OSA monitoring and early warning method and device integrating snore and blood oxygen characteristics

Abstract

The embodiment of the invention relates to an OSA monitoring and early warning method and device fusing snore and blood oxygen characteristics. The method comprises the steps that an additional early warning strategy is received; after the testee enters the sleep state, sound wave and blood oxygen signals of the testee are collected; in the collecting process, human voice and background separation is conducted on the sound wave signals, normalization is conducted on the two separated signals, snore frame feature extraction is conducted on the basis of the two normalized signals, and a first feature list is updated; the blood oxygen signals are filtered and calibrated, blood oxygen decreasing segment recognition features are extracted based on the calibration signals, and a second feature list is updated; the obstructive sleep apnea event is recognized regularly according to the first feature list and the second feature list, comprehensive scoring is carried out based on the recognition result, and the third feature list and the first scoring list are updated; and real-time early warning is carried out according to the third feature list, and additional early warning is carried out based on an additional early warning strategy and the first score list. According to the invention, the evaluation accuracy and the early warning flexibility can be improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of data processing, in particular to an OSA monitoring and early warning method and device fusing snoring and blood oxygen characteristics. BACKGROUND

[0002] Obstructive Sleep Apnea (OSA) is a common sleep disorder. The conventional OSA grading (mild, moderate, severe) monitoring and early warning processing scheme is roughly as follows: real-time recording is performed on the sleep process of a subject, and feature analysis is performed on the sound wave signals of the recording to complete three-level evaluation, and a fixed grading early warning mode is used to alarm according to the real-time evaluation level. In recent years, some studies have shown that the oxygen saturation (SpO2) can also be used as an OSA evaluation index, and if the sound wave and blood oxygen characteristics can be fused in the conventional monitoring and early warning mechanism, the evaluation accuracy can be further improved. In addition, the early warning process in the conventional monitoring and early warning mechanism is relatively fixed, and if a response process of adding a self-defined early warning strategy is added, the early warning flexibility can be further improved. SUMMARY

[0003] The purpose of the present application is to provide an OSA monitoring and early warning method and device fusing snoring and blood oxygen characteristics, an electronic device and a computer readable storage medium, which can improve the evaluation accuracy and the early warning flexibility. Before the subject enters the sleep state, the additional early warning strategy defined by the subject is used as the first early warning strategy; after the subject enters the sleep state, the synchronous sound wave signal and the oxygen saturation signal of the subject are continuously collected; during the synchronous collection process, a pair of synchronous sound wave signal and oxygen saturation signal is generated every preset collection time L0, the collected sound wave signal is subjected to human voice and background signal separation and normalization processing, the collected oxygen saturation signal is subjected to filtering and calibration processing, the first signal list is updated based on the two original collected signals and the corresponding three preprocessed signals, the snoring frame recognition and snoring frame feature extraction processing are performed based on the human voice and background signal, the first feature list is updated based on the feature extraction result, the oxygen drop segment recognition and segment feature extraction processing are performed based on the calibration signal, and the second feature list is updated based on the feature extraction result; during the synchronous collection process, the event characteristics of the obstructive sleep apnea event are identified based on the first and second feature lists, the severity of the obstructive sleep apnea is comprehensively scored based on the identification result, and the third feature list and the first score list are updated based on the event characteristic identification result and the comprehensive score result; during the synchronous collection process, the fixed process early warning is performed based on the third feature list, and the additional early warning is performed based on the first early warning strategy and the first score list when the first early warning strategy is not empty. Through the present application, the evaluation accuracy and the early warning flexibility can be improved.

[0004] To achieve the above objectives, a first aspect of the present invention provides an OSA monitoring and early warning method that integrates snoring and blood oxygenation characteristics, the method comprising:

[0005] Before the subject enters a sleep state, an additional early warning strategy defined by the monitoring party or the subject is received as the corresponding first early warning strategy.

[0006] After the subject enters a sleep state, synchronous sound wave signals and blood oxygen saturation signals are continuously collected from the subject.

[0007] During synchronous acquisition, at every preset acquisition duration L0, a pair of synchronized acoustic signals and blood oxygen saturation signals are generated as the corresponding first acoustic signal and first blood oxygen signal. The first acoustic signal is then separated into human voice signal and background signal, and the two separated signals are normalized to obtain the corresponding first human voice signal and first background signal. The first blood oxygen signal is then filtered and calibrated to obtain the corresponding first calibration signal. A preset first signal list is updated based on the two original acquisition signals and the corresponding three preprocessed signals. Snoring frame recognition and snoring frame feature extraction are performed based on the latest first human voice signal and first background signal, and a preset first feature list is updated based on the feature extraction results. Finally, blood oxygen drop segment recognition and segment feature extraction are performed based on the latest first calibration signal, and a preset second feature list is updated based on the feature extraction results.

[0008] During the synchronous data collection process, the event characteristics of obstructive sleep apnea events are periodically identified based on the first and second feature lists, and the severity of obstructive sleep apnea is comprehensively scored based on the identification results. The preset third feature list and first score list are updated based on the event characteristic identification results and the comprehensive score results.

[0009] During the synchronous acquisition process, real-time warnings are issued based on the third feature list; and when the first warning strategy is not empty, additional warnings are issued based on the first warning strategy and the first scoring list.

[0010] Preferably, the first warning strategy includes a first warning configuration and a first contact configuration; the first warning configuration includes multiple first sub-item configurations; each first sub-item configuration includes a sub-item switch configuration and a sub-item trigger condition configuration; the sub-item switch configuration includes on and off; the sub-item trigger condition configuration includes sub-item parameter type and sub-item logical judgment expression; the sub-item parameter type includes a first type, a second type, and a third type; the sub-item logical judgment expression consists of a logical judgment operator and a judgment threshold, the logical judgment operator includes less than, less than or equal to, equal to, greater than or equal to, and greater than; the first contact configuration includes one or more second sub-item configurations; the second sub-item configuration includes a contact name and a contact interface; the contact interface includes an SMS interface and an email interface;

[0011] The signal start time of the first acoustic signal and the first blood oxygen signal is denoted as t. start The signal end time is denoted as t. end , t end =t start +L0; The first acoustic signal consists of multiple sampling point amplitudes x i The first blood oxygen signal is composed of the amplitude y of multiple sampling points. i Composition; 1 ≤ sampling point index i ≤ total number of sampling points N0, △t is a preset sampling time interval, L0>△t>0, and the sampling duration L0 is an integer multiple of the sampling time interval △t;

[0012] The first signal list includes multiple first signal records; the first signal record includes the signal start time t. start The signal end time t end The first sound wave signal, the first blood oxygen signal, the first human voice signal, the first background signal, and the first calibration signal;

[0013] The first feature list includes multiple first feature records; each first feature record corresponds to a snoring frame; the first feature record includes a first start time, a first end time, a first short-time energy, a first short-time zero-crossing rate, a first short-time autocorrelation coefficient, a first spectral centroid, a first frequency band energy ratio, and a first dominant frequency;

[0014] The second feature list includes multiple second feature records; each second feature record corresponds to a blood oxygen decline segment; the second feature record includes a second start time, a second end time, a first segment duration, and a first decline rate;

[0015] The third feature list includes multiple third feature records; each third feature record corresponds to one obstructive sleep apnea event; the third feature record includes a third start time, a third end time, a first event duration, and a first event type; the first event type includes mild events, moderate events, and severe events;

[0016] The first scoring list includes multiple first scoring records; each first scoring record corresponds to a continuous monitoring period in which one or more obstructive sleep apnea events occurred; the first scoring record includes a fourth start time, a fourth end time, a first abnormal parameter, a second abnormal parameter, a third abnormal parameter, and a first time period score; the first abnormal parameter is a normalized value of the frequency of obstructive sleep apnea events in the current continuous monitoring period, the second abnormal parameter is a normalized value of the rate of decrease in blood oxygen saturation during obstructive sleep apnea events in the current continuous monitoring period, and the third abnormal parameter is a normalized value of the duration of obstructive sleep apnea events in the current continuous monitoring period.

[0017] Preferably, the step of separating the human voice signal from the background signal in the first sound wave signal and normalizing the two separated signals to obtain the corresponding first human voice signal and first background signal specifically includes:

[0018] Step 31: Perform background noise removal on the first sound wave signal based on a bandpass filter in the 50-500Hz frequency band to obtain the corresponding initial human voice signal; and extract background noise from the first sound wave signal based on a bandstop filter in the 50-500Hz frequency band to obtain the corresponding initial background signal.

[0019] The initial human voice signal retains only the sound wave signal in the 50-500Hz frequency band, and is composed of N0 sampling point amplitudes. Composition; the initial background signal retains only sound wave signals below 50Hz or above 500Hz, composed of N0 sampling point amplitudes. composition;

[0020] Step 32: Normalize the initial human voice signal and the initial background signal to obtain the corresponding first human voice signal and the first background signal;

[0021] The first human voice signal includes N0 sampling point amplitudes. The first background signal includes the amplitude of N0 sampling points.

[0022]

[0023] u1 and σ1 are the average signal value and standard deviation of the initial human voice signal, respectively; u2 and σ2 are the average signal value and standard deviation of the initial background signal, respectively.

[0024] Preferably, the step of filtering and calibrating the first blood oxygen signal to obtain the corresponding first calibration signal specifically includes:

[0025] The first filtered signal Y is obtained by performing pulse noise cancellation on the first blood oxygen signal using a five-point median filter. 1 ; and the first filtered signal Y is processed by a three-point moving average filter. 1 The first smoothed signal Y is obtained by performing signal smoothing processing. 2 ; and for the first smoothed signal Y 2 Data calibration is performed to obtain the corresponding first calibration signal; wherein, Y 3 =aY 2 +b, Y 3 Here, a and b are the first calibration signal, and a and b are two calibration parameters of the signal acquisition device corresponding to the first blood oxygen signal.

[0026] Preferably, updating the preset first signal list based on two original acquired signals and three corresponding preprocessed signals specifically includes:

[0027] The two original acquired signals and their corresponding signal start times t start The signal end time t end The three preprocessed signals are combined to form a corresponding first signal record and added to the first signal list; the two original acquired signals include the first acoustic signal and the first blood oxygen signal; the three preprocessed signals include the first human voice signal, the first background signal and the first calibration signal.

[0028] Preferably, the step of performing snoring frame recognition and snoring frame feature extraction processing based on the latest first human voice signal and the first background signal, and updating the preset first feature list based on the feature extraction results, specifically includes:

[0029] Step 61: Based on the preset short frame duration w and sliding duration s, perform sliding frame segmentation processing on the first human voice signal and the first background signal respectively to obtain the corresponding human voice frame sequence F. 1 and background frame sequence F 2 ;

[0030] Wherein, L0>w>s>△t, the short frame duration w and the sliding duration s are both integer multiples of the sampling time interval △t, and the difference between the acquisition duration L0 and the short frame duration w, L0-w, is an integer multiple of the sliding duration s;

[0031] The human voice frame sequence F 1 Includes multiple voice frames The human voice frame Including amplitude of multiple single-frame sampling points The background frame sequence F 2 Includes multiple background frames The background frame Including amplitude of multiple single-frame sampling points 1 ≤ frame index j ≤ total number of frames N1, 1 ≤ single frame sampling point index k ≤ total number of single frame sampling points N2;

[0032]

[0033] Step 62: Sequentially extract each of the aforementioned voice frames in chronological order. As the corresponding current voice frame, and the background frame corresponding to the current voice frame. As the corresponding current background frame; and calculate the corresponding short-time energy for the current voice frame and the current background frame respectively. and short-term energy and the current short-time energy As a corresponding historical background frame short-time energy, it is pushed into a preset short-time energy queue; and the median of all the historical background frame short-time energies in the short-time energy queue is identified to obtain the corresponding current median energy, and the product of the current median energy and a preset energy weighting coefficient is used as the corresponding current short-time energy threshold; and the short-time energy is... The system identifies whether the current short-term energy threshold is exceeded; if so, it sets the corresponding human voice frame type g. j If it is a snoring frame, then set the corresponding human voice frame type g. j Non-snoring frames;

[0034] The short-time energy queue is a first-in-first-out circular queue with a fixed queue length L1, which is a preset positive integer; the energy value queue includes at most L1 of the most recent short-time energies of the historical background frames; and the energy weighting coefficient is a preset coefficient greater than 1.

[0035] Step 63, from the obtained N1 human voice frame types g j The corresponding frame type sequence G is formed; and the human voice frame type g in the frame type sequence G, where adjacent frame types are all non-snoring frames, is also included. j Reset to a non-snoring frame;

[0036] Step 64, select each of the human voice frame types g in the frame type sequence G. j The human voice frame of the snoring frame The current snoring frame is used as the corresponding current snoring frame; and the frame start time and frame end time corresponding to the current snoring frame are used as the corresponding first start time and first end time; and the short-time energy corresponding to the current snoring frame is used as the first start time and first end time. The first short-time energy is used as the corresponding first short-time energy; the short-time zero-crossing rate and short-time autocorrelation coefficient of the current snoring frame are calculated to obtain the corresponding first short-time zero-crossing rate and the first short-time autocorrelation coefficient; the current snoring frame is subjected to a fast Fourier transform to obtain the corresponding current frame spectrum; the spectral centroid, frequency band energy ratio and dominant frequency are identified based on the current frame spectrum to obtain the corresponding first spectral centroid, first frequency band energy ratio and first dominant frequency; the first start time, first end time, first short-time energy, first short-time zero-crossing rate, first short-time autocorrelation coefficient, first spectral centroid, first frequency band energy ratio and first dominant frequency corresponding to the current snoring frame are combined to form a corresponding first feature record; and when it is confirmed that there are no other records in the first feature list that are duplicated with the content of the current first feature record, the current first feature record is added to the first feature list.

[0037] Preferably, the step of identifying blood oxygen-decrease segments and extracting segment features based on the latest first calibration signal, and updating the preset second feature list based on the feature extraction results, specifically includes:

[0038] Step 71: Push the first calibration signal into a preset calibration signal queue as a corresponding historical calibration signal; calculate the average signal amplitude of all the historical calibration signals in the calibration signal queue and use the calculation result as the corresponding current blood oxygen threshold.

[0039] The calibration signal queue is a first-in-first-out circular queue with a fixed queue length L2, which is a preset positive integer; the calibration signal queue contains at most L2 of the most recent historical calibration signals.

[0040] Step 72: Record the amplitude of each sampling point on the first calibration signal that is less than the current blood oxygen threshold as the corresponding low-order amplitude; calculate the decrease rate of each low-order amplitude relative to the current blood oxygen threshold, decrease rate = (current blood oxygen threshold - low-order amplitude) / current blood oxygen threshold; record the signal sampling point corresponding to the decrease rate that exceeds a preset first decrease rate threshold as the corresponding decrease signal point; record the signal segment on the first calibration signal composed of multiple consecutive decrease signal points as the corresponding pre-selected segment, and record the pre-selected segment whose segment duration exceeds a preset first duration threshold as a corresponding blood oxygen decrease segment;

[0041] Step 73: The average decline rate of all decline signal points in each of the blood oxygen decline segments is taken as the corresponding first decline rate; the segment start time, segment end time, and segment duration corresponding to each blood oxygen decline segment are taken as the corresponding second start time, second end time, and first segment duration; and a corresponding second feature record is formed by the second start time, second end time, first segment duration, and first decline rate corresponding to each blood oxygen decline segment; and when it is confirmed that there are no other records in the second feature list that are duplicates of the current second feature record, the current second feature record is added to the second feature list.

[0042] Preferably, the step of periodically identifying event characteristics of obstructive sleep apnea events based on the first and second feature lists, comprehensively scoring the severity of obstructive sleep apnea based on the identification results, and updating the preset third feature list and first score list based on the event characteristic identification results and the comprehensive score results specifically includes:

[0043] Step 81: According to a preset first time frequency, periodically use the current time as the corresponding end time of the current period, and subtract a preset first recent duration from the end time of the current period as the corresponding start time of the current period; and combine the start time of the current period with the end time of the current period to form the corresponding current period; and extract all the second feature records in the second feature list that are in the current period to form the corresponding second feature record sequence;

[0044] Step 82, and when the second feature record sequence is not empty, perform snoring-blood oxygen pair sequence recognition based on the second feature record sequence to obtain the corresponding first sequence, specifically: take each second feature record of the second feature record sequence as the corresponding current record; take the second start time of the current record as the corresponding current timestamp; and take each first feature record in the first feature list whose first start time is before the current timestamp and whose time difference with the current timestamp does not exceed a preset first time difference range as the corresponding first record, and take the first record whose first start time is closest to the current timestamp as the corresponding current matching record; when the current matching record is not empty, form a corresponding snoring-blood oxygen pair by the current matching record and the current record; and sort all the obtained snoring-blood oxygen pairs in chronological order to form the corresponding first sequence;

[0045] Step 83: When the first sequence is not empty, the first sequence is binary-classified based on the time difference distribution of the start time difference between the snoring frame and the blood oxygen decline segment. Specifically, the time difference between the first start time and the second start time of each snoring-blood oxygen pair in the first sequence is recorded as the corresponding first time difference; the current mean u3 and the current standard deviation σ3 are calculated based on the average and standard deviation of all the obtained first time differences; the current mean u3 and the current standard deviation σ3 are identified; if the current mean u3 meets the first time difference range and the current standard deviation σ3 does not exceed the preset time difference standard deviation threshold, the first sequence is marked as a Class I sequence; if the current mean u3 does not meet the first time difference range or the current standard deviation σ3 exceeds the time difference standard deviation threshold, the first sequence is marked as a Class II sequence.

[0046] Step 84, and when the first sequence is labeled as a class of sequences, the correlation coefficient R between the snoring intensity of the first sequence and the decrease in blood oxygenation is calculated. E-r Perform calculations;

[0047]

[0048] H represents the total number of snoring-oxygen pairs in the first sequence, where 1 ≤ snoring-oxygen pair index h ≤ H; E avr E is the average of the H first short-time energies of the first sequence. h r is the first short-time energy of the h-th snoring-blood oxygen pair in the first sequence; avr r is the average of the H first decrease rates of the first sequence. h The first decrease rate is the h-th snoring-blood oxygen pair in the first sequence;

[0049] Step 85, and in the correlation coefficient R E-r When the correlation coefficient exceeds a preset threshold, obstructive sleep apnea event identification is performed on the first sequence to obtain the corresponding first event sequence. Specifically, the snoring-blood oxygen pair in the first sequence where the first band energy ratio exceeds a preset first band energy ratio threshold or the first dominant frequency meets a preset first dominant frequency threshold is taken as a corresponding first event; and all the obtained first events are sorted in chronological order to form the corresponding first event sequence.

[0050] Step 86, and when the first event sequence is not empty, identify the event type of each first event in the first event sequence to obtain the corresponding first identification type; and record each first event in the first event sequence whose first identification type is a mild event, a moderate event, or a severe event as the corresponding first determined event;

[0051] The first identification type includes uncertain events, minor events, moderate events, and severe events;

[0052] Step 87, and when the total number of the first determined events is not 0, a comprehensive score is obtained based on all the first determined events to assess the severity of obstructive sleep apnea and obtain the corresponding first score.

[0053] Step 88: The first start time and the second end time corresponding to each of the first determined events are taken as the corresponding third start time and the third end time, and the time difference between the current third start time and the third end time is taken as the corresponding first event duration; the first identification type corresponding to each of the first determined events is taken as the corresponding first event type; and a corresponding third feature record is formed by the third start time, the third end time, the first event duration, and the first event type corresponding to each of the first determined events; and when it is confirmed that there are no other records in the third feature list that are duplicated with the content of the current third feature record, the current third feature record is added to the third feature list;

[0054] Step 89, and when the first score is not empty, use the current time period start time, the current time period end time, and the first score as the corresponding fourth start time, fourth end time, and first time period score; and set the normalized event frequency f corresponding to the current first score as... NORM Normalized event decline rate r NORM Duration L of normalization event NORMAs corresponding to the first abnormal parameter, the second abnormal parameter, and the third abnormal parameter; and the fourth start time, the fourth end time, the first abnormal parameter, the second abnormal parameter, the third abnormal parameter, and the first time period score obtained this time, a corresponding first score record is added to the first score list.

[0055] Furthermore, the step of identifying the event type of each of the first events in the first event sequence to obtain the corresponding first identification type specifically includes:

[0056] Each of the first events in the first event sequence is taken as the corresponding current event; and the first short-time energy, the first short-time zero-crossing rate, the first short-time autocorrelation coefficient, the first spectral centroid, the first band energy ratio, the first segment duration, and the first descent rate of the current event are taken as the corresponding current short-time energy, current short-time zero-crossing rate, current short-time autocorrelation coefficient, current spectral centroid, current band energy ratio, current segment duration, and current descent rate;

[0057] And when the current decline rate is less than a preset second decline rate threshold, it is identified whether the current short-term energy is higher than a preset first energy threshold; if yes, the corresponding first identification type is set as a mild event; if no, the corresponding first identification type is set as an uncertain event; wherein, the second decline rate threshold is greater than the first decline rate threshold;

[0058] And when the current descent rate is greater than or equal to the second descent rate threshold and less than the preset third descent rate threshold, it is identified whether the current spectral centroid is greater than the preset first frequency threshold; if yes, the corresponding first identification type is set as a moderate event; if no, the corresponding first identification type is set as a mild event; wherein, the third descent rate threshold is greater than the second descent rate threshold;

[0059] When the current drop rate is greater than or equal to the third drop rate threshold or the current segment duration is greater than or equal to a preset second duration threshold, the current short-time zero-crossing rate, the current short-time autocorrelation coefficient, and the current frequency band energy ratio are identified. If the current short-time zero-crossing rate is less than a preset first zero-crossing rate threshold, or the current short-time autocorrelation coefficient is less than a preset first autocorrelation coefficient threshold, or the current frequency band energy ratio is greater than a preset second frequency band energy ratio threshold, then the corresponding first identification type is set as a severe event. If the current short-time zero-crossing rate is greater than or equal to the first zero-crossing rate threshold, the current short-time autocorrelation coefficient is greater than or equal to the first autocorrelation coefficient threshold, and the current frequency band energy ratio is less than or equal to the second frequency band energy ratio threshold, then the corresponding first identification type is set as a moderate event. Wherein, the second duration threshold > the first duration threshold, and the second frequency band energy ratio threshold > the first frequency band energy ratio threshold.

[0060] Furthermore, the first score, obtained by comprehensively assessing the severity of obstructive sleep apnea based on all the first determined events, specifically includes:

[0061] The total number of events N is obtained by counting the total number of the first determined events. * And calculate the corresponding event decline rate r by averaging the first decline rates of all the first determined events. * The time difference between the first start time and the second end time of each of the first determined events is taken as the corresponding event length, and the average of all event lengths is calculated to obtain the corresponding event duration L. * And record the first most recent duration as the corresponding L. ref ; and based on the total number of events N * And the first most recent duration L ref Calculate the corresponding event frequency f * =N * / L ref ;

[0062] And the event frequency f * The event decline rate r * The duration L of the event * Normalization is performed on each event to obtain the corresponding normalized event frequency f. NORM The normalized event decline rate r NORM and the duration L of the normalized event NORM ; and based on the normalized event frequency f NORM The normalized event decline rate r NORM and the duration L of the normalized event NORMCalculate the corresponding first score = w1 × f NORM +w2×r NORM +w3×L NORM Where w1+w2+w3=1, w1, w2, and w3 are the first, second, and third weighting coefficients preset respectively.

[0063] Preferably, the real-time early warning based on the third feature list specifically includes:

[0064] When a third feature record is added to the third feature list, the first event type of the newly added third feature record is taken as the corresponding current event type; and the current real-time alarm status is identified; if the real-time alarm status is a non-alarm status, a real-time alarm is triggered according to the mild, moderate, or severe alarm mode corresponding to the current event type, and during each real-time alarm process, the real-time alarm status is updated to the corresponding mild, moderate, or severe alarm status, and the alarm duration of the current mild, moderate, or severe alarm mode is taken as the corresponding current alarm duration; if the alarm duration exceeds the current alarm duration, the alarm is immediately stopped, the real-time alarm status is switched back to the non-alarm status, and the current alarm duration is cleared to zero; if the real-time alarm... If the alarm status is mild, moderate, or severe, the alarm severity of the current event type is compared with the real-time alarm status. If the alarm severity is lower than the real-time alarm status, the alarm is ignored. If the alarm severity matches the real-time alarm status, the current alarm duration is delayed based on the alarm duration of the mild, moderate, or severe alarm mode corresponding to the current event type. If the alarm severity is higher than the real-time alarm status, the current alarm is immediately stopped, and a real-time alarm is triggered according to the mild, moderate, or severe alarm mode corresponding to the current event type. Each of the mild, moderate, and severe alarm modes includes a set of corresponding alarm volume and alarm vibration configuration parameters. The alarm volume and alarm vibration intensity of the mild, moderate, and severe alarm modes increase progressively.

[0065] Preferably, the additional warning based on the first warning strategy and the first scoring list specifically includes:

[0066] When a new first rating record is added to the first rating list, the newly added first rating record is taken as the corresponding current record; an alarm counter initialized to 0 is set for the current record; a one-to-one correspondence is established between the first abnormal parameter, the second abnormal parameter, and the third abnormal parameter of the current record and the first sub-item configuration of the first type, the second type, and the third type of the sub-item parameter in the first early warning strategy; the first sub-item configuration with the sub-item switch set to "on" in the first early warning strategy is taken as the corresponding valid sub-item configuration; and when the number of valid sub-item configurations is not 0, the sub-item logic judgment expression of the current valid sub-item configuration is checked based on the first abnormal parameter, the second abnormal parameter, or the third abnormal parameter corresponding to any one of the valid sub-item configurations. If the condition is met, the alarm counter is incremented by 1. When the final alarm counter is greater than 0, a corresponding custom early warning activation record is formed by each valid sub-item configuration that is satisfied in this sub-item logic judgment and its corresponding first abnormal parameter, second abnormal parameter or third abnormal parameter. A corresponding custom early warning activation report is formed by all the obtained custom early warning activation records and sent to each contact interface of the first contact configuration of the first early warning strategy. When the final alarm counter is greater than 0, the current real-time alarm status is identified. If the real-time alarm status is a non-alarm status, a real-time alarm is triggered in a progressively advancing alarm mode of mild, moderate and severe. If the real-time alarm status is a mild, moderate or severe alarm status, the alarm is ignored.

[0067] A second aspect of the present invention provides an apparatus for implementing the OSA monitoring and early warning method that integrates snoring and blood oxygenation characteristics as described in the first aspect above. The apparatus includes: an additional early warning strategy receiving module, a monitoring signal acquisition module, a signal feature extraction module, an OSA event processing module, and an early warning module.

[0068] The additional warning strategy receiving module is used to receive additional warning strategies defined by the monitoring party or the subject before the subject enters a sleep state, as the corresponding first warning strategy.

[0069] The monitoring signal acquisition module is used to continuously acquire synchronous sound wave signals and blood oxygen saturation signals from the subject after the subject enters a sleep state.

[0070] The signal feature extraction module is used to generate a pair of synchronous sound wave signals and blood oxygen saturation signals as corresponding first sound wave signals and first blood oxygen signals at every preset acquisition time L0 during the synchronous acquisition process; and to separate the human voice signal and background signal from the first sound wave signal and normalize the two separated signals to obtain the corresponding first human voice signal and first background signal; and to filter and calibrate the first blood oxygen signal to obtain the corresponding first calibration signal; and to update the preset first signal list based on the two original acquisition signals and the corresponding three preprocessed signals; and to perform snoring frame recognition and snoring frame feature extraction processing based on the latest first human voice signal and first background signal and update the preset first feature list based on the feature extraction results; and to perform blood oxygen drop segment recognition and segment feature extraction processing based on the latest first calibration signal and update the preset second feature list based on the feature extraction results.

[0071] The OSA event processing module is used to periodically identify the event characteristics of obstructive sleep apnea events according to the first and second feature lists during the synchronous acquisition process, and to comprehensively score the severity of obstructive sleep apnea based on the identification results, and update the preset third feature list and first score list based on the event feature identification results and the comprehensive score results.

[0072] The early warning module is used to provide real-time early warnings based on the third feature list during the synchronous data acquisition process; and to provide additional early warnings based on the first early warning strategy and the first scoring list when the first early warning strategy is not empty.

[0073] A third aspect of the present invention provides an electronic device, including: a memory, a processor, and a transceiver;

[0074] The processor is used to couple with the memory, read and execute instructions in the memory to implement the steps of the method described in the first aspect above;

[0075] The transceiver is coupled to the processor, and the processor controls the transceiver to send and receive messages.

[0076] A fourth aspect of the present invention provides a computer-readable storage medium storing computer instructions that, when executed by a computer, cause the computer to perform the instructions described in the first aspect.

[0077] This invention provides an OSA monitoring and early warning method, device, electronic device, and computer-readable storage medium that integrates snoring and blood oxygenation characteristics. As described above, this invention uses a user-defined additional early warning strategy as the first early warning strategy before the subject enters a sleep state; and after the subject enters a sleep state, it continuously collects synchronous sound wave signals and blood oxygen saturation signals; during the synchronous collection process, a pair of synchronous sound wave signals and blood oxygen saturation signals are generated every preset collection duration L0, and the collected sound wave signals are processed by separating and normalizing human voice and background signals, and the collected blood oxygen saturation signals are filtered and calibrated. A first signal list is updated based on the two original collected signals and the corresponding three pre-processed signals, and snoring frame recognition and snoring frame analysis are performed based on human voice and background signals. Feature extraction processing is performed, and the first feature list is updated based on the feature extraction results. Segment identification and feature extraction processing of decreased blood oxygenation based on calibration signals are also performed, and the second feature list is updated based on the feature extraction results. During synchronous data acquisition, event characteristics of obstructive sleep apnea events are periodically identified based on the first and second feature lists, and the severity of obstructive sleep apnea is comprehensively scored based on the identification results. The third feature list and the first scoring list are updated based on the event feature identification results and the comprehensive scoring results. During synchronous data acquisition, a fixed-process early warning is implemented based on the third feature list, and additional early warnings are implemented based on the first early warning strategy and the first scoring list when the first early warning strategy is not empty. Through this embodiment of the invention, both assessment accuracy and early warning flexibility are improved. Attached Figure Description

[0078] Figure 1 This is a schematic diagram of an OSA monitoring and early warning method that integrates snoring and blood oxygenation characteristics, provided in Embodiment 1 of the present invention.

[0079] Figure 2 This is a module structure diagram of an OSA monitoring and early warning device that integrates snoring and blood oxygenation characteristics, provided in Embodiment 2 of the present invention;

[0080] Figure 3 This is a schematic diagram of the structure of an electronic device provided in Embodiment 3 of the present invention. Detailed Implementation

[0081] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings. Obviously, the described embodiments are merely some embodiments of this invention, and not all embodiments. Based on the embodiments of this invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this invention.

[0082] Embodiment 1 of the present invention provides an OSA monitoring and early warning method that integrates snoring and blood oxygenation characteristics, such as... Figure 1 The schematic diagram of an OSA monitoring and early warning method integrating snoring and blood oxygenation characteristics provided in Embodiment 1 of the present invention is shown. The method mainly includes the following steps:

[0083] Step 1: Before the subject enters a sleep state, receive an additional early warning strategy customized by the monitoring party or the subject as the corresponding first early warning strategy.

[0084] Here, the first early warning strategy implemented by the present invention includes a first early warning configuration and a first contact configuration; the first early warning configuration includes multiple first sub-item configurations; each first sub-item configuration includes a sub-item switch configuration and a sub-item trigger condition configuration; the sub-item switch configuration includes on and off; the sub-item trigger condition configuration includes sub-item parameter type and sub-item logical judgment expression; the sub-item parameter type includes a first type, a second type, and a third type; the sub-item logical judgment expression consists of a logical judgment symbol and a judgment threshold, and the logical judgment symbol includes less than, less than or equal to, equal to, greater than or equal to, and greater than; the first contact configuration includes one or more second sub-item configurations; the second sub-item configuration includes a contact name and a contact interface; the contact interface includes an SMS interface and an email interface.

[0085] Step 2: After the subject enters a sleep state, continuously collect synchronous sound wave signals and blood oxygen saturation signals from the subject.

[0086] Here, in this embodiment of the invention, a highly sensitive miniature microphone is used as a sound wave signal acquisition sensor to acquire the sound wave signal of the subject.

[0087] This invention employs a pulse oximeter, composed of a red LED, an infrared LED, and a photodetector, as the sensor for acquiring blood oxygen saturation signals. This pulse oximeter operates based on the photoplethysmography (PPG) principle, calculating the percentage of oxyhemoglobin in the blood (i.e., blood oxygen saturation) by measuring the difference in absorption rate of light of different wavelengths after passing through blood vessels. The pulse oximeter is fixed to the user's earlobe using a clamping arm, ensuring stable contact and accurate measurement.

[0088] Step 3: During the synchronous acquisition process, a pair of synchronized acoustic signals and blood oxygen saturation signals are generated every preset acquisition duration L0 as the corresponding first acoustic signal and first blood oxygen signal; the first acoustic signal is separated into human voice signal and background signal, and the two separated signals are normalized to obtain the corresponding first human voice signal and first background signal; the first blood oxygen signal is filtered and calibrated to obtain the corresponding first calibration signal; the preset first signal list is updated based on the two original acquisition signals and the corresponding three preprocessed signals; snoring frame recognition and snoring frame feature extraction are performed based on the latest first human voice signal and first background signal, and the preset first feature list is updated based on the feature extraction results; and blood oxygen decline segment recognition and segment feature extraction are performed based on the latest first calibration signal, and the preset second feature list is updated based on the feature extraction results.

[0089] Specifically, it includes: Step 31, during the synchronous acquisition process, every preset acquisition time L0, a pair of synchronous sound wave signals and blood oxygen saturation signals are generated as the corresponding first sound wave signal and first blood oxygen signal.

[0090] The signal start times of the first acoustic signal and the first blood oxygen signal are denoted as t. start The signal end time is denoted as t. end , t end =t start +L0; The first acoustic signal consists of multiple sampling point amplitudes x i The first blood oxygen signal is composed of the amplitude y of multiple sampling points. i Composition; 1 ≤ sampling point index i ≤ total number of sampling points N0, △t is the preset sampling time interval, L0>△t>0, and the sampling duration L0 is an integer multiple of the sampling time interval △t;

[0091] Step 32, and separate the human voice signal and background signal of the first sound wave signal, and normalize the two separated signals to obtain the corresponding first human voice signal and first background signal;

[0092] Specifically, it includes: step 321, using a bandpass filter in the 50-500Hz frequency band to remove background noise from the first sound wave signal to obtain the corresponding initial human voice signal; and using a bandstop filter in the 50-500Hz frequency band to extract background noise from the first sound wave signal to obtain the corresponding initial background signal.

[0093] The initial human voice signal retains only the sound wave signal in the 50-500Hz frequency band, composed of N0 sampling points. Composition; the initial background signal retains only sound wave signals below 50Hz or above 500Hz, consisting of N0 sampling point amplitudes. composition;

[0094] Step 322: Normalize the initial human voice signal and the initial background signal to obtain the corresponding first human voice signal and first background signal;

[0095] The first human voice signal includes N0 sampling point amplitudes. The first background signal includes the amplitude of N0 sampling points.

[0096]

[0097] u1 and σ1 are the average signal value and standard deviation of the initial human voice signal, respectively; u2 and σ2 are the average signal value and standard deviation of the initial background signal, respectively.

[0098] Step 33, and filter and calibrate the first blood oxygen signal to obtain the corresponding first calibration signal;

[0099] Specifically, this includes: performing pulse noise cancellation on the first blood oxygen signal using a five-point median filter to obtain the corresponding first filtered signal Y. 1 ; and the first filtered signal Y is processed by a three-point moving average filter. 1 The first smoothed signal Y is obtained by performing signal smoothing processing. 2 ; and the first smoothed signal Y 2 Perform data calibration to obtain the corresponding first calibration signal;

[0100] Among them, Y 3 =aY 2 +b, Y 3 Here, a and b are the first calibration signal, and a and b are two calibration parameters of the signal acquisition device corresponding to the first blood oxygen signal.

[0101] Step 34, and update the preset first signal list based on the two original acquired signals and the corresponding three preprocessed signals;

[0102] Specifically, this includes: two original acquired signals and their corresponding signal start times t. start Signal end time t end Each of the three preprocessed signals forms a corresponding first signal record, which is added to the first signal list.

[0103] The first signal list includes multiple first signal records; each first signal record includes a signal start time t. start Signal end time t endThe system comprises: a first acoustic signal, a first blood oxygen signal, a first human voice signal, a first background signal, and a first calibration signal; two raw acquisition signals, including the first acoustic signal and the first blood oxygen signal; and three preprocessed signals, including the first human voice signal, the first background signal, and the first calibration signal.

[0104] Step 35, and perform snoring frame recognition and snoring frame feature extraction processing based on the latest first human voice signal and first background signal, and update the preset first feature list based on the feature extraction results;

[0105] The first feature list includes multiple first feature records; each first feature record corresponds to a snoring frame; the first feature record includes a first start time, a first end time, a first short-time energy, a first short-time zero-crossing rate, a first short-time autocorrelation coefficient, a first spectral centroid, a first frequency band energy ratio, and a first dominant frequency;

[0106] Specifically, this includes: Step 351, performing sliding frame segmentation processing on the first human voice signal and the first background signal based on the preset short frame duration w and sliding duration s to obtain the corresponding human voice frame sequence F. 1 and background frame sequence F 2 ;

[0107] Where L0>w>s>△t, the short frame duration w and the sliding duration s are both integer multiples of the sampling time interval △t, and the difference between the sampling duration L0 and the short frame duration w, L0-w, is an integer multiple of the sliding duration s.

[0108] Human voice frame sequence F 1 Includes multiple voice frames Human voice frame Including amplitude of multiple single-frame sampling points Background frame sequence F 2 Includes multiple background frames Background Frame Including amplitude of multiple single-frame sampling points 1 ≤ frame index j ≤ total number of frames N1, 1 ≤ single frame sampling point index k ≤ total number of single frame sampling points N2;

[0109]

[0110] Step 352: Separate each voice frame in chronological order. As the corresponding current voice frame, and the background frame corresponding to the current voice frame. This serves as the corresponding current background frame; and the short-time energy of the current voice frame and the current background frame are calculated separately to obtain the corresponding short-time energy. and short-term energy and the current short-term energy As a corresponding historical background frame, the short-time energy is pushed into a preset short-time energy queue; the median of the short-time energy of all historical background frames in the short-time energy queue is identified to obtain the corresponding current median energy, and the product of the current median energy and a preset energy weighting coefficient is used as the corresponding current short-time energy threshold; and the short-time energy is then... If the current short-term energy threshold is exceeded, then the corresponding human voice frame type g is set. j If it is a snoring frame, then set the corresponding human voice frame type g. j Non-snoring frames;

[0111] The short-time energy queue is a first-in-first-out circular queue with a fixed queue length L1, which is a preset positive integer. The energy value queue includes at most L1 short-time energies from the most recent historical background frames. The energy weighting coefficient is a preset coefficient greater than 1.

[0112] Step 353, based on the obtained N1 personal voice frame type g j The corresponding frame type sequence G is formed; and the human voice frame type g in the frame type sequence G, where adjacent frame types are all non-snoring frames, is also formed. j Reset to a non-snoring frame;

[0113] Step 354, select each human voice frame type g in the frame type sequence G. j Human voice frames for snoring frames The current snoring frame is used as the corresponding frame start time and frame end time; the frame start time and frame end time corresponding to the current snoring frame are used as the corresponding first start time and first end time; and the short-time energy corresponding to the current snoring frame is used as the corresponding first start time and first end time. The first short-time energy is used as the corresponding first short-time energy; the short-time zero-crossing rate and short-time autocorrelation coefficient of the current snoring frame are calculated to obtain the corresponding first short-time zero-crossing rate and first short-time autocorrelation coefficient; the current snoring frame is subjected to a fast Fourier transform to obtain the corresponding current frame spectrum; the spectral centroid, frequency band energy ratio and dominant frequency are identified based on the current frame spectrum to obtain the corresponding first spectral centroid, first frequency band energy ratio and first dominant frequency; the first start time, first end time, first short-time energy, first short-time zero-crossing rate, first short-time autocorrelation coefficient, first spectral centroid, first frequency band energy ratio and first dominant frequency corresponding to the current snoring frame are combined to form a corresponding first feature record; and when it is confirmed that there are no other records in the first feature list that are duplicated with the content of the current first feature record, the current first feature record is added to the first feature list;

[0114] Step 36, and based on the latest first calibration signal, perform blood oxygenation decline segment identification and segment feature extraction processing, and update the preset second feature list based on the feature extraction results;

[0115] The second feature list includes multiple second feature records; each second feature record corresponds to a blood oxygen decline segment; the second feature record includes a second start time, a second end time, a first segment duration, and a first decline rate;

[0116] Specifically, it includes: step 361, pushing the first calibration signal into a preset calibration signal queue as a corresponding historical calibration signal; and calculating the average signal amplitude of all historical calibration signals in the calibration signal queue and using the calculation result as the corresponding current blood oxygen threshold;

[0117] The calibration signal queue is a first-in-first-out circular queue with a fixed queue length L2, which is a preset positive integer; the calibration signal queue contains at most L2 recent historical calibration signals.

[0118] Step 362: Record the amplitude of each sampling point on the first calibration signal that is less than the current blood oxygen threshold as the corresponding low-order amplitude; calculate the decrease rate of each low-order amplitude relative to the current blood oxygen threshold, decrease rate = (current blood oxygen threshold - low-order amplitude) / current blood oxygen threshold; record the signal sampling point corresponding to the decrease rate that exceeds the preset first decrease rate threshold as the corresponding decrease signal point; record the signal segment on the first calibration signal composed of multiple consecutive decrease signal points as the corresponding pre-selected segment, and record the pre-selected segment whose segment duration exceeds the preset first duration threshold as a corresponding blood oxygen decrease segment;

[0119] Here, the first drop rate threshold and the first duration threshold are two preset probability thresholds and duration thresholds, respectively; for example, the first drop rate threshold is set to 2.5%, and the first duration threshold is set to 10ms.

[0120] Step 363: The average decline rate of all decline signal points in each blood oxygen decline segment is taken as the corresponding first decline rate; the segment start time, segment end time, and segment duration corresponding to each blood oxygen decline segment are taken as the corresponding second start time, second end time, and first segment duration; and a corresponding second feature record is formed by the second start time, second end time, first segment duration, and first decline rate corresponding to each blood oxygen decline segment; and when it is confirmed that there are no other records in the second feature list that are duplicates of the current second feature record, the current second feature record is added to the second feature list.

[0121] Step 4: During the synchronous data acquisition process, the event characteristics of obstructive sleep apnea events are periodically identified according to the first and second feature lists, and the severity of obstructive sleep apnea is comprehensively scored based on the identification results. The preset third feature list and first score list are updated based on the event characteristic identification results and the comprehensive score results.

[0122] The third feature list includes multiple third feature records; each third feature record corresponds to one obstructive sleep apnea event; the third feature record includes the third start time, the third end time, the duration of the first event, and the type of the first event; the type of the first event includes mild events, moderate events, and severe events;

[0123] The first score list includes multiple first score records; each first score record corresponds to a continuous monitoring period in which one or more obstructive sleep apnea events occurred; the first score record includes a fourth start time, a fourth end time, a first abnormal parameter, a second abnormal parameter, a third abnormal parameter, and a first time period score; the first abnormal parameter is the normalized value of the frequency of obstructive sleep apnea events in the current continuous monitoring period, the second abnormal parameter is the normalized value of the rate of decrease in blood oxygen saturation during obstructive sleep apnea events in the current continuous monitoring period, and the third abnormal parameter is the normalized value of the duration of obstructive sleep apnea events in the current continuous monitoring period;

[0124] Specifically, it includes: Step 41, according to a preset first time frequency, periodically taking the current time as the corresponding end time of the current period, and taking the time obtained by subtracting the preset first recent duration from the end time of the current period as the corresponding start time of the current period; and forming the corresponding current period by the start time and end time of the current period; and extracting all second feature records in the second feature list that are in the current period to form the corresponding second feature record sequence;

[0125] Here, the first time frequency is a preset time frequency parameter; the first recent duration is a preset length parameter;

[0126] Step 42, and when the second feature recording sequence is not empty, perform snoring-blood oxygen pair sequence recognition based on the second feature recording sequence to obtain the corresponding first sequence;

[0127] Specifically, this includes: taking each second feature record in the second feature record sequence as the corresponding current record; taking the second start time of the current record as the corresponding current timestamp; taking each first feature record in the first feature list whose first start time is before the current timestamp and whose time difference with the current timestamp does not exceed a preset first time difference range as the corresponding first record; taking the first record whose first start time is closest to the current timestamp as the corresponding current matching record; and when the current matching record is not empty, forming a corresponding snoring-blood oxygen pair by the current matching record and the current record; and forming a corresponding first sequence by sorting all the obtained snoring-blood oxygen pairs in chronological order.

[0128] Here, the first time difference range is a pre-set numerical range; for example, the first time difference range is a range of values ​​[5 seconds, 15 seconds] centered at 10 seconds and extending forward and backward by 5 seconds respectively.

[0129] Step 43, and when the first sequence is not empty, perform binary classification labeling on the first sequence based on the time difference distribution of the start time difference between the snoring frame and the blood oxygen decline segment;

[0130] Specifically, this includes: recording the time difference between the first and second start times of each snoring-blood oxygen pair in the first sequence as the corresponding first time difference; calculating the current mean u3 and current standard deviation σ3 based on the average and standard deviation of all obtained first time differences; identifying the current mean u3 and current standard deviation σ3; if the current mean u3 meets the first time difference range and the current standard deviation σ3 does not exceed the preset time difference standard deviation threshold, then the first sequence is marked as a Class I sequence; if the current mean u3 does not meet the first time difference range or the current standard deviation σ3 exceeds the time difference standard deviation threshold, then the first sequence is marked as a Class II sequence.

[0131] Here, the time difference standard deviation threshold is a preset standard deviation value;

[0132] Step 44, and when the first sequence is labeled as a class of sequences, the correlation coefficient R between the snoring intensity of the first sequence and the decrease in blood oxygenation is calculated. E-r Perform calculations;

[0133] Here, the correlation coefficient R E-r The calculation method is as follows:

[0134]

[0135] Where H is the total number of snoring-blood oxygen pairs in the first sequence, 1 ≤ snoring-blood oxygen pair index h ≤ H; E avr E is the average of the H first short-time energies of the first sequence. h r is the first short-time energy of the h-th snoring-blood oxygen pair in the first sequence; avr r is the average of the H first decrease rates of the first sequence. h The first decrease rate of the h-th snoring-blood oxygen pair in the first sequence;

[0136] Step 45, and in the correlation coefficient R E-r When the correlation coefficient exceeds a preset threshold, obstructive sleep apnea event identification is performed on the first sequence to obtain the corresponding first event sequence;

[0137] Here, the correlation coefficient threshold is a pre-set threshold parameter;

[0138] Specifically, this includes: taking a snoring-blood oxygen pair in the first sequence whose first frequency band energy ratio exceeds a preset first frequency band energy ratio threshold or whose first dominant frequency meets a preset first dominant frequency threshold as a corresponding first event; and arranging all the obtained first events in chronological order to form a corresponding first event sequence;

[0139] Here, the first frequency band energy ratio threshold and the first main frequency threshold are two preset threshold parameters;

[0140] Step 46: When the first event sequence is not empty, identify the event type of each first event in the first event sequence to obtain the corresponding first identification type; and record each first event in the first event sequence whose first identification type is a mild event, a moderate event, or a severe event as the corresponding first determined event.

[0141] Specifically, this includes: step 461, and when the first event sequence is not empty, identifying the event type of each first event in the first event sequence to obtain the corresponding first identification type;

[0142] The first identification type includes uncertain events, minor events, moderate events, and severe events;

[0143] Specifically, this includes step 4611, where each first event in the first event sequence is taken as the corresponding current event; and the first short-time energy, first short-time zero-crossing rate, first short-time autocorrelation coefficient, first spectral centroid, first band energy ratio, first segment duration, and first descent rate of the current event are taken as the corresponding current short-time energy, current short-time zero-crossing rate, current short-time autocorrelation coefficient, current spectral centroid, current band energy ratio, current segment duration, and current descent rate.

[0144] Step 4612: When the current decline rate is less than the preset second decline rate threshold, identify whether the current short-term energy is higher than the preset first energy threshold; if yes, set the corresponding first identification type as mild event; if no, set the corresponding first identification type as uncertain event.

[0145] Here, the second rate of decline threshold and the first energy threshold are two preset threshold parameters, and the second rate of decline threshold is greater than the first rate of decline threshold;

[0146] Step 4613: When the current descent rate is greater than or equal to the second descent rate threshold and less than the preset third descent rate threshold, identify whether the current spectral centroid is greater than the preset first frequency threshold; if yes, set the corresponding first identification type as a moderate event; if no, set the corresponding first identification type as a mild event.

[0147] Here, the third descent rate threshold and the first frequency threshold are two preset threshold parameters, and the third descent rate threshold is greater than the second descent rate threshold;

[0148] Step 4614: When the current decline rate is greater than or equal to the third decline rate threshold or the current segment duration is greater than or equal to the preset second duration threshold, the current short-time zero-crossing rate, the current short-time autocorrelation coefficient, and the current frequency band energy ratio are identified; if the current short-time zero-crossing rate is less than the preset first zero-crossing rate threshold, or the current short-time autocorrelation coefficient is less than the preset first autocorrelation coefficient threshold, or the current frequency band energy ratio is greater than the preset second frequency band energy ratio threshold, then the corresponding first identification type is set as a severe event; if the current short-time zero-crossing rate is greater than or equal to the first zero-crossing rate threshold and the current short-time autocorrelation coefficient is greater than or equal to the first autocorrelation coefficient threshold and the current frequency band energy ratio is less than or equal to the second frequency band energy ratio threshold, then the corresponding first identification type is set as a moderate event.

[0149] Here, the second duration threshold, the first zero-crossing rate threshold, the first autocorrelation coefficient threshold, and the second frequency band energy ratio threshold are four preset threshold parameters, and the second duration threshold is greater than the first duration threshold, and the second frequency band energy ratio threshold is greater than the first frequency band energy ratio threshold.

[0150] Step 462, and record each first event in the first event sequence that is identified as a mild event, a moderate event, or a severe event as the corresponding first determined event;

[0151] Step 47, and when the total number of first determined events is not 0, perform a comprehensive score on the severity of obstructive sleep apnea based on all first determined events to obtain the corresponding first score;

[0152] Specifically, this includes: Step 471, calculating the total number of the first determined events to obtain the corresponding total number of events N. * The event decline rate r is obtained by averaging the first decline rates of all first deterministic events. * The time difference between the first start time and the second end time of each first determined event is taken as the corresponding event length, and the average of all event lengths is used to calculate the corresponding event duration L. * And record the first most recent duration as the corresponding L. ref Based on the total number of events N * The most recent duration L ref Calculate the corresponding event frequency f * =N * / L ref ;

[0153] Step 472, and the event frequency f * Event decline rate r* Duration of the event (L) * Normalization is performed on each event to obtain the corresponding normalized event frequency f. NORM Normalized event decline rate r NORM Duration L of normalization event NORM And based on the normalized event frequency f NORM Normalized event decline rate r NORM Duration L of normalization event NORM Calculate the corresponding first score = w1 × f NORM +w2×r NORM +w3×L NORM ;

[0154] Where w1+w2+w3=1, w1, w2, and w3 are the first weighting coefficient, the second weighting coefficient, and the third weighting coefficient, respectively.

[0155] Step 48: The first start time and second end time corresponding to each first determined event are taken as the corresponding third start time and third end time, and the time difference between the current third start time and third end time is taken as the corresponding first event duration; the first identification type corresponding to each first determined event is taken as the corresponding first event type; and a corresponding third feature record is formed by the third start time, third end time, first event duration, and first event type corresponding to each first determined event; and when it is confirmed that there are no other records in the third feature list that are duplicated with the content of the current third feature record, the current third feature record is added to the third feature list.

[0156] Step 49, and if the first score is not empty, use the current time period start time, current time period end time, and first score as the corresponding fourth start time, fourth end time, and first time period score; and set the normalized event frequency f corresponding to the current first score. NORM Normalized event decline rate r NORM Duration L of normalization event NORM The first, second, and third abnormal parameters are used as corresponding abnormal parameters; and the fourth start time, fourth end time, first abnormal parameter, second abnormal parameter, third abnormal parameter, and first time period score obtained this time are combined to form a corresponding first score record, which is added to the first score list.

[0157] Step 5: During the synchronous acquisition process, real-time warnings are issued based on the third feature list; and when the first warning strategy is not empty, additional warnings are issued based on the first warning strategy and the first scoring list.

[0158] Specifically, this includes: Step 51, during the synchronous acquisition process, real-time early warning is given based on the third feature list;

[0159] Specifically, this includes: when a third feature record is added to the third feature list, the first event type of the newly added third feature record is taken as the corresponding current event type; and the current real-time alarm status is identified.

[0160] If the real-time alarm status is non-alarm status, then a real-time alarm will be triggered according to the mild, moderate or severe alarm mode corresponding to the current event type. During each real-time alarm process, the real-time alarm status will be updated to the corresponding mild, moderate or severe alarm status, and the alarm duration of the current mild, moderate or severe alarm mode will be used as the corresponding current alarm duration. When the alarm duration exceeds the current alarm duration, the alarm will be stopped immediately, the real-time alarm status will be switched back to non-alarm status, and the current alarm duration will be cleared to zero.

[0161] If the real-time alarm status is mild, moderate, or severe, the alarm severity of the current event type is compared with the real-time alarm status. If the alarm severity is lower than the real-time alarm status, the alarm is ignored. If the alarm severity matches the real-time alarm status, the alarm duration is delayed based on the alarm duration of the mild, moderate, or severe alarm mode corresponding to the current event type. If the alarm severity is higher than the real-time alarm status, the current alarm is stopped immediately, and a real-time alarm is triggered according to the mild, moderate, or severe alarm mode corresponding to the current event type.

[0162] Here, the mild, moderate, and severe alarm modes of this invention each include a set of corresponding alarm volume and alarm vibration configuration parameters; the alarm volume and alarm vibration intensity of the mild, moderate, and severe alarm modes increase progressively.

[0163] Step 52, and when the first warning strategy is not empty, perform additional warnings based on the first warning strategy and the first score list;

[0164] Specifically, this includes: when a new first rating record is added to the first rating list, the newly added first rating record is taken as the corresponding current record; an alarm counter initialized to 0 is set for the current record; a one-to-one correspondence is established between the first abnormal parameter, the second abnormal parameter, and the third abnormal parameter of the current record and the first sub-item configuration of the first type, the second type, and the third type of sub-item parameter in the first early warning strategy; and the first sub-item configuration of the sub-item switch in the first early warning strategy that is turned on is taken as the corresponding valid sub-item configuration.

[0165] When the number of valid sub-item configurations is not 0, the sub-item logic judgment formula of the current valid sub-item configuration is identified based on the first abnormal parameter, second abnormal parameter or third abnormal parameter corresponding to any valid sub-item configuration. If it is satisfied, the alarm counter is incremented by 1.

[0166] When the final alarm counter is greater than 0, a corresponding custom alarm activation record is formed by each valid sub-item configuration that is satisfied by the sub-item logic judgment and its corresponding first abnormal parameter, second abnormal parameter or third abnormal parameter. A corresponding custom alarm activation report is formed by all the obtained custom alarm activation records and sent to each contact interface of the first contact configuration of the first alarm strategy.

[0167] When the final alarm counter is greater than 0, the current real-time alarm status is identified. If the real-time alarm status is a non-alarm status, a real-time alarm is triggered in three progressively higher levels: mild, moderate, and severe. If the real-time alarm status is a mild, moderate, or severe alarm status, the alarm is ignored.

[0168] Figure 2 This is a module structure diagram of an OSA monitoring and early warning device integrating snoring and blood oxygenation characteristics provided in Embodiment 2 of the present invention. This device can be a terminal device or server implementing the aforementioned method embodiments, or it can be a device that enables the aforementioned terminal device or server to implement the aforementioned method embodiments. For example, the device can be a device or chip system of the aforementioned terminal device or server. Figure 2 As shown, the device includes: an additional early warning strategy receiving module 201, a monitoring signal acquisition module 202, a signal feature extraction module 203, an OSA event processing module 204, and an early warning module 205.

[0169] The additional warning strategy receiving module 201 is used to receive additional warning strategies defined by the monitoring party or the subject before the subject enters a sleep state, as the corresponding first warning strategy.

[0170] The monitoring signal acquisition module 202 is used to continuously acquire synchronous sound wave signals and blood oxygen saturation signals from the subject after the subject enters a sleep state.

[0171] The signal feature extraction module 203 is used to generate a pair of synchronous sound wave signals and blood oxygen saturation signals as corresponding first sound wave signals and first blood oxygen signals at every preset acquisition time L0 during the synchronous acquisition process; and to separate the human voice signal and background signal of the first sound wave signal and normalize the two separated signals to obtain the corresponding first human voice signal and first background signal; and to filter and calibrate the first blood oxygen signal to obtain the corresponding first calibration signal; and to update the preset first signal list based on the two original acquisition signals and the corresponding three preprocessed signals; and to perform snoring frame recognition and snoring frame feature extraction processing based on the latest first human voice signal and first background signal and update the preset first feature list based on the feature extraction results; and to perform blood oxygen drop segment recognition and segment feature extraction processing based on the latest first calibration signal and update the preset second feature list based on the feature extraction results.

[0172] The OSA event processing module 204 is used to periodically identify the event characteristics of obstructive sleep apnea events according to the first and second feature lists during synchronous acquisition, and to comprehensively score the severity of obstructive sleep apnea based on the identification results, and update the preset third feature list and first score list based on the event feature identification results and the comprehensive score results.

[0173] The early warning module 205 is used to provide real-time early warning based on the third feature list during the synchronous acquisition process; and to provide additional early warning based on the first early warning strategy and the first scoring list when the first early warning strategy is not empty.

[0174] The OSA monitoring and early warning device that integrates snoring and blood oxygenation characteristics provided in this embodiment of the invention can execute the method steps in the above method embodiment. Its implementation principle and technical effect are similar, and will not be repeated here.

[0175] It should be noted that the division of the various modules in the above device is merely a logical functional division. In actual implementation, they can be fully or partially integrated into a single physical entity, or they can be physically separated. Furthermore, these modules can be implemented entirely in software via processing element calls; they can be fully implemented in hardware; or some modules can be implemented in software via processing element calls, while others are implemented in hardware. For example, the additional early warning strategy receiving module can be a separate processing element, or it can be integrated into a chip in the above device. Alternatively, it can be stored as program code in the memory of the above device, and called and executed by a processing element of the device. The implementation of other modules is similar. Moreover, these modules can be fully or partially integrated together, or they can be implemented independently. The processing element described here can be an integrated circuit with signal processing capabilities. In the implementation process, each step of the above method or each of the above modules can be completed through integrated logic circuits in the hardware of the processor element or through software instructions.

[0176] For example, these modules can be one or more integrated circuits configured to implement the above methods, such as one or more Application Specific Integrated Circuits (ASICs), one or more Digital Signal Processors (DSPs), or one or more Field Programmable Gate Arrays (FPGAs). As another example, when a module is implemented using processing element scheduler code, the processing element can be a general-purpose processor, such as a Central Processing Unit (CPU) or other processor capable of calling program code. Furthermore, these modules can be integrated together as a System-on-a-Chip (SOC).

[0177] In the above embodiments, implementation can be achieved, in whole or in part, through software, hardware, firmware, or any combination thereof. When implemented in software, it can be implemented, in whole or in part, as a computer program product. This computer program product includes one or more computer instructions. When these computer program instructions are loaded and executed on a computer, all or part of the processes or functions described in the foregoing method embodiments are generated. The computer described above can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The aforementioned computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the aforementioned computer instructions can be transmitted from one website, computer, server, or data center to another via wired (e.g., coaxial cable, fiber optic, Digital Subscriber Line (DSL)) or wireless (e.g., infrared, wireless, Bluetooth, microwave, etc.) means. The aforementioned computer-readable storage medium can be any available medium that a computer can access, or a data storage device such as a server or data center that integrates one or more available media. The aforementioned available media can be magnetic media (e.g., floppy disks, hard disks, magnetic tapes), optical media (e.g., DVDs), or semiconductor media (e.g., solid-state disks (SSDs)).

[0178] Figure 3 This is a schematic diagram of an electronic device provided in Embodiment 3 of the present invention. This electronic device can be a terminal device or server implementing the methods of the aforementioned embodiments, or it can be a terminal device or server connected to the aforementioned terminal device or server implementing the methods of the aforementioned embodiments. Figure 3As shown, the electronic device may include: a processor 301 (e.g., CPU), a memory 302, and a transceiver 303; the transceiver 303 is coupled to the processor 301, and the processor 301 controls the transmission and reception operations of the transceiver 303. The memory 302 may store various instructions for performing various processing functions and implementing the processing steps described in the foregoing embodiments. Preferably, the electronic device involved in the embodiments of the present invention further includes: a power supply 304, a system bus 305, and a communication port 306. The system bus 305 is used to realize communication connections between components. The communication port 306 is used for communication between the electronic device and other peripherals.

[0179] exist Figure 3 The system bus 305 mentioned can be a Peripheral Component Interconnect (PCI) bus or an Extended Industry Standard Architecture (EISA) bus, etc. This system bus can be divided into address bus, data bus, control bus, etc. For ease of representation, it is represented by only one thick line in the figure, but this does not indicate that there is only one bus or one type of bus. The communication interface is used to enable communication between the database access device and other devices (e.g., clients, read-write libraries, and read-only libraries). Memory may include Random Access Memory (RAM) and may also include non-volatile memory, such as at least one disk storage device.

[0180] The processors mentioned above can be general-purpose processors, including central processing units (CPUs), network processors (NPs), graphics processing units (GPUs), etc.; they can also be digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components.

[0181] It should be noted that the embodiments of the present invention also provide a computer-readable storage medium storing instructions that, when run on a computer, cause the computer to perform the methods and processes provided in the above embodiments.

[0182] This invention provides an OSA monitoring and early warning method, device, electronic device, and computer-readable storage medium that integrates snoring and blood oxygenation characteristics. As described above, this invention uses a user-defined additional early warning strategy as the first early warning strategy before the subject enters a sleep state; and after the subject enters a sleep state, it continuously collects synchronous sound wave signals and blood oxygen saturation signals; during the synchronous collection process, a pair of synchronous sound wave signals and blood oxygen saturation signals are generated every preset collection duration L0, and the collected sound wave signals are processed by separating and normalizing human voice and background signals, and the collected blood oxygen saturation signals are filtered and calibrated. A first signal list is updated based on the two original collected signals and the corresponding three pre-processed signals, and snoring frame recognition and snoring frame analysis are performed based on human voice and background signals. Feature extraction processing is performed, and the first feature list is updated based on the feature extraction results. Segment identification and feature extraction processing of decreased blood oxygenation based on calibration signals are also performed, and the second feature list is updated based on the feature extraction results. During synchronous data acquisition, event characteristics of obstructive sleep apnea events are periodically identified based on the first and second feature lists, and the severity of obstructive sleep apnea is comprehensively scored based on the identification results. The third feature list and the first scoring list are updated based on the event feature identification results and the comprehensive scoring results. During synchronous data acquisition, a fixed-process early warning is implemented based on the third feature list, and additional early warnings are implemented based on the first early warning strategy and the first scoring list when the first early warning strategy is not empty. Through this embodiment of the invention, both assessment accuracy and early warning flexibility are improved.

[0183] The steps of the methods or algorithms described in conjunction with the embodiments disclosed herein can be implemented in hardware, a software module executed by a processor, or a combination of both. The software module can be located in random access memory (RAM), main memory, read-only memory (ROM), electrically programmable ROM, electrically erasable programmable ROM, registers, hard disk, removable disk, CD-ROM, or any other form of storage medium known in the art.

[0184] The specific embodiments described above further illustrate the purpose, technical solution, and beneficial effects of the present invention. It should be understood that the above description is only a specific embodiment of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for monitoring and early warning of OSA (Obstructive Sleep Apnea) that integrates snoring and blood oxygenation characteristics, characterized in that, The method includes: Before the subject enters a sleep state, an additional early warning strategy defined by the monitoring party or the subject is received as the corresponding first early warning strategy. After the subject enters a sleep state, synchronous sound wave signals and blood oxygen saturation signals are continuously collected from the subject. During synchronous acquisition, at every preset acquisition duration L0, a pair of synchronized acoustic signals and blood oxygen saturation signals are generated as the corresponding first acoustic signal and first blood oxygen signal. The first acoustic signal is then separated into human voice signal and background signal, and the two separated signals are normalized to obtain the corresponding first human voice signal and first background signal. The first blood oxygen signal is then filtered and calibrated to obtain the corresponding first calibration signal. A preset first signal list is updated based on the two original acquisition signals and the corresponding three preprocessed signals. Snoring frame recognition and snoring frame feature extraction are performed based on the latest first human voice signal and first background signal, and a preset first feature list is updated based on the feature extraction results. Finally, blood oxygen drop segment recognition and segment feature extraction are performed based on the latest first calibration signal, and a preset second feature list is updated based on the feature extraction results. During the synchronous data collection process, the event characteristics of obstructive sleep apnea events are periodically identified based on the first and second feature lists, and the severity of obstructive sleep apnea is comprehensively scored based on the identification results. The preset third feature list and first score list are updated based on the event characteristic identification results and the comprehensive score results. During the synchronous acquisition process, real-time warnings are issued based on the third feature list; and when the first warning strategy is not empty, additional warnings are issued based on the first warning strategy and the first scoring list.

2. The OSA monitoring and early warning method integrating snoring and blood oxygenation characteristics according to claim 1, characterized in that, The first early warning strategy includes a first early warning configuration and a first contact configuration; the first early warning configuration includes multiple first sub-item configurations; each first sub-item configuration includes a sub-item switch configuration and a sub-item trigger condition configuration; the sub-item switch configuration includes on and off; the sub-item trigger condition configuration includes sub-item parameter type and sub-item logical judgment expression; the sub-item parameter type includes a first type, a second type, and a third type; the sub-item logical judgment expression consists of a logical judgment operator and a judgment threshold, the logical judgment operator includes less than, less than or equal to, equal to, greater than or equal to, and greater than; the first contact configuration includes one or more second sub-item configurations; the second sub-item configuration includes a contact name and a contact interface; The contact interface includes an SMS interface and an email interface; The signal start time of the first acoustic signal and the first blood oxygen signal is denoted as t. start The signal end time is denoted as t. end , t end =t start +L0; The first acoustic signal consists of multiple sampling point amplitudes x i The first blood oxygen signal is composed of the amplitude y of multiple sampling points. i Composition; 1 ≤ sampling point index i ≤ total number of sampling points N0, △t is a preset sampling time interval, L0>△t>0, and the sampling duration L0 is an integer multiple of the sampling time interval △t; The first signal list includes multiple first signal records; the first signal record includes the signal start time t. start The signal end time t end The first sound wave signal, the first blood oxygen signal, the first human voice signal, the first background signal, and the first calibration signal; The first feature list includes multiple first feature records; each first feature record corresponds to a snoring frame; the first feature record includes a first start time, a first end time, a first short-time energy, a first short-time zero-crossing rate, a first short-time autocorrelation coefficient, a first spectral centroid, a first frequency band energy ratio, and a first dominant frequency; The second feature list includes multiple second feature records; each second feature record corresponds to a blood oxygen decline segment; the second feature record includes a second start time, a second end time, a first segment duration, and a first decline rate; The third feature list includes multiple third feature records; each third feature record corresponds to one obstructive sleep apnea event; the third feature record includes a third start time, a third end time, a first event duration, and a first event type; the first event type includes mild events, moderate events, and severe events; The first scoring list includes multiple first scoring records; each first scoring record corresponds to a continuous monitoring period in which one or more obstructive sleep apnea events occurred; the first scoring record includes a fourth start time, a fourth end time, a first abnormal parameter, a second abnormal parameter, a third abnormal parameter, and a first time period score; the first abnormal parameter is a normalized value of the frequency of obstructive sleep apnea events in the current continuous monitoring period, the second abnormal parameter is a normalized value of the rate of decrease in blood oxygen saturation during obstructive sleep apnea events in the current continuous monitoring period, and the third abnormal parameter is a normalized value of the duration of obstructive sleep apnea events in the current continuous monitoring period.

3. The OSA monitoring and early warning method integrating snoring and blood oxygenation characteristics according to claim 2, characterized in that, The step of separating the human voice signal from the background signal from the first sound wave signal and normalizing the two separated signals to obtain the corresponding first human voice signal and first background signal specifically includes: Step 31: Perform background noise removal on the first sound wave signal based on a bandpass filter in the 50-500Hz frequency band to obtain the corresponding initial human voice signal; and extract background noise from the first sound wave signal based on a bandstop filter in the 50-500Hz frequency band to obtain the corresponding initial background signal. The initial human voice signal retains only the sound wave signal in the 50-500Hz frequency band, and is composed of N0 sampling point amplitudes. Composition; the initial background signal retains only sound wave signals below 50Hz or above 500Hz, composed of N0 sampling point amplitudes. composition; Step 32: Normalize the initial human voice signal and the initial background signal to obtain the corresponding first human voice signal and the first background signal; The first human voice signal includes N0 sampling point amplitudes. The first background signal includes the amplitude of N0 sampling points. u1 and σ1 are the average signal value and standard deviation of the initial human voice signal, respectively; u2 and σ2 are the average signal value and standard deviation of the initial background signal, respectively.

4. The OSA monitoring and early warning method integrating snoring and blood oxygenation characteristics according to claim 2, characterized in that, The step of filtering and calibrating the first blood oxygen signal to obtain the corresponding first calibration signal specifically includes: The first pulse noise is eliminated by applying a five-point median filter to the first blood oxygen signal to obtain the corresponding first filtered signal Y. 1 ; and the first filtered signal Y is processed by a three-point moving average filter. 1 The first smoothed signal Y is obtained by performing signal smoothing processing. 2 ; and for the first smoothed signal Y 2 Data calibration is performed to obtain the corresponding first calibration signal; wherein, Y 3 =aY 2 +b, Y 3 Here, a and b are the first calibration signal, and a and b are two calibration parameters of the signal acquisition device corresponding to the first blood oxygen signal.

5. The OSA monitoring and early warning method integrating snoring and blood oxygenation characteristics according to claim 2, characterized in that, The process of updating the preset first signal list based on two original acquired signals and three corresponding preprocessed signals specifically includes: The two original acquired signals and their corresponding signal start times t start The signal end time t end The three preprocessed signals are combined to form a corresponding first signal record and added to the first signal list; the two original acquired signals include the first acoustic signal and the first blood oxygen signal; the three preprocessed signals include the first human voice signal, the first background signal and the first calibration signal.

6. The OSA monitoring and early warning method integrating snoring and blood oxygenation characteristics according to claim 3, characterized in that, The process of performing snoring frame recognition and snoring frame feature extraction based on the latest first human voice signal and the first background signal, and updating the preset first feature list based on the feature extraction results, specifically includes: Step 61: Based on the preset short frame duration w and sliding duration s, perform sliding frame segmentation processing on the first human voice signal and the first background signal respectively to obtain the corresponding human voice frame sequence F. 1 and background frame sequence F 2 ; Wherein, L0>w>s>△t, the short frame duration w and the sliding duration s are both integer multiples of the sampling time interval △t, and the difference between the acquisition duration L0 and the short frame duration w, L0-w, is an integer multiple of the sliding duration s; The human voice frame sequence F 1 Includes multiple voice frames The human voice frame Including amplitude of multiple single-frame sampling points The background frame sequence F 2 Includes multiple background frames The background frame Including amplitude of multiple single-frame sampling points 1 ≤ frame index j ≤ total number of frames N1, 1 ≤ single frame sampling point index k ≤ total number of single frame sampling points N2; Step 62: Sequentially extract each of the aforementioned voice frames in chronological order. As the corresponding current voice frame, and the background frame corresponding to the current voice frame. As the corresponding current background frame; and calculate the corresponding short-time energy for the current voice frame and the current background frame respectively. and short-term energy and the current short-time energy As a corresponding historical background frame short-time energy, it is pushed into a preset short-time energy queue; and the median of all the historical background frame short-time energies in the short-time energy queue is identified to obtain the corresponding current median energy, and the product of the current median energy and a preset energy weighting coefficient is used as the corresponding current short-time energy threshold; and the short-time energy is... The system identifies whether the current short-term energy threshold is exceeded; if so, it sets the corresponding human voice frame type g. j If it is a snoring frame, then set the corresponding human voice frame type g. j Non-snoring frames; The short-time energy queue is a first-in-first-out circular queue with a fixed queue length L1, which is a preset positive integer; the energy value queue includes at most L1 of the most recent short-time energies of the historical background frames; and the energy weighting coefficient is a preset coefficient greater than 1. Step 63, from the obtained N1 human voice frame types g j The corresponding frame type sequence G is formed; and the human voice frame type g in the frame type sequence G, where adjacent frame types are all non-snoring frames, is also included. j Reset to a non-snoring frame; Step 64, select each of the human voice frame types g in the frame type sequence G. j The human voice frame of the snoring frame The current snoring frame is used as the corresponding current snoring frame; and the frame start time and frame end time corresponding to the current snoring frame are used as the corresponding first start time and first end time; and the short-time energy corresponding to the current snoring frame is used as the first start time and first end time. The first short-time energy is used as the corresponding first short-time energy; the short-time zero-crossing rate and short-time autocorrelation coefficient of the current snoring frame are calculated to obtain the corresponding first short-time zero-crossing rate and the first short-time autocorrelation coefficient; the current snoring frame is subjected to a fast Fourier transform to obtain the corresponding current frame spectrum; the spectral centroid, frequency band energy ratio and dominant frequency are identified based on the current frame spectrum to obtain the corresponding first spectral centroid, first frequency band energy ratio and first dominant frequency; the first start time, first end time, first short-time energy, first short-time zero-crossing rate, first short-time autocorrelation coefficient, first spectral centroid, first frequency band energy ratio and first dominant frequency corresponding to the current snoring frame are combined to form a corresponding first feature record; and when it is confirmed that there are no other records in the first feature list that are duplicated with the content of the current first feature record, the current first feature record is added to the first feature list.

7. The OSA monitoring and early warning method integrating snoring and blood oxygenation characteristics according to claim 4, characterized in that, The process of identifying blood oxygen-decrease segments and extracting segment features based on the latest first calibration signal, and updating the preset second feature list based on the feature extraction results, specifically includes: Step 71: Push the first calibration signal into a preset calibration signal queue as a corresponding historical calibration signal; calculate the average signal amplitude of all the historical calibration signals in the calibration signal queue and use the calculation result as the corresponding current blood oxygen threshold. The calibration signal queue is a first-in-first-out circular queue with a fixed queue length L2, which is a preset positive integer; the calibration signal queue contains at most L2 of the most recent historical calibration signals. Step 72: Record the amplitude of each sampling point on the first calibration signal that is less than the current blood oxygen threshold as the corresponding low-order amplitude; calculate the decrease rate of each low-order amplitude relative to the current blood oxygen threshold, decrease rate = (current blood oxygen threshold - low-order amplitude) / current blood oxygen threshold; record the signal sampling point corresponding to the decrease rate that exceeds a preset first decrease rate threshold as the corresponding decrease signal point; record the signal segment on the first calibration signal composed of multiple consecutive decrease signal points as the corresponding pre-selected segment, and record the pre-selected segment whose segment duration exceeds a preset first duration threshold as a corresponding blood oxygen decrease segment; Step 73: The average decline rate of all decline signal points in each of the blood oxygen decline segments is taken as the corresponding first decline rate; the segment start time, segment end time, and segment duration corresponding to each blood oxygen decline segment are taken as the corresponding second start time, second end time, and first segment duration; and a corresponding second feature record is formed by the second start time, second end time, first segment duration, and first decline rate corresponding to each blood oxygen decline segment; and when it is confirmed that there are no other records in the second feature list that are duplicates of the current second feature record, the current second feature record is added to the second feature list.

8. The OSA monitoring and early warning method integrating snoring and blood oxygenation characteristics according to claim 2, characterized in that, The process of periodically identifying event characteristics of obstructive sleep apnea events based on the first and second feature lists, comprehensively scoring the severity of obstructive sleep apnea based on the identification results, and updating the preset third feature list and first score list based on the event characteristic identification results and the comprehensive score results specifically includes: Step 81: According to a preset first time frequency, periodically use the current time as the corresponding end time of the current period, and subtract a preset first recent duration from the end time of the current period as the corresponding start time of the current period; and combine the start time of the current period with the end time of the current period to form the corresponding current period; and extract all the second feature records in the second feature list that are in the current period to form the corresponding second feature record sequence; Step 82, and when the second feature record sequence is not empty, perform snoring-blood oxygen pair sequence recognition based on the second feature record sequence to obtain the corresponding first sequence, specifically: take each second feature record of the second feature record sequence as the corresponding current record; take the second start time of the current record as the corresponding current timestamp; and take each first feature record in the first feature list whose first start time is before the current timestamp and whose time difference with the current timestamp does not exceed a preset first time difference range as the corresponding first record, and take the first record whose first start time is closest to the current timestamp as the corresponding current matching record; when the current matching record is not empty, form a corresponding snoring-blood oxygen pair by the current matching record and the current record; and sort all the obtained snoring-blood oxygen pairs in chronological order to form the corresponding first sequence; Step 83: When the first sequence is not empty, the first sequence is binary-classified based on the time difference distribution of the start time difference between the snoring frame and the blood oxygen decline segment. Specifically, the time difference between the first start time and the second start time of each snoring-blood oxygen pair in the first sequence is recorded as the corresponding first time difference; the current mean u3 and the current standard deviation σ3 are calculated based on the average and standard deviation of all the obtained first time differences; the current mean u3 and the current standard deviation σ3 are identified; if the current mean u3 meets the first time difference range and the current standard deviation σ3 does not exceed the preset time difference standard deviation threshold, the first sequence is marked as a Class I sequence; if the current mean u3 does not meet the first time difference range or the current standard deviation σ3 exceeds the time difference standard deviation threshold, the first sequence is marked as a Class II sequence. Step 84, and when the first sequence is labeled as a class of sequences, the correlation coefficient R between the snoring intensity of the first sequence and the decrease in blood oxygenation is calculated. E-r Perform calculations; H represents the total number of snoring-oxygen pairs in the first sequence, where 1 ≤ snoring-oxygen pair index h ≤ H; E avr E is the average of the H first short-time energies of the first sequence. h r is the first short-time energy of the h-th snoring-blood oxygen pair in the first sequence; avr r is the average of the H first decrease rates of the first sequence. h The first decrease rate is the h-th snoring-blood oxygen pair in the first sequence; Step 85, and in the correlation coefficient R E-r When the correlation coefficient exceeds a preset threshold, obstructive sleep apnea event identification is performed on the first sequence to obtain the corresponding first event sequence. Specifically, the snoring-blood oxygen pair in the first sequence where the first band energy ratio exceeds a preset first band energy ratio threshold or the first dominant frequency meets a preset first dominant frequency threshold is taken as a corresponding first event; and all the obtained first events are sorted in chronological order to form the corresponding first event sequence. Step 86, and when the first event sequence is not empty, identify the event type of each first event in the first event sequence to obtain the corresponding first identification type; and record each first event in the first event sequence whose first identification type is a mild event, a moderate event, or a severe event as the corresponding first determined event; The first identification type includes uncertain events, minor events, moderate events, and severe events; Step 87, and when the total number of the first determined events is not 0, a comprehensive score is obtained based on all the first determined events to assess the severity of obstructive sleep apnea and obtain the corresponding first score. Step 88: The first start time and the second end time corresponding to each of the first determined events are taken as the corresponding third start time and the third end time, and the time difference between the current third start time and the third end time is taken as the corresponding first event duration; the first identification type corresponding to each of the first determined events is taken as the corresponding first event type; and a corresponding third feature record is formed by the third start time, the third end time, the first event duration, and the first event type corresponding to each of the first determined events; and when it is confirmed that there are no other records in the third feature list that are duplicated with the content of the current third feature record, the current third feature record is added to the third feature list; Step 89, and when the first score is not empty, use the current time period start time, the current time period end time, and the first score as the corresponding fourth start time, fourth end time, and first time period score; and set the normalized event frequency f corresponding to the current first score as... NORM Normalized event decline rate r NORM Duration L of normalization event NORM As corresponding to the first abnormal parameter, the second abnormal parameter, and the third abnormal parameter; and the fourth start time, the fourth end time, the first abnormal parameter, the second abnormal parameter, the third abnormal parameter, and the first time period score obtained this time, a corresponding first score record is added to the first score list.

9. The OSA monitoring and early warning method integrating snoring and blood oxygenation characteristics according to claim 8, characterized in that, The step of identifying the event type of each of the first events in the first event sequence to obtain the corresponding first identification type specifically includes: Each of the first events in the first event sequence is taken as the corresponding current event; and the first short-time energy, the first short-time zero-crossing rate, the first short-time autocorrelation coefficient, the first spectral centroid, the first band energy ratio, the first segment duration, and the first descent rate of the current event are taken as the corresponding current short-time energy, current short-time zero-crossing rate, current short-time autocorrelation coefficient, current spectral centroid, current band energy ratio, current segment duration, and current descent rate; And when the current decline rate is less than a preset second decline rate threshold, it is identified whether the current short-term energy is higher than a preset first energy threshold; if yes, the corresponding first identification type is set as a mild event; if no, the corresponding first identification type is set as an uncertain event; wherein, the second decline rate threshold is greater than the first decline rate threshold; And when the current descent rate is greater than or equal to the second descent rate threshold and less than the preset third descent rate threshold, it is identified whether the current spectral centroid is greater than the preset first frequency threshold; if yes, the corresponding first identification type is set as a moderate event; if no, the corresponding first identification type is set as a mild event; wherein, the third descent rate threshold is greater than the second descent rate threshold; When the current drop rate is greater than or equal to the third drop rate threshold or the current segment duration is greater than or equal to a preset second duration threshold, the current short-time zero-crossing rate, the current short-time autocorrelation coefficient, and the current frequency band energy ratio are identified. If the current short-time zero-crossing rate is less than a preset first zero-crossing rate threshold, or the current short-time autocorrelation coefficient is less than a preset first autocorrelation coefficient threshold, or the current frequency band energy ratio is greater than a preset second frequency band energy ratio threshold, then the corresponding first identification type is set as a severe event. If the current short-time zero-crossing rate is greater than or equal to the first zero-crossing rate threshold, the current short-time autocorrelation coefficient is greater than or equal to the first autocorrelation coefficient threshold, and the current frequency band energy ratio is less than or equal to the second frequency band energy ratio threshold, then the corresponding first identification type is set as a moderate event. Wherein, the second duration threshold > the first duration threshold, and the second frequency band energy ratio threshold > the first frequency band energy ratio threshold.

10. The OSA monitoring and early warning method integrating snoring and blood oxygenation characteristics according to claim 8, characterized in that, The first score is obtained by comprehensively scoring the severity of obstructive sleep apnea based on all the first determined events, specifically including: The total number of events N is obtained by counting the total number of the first determined events. * And calculate the corresponding event decline rate r by averaging the first decline rates of all the first determined events. * The time difference between the first start time and the second end time of each of the first determined events is taken as the corresponding event length, and the average of all event lengths is calculated to obtain the corresponding event duration L. * And record the first most recent duration as the corresponding L. ref ; and based on the total number of events N * And the first most recent duration L ref Calculate the corresponding event frequency f * =N * / L ref ; And the event frequency f * The event decline rate r * The duration L of the event * Normalization is performed on each event to obtain the corresponding normalized event frequency f. NORM The normalized event decline rate r NORM and the duration L of the normalized event NORM ; and based on the normalized event frequency f NORM The normalized event decline rate r NORM and the duration L of the normalized event NORM Calculate the corresponding first score = w1 × f NORM +w2×r NORM +w3×L NORM Where w1+w2+w3=1, w1, w2, and w3 are the first, second, and third weighting coefficients preset respectively.

11. The OSA monitoring and early warning method integrating snoring and blood oxygenation characteristics according to claim 2, characterized in that, The real-time early warning based on the third feature list specifically includes: When a third feature record is added to the third feature list, the first event type of the newly added third feature record is taken as the corresponding current event type; and the current real-time alarm status is identified; if the real-time alarm status is a non-alarm status, a real-time alarm is triggered according to the mild, moderate, or severe alarm mode corresponding to the current event type, and during each real-time alarm process, the real-time alarm status is updated to the corresponding mild, moderate, or severe alarm status, and the alarm duration of the current mild, moderate, or severe alarm mode is taken as the corresponding current alarm duration; if the alarm duration exceeds the current alarm duration, the alarm is immediately stopped, the real-time alarm status is switched back to the non-alarm status, and the current alarm duration is cleared to zero; if the real-time alarm... If the alarm status is mild, moderate, or severe, the alarm severity of the current event type is compared with the real-time alarm status. If the alarm severity is lower than the real-time alarm status, the alarm is ignored. If the alarm severity matches the real-time alarm status, the current alarm duration is delayed based on the alarm duration of the mild, moderate, or severe alarm mode corresponding to the current event type. If the alarm severity is higher than the real-time alarm status, the current alarm is immediately stopped, and a real-time alarm is triggered according to the mild, moderate, or severe alarm mode corresponding to the current event type. Each of the mild, moderate, and severe alarm modes includes a set of corresponding alarm volume and alarm vibration configuration parameters. The alarm volume and alarm vibration intensity of the mild, moderate, and severe alarm modes increase progressively.

12. The OSA monitoring and early warning method integrating snoring and blood oxygenation characteristics according to claim 2, characterized in that, The additional warning based on the first warning strategy and the first scoring list specifically includes: When a new first rating record is added to the first rating list, the newly added first rating record is taken as the corresponding current record; an alarm counter initialized to 0 is set for the current record; a one-to-one correspondence is established between the first abnormal parameter, the second abnormal parameter, and the third abnormal parameter of the current record and the first sub-item configuration of the first type, the second type, and the third type of the sub-item parameter in the first early warning strategy; the first sub-item configuration with the sub-item switch set to "on" in the first early warning strategy is taken as the corresponding valid sub-item configuration; and when the number of valid sub-item configurations is not 0, the sub-item logic judgment expression of the current valid sub-item configuration is checked based on the first abnormal parameter, the second abnormal parameter, or the third abnormal parameter corresponding to any one of the valid sub-item configurations. If the condition is met, the alarm counter is incremented by 1. When the final alarm counter is greater than 0, a corresponding custom early warning activation record is formed by each valid sub-item configuration that is satisfied in this sub-item logic judgment and its corresponding first abnormal parameter, second abnormal parameter or third abnormal parameter. A corresponding custom early warning activation report is formed by all the obtained custom early warning activation records and sent to each contact interface of the first contact configuration of the first early warning strategy. When the final alarm counter is greater than 0, the current real-time alarm status is identified. If the real-time alarm status is a non-alarm status, a real-time alarm is triggered in a progressively advancing alarm mode of mild, moderate and severe. If the real-time alarm status is a mild, moderate or severe alarm status, the alarm is ignored.

13. An apparatus for performing the OSA monitoring and early warning method integrating snoring and blood oxygenation characteristics as described in any one of claims 1-12, characterized in that, The device includes: an additional early warning strategy receiving module, a monitoring signal acquisition module, a signal feature extraction module, an OSA event processing module, and an early warning module; The additional warning strategy receiving module is used to receive additional warning strategies defined by the monitoring party or the subject before the subject enters a sleep state, as the corresponding first warning strategy. The monitoring signal acquisition module is used to continuously acquire synchronous sound wave signals and blood oxygen saturation signals from the subject after the subject enters a sleep state. The signal feature extraction module is used to generate a pair of synchronous sound wave signals and blood oxygen saturation signals as corresponding first sound wave signals and first blood oxygen signals at every preset acquisition time L0 during the synchronous acquisition process; and to separate the human voice signal and background signal from the first sound wave signal and normalize the two separated signals to obtain the corresponding first human voice signal and first background signal; and to filter and calibrate the first blood oxygen signal to obtain the corresponding first calibration signal; and to update the preset first signal list based on the two original acquisition signals and the corresponding three preprocessed signals; and to perform snoring frame recognition and snoring frame feature extraction processing based on the latest first human voice signal and first background signal and update the preset first feature list based on the feature extraction results; and to perform blood oxygen drop segment recognition and segment feature extraction processing based on the latest first calibration signal and update the preset second feature list based on the feature extraction results. The OSA event processing module is used to periodically identify the event characteristics of obstructive sleep apnea events according to the first and second feature lists during the synchronous acquisition process, and to comprehensively score the severity of obstructive sleep apnea based on the identification results, and update the preset third feature list and first score list based on the event feature identification results and the comprehensive score results. The early warning module is used to provide real-time early warnings based on the third feature list during the synchronous data acquisition process; and to provide additional early warnings based on the first early warning strategy and the first scoring list when the first early warning strategy is not empty.

14. An electronic device, characterized in that, include: Memory, processor, and transceiver; The processor is configured to be coupled to the memory, read and execute instructions in the memory to implement the method according to any one of claims 1-12; The transceiver is coupled to the processor, and the processor controls the transceiver to send and receive messages.

15. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer instructions that, when executed by a computer, cause the computer to perform the method according to any one of claims 1-12.

Images