[0027] The specific embodiments of the present invention will be described in further detail below in conjunction with the accompanying drawings of the specification.
[0028] Such as figure 1 As shown, a sleep breathing monitoring method based on snoring signals designed by the present invention is implemented in practical applications with reference to the following steps:
[0029] Step 001. Since the human voice frequency is 300~3400Hz, the client (smartphone) is used to continuously record in real time according to the preset sampling frequency of 8000Hz to obtain the continuous breathing signal of a single sleeper to be tested. At the same time, in order to save data storage space, Also in order to save the data traffic of the user uploading to the server, the data is quantified as 8bit, and step 002 is entered.
[0030] Regarding the operation in step 001 above, there are two situations in practical applications. One is to simultaneously process multiple sleepers to be tested and the sleeper to be tested actively cooperates with the pre-recorded breathing signal, and the other is to record the continuous breathing signal of the sleeper to be tested In the environment, there are other people who are not asleep or those who are to be tested do not actively cooperate with the pre-recorded breathing signal. For these two situations, step 001 can specifically do the following operations:
[0031] When processing the breathing signals of multiple sleeping persons to be measured at the same time, firstly record the breathing signals of the sleeping persons to be measured separately, extract the acoustic features from them, and train to obtain the breathing sound models corresponding to the sleeping persons to be measured; Acquire the continuous breathing signals of multiple sleepers to be tested, and obtain the continuous breathing sounds corresponding to each sleeper to be tested by respectively corresponding to the breathing sound models of the sleepers to be tested, and finally the continuous breathing for each sleeper to be tested Breathing sound, perform all subsequent steps separately.
[0032] When the continuous breathing signal of the sleeping person to be tested is recorded in the environment, there are other people who are not asleep or the sleeping person to be tested does not actively cooperate with the pre-recorded breathing signal, according to the preset number of sleeping persons to be tested, the continuous breathing obtained by continuous recording Signal, realize the automatic clustering of the breathing signals of each sleeping person to be tested, obtain the continuous breathing sound corresponding to each sleeping person to be tested, and finally perform all subsequent steps separately for the continuous breathing sound of each sleeping person to be tested.
[0033] Step 002. Regarding the continuous breathing signal of the person to be tested for falling asleep, each time starting from the 0th time and successively adjacent to each other for a duration of 2 minutes is used as each starting time, and each starting time corresponds to each of the 4 minutes long Respiration signals constitute each respiration signal segment, and step 003 is entered.
[0034] Step 003. Considering that when continuous breathing signals are collected, wind sounds, car passing sounds, noises, etc. may be collected. These sound signals are obviously different from breathing signals. Independent component analysis (ICA) can be used to first target the Each respiratory signal segment of a sleeping person is pre-processed separately to remove the non-respiratory signal part; then, in real time at each time starting from the 4th minute and successively adjacent to each other for 2 minutes, for each time as the end time The corresponding respiratory signal segment is processed in real time, that is, all subsequent steps are continued to realize real-time processing for each respiratory signal segment. Among them, it is determined whether each respiratory signal segment contains a snoring signal, if each respiratory signal segment does not contain snoring sound Signal, it is determined that there is no obstructive respiration in the continuous breathing signal acquisition process of step 001, and the monitoring method ends; if there is at least one breathing signal segment in each breathing signal segment that contains a snoring signal, it is obtained The position of the snoring signal in all the corresponding breathing signal segments and the duration of the snoring signal enter step 004; among them, the operations for each breathing signal segment are as follows:
[0035] Step 003-1. Obtain the average energy of the respiratory signal segment as the reference energy, and dynamically set the energy threshold of the respiratory signal segment; at the same time, obtain the highest zero-crossing rate of the respiratory signal segment as the reference zero-crossing rate, and dynamically set the respiratory signal segment The zero-crossing rate threshold.
[0036] Step 003-2. Use a smoothing window to calculate the short-term average energy in the smoothing window as the current actual energy; at the same time, use a smoothing window to calculate the short-term average energy in the smoothing window for the respiratory signal. The zero rate is used as the current actual zero-crossing rate.
[0037] Step 003-3. Perform framing for the respiratory signal segment, and for each respiratory signal frame on the respiratory signal, determine whether the current actual energy and the current actual zero-crossing rate of the respiratory signal frame respectively correspond to exceed the respiratory signal segment The energy threshold and the zero-crossing rate threshold are determined to be the snoring signal frame, and the time corresponding to the first frame of the snoring signal is regarded as the starting time of the snoring signal, and the time corresponding to the last frame is regarded as the snoring signal At the end time of the signal, the position of the snoring signal in the corresponding breathing signal segment and the duration of the snoring signal are obtained; otherwise, it is determined that the breathing signal frame is not a snoring signal.
[0038] Step 004. According to the position of the snore signal and the duration of the snore signal, determine whether the interval between two adjacent snore signals exceeds the lower limit of the preset breathing interval by 10 seconds but is less than the preset breathing interval The upper limit is 100 seconds, and there is a continuous snoring signal within 30 seconds of the preset judgment time before the two pieces of snoring signal, and there is a continuous snoring signal within 30 seconds of the preset judgment time after the two pieces of snoring signal. It is detected that there is no obstructive breathing in the continuous breathing signal acquisition process of the sleeping person in step 001, and the monitoring method ends; if there is, go to step 005;
[0039] Step 005. For each group of adjacent snoring signals that meet the conditions in step 004, according to the method of judging whether the respiratory signal segment contains a snoring signal in step 003, determine whether there is a breathing signal in each group of adjacent snoring signals. Each group of adjacent snoring signals contains a breathing sound signal, it is determined that the sleeper to be tested does not have obstructive breathing during the continuous breathing signal acquisition process in step 001, and the monitoring method ends; if each group of adjacent snoring signals, If there is at least one set of adjacent snoring signals that do not contain breathing signals, it is determined that there is primary obstructive breathing in the corresponding adjacent snoring signals, and data is merged for the repeated parts in the corresponding adjacent snoring signals to obtain all existing primary obstructive respirations. Adjacent snoring signals of obstructive respiration, the client stores all adjacent snoring signals with primary obstructive respiration, and transmits them to the server through the wireless communication network, and step 006 is performed.
[0040] Step 006. The server receives adjacent snoring signals with primary obstructive respiration, and extracts the acoustic characteristics of each adjacent snoring signal for all adjacent snoring signals with primary obstructive respiration, and based on the pre-established obstructive respiration The model determines whether there is obstructive respiration in each adjacent snoring signal. If it does not exist, it is determined that there is no obstructive respiration in the continuous breathing signal recording process of step 001; if there is, continue to obtain statistics. The number of obstructive respirations and the single duration of the adjacent snoring signals are further based on the statistics of the number of obstructive respirations and the duration of single respiration of the person to be tested during the continuous breathing signal registration process in step 001, and based on this The sleep quality judgment rule is preset, and it is determined to obtain the sleep quality of the to-be-tested sleeper during the continuous breathing signal recording process in step 001.
[0041] Among them, the method for the server to establish an obstructive breathing model is to collect multiple sets of blood oxygen and snoring signals based on a polysomnography, and refer to the indicators of blood oxygen to extract the snoring signal when the blood oxygen is reduced, and use the characteristics of the snoring signal, To establish an obstructive breathing model, the specific process is: first use the blood oxygen index to obtain the snoring data when blood oxygen decreases, and then extract the acoustic characteristics of the snoring data, including: energy, frequency, zero-crossing rate, mfcc, lpcc, etc., and then use Models, including: gmm, svdd, etc., model the extracted acoustic features to realize intelligent judgment of whether there is obstructive breathing in the snoring signal. In the early stage of obstructive breathing model, blood oxygen data and snoring data collected from hospitals were mainly used to establish a primary obstructive breathing model model. When the system is running, the model will be continuously revised and improved by using the data uploaded by users to improve the obstructive breathing model. The specific method of recognition accuracy is: use manual monitoring to mark obstructive respiratory events, compare and correlate with obstructive respiratory events recognized by the system, and continuously modify the snoring model parameters to improve the recognition accuracy of obstructive respiratory models.
[0042] Sleeping person’s sleep breathing status, including quiet sleep time, snoring time, time of obstructive breathing event, total number of times, and single duration statistics; the preset sleep quality judgment rules are as follows:
[0043] Good quality of sleep: no obstructive breathing;
[0044] Good sleep quality: The number of obstructive breaths is less than or equal to 15 times, and the duration of each time is less than or equal to 20 seconds;
[0045] Sleep quality: the number of obstructive breaths is greater than 15 times and less than or equal to 30 times, and the duration of each time is less than or equal to 40 seconds;
[0046] Poor sleep quality: The number of obstructive breaths is greater than 30, or the duration of a single breath is greater than 40 seconds.
[0047] As a preferred technical solution of the present invention: In step 006, it is determined whether there is obstructive breathing in each adjacent snoring signal, and if not, it is determined that the person to be tested for falling asleep is continuously recorded in step 001. There is no obstructive respiration in the process; if it exists, continue to count the number of obstructive respirations in the adjacent snoring signal and the single duration, and then according to the statistics, the sleeper under test is in the continuous breathing signal recording process in step 001 The number of obstructive respirations and the duration of a single respiration are used to determine the sleep quality of the person to be tested during the continuous breathing signal recording process in step 001 according to the preset sleep quality determination rule.
[0048] As a preferred technical solution of the present invention: the sleep breathing monitoring method is implemented based on the client and the server, wherein the steps 001 to 005 are executed in the client, and the step 006 is executed in the server. The server communicates with each other through a wireless communication network; wherein, in step 006, after the server obtains the sleep quality of the sleeper to be tested during the continuous breathing signal acquisition process in step 001, the server feeds back the sleep quality through the wireless communication network To the client, the client feeds back the received sleep quality to the user in combination with the stored adjacent snoring signal of primary obstructive breathing.
[0049] Among them, the content of sleep quality includes: including the number of snoring, the ratio of snoring to sleep time, the number of obstructions, the duration of a single obstruction, the cumulative obstruction duration, the period of occurrence of obstructed breathing, sleep quality scores, suggestions, etc., and in actual applications, it is targeted at the client , It is designed as a smart phone with corresponding software installed.
[0050] In the practical application of the present invention, if step 001 realizes the continuous breathing signal acquisition for the sleeper to be tested throughout the night, it can be determined that the sleep quality of the sleeper to be tested for the entire night can be finally obtained.
[0051] In practical applications of the present invention, hypnosis and wake-up functions can also be designed for the client. The hypnosis function is mainly to play soft music, voice, etc. at regular intervals to help the user fall asleep, and automatically stop the sound playback after falling asleep; the wake-up function is mainly to sleep through the ringtone as needed Patients in the state or obstructive breathing state wake up and change their current state.
[0052] When an obstructive respiratory event is detected several times in a sleeping person, and an obstructive respiratory event is detected again, the cell phone can be set to automatically dial the guardian’s cell phone, and the guardian can monitor the sleeper’s snoring in real time. When the sleeping person is confirmed as When severely obstructive breathing (more times or a single time of more than 40 seconds), choose to send instructions to the sleeping person’s mobile phone to wake up the sleeping person; among them, the guardian can be the sleeping person’s relatives or the background of the design system of the present invention Managers, guardians, or other smart devices (the smart device has the ability to recognize obstructive breathing online).
[0053] The sleep breathing monitoring method based on the snoring signal designed by the present invention has clear logic. It is determined whether the sleeper has obstructive breathing based on the snoring signal, can realize real-time monitoring for the sleeper, and provide the sleeper with scientific and accurate sleep breathing monitoring. Pay attention to the health of sleepers during sleep; and based on the client and server as the hardware carrier, the design method of the present invention is realized, which effectively controls the cost in the practical application of the method. Among them, not only the layered processing architecture of data analysis is realized, but also at the same time It ensures the quality of data communication between the data layered processing and is easy to promote in the market.
[0054] The embodiments of the present invention are described in detail above with reference to the accompanying drawings, but the present invention is not limited to the above-mentioned embodiments, and can be made without departing from the purpose of the present invention within the scope of knowledge possessed by those of ordinary skill in the art. Various changes.
