, including a heartbeat event time sequence { t_k }, a respiration source signal S_resp (t) and a body energy index A (t), into continuous preset time periods, extracting heart rate variability characteristics, body movement characteristics, respiration characteristics and event marks of each time period, inputting a hidden Markov model for automatic stage, and outputting sleep period labels of each time period: The heart rate variability feature is that based on the heart beat event time sequence obtained in the step S1025, the time domain and frequency domain features in the time period are calculated; Body movement characteristics, namely calculating average body energy in the period based on the body energy index A (t); respiratory characteristics, namely calculating average respiratory rate and respiratory amplitude variability of the period based on a respiratory source signal S_resp (t); Event markers, namely whether the period contains the effective respiratory event judged in the step S104, wherein the effective respiratory event refers to the adopted event which is judged in the step S1045, has the duration of more than or equal to 10 seconds and has the comprehensive confidence score C of more than or equal to a threshold value, and comprises central apneas, obstructive apneas, mixed apneas and hypopneas; S1052, searching backwards from a starting point in the sleep period label sequence, judging the starting time of the first period in at least 3 non-awake period periods continuously appearing as the sleep time, searching backwards forwards from an ending point, and judging the ending time of the last sleep period before the last period in at least 5 awake periods continuously appearing as the wake time; S1053, determining total sleep time TST according to the determined sleeping time and wake-up time, counting the total number of effective respiratory events occurring in the total sleep time TST according to the event markers, and calculating sleep apnea-hypopnea index AHI=the total number of effective respiratory events/total sleep time TST.
  8. 8. A computer readable storage medium storing a computer program, characterized in that the computer program when executed by a processor implements the steps of the noncontact sleep respiration monitoring method of any one of claims 1 to 7.
  9. 9. A non-contact sleep respiration monitoring apparatus, comprising: A communication port; A processor for executing computer instructions comprising the steps of performing the non-contact sleep respiration monitoring method according to any of claims 1 to 7; an internal communication bus for inter-system communication; a memory configured to store data and instructions; An input/output component configured to support data input/output.
  10. 10. A non-contact sleep respiration monitoring system, comprising: one or more first fiber optic sensors for acquiring chest vibration data of a subject; one or more second fiber optic sensors for acquiring abdominal vibration data of the subject; One or more PPG sensors for synchronously acquiring pulse wave data of the subject, and The non-contact sleep respiration monitoring apparatus of claim 9.

Description

Non-contact sleep respiration monitoring method, device and system Technical Field The invention belongs to the technical field of sleep monitoring, and particularly relates to a non-contact sleep respiration monitoring method, equipment and a system. Background Sleep apnea-hypopnea syndrome (SLEEP APNEA Hypopnea Syndrome, SAHS) refers to a clinical syndrome in which a series of pathophysiological changes occur in the body due to repeated occurrence of apnea and/or hypopnea in a sleep state caused by various causes, such as hypoxia, hypercarbonated blood, or sleep interruption. The complications such as pulmonary arterial hypertension, pulmonary heart disease, respiratory failure, hypertension, arrhythmia, cerebrovascular accident and the like can appear in the gradual progress of the disease. Sleep apnea includes central, obstructive, and a mixture of both. The continuous monitoring of sleep apnea and hypopnea can diagnose the illness state, provide basis for the subsequent treatment, and can be used for dynamically evaluating the illness state of a patient in the treatment process so as to evaluate the treatment effect and the like. AHI (Apnea-Hypopnea Index) is used in the industry to mark the severity of sleep Apnea hypopnea syndrome, meaning the number of apneas and hypopneas in one minute, typically AHI <5 is normal, 5≤AHI <15 is mild, 15≤AHI <30 is moderate, and AHI≤30 is severe. Currently, the gold standard in the industry is polysomnography (Polysomnography, PSG), i.e. continuously and synchronously tracing multiple parameters such as electroencephalogram (analyzing sleep structure), electrooculogram, mandibular myoelectricity, oronasal airflow and respiratory motion, electrocardiography, blood oxygen, snoring, limb movement, body position, etc. during the whole night sleep, and the professional technician monitors the whole night and manually interprets the next day and reports the sleep. The PSG usage field Jing Shouxian is only suitable for medical institutions such as hospitals and clinics, and needs professionals to participate, and in the monitoring process, a plurality of electrodes are attached to the patient to cause poor comfort, which can change the sleeping habit of the patient, and the patient needs to be in the medical bed overnight, which can influence a part of people sensitive to the sleeping environment, and the monitoring result can possibly not reflect the real situation of the patient, so that the requirements of the household and self-operable sleeping monitoring equipment for the patient are induced, and the sleeping and breathing situation of the patient can be conveniently and accurately monitored in the household environment. Disclosure of Invention The invention aims to provide a non-contact sleep respiration monitoring method, equipment and a system, and aims to solve the problems that a sleep respiration monitoring device PSG in the prior art is only suitable for medical institutions such as hospitals and clinics, and a plurality of electrodes are attached to a patient in the monitoring process, so that the monitoring result possibly cannot reflect the real situation of the patient. In a first aspect, the present invention provides a method of non-contact sleep respiration monitoring, the method comprising: s101, respectively acquiring a chest vibration signal and an abdomen vibration signal through a first optical fiber sensor arranged below the chest of a lying and supine subject and a second optical fiber sensor arranged below the abdomen of the lying and supine subject, and synchronously acquiring pulse wave signals and blood oxygen saturation values from fingers of the lying and supine subject through a PPG sensor; s102, blind source separation is carried out on chest vibration signals and abdomen vibration signals, then respiratory source signals, heart beat source signals and body movement noise source signals are respectively extracted, and the heart beat source signals are subjected to templated enhancement to obtain a heart beat event time sequence; s103, calculating a respiration amplitude reduction ratio, a blood oxygen saturation reduction value, a body energy index, a respiration effort index and a chest-abdomen contradiction index based on a respiration source signal, a pulse wave signal, a body movement noise source signal, a blood oxygen saturation value, a chest vibration signal and an abdomen vibration signal; S104, constructing a multi-feature fuzzy inference system, and judging event types based on five features of a respiratory amplitude reduction ratio, a blood oxygen saturation reduction value, a body energy index, a respiratory effort index and a chest-abdomen contradiction index; S105, obtaining a sleep period label of each period according to the heartbeat event time sequence, the respiration source signal and the physical energy index, calculating total sleep time according to the sleep period label, counting the total number of effective res