Sleep apnea detection method and device and electronic equipment

Abstract

The invention discloses a sleep apnea detection method and device and electronic equipment. The method comprises the following steps: acquiring target physiological signal data corresponding to a target object; extracting target respiratory signal data from the target physiological signal data; calling a reference frequency amplitude fluctuation index; according to the target respiratory signal data, multiple pieces of sub respiratory signal data are determined; determining sub-respiration frequency amplitude characteristics corresponding to the multiple pieces of sub-respiration signal data; and determining an apnea detection result of the target object according to the reference frequency amplitude fluctuation index and the sub-respiration frequency amplitude characteristics corresponding to the multiple pieces of sub-respiration signal data. The sleep apnea detection method and the sleep apnea detection device solve the technical problem that sleep apnea detection is inaccurate when sleep apnea detection is carried out according to physiological signals.

Description

Technical Field

[0001] The present invention relates to the field of data processing, and in particular to a sleep apnea detection method, device and electronic equipment. Background Art

[0002] In the related art, when performing sleep apnea detection, the complexity of the sleeping environment, such as ambient noise, may interfere with physiological signals, resulting in inaccurate sleep apnea detection when performing sleep apnea detection based on physiological signals.

[0003] To address the above-mentioned problems, no effective solutions have been proposed so far. Summary of the Invention

[0004] Embodiments of the present invention provide a sleep apnea detection method, apparatus, and electronic device to at least solve the technical problem of inaccurate sleep apnea detection when performing sleep apnea detection based on physiological signals.

[0005] According to one aspect of an embodiment of the present invention, a sleep apnea detection method is provided, comprising: acquiring target physiological signal data corresponding to a target object; extracting target respiratory signal data from the target physiological signal data; retrieving a reference frequency-amplitude fluctuation index, wherein the reference frequency-amplitude fluctuation index is obtained based on a plurality of respiratory frequency-amplitude distributions, which are respectively obtained based on corresponding sample respiratory signal data, which are respectively obtained based on corresponding sample physiological signal data; determining a plurality of sub-respiratory signal data based on the target respiratory signal data; determining sub-respiratory frequency-amplitude features respectively corresponding to the plurality of sub-respiratory signal data; and determining an apnea detection result for the target object based on the reference frequency-amplitude fluctuation index and the sub-respiratory frequency-amplitude features respectively corresponding to the plurality of sub-respiratory signal data.

[0006] Optionally, before retrieving the reference frequency-amplitude fluctuation index, the method further includes: when the sample respiratory signal data includes multiple respiratory frequency-amplitude data, dividing the sample respiratory signal data to obtain multiple sample signal data segments, wherein the multiple sample signal data segments each include at least one respiratory frequency-amplitude data; based on the multiple respiratory frequency-amplitude data, determining the sample respiratory frequency-amplitude characteristics corresponding to the multiple sample signal data segments respectively; based on the sample respiratory frequency-amplitude characteristics corresponding to the multiple sample signal data segments respectively, determining the respiratory frequency-amplitude distribution corresponding to the sample respiratory signal data.

[0007] Optionally, dividing the sample respiratory signal data to obtain a plurality of sample signal data segments includes: determining a time overlap index; and dividing the sample respiratory signal data according to the time overlap index to obtain a plurality of sample signal data segments.

[0008] Optionally, determining the respiratory amplitude distribution corresponding to the sample respiratory signal data based on the sample respiratory amplitude characteristics corresponding to the multiple sample signal data segments also includes: when the sample respiratory amplitude characteristics include an average deviation index and the respiratory amplitude distribution includes a average deviation histogram, determining the number of occurrences corresponding to multiple target deviation values, wherein the multiple average deviation indices include the multiple target deviation values; and obtaining the average deviation histogram corresponding to the sample respiratory signal data based on the multiple target deviation values ​​and the number of occurrences corresponding to the multiple target deviation values.

[0009] Optionally, the time overlap index includes any one of the following: a time overlap ratio of 50%.

[0010] Optionally, before retrieving the reference frequency amplitude fluctuation index, it also includes: determining the amplitude distribution interval corresponding to the multiple respiratory frequency amplitude distributions based on the multiple matching indexes corresponding to the multiple respiratory frequency amplitude distributions; determining the selected fluctuation indexes corresponding to the multiple respiratory frequency amplitude distributions based on the amplitude distribution interval; determining the sample evaluation results corresponding to the multiple respiratory frequency amplitude distributions based on the multiple selected fluctuation indices; and determining the reference frequency amplitude fluctuation index from the multiple selected fluctuation indices based on the sample evaluation results corresponding to the multiple respiratory frequency amplitude distributions.

[0011] Optionally, determining a reference frequency amplitude fluctuation index from the multiple fluctuation indices to be selected based on the sample evaluation results corresponding to the multiple respiratory frequency amplitude distributions includes: determining predetermined detection results corresponding to the multiple respiratory frequency amplitude distributions; determining multiple matching indices corresponding to the multiple respiratory frequency amplitude distributions, wherein the multiple matching indices represent the degree of matching between the sample evaluation results and predetermined detection results of the corresponding respiratory frequency amplitude distributions under the multiple fluctuation indices to be selected; determining a reference frequency amplitude fluctuation index from the multiple fluctuation indices to be selected based on the multiple matching indices corresponding to the multiple respiratory frequency amplitude distributions, wherein the reference frequency amplitude fluctuation index is the selected fluctuation index with the largest matching index among the multiple fluctuation indices to be selected.

[0012] Optionally, when the respiratory frequency amplitude distribution includes a mean deviation histogram, the amplitude distribution interval includes any one of the following: the first 20% of multiple mean deviation indices arranged from small to large in the mean deviation histogram, and the reference frequency amplitude fluctuation index includes any one of the following: 65%.

[0013] Optionally, extracting target respiratory signal data from the target physiological signal data includes: determining a respiratory frequency range; determining a cutoff frequency parameter based on the respiratory frequency range; filtering the target physiological signal data based on the cutoff frequency parameter to obtain initial respiratory signal data; and determining target respiratory signal data based on the initial respiratory signal data.

[0014] Optionally, determining the target respiratory signal data based on the initial respiratory signal data includes: dividing the initial respiratory signal data into multiple initial sub-data segments in chronological order; performing windowing processing on the multiple initial sub-data segments to obtain multiple target sub-data segments, wherein the windowing processing is used to make the signal data value of the first time point and the signal data value of the last time point in the corresponding initial sub-data segment smoothly decay to a predetermined value, and the target sub-data segment corresponds one-to-one to the multiple initial sub-data segments; determining the frequency and amplitude parameters corresponding to the multiple target sub-data segments respectively; and determining the target respiratory signal data based on the frequency and amplitude parameters corresponding to the multiple target sub-data segments respectively.

[0015] Optionally, determining the apnea detection result for the target object based on the reference frequency fluctuation index and the sub-respiration frequency characteristics corresponding to the multiple sub-respiration signal data includes: determining multiple candidate signal data from the multiple sub-respiration signal data based on the sub-respiration frequency characteristics corresponding to the multiple sub-respiration signal data; dividing each signal data in the multiple candidate signal data to obtain multiple signal segments corresponding to the multiple candidate signal data; determining the segment frequency characteristics corresponding to the multiple signal segments corresponding to each candidate signal data; determining the apnea detection result for the target object based on the segment frequency characteristics corresponding to the multiple signal segments corresponding to each candidate signal data, wherein the apnea detection result includes the signal detection results corresponding to the multiple candidate signal data.

[0016] Optionally, the apnea detection result for the target object is determined based on the segment frequency and amplitude characteristics corresponding to the multiple signal segments corresponding to each candidate signal data, including: when the multiple signal segments corresponding to each candidate signal data are multiple signal segments arranged in a predetermined order, determining the arrangement positions corresponding to the multiple signal segments; determining the feature difference index corresponding to each candidate signal data based on the arrangement positions and segment frequency and amplitude characteristics corresponding to the multiple signal segments, wherein the feature difference index is used to represent the degree of feature difference between signal segments of adjacent arrangement positions; determining the apnea detection result for the target object based on the feature difference index corresponding to each candidate signal data.

[0017] Optionally, based on the target respiratory signal data, multiple sub-respiratory signal data are determined, including: determining a division index corresponding to the target respiratory signal data, wherein the division index includes: data time span, time overlap rate, and division strategy, and the division strategy is determined based on data features corresponding to the target respiratory signal data; based on the division index, the target respiratory signal data is divided to obtain multiple sub-respiratory signal data.

[0018] According to one aspect of an embodiment of the present invention, a sleep apnea detection device is provided, comprising: an acquisition module for acquiring target physiological signal data corresponding to a target object; an extraction module for extracting target respiratory signal data from the target physiological signal data; a calling module for calling a reference frequency-amplitude fluctuation index, wherein the reference frequency-amplitude fluctuation index is obtained based on multiple respiratory frequency-amplitude distributions, which are respectively obtained based on corresponding sample respiratory signal data, and the corresponding sample respiratory signal data are obtained based on corresponding sample physiological signal data; a first determination module for determining multiple sub-respiratory signal data based on the target respiratory signal data; a second determination module for determining sub-respiratory frequency-amplitude features corresponding to the multiple sub-respiratory signal data; and a third determination module for determining an apnea detection result for the target object based on the reference frequency-amplitude fluctuation index and the sub-respiratory frequency-amplitude features corresponding to the multiple sub-respiratory signal data.

[0019] According to one aspect of an embodiment of the present invention, an electronic device is provided, comprising: a processor; and a memory for storing processor-executable instructions; wherein the processor is configured to execute the instructions to implement any one of the above sleep apnea detection methods.

[0020] According to one aspect of an embodiment of the present invention, a computer-readable storage medium is provided. When instructions in the computer-readable storage medium are executed by a processor of an electronic device, the electronic device is enabled to perform any of the above sleep apnea detection methods.

[0021] In an embodiment of the present invention, target physiological signal data corresponding to a target subject is acquired; target respiratory signal data is extracted from the target physiological signal data; a reference frequency-amplitude fluctuation index is retrieved, wherein the reference frequency-amplitude fluctuation index is obtained based on multiple respiratory frequency-amplitude distributions, each of which is obtained based on corresponding sample respiratory signal data, each of which is obtained based on corresponding sample physiological signal data; multiple sub-respiratory signal data are determined based on the target respiratory signal data; sub-respiratory frequency-amplitude characteristics corresponding to each of the multiple sub-respiratory signal data are determined; and an apnea detection result for the target subject is determined based on the reference frequency-amplitude fluctuation index and the sub-respiratory frequency-amplitude characteristics corresponding to each of the multiple sub-respiratory signal data. By extracting the target respiratory signal data from the acquired target physiological signal data and retrieving the reference frequency-amplitude fluctuation index, since the reference frequency-amplitude fluctuation index is obtained based on the respiratory frequency-amplitude distributions corresponding to each of the multiple sample respiratory signal data, it can adapt to the differences in the detected subjects, thereby analyzing the sub-respiratory frequency-amplitude characteristics of the respiratory signal data segment by segment, and combining the reference frequency-amplitude fluctuation index, the accuracy of sleep apnea detection can be effectively improved, thereby solving the technical problem of inaccurate sleep apnea detection when performing sleep apnea detection based on physiological signals. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] The drawings described herein are used to provide a further understanding of the present invention and constitute a part of this application. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute an improper limitation of the present invention. In the drawings:

[0023] Figure 1 is a flow chart of a sleep apnea detection method according to an embodiment of the present invention;

[0024] Figure 2 is a flow chart of the sleep apnea detection method of the present invention;

[0025] Figure 3 is a visualization diagram of the mixed original signal in an optional embodiment of the present invention;

[0026] Figure 4 is a schematic diagram of a filtered result in an optional embodiment of the present invention;

[0027] Figure 5 is a schematic diagram of a time window for an apnea event in an optional embodiment of the present invention;

[0028] Figure 6 is a histogram based on the mean absolute deviation of respiratory waveform amplitudes in an optional embodiment of the present invention;

[0029] Figure 7 is a graph showing the relationship between AUC value and threshold selection percentage in an optional embodiment of the present invention;

[0030] Figure 8 is a schematic diagram of a systematic measured evaluation result in an optional embodiment of the present invention;

[0031] Figure 9 is a schematic diagram of a sleep apnea detection system in an optional embodiment of the present invention;

[0032] Figure 10 FIG. 4 is a structural block diagram of a sleep apnea detection device according to an embodiment of the present invention. DETAILED DESCRIPTION

[0033] In order to enable those skilled in the art to better understand the solutions of the present invention, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts should fall within the scope of protection of the present invention.

[0034] It should be noted that the terms "first", "second", etc. in the description and claims of the present invention and the above-mentioned drawings are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that the numbers used in this way can be interchanged where appropriate, so that the embodiments of the present invention described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions. For example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0035] First, some nouns or terms that appear in the description of the embodiments of the present application are subject to the following interpretations:

[0036] Fiber optic BCG sensor: A fiber optic BCG sensor is a physiological sensor that uses fiber optic technology to detect ballistocardiogram (BCG) signals.

[0037] Example 1

[0038] According to an embodiment of the present invention, an embodiment of a sleep apnea detection method is provided. It should be noted that the steps shown in the flowcharts of the accompanying drawings can be executed in a computer system, such as a set of computer-executable instructions, and although a logical order is shown in the flowcharts, in some cases, the steps shown or described may be executed in an order different from that shown.

[0039] Figure 1 FIG. 1 is a flow chart of a sleep apnea detection method according to an embodiment of the present invention. Figure 1 As shown, the method includes the following steps:

[0040] S102, acquiring target physiological signal data corresponding to the target object;

[0041] In step S102 provided in the present application, target physiological signal data corresponding to the target object is acquired.

[0042] The target object is a person who needs to undergo sleep apnea testing. For example, the target object may be a person who needs to undergo sleep apnea testing, and the person may be a person of different ages.

[0043] This involves target physiological signal data, which is mixed signal data collected from the target subject and reflects their physiological state during sleep, including heartbeat signal data, respiratory signal data, body movement signal data, etc. This target physiological signal data can be collected through non-contact signal collection methods. For example, a fiber optic BCG sensor can be used to collect non-contact signals from the target subject to obtain target physiological signal data.

[0044] The acquired target physiological signal data includes signal data of multiple physiological components such as heartbeat activity, respiratory movement, and body movement, which can comprehensively reflect the physiological activities of the target object during sleep and provide a reliable analysis basis for subsequent apnea detection of the target object.

[0045] S104, extracting target respiratory signal data from the target physiological signal data;

[0046] In step S104 provided in the present application, target respiratory signal data is extracted from target physiological signal data.

[0047] This involves target respiratory signal data, which is extracted from target physiological signal data and is used to reflect the respiratory characteristics of the target subject. For example, the target respiratory signal data can be extracted from the target physiological signal data using a bandpass filter to remove ambient noise and other physiological component data unrelated to apnea detection, such as high-frequency heartbeat interference and low-frequency noise.

[0048] By extracting the target respiratory signal data from the target physiological signal data, the respiratory state of the target subject can be reflected in a purer form, thereby avoiding the interference of environmental noise and other irrelevant physiological signals on subsequent apnea detection.

[0049] S106, retrieving a reference frequency-amplitude fluctuation index, wherein the reference frequency-amplitude fluctuation index is obtained based on a plurality of respiratory frequency-amplitude distributions, wherein the plurality of respiratory frequency-amplitude distributions are respectively obtained based on corresponding sample respiratory signal data, wherein the corresponding sample respiratory signal data is obtained based on corresponding sample physiological signal data;

[0050] In step S106 provided in the present application, a reference frequency fluctuation index is retrieved.

[0051] The present invention involves a reference frequency fluctuation index, which is an index used to distinguish between non-apneic events and apneic events. The reference frequency fluctuation index is obtained based on the respiratory frequency distribution of multiple sample subjects. For example, by acquiring sample signal data from multiple sample subjects and analyzing the sample signal data of each sample subject separately to obtain the respiratory frequency distribution of each sample signal data, the respiratory frequency distribution of each sample signal data is then comprehensively analyzed to determine a reference frequency fluctuation index for use in determining whether an apneic event occurred within a corresponding time period. Specifically, when the respiratory frequency distribution includes a normalized mean absolute deviation (NMAAD) histogram (i.e., a mean deviation histogram), the NMAAD histograms of the multiple sample subjects are analyzed and the top 20% histogram value interval is selected as the optimal histogram parameter. The AUC (Area Under the Curve) is calculated. The threshold selection percentage corresponding to the highest AUC is 65%, which achieves the best overall detection performance. In this case, the reference frequency fluctuation index is 65%. The horizontal axis of the NMAAD histogram is the NMAAD value of the respiratory amplitude, and the vertical axis is the frequency of occurrence of the corresponding NMAAD value, and the NMAAD values ​​on the horizontal axis are arranged from small to large.

[0052] This involves multiple respiratory rate-amplitude distributions, which are obtained based on sample respiratory signal data from multiple sample subjects and are used to reflect the distribution of respiratory rate-amplitude characteristics (such as signal strength) of the corresponding sample subjects. For example, when the multiple respiratory rate-amplitude distributions are multiple NMAAD histograms, they can reflect the respiratory signal strength characteristics of the corresponding sample subjects, including the mean, variance, and distribution shape of the signal amplitude.

[0053] The sample respiratory signal data is extracted from the sample physiological signal data of the corresponding sample subject and is used to reflect the respiratory state of the sample subject. The acquisition of the sample respiratory signal data can exclude interference from other physiological signals (such as heartbeat signal data) and environmental noise.

[0054] The sample physiological signal data is collected from the corresponding sample subject and is mixed signal data reflecting the physiological state of the subject during sleep.

[0055] By analyzing the respiratory frequency and amplitude distribution of multiple sample subjects, the general patterns and individual differences in respiratory amplitude changes can be captured, and a reference frequency fluctuation index can be determined based on multiple respiratory frequency and amplitude distributions. This can better adapt to the physiological differences and environmental changes between subjects, and thus help improve the accuracy of sleep apnea detection.

[0056] S108, determining a plurality of sub-respiration signal data according to the target respiration signal data;

[0057] In step S108 provided in the present application, a plurality of sub-respiration signal data are determined according to the target respiration signal data.

[0058] This involves multiple sub-respiratory signal data sets, which are further divided from the target respiratory signal data and represent different sub-time periods in the time series. For example, the total time period corresponding to the target respiratory signal data is divided into a series of fixed-length time windows. The time window size can be 20 seconds, 60 seconds, etc., and can be determined based on the specific scenario. Within each time window, the corresponding respiratory signal segment is extracted from the target respiratory signal data to obtain the corresponding sub-respiratory signal data.

[0059] Multiple sub-respiratory signal data provide the target subject's respiratory status information on a shorter time scale, which helps to analyze the respiratory status in detail and is conducive to accurately locating the occurrence time of respiratory arrest events.

[0060] In addition, during the division of time windows, window overlap processing can be performed. For example, a certain overlap (such as a 50% overlap rate) will be designed between adjacent time windows to ensure the continuity and accuracy of apnea detection and avoid missing respiratory disorder events due to time window division.

[0061] S110, determining sub-respiration frequency and amplitude features corresponding to the plurality of sub-respiration signal data respectively;

[0062] In step S110 provided in the present application, sub-respiration frequency-amplitude features corresponding to a plurality of sub-respiration signal data are determined.

[0063] Among them, the sub-respiratory frequency and amplitude feature is involved, which reflects the respiratory intensity feature of the corresponding sub-respiratory signal data, such as the normalized mean absolute deviation (NMAAD) value (denoted as the mean deviation value), which is used to evaluate and reflect the stability and volatility of the target object's respiratory state within the time period (i.e., time window) corresponding to the sub-respiratory signal data.

[0064] By determining the sub-respiratory frequency and amplitude features corresponding to multiple sub-respiratory signal data, it is possible to quantitatively evaluate the intensity changes and stability characteristics of the respiratory signal within the time window. Each sub-respiratory frequency and amplitude feature corresponds to a specific time window, which can help to accurately locate the time point of potential apnea events.

[0065] S112 , determining an apnea detection result of the target subject based on the reference frequency fluctuation index and the sub-respiration frequency characteristics corresponding to the plurality of sub-respiration signal data.

[0066] In step S112 provided in the present application, an apnea detection result of the target object is determined based on the reference frequency fluctuation index and the sub-respiration frequency characteristics corresponding to the plurality of sub-respiration signal data.

[0067] Among them, the apnea detection result is involved. The apnea detection result is obtained by comparing and analyzing the sub-respiratory frequency and amplitude characteristics corresponding to the reference frequency fluctuation index and multiple sub-respiratory signal data, and obtaining the detection results related to apnea, including whether an apnea event occurs, the time point corresponding to the apnea event, the severity of the apnea, etc.

[0068] By comparing and analyzing the sub-respiratory frequency characteristics corresponding to multiple sub-respiratory signal data with the reference frequency fluctuation index, it is possible to quickly and accurately identify which sub-time periods have experienced respiratory arrest events, thereby improving the accuracy and timeliness of detection.

[0069] Through the above steps S102-S112, target physiological signal data corresponding to the target object is obtained; target respiratory signal data is extracted from the target physiological signal data; a reference frequency-amplitude fluctuation index is retrieved, wherein the reference frequency-amplitude fluctuation index is obtained based on multiple respiratory frequency-amplitude distributions, and the multiple respiratory frequency-amplitude distributions are respectively obtained based on corresponding sample respiratory signal data, and the corresponding sample respiratory signal data are obtained based on corresponding sample physiological signal data; based on the target respiratory signal data, multiple sub-respiratory signal data are determined; sub-respiratory frequency-amplitude features corresponding to the multiple sub-respiratory signal data are determined; based on the reference frequency-amplitude fluctuation index and the sub-respiratory frequency-amplitude features corresponding to the multiple sub-respiratory signal data, an apnea detection result corresponding to the target object is determined. By extracting target respiratory signal data from the acquired target physiological signal data and retrieving the reference frequency fluctuation index, since the reference frequency fluctuation index is obtained based on the respiratory frequency distribution corresponding to multiple sample respiratory signal data, it can adapt to the differences in the detection objects, and then analyze the sub-respiratory frequency characteristics of the respiratory signal data segment by segment, and combine with the reference frequency fluctuation index, it can effectively improve the accuracy of sleep apnea detection, thereby solving the technical problem of inaccurate sleep apnea detection when performing sleep apnea detection based on physiological signals.

[0070] As an optional embodiment, before calling the reference frequency-amplitude fluctuation index, it also includes: when the sample respiratory signal data includes multiple respiratory frequency-amplitude data, dividing the sample respiratory signal data to obtain multiple sample signal data segments, wherein the multiple sample signal data segments each include at least one respiratory frequency-amplitude data; based on the multiple respiratory frequency-amplitude data, determining the sample respiratory frequency-amplitude characteristics corresponding to the multiple sample signal data segments respectively; based on the sample respiratory frequency-amplitude characteristics corresponding to the multiple sample signal data segments respectively, determining the respiratory frequency-amplitude distribution corresponding to the sample respiratory signal data.

[0071] In this embodiment, the specific steps before calling the reference frequency fluctuation index are described.

[0072] Specifically, based on the sample respiratory amplitude characteristics corresponding to multiple sample signal data segments, the respiratory amplitude distribution corresponding to the sample respiratory signal data is determined, and it also includes: when the sample respiratory amplitude characteristics include an average deviation index and the respiratory amplitude distribution includes an average deviation histogram, determining the number of occurrences corresponding to multiple target deviation values, wherein the multiple average deviation indices include multiple target deviation values; based on the multiple target deviation values ​​and the number of occurrences corresponding to the multiple target deviation values, obtaining the average deviation histogram corresponding to the sample respiratory signal data.

[0073] Wherein, a plurality of respiratory frequency and amplitude data are involved. The plurality of respiratory frequency and amplitude data are a series of data included in the sample respiratory signal data and arranged in time sequence, reflecting the change of the respiratory signal intensity of the sample object during the detection time period.

[0074] This involves multiple sample signal data segments, which are divided in chronological order based on characteristics such as the sample subject's respiratory cycle. The signal data segments corresponding to each time period are obtained, including respiratory waveform amplitude, etc. For example, taking the sample subject's respiratory cycle as 60 seconds, the sample respiratory signal data is divided into equal 60-second time windows to obtain multiple sample signal data segments, each corresponding to a 60-second time window.

[0075] This involves sample respiratory frequency and amplitude characteristics, which are obtained through statistical analysis of the respiratory frequency and amplitude data in each sample signal data segment and can reflect the strength and stability of the respiratory signal within that time period. These characteristics include the mean deviation index (i.e., NMAAD), such as the mean, variance, maximum, minimum, or standardized mean absolute deviation (NMAAD) value obtained based on the respiratory waveform amplitude in each time window. Taking the NMAAD value as an example, the NMAAD histogram (i.e., respiratory frequency and amplitude distribution) corresponding to the sample respiratory signal data is determined based on the NMAAD values ​​corresponding to multiple sample signal data segments.

[0076] Multiple target deviation values ​​are involved, and these target deviation values ​​are non-repeating average deviation indices that appear in multiple average deviation indices. For example, if the average deviation indices are arranged in ascending order [10, 10, 20, 30, 20, 50], then the target deviation values ​​are arranged in ascending order [10, 20, 30, 50].

[0077] The number of occurrences is used to represent the frequency of each target deviation value. For example, if the average deviation indices are arranged from small to large [10, 10, 20, 30, 20, 50], and the target deviation values ​​are arranged from small to large [10, 20, 30, 50], then the number of occurrences of the target deviation value 10 is 2, and the number of occurrences of the target deviation value 30 is 1. The number of occurrences can be represented by the vertical axis of the average deviation histogram.

[0078] Through the above steps, the sample respiratory signal data is divided into multiple sample signal data segments, and the sample respiratory frequency-amplitude characteristics corresponding to each sample signal data segment are determined. This can quantify the strength and stability of the respiratory signal, thereby helping to construct a respiratory frequency-amplitude distribution that can accurately reflect the respiratory state of the sample subject.

[0079] As an optional embodiment, dividing the sample respiratory signal data to obtain multiple sample signal data segments includes: determining a time overlap index; and dividing the sample respiratory signal data according to the time overlap index to obtain multiple sample signal data segments.

[0080] In this embodiment, the specific steps of dividing the sample respiratory signal data to obtain a plurality of sample signal data segments are described.

[0081] Among them, the time overlap index is involved. The time overlap index is used to indicate the size of the overlapping time period between adjacent sample signal data segments when dividing the sample respiratory signal data, including the time overlap ratio (that is, the time overlap rate), etc. Specifically, the time overlap index includes any of the following: a time overlap ratio of 50%. Taking the time overlap index as the time overlap ratio as an example, the time window based on which the sample respiratory signal data is divided is 60 seconds. For example, if the overlap rate (that is, the time overlap ratio) is set to 50%, then the two adjacent sample signal data segments will share 30 seconds of data (that is, 50% of the window length). Under this method, each new data segment starts at the midpoint of the previous data segment, ensuring the continuity of the data flow and the integrity of the analysis.

[0082] Through these steps, the temporal overlap index ensures that the analysis of sample signal segments takes into account the correlation between the data in the preceding and following time periods, avoiding the information fragmentation that can result from analyzing each data segment completely independently. By allowing overlap between sample signal segments, it helps capture transitional features of the breathing pattern, particularly those that cross time boundaries, such as gradual or sudden changes in breathing rate.

[0083] As an optional embodiment, before calling the reference frequency amplitude fluctuation index, it includes: determining the amplitude distribution interval corresponding to the multiple respiratory frequency amplitude distributions based on the multiple matching indexes corresponding to the multiple respiratory frequency amplitude distributions; determining the candidate fluctuation indexes corresponding to the multiple respiratory frequency amplitude distributions based on the amplitude distribution intervals; determining the sample evaluation results corresponding to the multiple respiratory frequency amplitude distributions based on the multiple candidate fluctuation indexes; and determining the reference frequency amplitude fluctuation index from the multiple candidate fluctuation indices based on the sample evaluation results corresponding to the multiple respiratory frequency amplitude distributions.

[0084] In this embodiment, the specific steps before calling the reference frequency fluctuation index are described.

[0085] Specifically, the reference amplitude fluctuation index is determined from the plurality of candidate fluctuation indexes according to sample evaluation results corresponding to the plurality of respiratory amplitude distribution, including: determining the candidate fluctuation index corresponding to the plurality of respiratory amplitude distribution and the predetermined detection result; determining a plurality of matching indexes corresponding to the plurality of respiratory amplitude distribution, wherein the plurality of matching indexes represent the matching degree between the sample evaluation results and the predetermined detection result of the corresponding respiratory amplitude distribution under the plurality of candidate fluctuation indexes; and determining the reference amplitude fluctuation index from the plurality of candidate fluctuation indexes according to the plurality of matching indexes corresponding to the plurality of respiratory amplitude distribution, wherein the reference amplitude fluctuation index is the candidate fluctuation index with the largest matching index in the plurality of candidate fluctuation indexes.

[0086] The amplitude distribution interval is used to limit the screening range of the reference amplitude fluctuation index, thereby improving the calculation efficiency. For example, in the case of a respiratory amplitude distribution including an average deviation histogram, the amplitude distribution interval can be the top 20% of the average deviation indexes arranged in ascending order in the average deviation histogram. The candidate fluctuation index is predetermined and can be used to distinguish non-apnea events from apnea events to some extent, and further evaluation is needed to determine whether it can be used as the reference amplitude fluctuation index.

[0087] The predetermined detection result is predetermined and is the result of apnea detection corresponding to the plurality of respiratory amplitude distribution, which is used to measure the detection effect under different candidate fluctuation indexes and help determine the best parameter setting (i.e., the reference amplitude fluctuation index).

[0088] The plurality of matching indexes are quantitative indicators for measuring the degree of agreement between the evaluation results of the sample respiratory signal data and the predetermined detection result. Each matching index corresponds to a candidate fluctuation index and reflects the accuracy and reliability of the evaluation results obtained by detecting apnea in the sample object under the candidate fluctuation index.

[0089] The sample evaluation result is the evaluation result obtained by detecting apnea in the sample object according to the corresponding respiratory amplitude distribution under the corresponding candidate fluctuation index.

[0090] Through the above steps, the matching degree between the sample evaluation result and the predetermined detection result of each sample object under different candidate fluctuation indexes can be quantified, and the influence of each candidate fluctuation index on apnea detection can be systematically evaluated, so that the reference amplitude fluctuation index suitable for the plurality of sample objects can be accurately screened from the plurality of candidate fluctuation indexes, and the reference amplitude fluctuation index obtained can be ensured to be most suitable for the individual differences between objects.

[0091] That is, for each candidate fluctuation index, the degree of matching between its sample evaluation result and the predetermined detection result is calculated, and from all the calculated matching indices, the candidate fluctuation index with the highest matching degree (or meeting the predetermined standard) is selected as the reference frequency fluctuation index, ensuring that the determined reference frequency fluctuation index can adapt to the physiological characteristics of different subjects, avoiding the technical problem of low accuracy of sleep apnea detection due to individual differences.

[0092] As an optional embodiment, target respiratory signal data is extracted from target physiological signal data, including: determining a respiratory frequency range; determining a cutoff frequency parameter based on the respiratory frequency range; filtering the target physiological signal data based on the cutoff frequency parameter to obtain initial respiratory signal data; and determining target respiratory signal data based on the initial respiratory signal data.

[0093] In this embodiment, specific steps of extracting target respiratory signal data from target physiological signal data are described.

[0094] The respiratory frequency range is used to represent the frequency range of the respiratory signal. The respiratory frequency range can be determined based on the target subject's respiratory rate. For example, the frequency range of the respiratory signal is generally concentrated between 0.1 Hz and 0.6 Hz.

[0095] Among them, the cutoff frequency parameter is involved. The cutoff frequency parameter is two frequency points used to define the pass frequency range in the design of the bandpass filter. Signals below the low cutoff frequency and above the high cutoff frequency will be filtered out, while signals between these two frequencies will be retained.

[0096] The initial respiratory signal data is involved, and the initial respiratory signal data is obtained from the target physiological signal data after bandpass filtering and preliminarily contains respiratory signal characteristics.

[0097] By determining the respiratory frequency range, the target interval for signal processing can be clarified, and then the cutoff frequency parameters can be set to design a bandpass filter. This can ensure that the filtered signal only contains signal components that match the respiratory frequency, so as to effectively separate the respiratory signal and reduce the interference of other physiological signals and environmental noise, providing a high-quality data basis for subsequent signal analysis and feature extraction. The target physiological signal data is then filtered to obtain initial respiratory signal data, and the initial respiratory signal data is further processed and optimized, such as fast Fourier transform (FFT), to determine the target respiratory signal data, which can effectively separate and extract the respiratory signal from complex physiological signals.

[0098] As an optional embodiment, the target respiratory signal data is determined according to the initial respiratory signal data, including: dividing the initial respiratory signal data into a plurality of initial sub-data segments in time sequence; performing windowing processing on the plurality of initial sub-data segments respectively to obtain a plurality of target sub-data segments, wherein the windowing processing is used to smooth the signal data values at the first time point and the last time point in the corresponding initial sub-data segment to a predetermined value, and the target sub-data segments correspond to the plurality of initial sub-data segments one by one; determining a frequency amplitude parameter corresponding to each of the plurality of target sub-data segments; and determining the target respiratory signal data according to the frequency amplitude parameters corresponding to the plurality of target sub-data segments.

[0099] In this embodiment, the specific steps of determining the target respiratory signal data according to the initial respiratory signal data are described.

[0100] Among them, the plurality of initial sub-data segments are a plurality of continuous subsets obtained by dividing the initial respiratory signal data according to a predetermined time length (such as 60 seconds or 120 seconds), and each sub-data segment contains respiratory signal information within the predetermined time length.

[0101] Among them, the windowing processing is a processing for smoothing the edges of the signal to avoid the appearance of pseudo-spectrum components caused by the discontinuity of the signal edges in the frequency domain analysis. In signal processing, the windowing processing can be applying a specific window function to the signal data segment.

[0102] Among them, the plurality of target sub-data segments are the initial sub-data segments after the windowing processing, which maintain a one-to-one correspondence with the initial sub-data segments in time sequence, but are smoothed in value to eliminate the influence of signal mutation on the frequency spectrum analysis.

[0103] Among them, the frequency amplitude parameter is obtained in the frequency domain analysis, reflecting the strength or energy of the signal at different frequency components, to quantify the frequency domain characteristics of the respiratory signal, for example, performing FFT operation on each target sub-data segment to obtain a frequency spectrum diagram (i.e. frequency amplitude parameter).

[0104] Through the above steps, the windowing processing on the plurality of initial sub-data segments can reduce the spectrum leakage to ensure that the frequency amplitude parameters determined subsequently can truly reflect the frequency domain characteristics of the respiratory signal, so that the target respiratory signal data can more accurately reflect the respiratory condition of the target object.

[0105] As an optional embodiment, based on a reference frequency-amplitude fluctuation index and sub-respiration frequency-amplitude characteristics corresponding to a plurality of sub-respiration signal data, determining an apnea detection result for a target object includes: determining a plurality of candidate signal data from the plurality of sub-respiration signal data based on the sub-respiration frequency-amplitude characteristics corresponding to the plurality of sub-respiration signal data; dividing each signal data in the plurality of candidate signal data to obtain a plurality of signal segments corresponding to the plurality of candidate signal data; determining the segment frequency-amplitude characteristics corresponding to the plurality of signal segments corresponding to each candidate signal data; determining the apnea detection result for the target object based on the segment frequency-amplitude characteristics corresponding to the plurality of signal segments corresponding to each candidate signal data, wherein the apnea detection result includes the signal detection results corresponding to the plurality of candidate signal data.

[0106] In this embodiment, specific steps of determining the apnea detection result of the target object based on the reference frequency fluctuation index and the sub-respiration frequency and amplitude characteristics corresponding to a plurality of sub-respiration signal data are described.

[0107] This involves multiple candidate signal data sets, which are determined from the multiple sub-respiratory signal data sets based on the sub-respiratory frequency and amplitude characteristics of each respiratory signal data set, and are used to limit the analysis scope. For example, taking the sub-respiratory frequency and amplitude characteristics as the mean deviation index, after constructing the aforementioned mean deviation histogram, the multiple candidate signal data sets are the sub-respiratory signal data sets corresponding to the mean deviation indices in the top 20% of the mean deviation histogram.

[0108] Multiple signal segments are involved, each of which is obtained by further segmenting each candidate signal data. For example, after segmenting the target signal data into 60-second windows (50% overlap) to obtain multiple sub-respiration signal data, the sub-respiration signal data that meets the histogram threshold condition (the top 20% of the mean deviation histogram) is selected as the candidate signal data. Each candidate signal data is then further divided into three 20-second sub-segments, which are the multiple signal segments mentioned above.

[0109] Specifically, based on the segment frequency and amplitude characteristics corresponding to the multiple signal segments corresponding to each candidate signal data, the apnea detection result of the target object is determined, including: when the multiple signal segments corresponding to each candidate signal data are multiple signal segments arranged in a predetermined order, determining the arrangement positions corresponding to the multiple signal segments; based on the arrangement positions and segment frequency and amplitude characteristics corresponding to the multiple signal segments, determining the feature difference index corresponding to each candidate signal data, wherein the feature difference index is used to represent the degree of feature difference between signal segments of adjacent arrangement positions; based on the feature difference index corresponding to each candidate signal data, the apnea detection result of the target object is determined.

[0110] For example, after the target signal data is divided into 60-second windows (50% overlap) to obtain multiple sub-respiratory signal data, the sub-respiratory signal data that meets the histogram threshold condition (the first 20% in the mean deviation histogram) is used as the selected signal data, and then each candidate signal number is further divided into three 20-second sub-segments to obtain multiple signal segments of each candidate signal data. Next, the segment frequency and amplitude characteristics (such as NMAAD value) of each signal segment are determined, and sorted in ascending order to determine the ranking of each signal segment. According to the characteristic difference index corresponding to each candidate signal data (such as the difference in NMAAD value), the apnea detection result of the target object is determined.

[0111] As an optional embodiment, determining multiple sub-respiration signal data based on the target respiration signal data includes: determining a division index corresponding to the target respiration signal data; and dividing the target respiration signal data based on the division index to obtain multiple sub-respiration signal data.

[0112] In this embodiment, the specific steps of determining a plurality of sub-respiration signal data according to the target respiration signal data are described.

[0113] This involves a partitioning index, a parameter used in signal processing to determine how signal data is segmented into subsegments. The partitioning index includes the data time span (i.e., the partitioning time length), the time overlap ratio, and the partitioning strategy. The partitioning strategy is determined based on the data characteristics corresponding to the target respiratory signal data, that is, a segmentation strategy dynamically determined based on the signal characteristics. For example, the partitioning time length can be 20 seconds, and the overlap ratio can be 50%.

[0114] The partition index affects the length and continuity of subsequent sub-signal data. Reasonable setting of the partition index can optimize the structure of the sub-signal data, making it more suitable for subsequent frequency domain analysis and feature extraction, thereby helping to improve the accuracy of apnea detection.

[0115] Based on the above embodiment and optional embodiment, an optional implementation manner is provided, which is described in detail below.

[0116] In related technologies, when performing sleep apnea detection, the complexity of the sleeping environment, such as ambient noise, may interfere with physiological signals, resulting in inaccurate sleep apnea detection when performing sleep apnea detection based on physiological signals.

[0117] To address the above-mentioned problems, no effective solutions have been proposed so far.

[0118] In view of this, an optional embodiment of the present invention provides a sleep apnea detection method and system, which can also be called a sleep apnea detection method and system based on optical fiber BCG signals, which can effectively solve the technical problem of inaccurate sleep apnea detection when sleep apnea detection is performed based on physiological signals.

[0119] (1) Sleep apnea detection method:

[0120] Figure 2 is a flow chart of the sleep apnea detection method of the present invention, such as Figure 2 As shown, a detailed description is given below.

[0121] S1. Acquire target physiological signal data corresponding to the target object;

[0122] For example, signal acquisition and preprocessing.

[0123] The BCG signal was collected and the respiratory frequency band was separated using a bandpass filter.

[0124] Specifically, fiber optic BCG sensors are used to collect non-contact vital signs from the target subject, such as by embedding slightly bent fiber optic sensors in mattresses. The raw BCG signals obtained contain multiple physiological components such as heart rate, respiratory movement, and body movements.

[0125] The fiber optic BCG sensor records the mixed raw signal (i.e., the raw BCG signal) including the respiratory component during the signal acquisition phase. Figure 3 is a visualization diagram of the mixed original signal in an optional embodiment of the present invention, such as Figure 3 As shown, the curve exhibits a regular undulating pattern, and multiple respiratory cycles can be observed in the signal. For ease of observation, the original signal has been normalized, and the respiratory peak positions (black dots) are marked. A preliminary smoothed respiratory curve (red dashed line) is also plotted for reference and comparison.

[0126] S2. extracting target respiratory signal data from target physiological signal data;

[0127] Specifically, because the mixed raw signal (the same as the target physiological signal data, referred to as the raw signal) is mixed with high-frequency components such as heartbeats, as well as possible low-frequency drift and motion artifacts, directly extracting the respiratory frequency from the raw signal will result in insufficient accuracy. Further processing is required to purify the respiratory information. Specifically, this includes bandpass filtering and fast Fourier transform, as follows:

[0128] (1) Respiratory component extraction: bandpass filtering;

[0129] Specifically, the method comprises: determining a breathing frequency range; determining a cutoff frequency parameter according to the breathing frequency range; performing filtering processing on the target physiological signal data according to the cutoff frequency parameter to obtain initial respiratory signal data; and determining target respiratory signal data according to the initial respiratory signal data.

[0130] For example, in order to effectively extract a stable and continuous respiratory waveform from the original signal, a band-pass filter (Band-Pass Filter) is used to process the signal. The frequency range of the respiratory signal (as described above) is generally concentrated between 0.1 Hz and 0.6 Hz, depending on the target object's breathing rate (i.e. about 6-36 times per minute).

[0131] Therefore, the passband range of the filter (as described above) is set as:

[0132] Low cutoff frequency: 0.1 Hz;

[0133] High cutoff frequency: 0.6 Hz.

[0134] In the filter design, a Butterworth filter is selected to maintain the smoothness of the signal and reduce phase distortion. Figure 4 is a schematic diagram of the filtered result in an optional embodiment of the present application, as shown in Figure 4 The signal (as described above) presents a clear periodic fluctuation, which is highly related to the morphological characteristics of the respiratory movement, and the high-frequency heartbeat interference and low-frequency noise are removed, thereby obtaining a relatively pure respiratory component waveform. Similarly, the respiratory peak value is also extracted and labeled on the filtered signal for subsequent periodic analysis and respiratory rate estimation.

[0135] (2) Respiratory frequency extraction based on Fast Fourier Transform (FFT).

[0136] Specifically, the method comprises: dividing the initial respiratory signal data into a plurality of initial sub-data segments in time sequence; performing windowing processing on the plurality of initial sub-data segments respectively to obtain a plurality of target sub-data segments, wherein the windowing processing is used to smoothly attenuate the signal data value at the first time point and the signal data value at the last time point in the corresponding initial sub-data segment to a predetermined value, and the target sub-data segment corresponds to the initial sub-data segment one by one; determining a frequency amplitude parameter corresponding to each of the plurality of target sub-data segments; and determining the target respiratory signal data according to the frequency amplitude parameters corresponding to the plurality of target sub-data segments.

[0137] For example, the respiratory signal has clear periodicity, so Fast Fourier Transform (FFT) is further used to perform frequency domain analysis on the filtered respiratory signal. The main steps include:

[0138] F1. Signal segmentation processing:

[0139] The filtered respiratory signal (the same as the above-mentioned initial respiratory signal data) is divided into multiple time windows according to a fixed length (such as 60 seconds or 120 seconds) to obtain multiple initial sub-data segments.

[0140] F2, window function application:

[0141] To reduce spectrum leakage, a Hanning window is used to perform windowing processing on each segment to obtain multiple target sub-data segments.

[0142] F3, FFT transform:

[0143] Perform FFT operation on each signal segment (same as the target sub-data segment mentioned above) to obtain the corresponding spectrum diagram (same as the frequency and amplitude parameters mentioned above).

[0144] F4, main frequency peak detection:

[0145] Find the frequency component corresponding to the main frequency peak in the spectrum, that is, the frequency point with the largest amplitude in the signal. This point is the estimated respiratory frequency.

[0146] F5, unit conversion:

[0147] Multiply the frequency value (in Hz) by 60 to get the respiratory rate per minute (in bpm), and then get the target respiratory signal data.

[0148] If the FFT main frequency is 0.25Hz ( Figure 4 For example), the respiratory rate is converted to 0.25×60=15 times / minute.

[0149] Using this method, the raw signal output by the fiber-optic BCG sensor is bandpass filtered to successfully extract a clear respiratory component, and a stable and accurate respiratory rate estimation is achieved using an FFT algorithm. This method not only exhibits excellent real-time and non-contact characteristics, but is also suitable for long-term, continuous sleep monitoring and assessment. The entire process from raw BCG signal to respiratory rate extraction provides important theoretical support and experimental basis for the engineering implementation of non-contact respiratory monitoring systems.

[0150] S3. Retrieving a reference frequency-amplitude fluctuation index, wherein the reference frequency-amplitude fluctuation index is obtained based on a plurality of respiratory frequency-amplitude distributions, wherein the plurality of respiratory frequency-amplitude distributions are respectively obtained based on corresponding sample respiratory signal data, and the corresponding sample respiratory signal data are respectively obtained based on corresponding sample physiological signal data;

[0151] For example, the definition of an apnea event requires that the interruption of breathing lasts for at least 10 seconds, or longer. These apneas are usually accompanied by increased body activity, intensified respiratory movements, and disordered respiratory waveforms / disappearance of peaks. By comparing the manually annotated apnea events in the electrocardiogram (ESG) with the extracted respiratory signals, it was found that most apnea events occurred in segments marked as having reduced respiratory rate and the presence of body motion artifacts. Therefore, during the apnea event detection process, these body motion artifacts were not removed to retain their potential identification value. In apnea detection, the goal is to distinguish apnea events from normal respiratory events through the extracted respiratory signals.

[0152] Figure 5 Schematic diagram of the time window of apnea event in an optional embodiment of the present invention. Figure 5 As shown in the figure, during the time window of apnea events, the respiratory waveform amplitude collected by the fiber-optic BCG sensor experiences significant attenuation. At time t = 36.3 s in the figure, the absolute amplitude of the respiratory waveform mostly exceeds 0.1 before this time point, and after this time point, the amplitude is less than 0.03.

[0153] To detect pause events based on changes in the general respiratory waveform amplitude, we first constructed a histogram of the Normalized Mean Amplitude Absolute Deviation (NMAAD) of the respiratory waveform amplitude for the sample respiratory signal data of multiple subjects (similar to the respiratory frequency and amplitude distribution corresponding to the sample respiratory signal data described above). This histogram was constructed by dividing the respiratory signal into equal 30-second time windows (with a 50% overlap between the windows) and calculating the mean absolute deviation of the amplitude in each time window.

[0154] Figure 6 is an optional embodiment of the present invention based on the respiratory waveform amplitude mean absolute deviation histogram, such as Figure 6 As shown, the resulting histogram is sorted in descending order. In other words, the maximum value of the histogram (i.e., the number of times this value occurs) corresponds to the most common value in the NMAAD value, that is, the position of the modal value, that is, the NMAAD value with the highest frequency mainly corresponds to normal respiratory events.

[0155] The NMAAD values ​​corresponding to the top 20% of the histogram values ​​are selected as the threshold for determining apnea events (same as the amplitude distribution interval mentioned above). The selection criteria for the top 20% of the histogram values ​​are as follows:

[0156] The determination of the reference frequency fluctuation index requires the optimization of two key parameters: one is the serial number value of the histogram (to determine the amplitude distribution range), and the other is the threshold parameter (after sorting, the serial number value of the histogram can confirm the threshold parameter, which is the same as the reference frequency fluctuation index mentioned above).

[0157] Specifically, based on multiple matching indexes corresponding to multiple respiratory frequency and amplitude distributions, the amplitude distribution intervals corresponding to the multiple respiratory frequency and amplitude distributions are determined; based on the amplitude distribution intervals, the selected fluctuation indices corresponding to the multiple respiratory frequency and amplitude distributions are determined; based on the multiple fluctuation indices to be selected, the sample evaluation results corresponding to the multiple respiratory frequency and amplitude distributions are determined; the predetermined detection results corresponding to the multiple respiratory frequency and amplitude distributions are determined; multiple matching indexes corresponding to the multiple respiratory frequency and amplitude distributions are determined, wherein the multiple matching indices represent the degree of matching between the sample evaluation results and the predetermined detection results of the corresponding respiratory frequency and amplitude distributions under the multiple fluctuation indices to be selected; based on the multiple matching indices corresponding to the multiple respiratory frequency and amplitude distributions, a reference frequency and amplitude fluctuation index is determined from the multiple fluctuation indices to be selected, wherein the reference frequency and amplitude fluctuation index is the selected fluctuation index with the largest matching index among the multiple fluctuation indices to be selected.

[0158] For example, in order to complete this parameter optimization process, an improved leave-one-out cross-validation method is used. Traditionally, the evaluation method of leave-one-out cross-validation is usually to record the performance indicators of the system after detecting the left-out samples (i.e., the sample respiratory signal data of a sample object), and finally obtain an estimate of the overall performance by averaging the evaluation indicators of all iterative results. Different from this, the improved leave-one-out cross-validation method does not evaluate the system performance on the data of a certain sample object that has been left out, but in each iteration (for example, for the sample respiratory signal data of 24 sample objects, a total of 24 iterations are performed), the data of the remaining 23 sample objects are used to test the influence of the two parameters on the performance of apnea event detection, that is, the optimal value of this key parameter is selected according to the multiple matching indices corresponding to the multiple respiratory frequency distributions.

[0159] First, the main goal is to determine the best histogram value for any threshold (same as the volatility index to be selected above). A series of different thresholds are predetermined, with values ​​ranging from 20 to 120 and a step size of 0.5. At the same time, for each threshold, the impact of the 2nd to 12th histogram values ​​on the detection performance is evaluated respectively. In other words, for a specific threshold, three evaluation indicators, namely sensitivity, specificity and accuracy, are measured through 11 different histogram numbers. This evaluation process is repeated 10 times for each group of data consisting of 23 different sample objects, and the average value of each indicator in each iteration is recorded. The ultimate goal of this process is to find an optimal histogram number value (same as the amplitude distribution range mentioned above) that can be applied to any threshold.

[0160] Through this process, it can be seen that under any threshold conditions, sensitivity is inversely proportional to the histogram value, while specificity and accuracy are directly proportional to the histogram value.

[0161] The results revealed that the top 20% of histogram values ​​represent a critical turning point. Across all tests, the NMAAD threshold for apnea onset occurred within the top 20% of the histogram ranking. This pattern of variation was consistent across all threshold conditions. Therefore, the top 20% of histogram values ​​were ultimately selected as the optimal histogram parameter interval for detecting apnea events (similar to the amplitude distribution interval described above).

[0162] On this basis, we further explored the optimal threshold value that matches the selected histogram value interval (the same as the frequency fluctuation index mentioned above). To achieve this goal, we calculated the average area under the curve (AUC) corresponding to each threshold value in 10 iterations, also known as balanced accuracy. The results showed that the 65% threshold value (the same as the frequency fluctuation index mentioned above) had the highest AUC, that is, the best overall detection performance.

[0163] Finally, the threshold that gives the highest AUC value is selected as the optimal value (same as the reference frequency fluctuation index mentioned above). Figure 7 : is a graph showing the relationship between the AUC value and the threshold selection percentage in an optional embodiment of the present invention, such as Figure 7As shown, the maximum AUC occurs at the 65% threshold (same as the frequency fluctuation index mentioned above), with a value of 51.7%. For every subject to be detected, the 65% threshold can be adapted. Therefore, the 65% threshold is determined to be the optimal threshold for apnea event detection. That is, if the total number of NMAAD occurrences in the first 20% of histogram values ​​exceeds 65% of the total number of occurrences, the NMAAD value corresponding to the histogram column exceeding 65% is used as the apnea detection threshold (adaptive threshold calculation).

[0164] In summary, the threshold definition for adaptive apnea amplitude detection was systematically optimized through the leave-one-out cross-validation method, providing a robust and reliable parameter basis for the subsequent automatic detection of apnea events.

[0165] The above detection algorithm performs sleep apnea detection based on the acquisition of fiber optic BCG respiratory waveforms, detection of their amplitudes, and adaptive threshold calculation. The sleep apnea detection algorithm is implemented in an embedded hardware device integrated with a specific fiber optic BCG sensor.

[0166] Experiments have shown that this threshold shows good stability and consistency among different objects to be detected (including sample objects and target objects).

[0167] S4. Determine a plurality of sub-respiration signal data based on the target respiration signal data;

[0168] Specifically, it includes: determining a division index corresponding to the target respiratory signal data; and dividing the target respiratory signal data according to the division index to obtain a plurality of sub-respiratory signal data.

[0169] For example, after determining the detection threshold (similar to the reference frequency fluctuation index, i.e., 65%), the partitioning index corresponding to the target respiratory signal data is determined, namely, a 60-second time window and a 50% overlap ratio. Next, the respiratory signal (similar to the target respiratory signal data) is divided into equal 60-second time windows with a 50% overlap ratio. The NMAAD value is calculated for each segment to measure the amplitude stability and size of the respiratory waveform in that segment.

[0170] S5, determining sub-respiration frequency and amplitude features corresponding to the plurality of sub-respiration signal data respectively;

[0171] For example, the NMAAD value is calculated for each respiratory signal segment to measure the amplitude stability and size of the respiratory waveform. All NMAAD values ​​are constructed into a histogram and arranged from smallest to largest. The maximum value (modal value) of the histogram corresponds to the most common NMAAD value, that is, a normal respiratory event.

[0172] S6, determining the apnea detection result of the target object according to the reference amplitude fluctuation index and the sub-breathing amplitude features corresponding to the plurality of sub-breathing signal data.

[0173] For example, for each 60-second segment, the NMAAD values of three 20-second segments (same as the sub-breathing amplitude features described above) are calculated and sorted in ascending order. Each 60-second time period is further divided into three 20-second sub-time segments (same as the plurality of signal segments corresponding to each candidate signal data described above), and the NMAAD values of each sub-time segment are determined (same as the segment amplitude features described above). The NMAAD values are arranged in ascending order, and if the difference between the third and second NMAAD values after sorting (same as the feature difference index described above) exceeds 45% of the histogram threshold (same as the reference amplitude fluctuation index described above), the 60-second segment is marked as an apnea event; otherwise, it is marked as a non-apnea event, and the apnea detection result of the target object is finally obtained.

[0174] In addition, in order to comprehensively verify the effectiveness and stability of the above steps in sleep apnea detection, systematic measurement and evaluation were carried out on a plurality of target objects, for example, the 26 target objects (same as the plurality of target objects described above) not used for parameter optimization (a total of 78 experiments).

[0175] Figure 8 is a schematic diagram of the system measurement and evaluation result in the optional embodiment of the present application, as shown in Figure 8 The experimental results show that the above steps achieve a 100% detection rate in all 78 simulated sleep apnea experiments, i.e. no false negatives, and exhibit good sensitivity and stability.

[0176] In terms of specific detection timeliness, the average detection delay of the above steps for apnea events is 20.38 seconds, and the average level can be clearly observed in the figure indicated by the red dashed line ("Average time (20.38s)"). All detection delay times are distributed between about 17 seconds and 24 seconds with small fluctuations, indicating that the algorithm has good robustness under different individuals, physiological states and noise backgrounds, and the red wide line represents the 95% confidence interval.

[0177] As can be seen from the blue curve in the figure (representing the detection time of each experiment), most of the detection time points fluctuate closely around the average value with good concentration. A small number of extreme values are slightly higher or lower, which can be attributed to physiological and environmental factors such as respiratory feature variation caused by individual differences, body movement interference or background signal changes.

[0178] Furthermore, the adaptability of the aforementioned steps across diverse target subjects was also demonstrated from another perspective. The 26 target subjects included individuals ranging in weight from 50kg to 110kg and in age from 18 to 42 years old. Despite extreme body shapes and physiological parameters, simulated apnea events were still accurately and rapidly detected. This demonstrates that the aforementioned steps are effective in accurately detecting sleep apnea, and their universal applicability across diverse individuals avoids the technical issue of inaccurate detection due to individual differences.

[0179] To more accurately assess the real-time performance of the above steps in detecting sleep apnea events, we statistically analyzed the detection latency data from all 78 experiments. The mean detection latency was 20.38 seconds, with a standard deviation of 2.14 seconds, indicating a relatively concentrated distribution around the mean.

[0180] After the apnea event is identified, the early warning mechanism is also determined based on the apnea detection results of the target object. If there are multiple consecutive low-amplitude breathing segments in multiple signal segments, an early warning or remote upload will be triggered.

[0181] The above content can be summarized into the following framework steps:

[0182] M1, preprocessing: collect BCG signals and separate respiratory frequency bands using a bandpass filter;

[0183] M2. Feature calculation: Segment the signal and calculate the NMAAD value of each respiratory signal segment;

[0184] M3, build histogram model: extract NMAAD distribution law and determine adaptive detection threshold;

[0185] M4, event recognition: determine whether each sub-segment is below the threshold and identify it as a candidate apnea event;

[0186] M5. Early warning mechanism: Multiple consecutive low-amplitude breathing segments trigger an early warning or remote upload.

[0187] In summary, the above content fully verifies that the sleep apnea detection method based on fiber-optic BCG signals can accurately detect sleep apnea, showing both technical feasibility and engineering practicality. Furthermore, through unsupervised statistical feature recognition, it not only overcomes the traditional method's reliance on labeled data and complex models, but also provides a theoretical basis and methodological foundation for the realization of a low-cost, highly deployable sleep monitoring system. Furthermore, the above steps can be further combined with multi-sensor information fusion and dynamic model adaptation mechanisms to enhance adaptability and robustness in real-world complex environments.

[0188] (2) Sleep apnea detection system:

[0189] Figure 9 FIG is a schematic diagram of a sleep apnea detection system in an optional embodiment of the present invention. Figure 9 As shown, the sleep apnea detection system includes an embedded acquisition module A1, a signal processing module A2, and an algorithm recognition module A3.

[0190] Among them, the embedded acquisition module A1, including a slightly bent optical fiber sensor and a signal conditioning circuit, is embedded in the mattress structure and is used to non-contactly obtain the weak mechanical vibrations of the human chest and abdomen caused by breathing.

[0191] The signal processing module A2 is used to obtain target physiological signal data corresponding to the target subject and extract target respiratory signal data from the target physiological signal data. For example, the BCG analog signal is obtained through A / D conversion and then bandpass filtered (0.1-0.5 Hz) to extract the respiratory signal (similar to the target respiratory signal data described above).

[0192] Among them, the algorithm identification module A3 is used to perform the following functions: calling a reference frequency-amplitude fluctuation index, wherein the reference frequency-amplitude fluctuation index is obtained based on multiple respiratory frequency-amplitude distributions, and the multiple respiratory frequency-amplitude distributions are respectively obtained based on corresponding sample respiratory signal data, and the corresponding sample respiratory signal data are obtained based on corresponding sample physiological signal data; determining multiple sub-respiratory signal data based on the target respiratory signal data; determining the sub-respiratory frequency-amplitude characteristics corresponding to the multiple sub-respiratory signal data; and determining the apnea detection result of the target object based on the reference frequency-amplitude fluctuation index and the sub-respiratory frequency-amplitude characteristics corresponding to the multiple sub-respiratory signal data.

[0193] For example, the algorithm identification module A3 is used to perform the following functions:

[0194] (1) Extract the normalized mean absolute deviation (NMAAD) of each respiratory signal segment in a 60-second sliding window;

[0195] (2) Construct its histogram distribution and dynamically set the detection threshold;

[0196] (3) Combined with the fluctuation characteristics of sub-segments within the window, abnormal conditions such as sudden drop in respiratory amplitude and waveform interruption are identified.

[0197] Through the above optional implementation, at least the following beneficial effects can be achieved:

[0198] (1) Compared with the related art, the present application can improve the accuracy of sleep apnea detection by extracting target respiratory signal data from the acquired target physiological signal data and calling a reference amplitude fluctuation index, since the reference amplitude fluctuation index is obtained based on the respiratory amplitude distribution of a plurality of sample respiratory signal data, thereby adapting to the differences of the detection object, and then analyzing the sub-respiratory amplitude characteristics of the respiratory signal data in segments, and combining the reference amplitude fluctuation index, thereby effectively improving the accuracy of sleep apnea detection, and solving the technical problem of inaccurate sleep apnea detection when detecting sleep apnea based on physiological signals.

[0199] (2) Compared with the related art, for each candidate fluctuation index, the matching degree between the sample evaluation result and the predetermined detection result is calculated, and the candidate fluctuation index with the highest matching degree (or meeting the predetermined standard) is selected from all calculated matching indexes as the reference amplitude fluctuation index, which ensures that the determined reference amplitude fluctuation index can adapt to the physiological characteristics of different objects, and avoids the technical problem of low accuracy of sleep apnea detection caused by individual differences.

[0200] (3) Compared with the related art, the present application can reduce spectral leakage by windowing a plurality of initial sub-data segments, to ensure that the subsequent determined amplitude parameters can truly reflect the frequency domain characteristics of the respiratory signal, so that the target respiratory signal data can more accurately reflect the respiratory condition of the target object.

[0201] (4) Compared with the related art, the present application combines adaptive amplitude threshold and respiratory spectrum statistical analysis, which significantly improves the detection accuracy compared with the traditional fixed threshold method, and realizes effective identification and analysis of respiratory signals under non-contact conditions, improves the sensitivity and accuracy of sleep apnea detection without introducing redundant configuration of multiple sensors.

[0202] (5) Compared with the related art, the present application can support local processing and alarm on an embedded device, reducing system integration and use threshold. The above steps can realize adaptive threshold setting, avoid dependence on large-scale labeling and supervised models, improve generalization ability, and based on statistical distribution dynamic adjustment, ensure stable applicability between individuals of different body types through histogram threshold selection, the overall detection speed meets the early warning demand, in addition, the calculation complexity is low, and supports contactless sleep monitoring system application.

[0203] It should be noted that for the aforementioned method embodiments, for simplicity of description, they are all expressed as a series of action combinations. However, those skilled in the art should be aware that the present invention is not limited by the order of the actions described, because according to the present invention, certain steps can be performed in other orders or simultaneously. Secondly, those skilled in the art should also be aware that the embodiments described in this specification are all preferred embodiments, and the actions and modules involved are not necessarily required by the present invention.

[0204] Through the description of the above embodiments, those skilled in the art can clearly understand that the method according to the above embodiment can be implemented by means of software plus the necessary general hardware platform, and of course it can also be implemented by hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, can be embodied in the form of a software product, which is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk), and includes a number of instructions for enabling a terminal device (which can be a mobile phone, computer, server, or network device, etc.) to execute the methods of various embodiments of the present invention.

[0205] Example 2

[0206] According to an embodiment of the present invention, a device for implementing the above sleep apnea detection method is also provided. Figure 10 FIG. 1 is a block diagram of a sleep apnea detection device according to an embodiment of the present invention. Figure 10 As shown, the device includes: an acquisition module 1002, an extraction module 1004, a call module 1006, a first determination module 1008, a second determination module 1010 and a third determination module 1012. The device is described in detail below.

[0207] The acquisition module 1002 is configured to acquire target physiological signal data corresponding to a target object; the extraction module 1004 is connected to the acquisition module 1002 and is configured to extract target respiratory signal data from the target physiological signal data; the calling module 1006 is connected to the extraction module 1004 and is configured to call a reference amplitude fluctuation index, wherein the reference amplitude fluctuation index is obtained according to a plurality of respiratory amplitude distributions, the plurality of respiratory amplitude distributions are respectively obtained according to corresponding sample respiratory signal data, and the corresponding sample respiratory signal data is obtained according to corresponding sample physiological signal data; the first determination module 1008 is connected to the calling module 1006 and is configured to determine a plurality of sub respiratory signal data according to the target respiratory signal data; the second determination module 1010 is connected to the first determination module 1008 and is configured to determine sub respiratory amplitude features corresponding to the plurality of sub respiratory signal data respectively; and the third determination module 1012 is connected to the second determination module 1010 and is configured to determine an apnea detection result of the target object according to the reference amplitude fluctuation index and the sub respiratory amplitude features corresponding to the plurality of sub respiratory signal data respectively.

[0208] Optionally, the calling module 1006 further includes: a first division unit configured to divide the sample respiratory signal data to obtain a plurality of sample signal data segments in a case where the sample respiratory signal data includes a plurality of respiratory amplitude data, wherein the plurality of sample signal data segments respectively include at least one respiratory amplitude data; a first feature determination unit configured to determine sample respiratory amplitude features corresponding to the plurality of sample signal data segments respectively according to the plurality of respiratory amplitude data; and a first distribution determination unit configured to determine a respiratory amplitude distribution corresponding to the sample respiratory signal data according to the sample respiratory amplitude features corresponding to the plurality of sample signal data segments respectively.

[0209] Optionally, the first division unit further includes: an overlap index determination unit configured to determine a time overlap index, wherein the time overlap index includes any one of the following: a time overlap ratio of 50%; and a second division unit configured to divide the sample respiratory signal data to obtain the plurality of sample signal data segments according to the time overlap index.

[0210] Optionally, the first distribution determination unit further includes: a frequency determination unit configured to determine occurrence frequencies of a plurality of target deviation values respectively in a case where the sample respiratory amplitude features include average deviation indexes and the respiratory amplitude distribution includes an average deviation histogram, wherein the plurality of average deviation indexes include the plurality of target deviation values; and a histogram determination unit configured to obtain the average deviation histogram corresponding to the sample respiratory signal data according to the plurality of target deviation values and the occurrence frequencies of the plurality of target deviation values respectively.

[0211] Optionally, the retrieval module 1006 also includes: an interval determination unit, which is used to determine the amplitude distribution interval corresponding to the multiple respiratory frequency and amplitude distributions based on the multiple matching indexes corresponding to the multiple respiratory frequency and amplitude distributions, wherein, in the case where the respiratory frequency and amplitude distribution includes a mean deviation histogram, the amplitude distribution interval includes any one of the following: the first 20% of the multiple mean deviation indices arranged from small to large in the mean deviation histogram; a first index determination unit, which is used to determine the candidate fluctuation indices corresponding to the multiple respiratory frequency and amplitude distributions based on the amplitude distribution interval; an evaluation result determination unit, which is used to determine the sample evaluation results corresponding to the multiple respiratory frequency and amplitude distributions based on the multiple candidate fluctuation indices; a second index determination unit, which is used to determine a reference frequency and amplitude fluctuation index from the multiple candidate fluctuation indices based on the sample evaluation results corresponding to the multiple respiratory frequency and amplitude distributions.

[0212] Optionally, the second index determination unit further includes: a predetermined result determination unit, used to determine predetermined detection results corresponding to multiple respiratory frequency and amplitude distributions; a matching index determination unit, used to determine multiple matching indices corresponding to multiple respiratory frequency and amplitude distributions, wherein the multiple matching indices represent the degree of matching between the sample evaluation results of the corresponding respiratory frequency and amplitude distributions under multiple selected fluctuation indices and the predetermined detection results; an index screening determination unit, used to determine a reference frequency and amplitude fluctuation index from multiple fluctuation indices to be selected based on the multiple matching indices corresponding to the multiple respiratory frequency and amplitude distributions, wherein the reference frequency and amplitude fluctuation index is the selected fluctuation index with the largest matching index among the multiple fluctuation indices to be selected, and the reference frequency and amplitude fluctuation index includes any one of the following: 65%.

[0213] Optionally, the extraction module 1004 also includes: a range determination unit for determining the respiratory frequency range; a filtering parameter determination unit for determining the cutoff frequency parameter based on the respiratory frequency range; a filtering processing unit for filtering the target physiological signal data based on the cutoff frequency parameter to obtain initial respiratory signal data; and a first target data determination unit for determining the target respiratory signal data based on the initial respiratory signal data.

[0214] Optionally, the first target data determination unit also includes: a third division unit, used to divide the initial respiratory signal data into multiple initial sub-data segments in time sequence; a windowing processing unit, used to perform windowing processing on the multiple initial sub-data segments respectively to obtain multiple target sub-data segments, wherein the windowing processing is used to make the signal data value of the first time point and the signal data value of the last time point in the corresponding initial sub-data segment smoothly decay to a predetermined value, and the target sub-data segment corresponds to the multiple initial sub-data segments one by one; a parameter determination unit, used to determine the frequency and amplitude parameters corresponding to the multiple target sub-data segments respectively; a second target data determination unit, used to determine the target respiratory signal data based on the frequency and amplitude parameters corresponding to the multiple target sub-data segments respectively.

[0215] Optionally, the third determination module 1012 further includes: a candidate signal determination unit, used to determine multiple candidate signal data from multiple sub-respiration signal data based on the sub-respiration frequency-amplitude characteristics corresponding to the multiple sub-respiration signal data; a signal segment determination unit, used to divide each signal data in the multiple candidate signal data, respectively, to obtain multiple signal segments corresponding to the multiple candidate signal data; a segment feature determination unit, used to determine the segment frequency-amplitude characteristics corresponding to the multiple signal segments corresponding to each candidate signal data; a first detection result determination unit, used to determine the apnea detection result of the target object based on the segment frequency-amplitude characteristics corresponding to the multiple signal segments corresponding to each candidate signal data, wherein the apnea detection result includes the signal detection results corresponding to the multiple candidate signal data.

[0216] Optionally, the first detection result determination unit further includes: an arrangement rank determination unit, which is used to determine the arrangement ranks corresponding to the multiple signal segments respectively when the multiple signal segments corresponding to each candidate signal data are multiple signal segments arranged in a predetermined order; a difference index determination unit, which is used to determine the feature difference index corresponding to each candidate signal data based on the arrangement ranks and segment frequency characteristics corresponding to the multiple signal segments, wherein the feature difference index is used to represent the degree of feature difference between signal segments of adjacent arrangement ranks; and a second detection result determination unit, which is used to determine the apnea detection result of the target object based on the feature difference index corresponding to each candidate signal data.

[0217] Optionally, the first determination module 1008 further includes: a division index determination unit, which determines a division index corresponding to the target respiratory signal data, wherein the division index includes: data time span, time overlap rate, and division strategy, and the division strategy is determined based on data features corresponding to the target respiratory signal data; a sub-data determination unit, which is used to divide the target respiratory signal data according to the division index to obtain multiple sub-respiratory signal data.

[0218] It should be noted that the acquisition module 1002, extraction module 1004, retrieval module 1006, first determination module 1008, second determination module 1010, and third determination module 1012 correspond to steps S102 to S112 in the sleep apnea detection method. The examples and application scenarios implemented by the multiple modules and the corresponding steps are the same, but are not limited to the contents disclosed in the above-mentioned embodiment 1.

[0219] Example 3

[0220] According to another aspect of an embodiment of the present invention, an electronic device is provided, including: a processor; and a memory for storing processor-executable instructions, wherein the processor is configured to execute the instructions to implement any of the above sleep apnea detection methods.

[0221] Example 4

[0222] According to another aspect of an embodiment of the present invention, a computer-readable storage medium is provided. When instructions in the computer-readable storage medium are executed by a processor of an electronic device, the electronic device can perform any of the above sleep apnea detection methods.

[0223] The serial numbers of the above embodiments of the present invention are for description only and do not represent the advantages or disadvantages of the embodiments.

[0224] In the above embodiments of the present invention, the description of each embodiment has its own focus. For parts that are not described in detail in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.

[0225] In the several embodiments provided in this application, it should be understood that the disclosed technical content can be implemented in other ways. Among them, the device embodiments described above are only exemplary. For example, the division of the units can be a logical function division. In actual implementation, there may be other division methods, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of units or modules, which can be electrical or other forms.

[0226] The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed across multiple units. Some or all of the units may be selected according to actual needs to achieve the purpose of the present embodiment.

[0227] In addition, the functional units in the various embodiments of the present invention may be integrated into a single processing unit, each unit may exist physically separately, or two or more units may be integrated into a single unit. The aforementioned integrated units may be implemented in the form of hardware or software functional units.

[0228] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, or all or part of the technical solution can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes several instructions for enabling a computer device (which can be a personal computer, server or network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), mobile hard disk, magnetic disk or optical disk, etc. Various media that can store program codes.

[0229] The above is only a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principles of the present invention. These improvements and modifications should also be regarded as within the scope of protection of the present invention.

Claims

1. A sleep apnea detection method, characterized in that: include: Acquiring target physiological signal data corresponding to the target object; extracting target respiratory signal data from the target physiological signal data; Retrieving a reference frequency-amplitude fluctuation index, wherein the reference frequency-amplitude fluctuation index is obtained based on a plurality of respiratory frequency-amplitude distributions, wherein the plurality of respiratory frequency-amplitude distributions are respectively obtained based on corresponding sample respiratory signal data, wherein the corresponding sample respiratory signal data is obtained based on corresponding sample physiological signal data; Determining a plurality of sub-respiration signal data according to the target respiration signal data; Determining sub-respiration frequency and amplitude features corresponding to the plurality of sub-respiration signal data respectively; An apnea detection result of the target object is determined based on the reference frequency fluctuation index and the sub-respiration frequency characteristics corresponding to the plurality of sub-respiration signal data.

2. The method according to claim 1, characterized in that Before retrieving the reference frequency fluctuation index, the method further includes: In a case where the sample respiratory signal data includes a plurality of respiratory frequency and amplitude data, dividing the sample respiratory signal data to obtain a plurality of sample signal data segments, wherein each of the plurality of sample signal data segments includes at least one respiratory frequency and amplitude data; Determining, based on the plurality of respiratory frequency and amplitude data, sample respiratory frequency and amplitude features corresponding to the plurality of sample signal data segments respectively; The respiratory frequency-amplitude distribution corresponding to the sample respiratory signal data is determined based on the sample respiratory frequency-amplitude characteristics corresponding to the plurality of sample signal data segments.

3. The method according to claim 2, characterized in that The dividing the sample respiratory signal data to obtain a plurality of sample signal data segments includes: Determine the temporal overlap index; The sample respiratory signal data is divided according to the time overlap index to obtain a plurality of sample signal data segments.

4. The method according to claim 2, characterized in that The step of determining the respiratory frequency and amplitude distribution corresponding to the sample respiratory signal data based on the sample respiratory frequency and amplitude characteristics corresponding to the plurality of sample signal data segments further includes: When the sample respiratory frequency and amplitude feature includes a mean deviation index and the respiratory frequency and amplitude distribution includes a mean deviation histogram, determining the number of occurrences corresponding to a plurality of target deviation values, wherein the plurality of mean deviation indices include the plurality of target deviation values; An average deviation histogram corresponding to the sample respiratory signal data is obtained according to the multiple target deviation values ​​and the number of occurrences corresponding to the multiple target deviation values.

5. The method according to claim 3, characterized in that in, The time overlap index includes any one of the following: a time overlap ratio of 50%.

6. The method according to claim 1, characterized in that Before retrieving the reference frequency fluctuation index, the method further includes: Determining amplitude distribution intervals corresponding to the multiple respiratory frequency-amplitude distributions based on multiple matching indices corresponding to the multiple respiratory frequency-amplitude distributions; Determining, based on the amplitude distribution interval, fluctuation indices to be selected that correspond respectively to the plurality of respiratory frequency amplitude distributions; Determining sample evaluation results corresponding to a plurality of respiratory frequency amplitude distributions according to a plurality of selected fluctuation indices; According to the sample evaluation results corresponding to the plurality of respiratory frequency-amplitude distributions, a reference frequency-amplitude fluctuation index is determined from the plurality of fluctuation indices to be selected.

7. The method according to claim 6, characterized in that Determining a reference frequency fluctuation index from the plurality of candidate fluctuation indices based on the sample evaluation results corresponding to the plurality of respiratory frequency and amplitude distributions includes: Determining predetermined detection results corresponding to the plurality of respiratory frequency and amplitude distributions respectively; Determining a plurality of matching indexes corresponding to the plurality of respiratory frequency-amplitude distributions, wherein the plurality of matching indexes represent a degree of matching between sample evaluation results of the corresponding respiratory frequency-amplitude distribution under a plurality of selected fluctuation indices and a predetermined detection result; According to the multiple matching indices corresponding to the multiple respiratory frequency and amplitude distributions, a reference frequency and amplitude fluctuation index is determined from multiple candidate fluctuation indices, wherein the reference frequency and amplitude fluctuation index is the candidate fluctuation index with the largest matching index among the multiple candidate fluctuation indices.

8. The method according to claim 6, characterized in that When the respiratory frequency amplitude distribution includes a mean deviation histogram, the amplitude distribution interval includes any one of the following: the first 20% of multiple mean deviation indices arranged from small to large in the mean deviation histogram, and the reference frequency amplitude fluctuation index includes any one of the following: 65%.

9. The method according to claim 1, characterized in that The step of extracting target respiratory signal data from the target physiological signal data includes: Determine respiratory rate range; Determining a cutoff frequency parameter according to the respiratory frequency range; Filtering the target physiological signal data according to the cutoff frequency parameter to obtain initial respiratory signal data; Target breathing signal data is determined based on the initial breathing signal data.

10. The method according to claim 9, characterized in that Determining target respiratory signal data based on the initial respiratory signal data includes: Dividing the initial respiratory signal data into a plurality of initial sub-data segments in time sequence; performing windowing processing on each of the multiple initial sub-data segments to obtain multiple target sub-data segments, wherein the windowing processing is used to smoothly decay the signal data value at the first time point and the signal data value at the last time point in the corresponding initial sub-data segment to a predetermined value, and the target sub-data segments correspond one-to-one to the multiple initial sub-data segments; Determining frequency and amplitude parameters corresponding to the plurality of target sub-data segments respectively; Target respiratory signal data is determined according to the frequency and amplitude parameters respectively corresponding to the plurality of target sub-data segments.

11. The method according to claim 1, wherein Determining the apnea detection result of the target object based on the reference frequency fluctuation index and the sub-respiration frequency characteristics corresponding to the plurality of sub-respiration signal data includes: Determining a plurality of candidate signal data from the plurality of sub-respiration signal data according to the sub-respiration frequency and amplitude characteristics respectively corresponding to the plurality of sub-respiration signal data; For each signal data of the plurality of signal data to be selected, dividing it respectively to obtain a plurality of signal segments respectively corresponding to the plurality of signal data to be selected; Determining the segment frequency and amplitude characteristics corresponding to each of the plurality of signal segments corresponding to each of the candidate signal data; The apnea detection result of the target object is determined based on the segment frequency and amplitude characteristics corresponding to the multiple signal segments corresponding to each candidate signal data, wherein the apnea detection result includes the signal detection results corresponding to the multiple candidate signal data.

12. The method according to claim 11, characterized in that The step of determining the apnea detection result of the target object based on the frequency and amplitude characteristics of the multiple signal segments corresponding to each of the selected signal data includes: In a case where the plurality of signal segments corresponding to each candidate signal data are a plurality of signal segments arranged in a predetermined order, determining an arrangement order corresponding to each of the plurality of signal segments; Determining a characteristic difference index corresponding to each of the candidate signal data based on the respective rankings and segment frequency characteristics corresponding to the plurality of signal segments, wherein the characteristic difference index is used to indicate the degree of characteristic difference between signal segments of adjacent rankings; The apnea detection result of the target object is determined according to the characteristic difference index corresponding to each of the candidate signal data.

13. The method according to any one of claims 1 to 12, characterized in that The step of determining a plurality of sub-respiration signal data based on the target respiration signal data includes: Determining a partition index corresponding to the target respiratory signal data, wherein the partition index includes: a data time span, a time overlap rate, and a partition strategy, wherein the partition strategy is determined based on data features corresponding to the target respiratory signal data; The target respiratory signal data is divided according to the division index to obtain a plurality of sub-respiratory signal data.

14. A sleep apnea detection device, characterized in that: include: An acquisition module, configured to acquire target physiological signal data corresponding to a target object; an extraction module, configured to extract target respiratory signal data from the target physiological signal data; a retrieving module, configured to retrieve a reference frequency-amplitude fluctuation index, wherein the reference frequency-amplitude fluctuation index is obtained based on a plurality of respiratory frequency-amplitude distributions, wherein the plurality of respiratory frequency-amplitude distributions are respectively obtained based on corresponding sample respiratory signal data, wherein the corresponding sample respiratory signal data is obtained based on corresponding sample physiological signal data; A first determining module is configured to determine a plurality of sub-respiration signal data according to the target respiration signal data; A second determining module is used to determine the sub-respiration frequency and amplitude features corresponding to the plurality of sub-respiration signal data respectively; The third determining module is configured to determine an apnea detection result of the target object based on the reference frequency fluctuation index and the sub-respiration frequency characteristics corresponding to the plurality of sub-respiration signal data.

15. An electronic device, characterized in that: include: processor; a memory for storing instructions executable by the processor; The processor is configured to execute the instructions to implement the sleep apnea detection method according to any one of claims 1 to 13.

16. A computer-readable storage medium, characterized in that When the instructions in the computer-readable storage medium are executed by a processor of an electronic device, the electronic device is enabled to perform the sleep apnea detection method according to any one of claims 1 to 13.

Images