Infant sleep state monitoring method and system based on sound analysis

Abstract

The invention relates to the technical field of infant sleep monitoring, in particular to an infant sleep state monitoring method and system based on sound analysis, and the method comprises the following steps: based on an intelligent infant bed, collecting sound signals sent by an infant in real time through a microphone sensor on the inner side of the bed, recognizing sound in a deep sleep period, sound in a light sleep period and sound in a rapid eye movement period, and transmitting the sound signals to the intelligent infant bed; and acquiring a sleep cycle sound category. According to the invention, the sound signals are collected in real time and are accurately classified according to the frequency range, so that the states of deep sleep, light sleep and rapid eye movement periods of the baby can be more accurately identified, and through careful comparison with reference modes in the audio database of the baby crib, the matching precision of data is improved, and the accuracy of the baby crib is improved. And the utilization efficiency of the sound data is optimized, so that the response capability of the whole monitoring system is improved, the monitoring sensitivity is dynamically adjusted, the sampling frequency is adjusted according to real-time data, the potential of misinformation is reduced, and the practicability of the monitoring system is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of infant sleep monitoring, and in particular to a method and system for monitoring an infant's sleep state based on sound analysis. Background Art

[0002] Infant sleep monitoring technology focuses on developing various methods and equipment to ensure the safety and health of infants during sleep. This technology involves the use of sensors, cameras, microphones and other equipment to collect data, including the baby's breathing, heart rate, body movement and sound information. By analyzing the data, the monitoring system can identify potential health problems of the baby, such as apnea, poor sleep quality, etc. In addition, this technology can also help parents better understand their baby's sleep patterns and habits, thereby improving baby's sleep safety and reducing parents' stress.

[0003] Among them, the baby sleep status monitoring method of sound analysis is a technology developed in the application environment of smart cribs. Smart cribs create a comfortable sleeping space through sound, temperature and shaking. At the same time, smart cribs also have monitoring functions. They can monitor the baby's sleep status through sound analysis technology while the baby is sleeping. By analyzing the baby's crying, breathing and other audio signals, the baby's sleep quality and whether there are potential sleep problems can be judged. The core is to use sound recognition and data analysis technology to evaluate the baby's safety and health in real time, provide parents with instant sleep status feedback, and intervene when necessary, thereby greatly improving the baby's sleep safety.

[0004] Existing technologies fail to adequately distinguish between different sound types and sleep stages, resulting in incomplete assessments of infants' sleep states. This limitation leads to misunderstandings of the infant's true sleep state and an inability to promptly identify and respond to important signals such as apnea or abnormal crying. In addition, the lack of effective sound analysis can easily lead to the monitoring system's inability to effectively filter out routine and abnormal sounds, increasing the risk of false alarms or missed alarms, directly affecting guardians' confidence and the timeliness of intervention, leading to unnecessary anxiety or neglect of real crises. Summary of the Invention

[0005] The purpose of the present invention is to solve the shortcomings of the prior art and to propose a method and system for monitoring the sleeping state of an infant based on sound analysis.

[0006] In order to achieve the above object, the present invention adopts the following technical solution: a method for monitoring infant sleep status based on sound analysis, comprising the following steps:

[0007] S1: Based on the smart crib, the microphone sensor inside the crib collects the sound signals emitted by the baby in real time, identifies the sounds during deep sleep, light sleep, and rapid eye movement, and obtains the sound categories of the sleep cycle;

[0008] S2: Based on the sleep cycle sound category, compare the sound data with the standard sleep pattern, determine the category to which the current sound data belongs, and obtain a sleep pattern matching score;

[0009] S3: Based on the sleep pattern matching score, monitor the current infant's sleep status in real time, analyze the sound fluctuation amplitude and mutation frequency of the current sleep stage, control the sampling frequency change of the sound data, match the differentiated sleep stages, and obtain the sleep monitoring adjustment configuration;

[0010] S4: adjusting the configuration according to the sleep monitoring, collecting the infant's crying characteristics, breathing sound pattern, and throat sound change rate, calculating the time distribution and intensity fluctuation of the sound, comparing it with the infant's normal sleep sound, detecting continuous or sudden abnormal sounds, and obtaining an abnormal sound identification record;

[0011] S5: Based on the abnormal sound recognition record, the frequency and duration of the abnormal sound are counted to evaluate whether the infant's current sleep state is abnormal, and obtain a sleep stability evaluation result.

[0012] The improvements of the present invention are that the sleep cycle sound categories include deep sleep sound types, light sleep sound types, and rapid eye movement sound types; the sleep pattern matching score includes a frequency matching score, an energy matching score, and a time matching score; the sleep monitoring adjustment configuration includes a fluctuation amplitude adjustment parameter, a sampling frequency adjustment parameter, and a sensitivity adjustment parameter; the abnormal sound identification record includes a crying feature record, a breathing pattern record, and a laryngeal sound change record; and the sleep stability assessment result includes a stability level, an abnormality occurrence frequency, and an abnormality duration.

[0013] The present invention is improved in that the step of obtaining the sleep cycle sound category is specifically as follows:

[0014] S111: Based on the smart crib, the microphone sensor inside the crib collects the baby's sound signals in real time, records the sound frequency data and volume change rate, analyzes the time series information, and then extracts the signal amplitude, frequency range and change trend characteristics to generate sound time series feature values;

[0015] S112: Based on the sound time series feature value, the sound signal of each time period is classified, the sound signal is divided into fixed time windows, the sound amplitude within the window is collected, and the formula is used:

[0016]

[0017] Calculate the average sound energy EF of the current window i , and then distinguish deep sleep, light sleep and rapid eye movement period to generate sleep cycle sound categories, where T is the number of sampling points, AFt is the instantaneous voltage amplitude of the sound signal collected at time t.

[0018] The present invention is improved in that the steps of obtaining the sleep pattern matching score are specifically as follows:

[0019] S211: Based on the sleep cycle sound category, perform frequency domain and time domain analysis on the sound data, calculate the frequency distribution, energy characteristics and time domain characteristics in each time period, and establish a sound feature parameter set;

[0020] S212: Based on the sound feature parameter set, using the sleep reference sound in the crib audio database, the formula:

[0021]

[0022] Calculate the sound matching degree MR between the current sound data and the sleep state sample s , where Pq i is the energy value of the current sound data in the i-th frequency segment, Rq i is the energy value of the sleep reference sound in the database at the i-th frequency segment, N q is the total number of frequency segments;

[0023] S213: Compare the sound matching degree with the sound data according to the reference benchmark of the sleep pattern, calculate the ratio in each matching degree interval, determine the category of the current sound data, and obtain the sleep pattern matching score.

[0024] The present invention is improved in that the steps of obtaining the sleep monitoring adjustment configuration are specifically as follows:

[0025] S311: Based on the sleep pattern matching score, the infant's current sleep condition is monitored in real time, the sound fluctuation amplitude and mutation frequency of the current sleep stage are analyzed, the time series change trend thereof is calculated, and it is determined whether the sound fluctuation amplitude exceeds a stable range, thereby obtaining a sound fluctuation deviation;

[0026] S312: Based on the sound fluctuation deviation, in case of exceeding the stable range, adjusting the sound monitoring sensitivity of the crib, controlling the sampling frequency of the sound data, and matching the different sleep stages in combination with the sound feature intervals of the different sleep stages;

[0027] S313: Based on the differentiated sleep stages, the sound fluctuation amplitudes of the differentiated sleep stages are classified according to the adjusted sound monitoring sensitivity and sampling frequency, using the formula:

[0028]

[0029] Calculate the monitoring adjustment index SY m, get the sleep monitoring adjustment configuration, where Vy i Represents the sound fluctuation amplitude at the i-th moment, Vy i-1 Represents the sound fluctuation amplitude at the i-1th moment, n y Represents the number of sampling points, Fy s Represents the adjusted sound detection sensitivity, Fy b Represents baseline sound detection sensitivity.

[0030] The present invention is improved in that the steps of obtaining the abnormal sound recognition record are specifically as follows:

[0031] S411: According to the sleep monitoring configuration, the infant's crying characteristics, breathing sound pattern, and throat sound change rate are collected, and the collected sound data is subjected to feature separation to extract the time distribution, intensity fluctuation, and frequency change of the sound. The short-term energy values and peak time intervals of multiple types of sounds are calculated to obtain a sound feature parameter set.

[0032] S412: Calling the sound feature parameter set, comparing it with the sound fluctuation range of the infant in the normal sleeping state, screening abnormal audio segments, detecting their duration and sudden change rate, and analyzing the distribution of abnormal audio segments on the time axis to obtain abnormal sound feature vectors;

[0033] S413: Based on the abnormal sound feature vector, calculate the features of each abnormal sound category using the formula:

[0034]

[0035] Get abnormal sound recognition record Ao r , among which, So i Represents the intensity of the ith abnormal sound segment, So norm Represents the average intensity of normal sleep sounds, To i Represents the duration of the i-th abnormal sound segment, To max Represents the longest duration of all abnormal sound segments, Fo i Represents the frequency of the i-th abnormal sound segment, Fo norm represents the average frequency of normal sleep sounds, n A Represents the total number of abnormal sound segments.

[0036] The present invention is improved in that the steps of obtaining the sleep stability evaluation result are specifically as follows:

[0037] S511: Based on the abnormal sound recognition record, calculate the frequency and duration of abnormal sounds during the monitoring period, count the total number of abnormal sound events during the time period, and accumulate the duration of abnormal sound events to obtain abnormal sound statistical parameters;

[0038] S512: Based on the abnormal sound statistical parameters and the sleep pattern matching score, the abnormal sound frequency and duration in each time period are compared, using the formula:

[0039]

[0040] Calculate the matching degree deviation value Sz d , and analyze the deviation between abnormal sound distribution and sleep pattern, evaluate whether the baby's current sleep state is abnormal, and obtain the sleep stability assessment result, among which Fz i Represents the number of abnormal sounds in the i-th time period, Tz i represents the duration of the i-th time period, Pz i represents the sleep pattern matching score corresponding to the i-th time period, n z Represents the total number of time periods.

[0041] A baby sleep state monitoring system based on sound analysis, the system comprising:

[0042] The sound collection and classification module is based on the smart crib. It uses the microphone sensor inside the crib to collect the baby's sound signals in real time, record the sound frequency and volume change rate, and analyze the timing information to identify deep sleep, light sleep, and rapid eye movement sounds, and obtain the sleep cycle sound categories.

[0043] The sound pattern matching module analyzes the frequency distribution, energy characteristics, and time domain characteristics of the sound data based on the sleep cycle sound category, compares the sound data with the standard sleep pattern, determines the category to which the current sound data belongs, and obtains a sleep pattern matching score;

[0044] The monitoring and adjustment module monitors the current infant's sleep status in real time based on the sleep pattern matching score, determines whether the sound fluctuation amplitude exceeds the stable range, controls the sampling frequency change of the sound data, matches the differentiated sleep stages, and obtains the sleep monitoring adjustment configuration;

[0045] The abnormal sound detection module adjusts the configuration according to the sleep monitoring, calculates the time distribution and intensity fluctuation of the sound, compares it with the normal sleep sound of the infant, detects continuous or sudden abnormal sounds, and obtains abnormal sound identification records;

[0046] The sleep stability assessment module counts the frequency and duration of abnormal sounds based on the abnormal sound recognition records, combines the sleep pattern matching score, analyzes the stability level of the current sleep state, assesses whether the infant's current sleep state is abnormal, and obtains a sleep stability assessment result.

[0047] Compared with the prior art, the advantages and positive effects of the present invention are:

[0048] In the present invention, by collecting sound signals in real time and accurately classifying them according to the frequency range, the deep sleep, light sleep and rapid eye movement states of the infant can be identified more accurately. By making a detailed comparison with the reference patterns in the crib audio database, not only the matching accuracy of the data is improved, but also the utilization efficiency of the sound data is optimized, thereby improving the responsiveness of the overall monitoring system. The monitoring sensitivity and sampling frequency are dynamically adjusted according to the real-time data, which reduces the potential for false alarms and improves the practicality of the monitoring system. The sensitive capture of abnormal sounds allows the system to detect and remind guardians in time, thereby enhancing the safety protection of the infant. BRIEF DESCRIPTION OF THE DRAWINGS

[0049] Figure 1 The present invention proposes a flow chart of a method for monitoring infant sleep status based on sound analysis;

[0050] Figure 2 This is a flowchart for obtaining sleep cycle sound categories in the present invention;

[0051] Figure 3 A flowchart for obtaining a sleep pattern matching score in the present invention;

[0052] Figure 4 A flowchart for obtaining the sleep monitoring adjustment configuration in the present invention;

[0053] Figure 5 This is a flowchart for obtaining abnormal sound recognition records in the present invention;

[0054] Figure 6 This is a flowchart for obtaining sleep stability evaluation results in the present invention. DETAILED DESCRIPTION

[0055] In order to make the purpose, technical solutions and advantages of the present invention more clearly understood, the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0056] In the description of the present invention, it should be understood that the terms "length", "width", "up", "down", "front", "back", "left", "right", "vertical", "horizontal", "top", "bottom", "inside", "outside", etc., indicating directions or positional relationships, are based on the directions or positional relationships shown in the accompanying drawings and are only for the convenience of describing the present invention and simplifying the description. They do not indicate or imply that the devices or elements referred to must have a specific direction, be constructed and operate in a specific direction, and therefore should not be understood as limiting the present invention. In addition, in the description of the present invention, the meaning of "plurality" is two or more, unless otherwise clearly and specifically defined.

[0057] Example

[0058] See also Figure 1 The present invention provides a technical solution: a method for monitoring infant sleep status based on sound analysis, comprising the following steps:

[0059] S1: Based on the smart crib, the system collects the baby's sound signals in real time through the microphone sensor inside the crib, records the sound frequency and volume change rate, analyzes the timing information, and then classifies the data according to the sound frequency range to identify deep sleep, light sleep, and rapid eye movement sounds, and obtain the sleep cycle sound categories;

[0060] S2: Based on the sleep cycle sound category, the sound data is analyzed using the sleep reference sounds in the crib audio database to determine the degree of match. The sound data is then compared with the standard sleep pattern to determine the category to which the current sound data belongs and obtain a sleep pattern match score.

[0061] S3: Based on the sleep pattern matching score, the current infant's sleep status is monitored in real time. The sound fluctuation amplitude and mutation frequency of the current sleep stage are analyzed to determine whether the sound fluctuation amplitude exceeds the stable range. If it exceeds the range, the sound monitoring sensitivity of the crib is adjusted, the sampling frequency of the sound data is controlled, the baby's sleep stage is matched, and the sleep monitoring adjustment configuration is obtained.

[0062] S4: Adjust the configuration based on sleep monitoring to collect the baby's crying characteristics, breathing sound patterns, and laryngeal sound change rate. Calculate the temporal distribution and intensity fluctuations of the sounds, compare them with the baby's normal sleep sounds, detect persistent or sudden abnormal sounds, including prolonged crying, rapid gasping, and abnormal breathing pauses, and obtain abnormal sound identification records.

[0063] S5: Based on the abnormal sound recognition records, the frequency and duration of abnormal sounds are counted, and combined with the sleep pattern matching score, the stability level of the current sleep state is analyzed to evaluate whether the infant's current sleep state is abnormal, and the sleep stability assessment result is obtained.

[0064] Sleep cycle sound categories include deep sleep sound types, light sleep sound types, and rapid eye movement sound types. Sleep pattern matching scores include frequency matching scores, energy matching scores, and time matching scores. Sleep monitoring adjustment configurations include fluctuation amplitude adjustment parameters, sampling frequency adjustment parameters, and sensitivity adjustment parameters. Abnormal sound identification records include crying feature records, breathing pattern records, and laryngeal sound change records. Sleep stability assessment results include stability level, frequency of abnormal occurrence, and duration of abnormality.

[0065] See also Figure 2 ,The steps for obtaining the sleep cycle sound category are as follows:

[0066] S111: Based on the smart crib, the microphone sensor inside the crib collects the baby's sound signals in real time, records the sound frequency data and volume change rate, analyzes the time series information, and then extracts the signal amplitude, frequency range and change trend characteristics to generate sound time series feature values;

[0067] The microphone sensor sets up multiple collection points at different positions inside the bed. Each collection point records the waveform data of the sound signal at a fixed time interval and stores the amplitude information of the sound in a numerical form. The sound signal undergoes preprocessing, including noise filtering and signal enhancement, to remove background noise and improve signal clarity. In practical applications, for example, there are interference sources such as fans and air conditioners in the room. The preprocessing step can use filtering technology to remove the interference frequency. The collected signal undergoes spectral analysis to extract the sound frequency data. The frequency at each time point is obtained by Fourier transform through the specified window size to ensure that the time domain data is converted into frequency domain data. At the same time, the volume change rate is recorded, that is, the amplitude change rate is calculated at adjacent time points. For example, the time interval is set to 10 milliseconds. The amplitude change rate can be calculated by dividing the difference between the current amplitude value and the amplitude value at the previous time point. It is obtained in time intervals, and the unit of volume change rate can be decibels per second. When recording time series information, each sound data is accompanied by a timestamp, so that subsequent data analysis can establish data associations on the time axis. In practical applications, if an infant falls asleep at night, his or her vocalization pattern will show periodic changes, so the recording of timestamps can help identify his or her vocalization rhythm. When further analyzing the sound signal, the amplitude, frequency range and change trend characteristics of the signal are extracted. The signal amplitude is calculated by taking the maximum and minimum values of the amplitude at each time point. The frequency range can be filtered out by setting a threshold to filter out common sound bands for infants. For example, sounds in the range of 200Hz to 1200Hz are more common. The change trend uses the sliding average method to calculate the average frequency value of the last N time points to determine how the frequency changes over time and form a sound time series feature value.

[0068] S112: Based on the sound time series feature value, the sound signal of each time period is classified, the sound signal is divided into fixed time windows, the sound amplitude within the window is collected, and the formula is used:

[0069]

[0070] Calculate the average sound energy EF of the current window i , and then distinguish deep sleep, light sleep and rapid eye movement period to generate sleep cycle sound categories, where T is the number of sampling points, AF t is the instantaneous voltage amplitude of the sound signal collected at time t, the voltage collected by the microphone. The microphone is a sensor that converts sound waves into electrical signals;

[0071] The sound data is divided into multiple time periods according to the time window, and each time period is usually 5 seconds to ensure the stability of classification. In each time period, the amplitude data of the sound signal is collected and the average sound energy of the segment is calculated. This energy can be used as an important basis for sleep state identification. For each time period, the average sound energy is calculated. For example, if the sampling time T = 5 seconds and the sampling is once per second, the sampling data is as follows:

[0072] Sound Amplitude AF t (Unit V): [0.2, 0.4, 0.6, 0.5, 0.3], then the average sound energy is:

[0073]

[0074] The result shows that in the current 5-second time window, the average energy of the sampled data is 0.18V 2 , the energy value can be compared with the energy baseline of each sleep stage to determine whether the time period belongs to deep sleep, light sleep or rapid eye movement period.

[0075] See also Figure 3 ,The steps for obtaining the sleep pattern matching score are as follows:

[0076] S211: Based on the sleep cycle sound categories, perform frequency domain and time domain analysis on the sound data to calculate the frequency distribution, energy characteristics, and time domain characteristics within each time period. The frequency distribution is obtained by calculating the energy proportion of the sound signal within different frequency ranges, the energy characteristics are calculated by signal power spectral density, and the time domain characteristics are obtained by amplitude envelope analysis to establish a sound feature parameter set.

[0077] The sound data is divided into time segments, and the duration of each segment is set to 5 seconds to ensure the stability of the analysis. In each segment, the sound signal is analyzed in the frequency domain and time domain to calculate its frequency distribution, energy characteristics and time domain characteristics. The frequency distribution is obtained by calculating the energy proportion of the sound signal in different frequency ranges. The amplitude of each frequency component is extracted by Fourier transform and normalized. For example, signals of 300Hz, 500Hz and 800Hz are detected in a certain segment. After normalization, their energy proportions are calculated to be 0.3, 0.5 and 0.6 respectively. 0.2, the energy feature is obtained by calculating the power spectrum density, that is, obtaining the average power of the signal at each frequency. For example, the power spectrum density of the signal in a certain time period is 0.1W / Hz at 200Hz, 0.3W / Hz at 400Hz, and 0.2W / Hz at 600Hz. The time domain feature is obtained through amplitude envelope analysis, that is, the envelope of the signal is extracted and the trend of its amplitude change is calculated. For example, the amplitude envelope value of a certain segment fluctuates between 0.2-0.8. By calculating the data of all segments, a sound feature parameter set is generated.

[0078] S212: Based on the sound feature parameter set, using the sleep reference sound in the crib audio database, the formula:

[0079]

[0080] Calculate the sound matching degree MR between the current sound data and the sleep state sample s , where Pq i is the energy value of the current sound data in the i-th frequency segment, Rq i is the energy value of the sleep reference sound in the database at the i-th frequency segment, N q is the total number of frequency segments;

[0081] Call the sleep reference sound in the crib audio database, calculate the matching degree between the current sound data and each sleep state sample in the database, and set the frequency segment to be N q = 10 intervals, each interval corresponds to a different frequency range, such as 200Hz-400Hz and 400Hz-600Hz, etc. The energy value difference between the current sound data and the reference sound in each frequency band is calculated respectively, and the sound matching degree is calculated using the formula, Pq i For example, the energy value of the 200Hz frequency band is 0.3dB, Rq i For example, the energy value of the 200Hz frequency band is 0.2dB. Given the data:

[0082] Current sound data Pq i :0.3,0.5,0.4,0.6,0.2;

[0083] Sleep Reference Sounds Rq i :0.2,0.6,0.3,0.7,0.3;

[0084] Compute the sum of absolute deviations:

[0085] |0.3-0.2|+|0.5-0.6|+|0.4-0.3|+|0.6-0.7|+|0.2-0.3|=0.1+0.1+0.1+0.1+0.1=0.5;

[0086] Calculate square roots of sums of squares:

[0087]

[0088] Calculate the matching degree:

[0089]

[0090] The sound matching degree of the current sound data is 0.253.

[0091] S213: Comparing the sound data with the sound matching degree based on the reference standard of the sleep pattern, calculating the ratio within each matching degree interval, determining the category of the current sound data, and obtaining the sleep pattern matching score;

[0092] Based on the reference benchmark of the sleep pattern, the sound data is compared to calculate the proportion of matching in different intervals, and the matching degree is divided into three intervals, such as 0≤MR s <0.1 is classified as high match, 0.1≤MR s <0.3 is classified as moderate match, MR s A match score of ≥0.3 is classified as a low match. The proportion of the matching scores of all time segments falling into different intervals is counted. For example, among 10 segments, 6 segments fall into the high match interval, 3 segments fall into the medium match interval, and 1 segment falls into the low match interval. The high match ratio is 60%, and the matching score of 0.253 falls into the medium match interval, indicating that the current sound data has a medium match with the standard sleep pattern sound in the crib audio database, indicating that the baby's sleep state is good and basically consistent with the expected healthy sleep pattern, and the sleep pattern matching score is obtained.

[0093] See also Figure 4 , the steps to obtain the sleep monitoring adjustment configuration are as follows:

[0094] S311: Based on the sleep pattern matching score, the infant's current sleep status is monitored in real time, the sound fluctuation amplitude and mutation frequency of the current sleep stage are analyzed, its time series change trend is calculated, and the sound fluctuation amplitude is determined to be beyond the stable range, thereby obtaining the sound fluctuation deviation;

[0095] The current sleep state of the baby is collected in real time, including body activity, respiratory rate and sound signals. After obtaining the sound data, the data needs to be preprocessed, including removing background noise and outliers to ensure the accuracy of the analyzed data. The characteristics of the sound signal mainly include sound fluctuation amplitude and mutation frequency. Among them, the sound fluctuation amplitude refers to the degree of change of sound intensity per unit time, and the mutation frequency refers to the severity of the change in sound amplitude and the number of mutation points. For example, if a baby suddenly cries from a quiet state, multiple mutation points will appear in a short period of time. Subsequently, based on the sound fluctuation amplitude and mutation frequency in continuous time, a time series model is established to calculate the change of the signal. Trend, which can be calculated by sliding window calculation to obtain the local average change rate, thereby reflecting the stability of the sound signal during the sleep stage. For example, within a 10-second time window, the mean of the sound fluctuation amplitude at each time point and its difference with the mean of the previous time window are calculated. If the difference exceeds the set stability range threshold, it means that the sound fluctuation has abnormal fluctuations. The threshold of the stability range needs to be set based on a large amount of monitoring data. Usually, the quantile calculation method can be used to set a benchmark range. For example, the sound fluctuation range within the 90% quantile during normal sleep of the infant is taken as the threshold. If the actual fluctuation exceeds this range, it is judged as an abnormal fluctuation, and the sound fluctuation deviation is obtained.

[0096] S312: Based on the sound fluctuation deviation, if the sound exceeds the stable range, the sound monitoring sensitivity of the crib is adjusted, the sampling frequency of the sound data is controlled, and the sound characteristic intervals of different sleep stages are combined to match the different sleep stages;

[0097] If the sound fluctuation deviation exceeds the set stable range, the crib's sound monitoring sensitivity needs to be adjusted to enable the system to more sensitively capture sound changes, or in specific cases, to reduce sensitivity to avoid false triggering. At the same time, the sampling frequency of the sound data is controlled to increase the sampling frequency during high-fluctuation phases and decrease the sampling frequency during stable phases to reduce the amount of computation. For example, when a small sound fluctuation deviation is detected, the sampling frequency can be set to 10Hz, while when a larger fluctuation is detected, the sampling frequency can be increased to 50Hz to ensure more granular capture of sound changes. Subsequently, the sound fluctuation data of different time periods is matched with the typical sound feature intervals of different sleep stages. The sound features of different sleep stages can be obtained based on a large amount of infant sleep data. For example, sound fluctuations in light sleep are typically around 10dB, less than 5dB in deep sleep, and greater than 15dB in wakefulness. Therefore, the real-time data is classified using the set threshold interval to determine the current differentiated sleep stage of the infant.

[0098] S313: Based on the differentiated sleep stages, the sound fluctuation amplitudes of the differentiated sleep stages are classified according to the adjusted sound monitoring sensitivity and sampling frequency using the formula:

[0099]

[0100] Calculate the monitoring adjustment index SY m ,The index is used to guide how to adjust the monitoring parameters of the smart crib and obtain the sleep monitoring adjustment configuration, where,Vy i Represents the sound fluctuation amplitude at the i-th moment, Vy i-1 Represents the sound fluctuation amplitude at the i-1th moment, n y Represents the number of sampling points, Fy s Represents the adjusted sound detection sensitivity, Fy b represents the baseline sound monitoring sensitivity;

[0101] During one monitoring cycle, five sampling points were collected, and the recorded sound fluctuation amplitude (unit: dB) is as follows:

[0102] Vy=[4.2,5.1,4.8,6.0,5.6];

[0103] Calculate the fluctuation between each adjacent point:

[0104] |Vy2-Vy1|=|5.1-4.2|=0.9;

[0105] |Vy3-Vy2|=|4.8-5.1|=0.3;

[0106] |Vy4-Vy3|=|6.0-4.8|=1.2;

[0107] |Vy5-Vy4|=|5.6-6.0|=0.4;

[0108] Calculate the sum of fluctuations:

[0109]

[0110] Assume the number of sampling points n y =5,

[0111] Adjusted sound detection sensitivity Fy s =1.3, reference sound monitoring sensitivity Fy b =1.0, then calculate the monitoring adjustment index SY m :

[0112]

[0113] SY m =0.7×1.3=0.91;

[0114] The results indicate that the sound fluctuations in the current sleep stage are more obvious. After adjusting the sensitivity, it is necessary to further evaluate whether the sleep monitoring adjustment configuration is suitable for the current stage to ensure the stability of the monitoring system.

[0115] See also Figure 5 The specific steps for obtaining abnormal sound recognition records are as follows:

[0116] S411: Adjusting the configuration based on sleep monitoring, collecting the infant's crying characteristics, breathing sound patterns, and throat sound change rates, performing feature separation on the collected sound data, extracting the temporal distribution, intensity fluctuations, and frequency changes of the sounds, and calculating the short-term energy values and peak time intervals of multiple types of sounds to obtain a sound feature parameter set;

[0117] For the collected sound data, feature separation is first performed to distinguish sound signals from different sources, including infant crying, normal breathing sounds, throat sounds and other environmental noises. The main basis for feature separation is the frequency component, duration and intensity distribution of the sound signal. For example, crying sounds are usually in the range of 400Hz to 600Hz, and breathing sounds are concentrated between 100Hz and 300Hz. For each type of sound signal, its time distribution characteristics are extracted, including the start time, duration and end time of each sound signal, to form complete time axis information. At the same time, the intensity fluctuation of the sound signal is measured. Calculate and record the short-time energy changes. The short-time energy value is calculated using a sliding window integration method. The instantaneous energy of the sound signal is calculated in each fixed time window. For example, the window length is set to 50ms, and the square sum of the sound signal is calculated in each window and the average is taken to obtain the short-time energy distribution curve. In addition, the peak time interval of the sound is detected, and the peak occurrence frequency of various sound signals is counted. For example, when a crying signal is detected, the time interval between two consecutive crying peaks is calculated, and its changing trend is calculated. If the crying peak time interval within a certain period of time is shortened, it may indicate that the baby is in an uncomfortable state.

[0118] S412: Calling the sound feature parameter set, comparing the sound fluctuation range of the infant in the normal sleeping state, screening abnormal audio segments, detecting their duration and sudden change rate, and analyzing the distribution of abnormal audio segments on the time axis to obtain abnormal sound feature vectors;

[0119] Compare the sound fluctuation range of the baby in the normal sleeping state and screen out abnormal audio segments. First, set the benchmark range of normal sleep sound fluctuations, including the sound intensity, frequency range and duration distribution of the baby in a quiet sleep state. For example, the normal breathing sound intensity is about 30dB to 40dB, and the frequency is mainly distributed between 150Hz and 250Hz. If the actual collected sound features exceed this range, it can be judged as an abnormal audio segment. The screened abnormal audio segment needs to further calculate its duration to detect whether the sound has long-term stable abnormal characteristics. For example, the minimum duration threshold of the abnormal sound is set to 500ms. If a certain segment is abnormal If the duration of a sound is less than 500ms, it is judged to be a short-term fluctuation rather than an abnormal event. At the same time, the sudden change rate is calculated using the short-term increase ratio of the abnormal signal, that is, comparing the change amplitude of the current sound intensity compared with the previous time window. For example, if the current sound intensity increases by more than 50% compared with the previous time window, the sound is considered to have sudden characteristics. In addition, the distribution of abnormal audio segments on the time axis is analyzed, and the frequency of abnormal events in different time periods is counted. For example, within a 10-minute monitoring time, if a certain type of abnormal sound occurs more than 5 times, its pattern can be further analyzed to obtain the abnormal sound feature vector.

[0120] S413: Based on the abnormal sound feature vector, calculate the features of each abnormal sound category using the formula:

[0121]

[0122] Get abnormal sound recognition record Ao r , among which, So i Represents the intensity of the ith abnormal sound segment, So norm Represents the average intensity of normal sleep sounds, To i Represents the duration of the i-th abnormal sound segment, To max Represents the longest duration of all abnormal sound segments, Fo i Represents the frequency of the i-th abnormal sound segment, Fo norm represents the average frequency of normal sleep sounds, n A represents the total number of abnormal sound segments;

[0123] During the monitoring period, 5 abnormal sound segments were detected, with intensities of 60, 55, 70, 50, and 65 dB. The average intensity of normal sleep sounds was So norm The duration of each abnormal sound segment is 2.1, 1.8, 2.5, 1.6, and 2.2 seconds respectively. The longest duration is To maxThe frequency of abnormal sound segments is 500, 480, 520, 460, 500 Hz, and the average frequency of normal sleep sound is Fo norm For 400Hz, the calculation process is:

[0124]

[0125] calculate

[0126]

[0127] calculate

[0128]

[0129] calculate

[0130]

[0131] Ao r =(0.333+0.84+1.25)+(0.222+0.72+1.2)+(0.556+1+1.3)+(0.111+0.64+1.15)+(0.444+0.88+1.25);

[0132] Ao r =2.423+2.142+2.856+1.901+2.574;

[0133] Ao r =11.896;

[0134] This result shows that the infant's abnormal sound activity has changed significantly during the monitoring period, and this value can be used to further analyze the pattern of abnormal sounds.

[0135] See also Figure 6 ,The specific steps for obtaining the sleep stability assessment results are:

[0136] S511: Based on the abnormal sound recognition records, calculate the frequency and duration of abnormal sounds during the monitoring period, count the total number of abnormal sound events during the time period, and accumulate the duration of abnormal sound events to obtain abnormal sound statistical parameters;

[0137] Calculate the frequency and duration of abnormal sounds during the monitoring period. Specifically, obtain the complete abnormal sound data record, split it into different time periods, and count the number of abnormal sound occurrences and duration in each time period. For example, in a certain monitoring, the monitoring equipment recorded 15 abnormal sound events in a 60-minute period. The duration of a single abnormal sound event was 2 seconds, 3 seconds, and 1 second, respectively, with a total duration of 30 seconds. The data was further broken down into the abnormal sound frequency in each time period, that is, the number of abnormal sound occurrences per minute, which can be calculated using the formula: Calculate the number of abnormal sounds per minute. In addition, the cumulative duration of abnormal sound needs to be calculated, that is, the sum of the duration of all abnormal sound events during the monitoring period, using the formula: Among them, DP i Represents the duration of a single abnormal sound event, seconds, thereby obtaining abnormal sound statistical parameters.

[0138] S512: Based on the abnormal sound statistical parameters and the sleep pattern matching score, the abnormal sound frequency and duration in each time period are compared using the formula:

[0139]

[0140] Calculate the matching degree deviation value Sz d , and analyze the deviation between abnormal sound distribution and sleep pattern, evaluate whether the baby's current sleep state is abnormal, and obtain the sleep stability assessment result, among which Fz i Represents the number of abnormal sounds in the i-th time period, Tz i represents the duration of the i-th time period, Pz i represents the sleep pattern matching score corresponding to the i-th time period, n z Represents the total number of time periods;

[0141] Combined with the sleep pattern matching score, it is necessary to compare the frequency and duration of abnormal sounds in each time period to determine the degree of matching with the sleep pattern. z =3, the abnormal sound frequency and matching score of each time period are:

[0142] Fz1=8,Tz1=20,Pz1=0.6;

[0143] Fz2=15, Tz2=30, Pz2=0.7;

[0144] Fz3=12, Tz3=25, Pz3=0.65;

[0145] calculate:

[0146]

[0147] Sz d =|0.4-0.6|+|0.5-0.7|+|0.48-0.65|;

[0148] Sz d =0.2+0.2+0.17=0.57;

[0149] The results show that the sleep stability deviation coefficient during the current monitoring period is 0.57. According to the sleep stability evaluation standard, if this value exceeds the set stability range, it can be determined that the baby's current sleep state is abnormal, thereby further judging the sleep quality problem.

[0150] A baby sleep status monitoring system based on sound analysis, the system comprising:

[0151] The sound collection and classification module is based on the smart crib. It uses the microphone sensor inside the crib to collect the baby's sound signals in real time, record the sound frequency and volume change rate, and analyze the timing information to identify deep sleep, light sleep, and rapid eye movement sounds, and obtain the sleep cycle sound categories.

[0152] The sound pattern matching module analyzes the frequency distribution, energy characteristics, and time domain characteristics of the sound data based on the sleep cycle sound category, compares the sound data with the standard sleep pattern, determines the category to which the current sound data belongs, and obtains the sleep pattern matching score;

[0153] The monitoring and adjustment module monitors the baby's current sleep status in real time based on the sleep pattern matching score, determines whether the sound fluctuation amplitude exceeds the stable range, controls the sampling frequency of the sound data, matches the differentiated sleep stages, and obtains the sleep monitoring adjustment configuration;

[0154] The abnormal sound detection module adjusts its configuration based on sleep monitoring, calculates the temporal distribution and intensity fluctuations of the sound, compares it with the baby's normal sleep sounds, detects persistent or sudden abnormal sounds, and generates abnormal sound identification records;

[0155] The sleep stability assessment module is based on abnormal sound recognition records, statistics of abnormal sound frequency and duration, combined with sleep pattern matching scores, to analyze the stability level of the current sleep state, evaluate whether the baby's current sleep state is abnormal, and obtain sleep stability assessment results.

[0156] The above are merely preferred embodiments of the present invention and do not limit the present invention in any other form. Any technician familiar with the profession may use the technical content disclosed above to change or modify it into an equivalent embodiment with equivalent changes and apply it to other fields. However, any simple modification, equivalent change and modification made to the above embodiment based on the technical essence of the present invention without departing from the content of the technical solution of the present invention shall still fall within the scope of protection of the technical solution of the present invention.

Claims

1. A method for monitoring infant sleep status based on sound analysis, characterized in that: The following steps are involved: S1: Based on the smart crib, the microphone sensor inside the crib collects the sound signals emitted by the baby in real time, identifies the sounds during deep sleep, light sleep, and rapid eye movement, and obtains the sound categories of the sleep cycle; S2: Based on the sleep cycle sound category, compare the sound data with the standard sleep pattern, determine the category to which the current sound data belongs, and obtain a sleep pattern matching score; S3: Based on the sleep pattern matching score, monitor the current infant's sleep status in real time, analyze the sound fluctuation amplitude and mutation frequency of the current sleep stage, control the sampling frequency change of the sound data, match the differentiated sleep stages, and obtain the sleep monitoring adjustment configuration; S4: adjusting the configuration according to the sleep monitoring, collecting the infant's crying characteristics, breathing sound pattern, and throat sound change rate, calculating the time distribution and intensity fluctuation of the sound, comparing it with the infant's normal sleep sound, detecting continuous or sudden abnormal sounds, and obtaining an abnormal sound identification record; S5: Based on the abnormal sound recognition record, the frequency and duration of the abnormal sound are counted to evaluate whether the infant's current sleep state is abnormal, and obtain a sleep stability evaluation result.

2. The method for monitoring infant sleep status based on sound analysis according to claim 1, characterized in that: The sleep cycle sound categories include deep sleep sound types, light sleep sound types, and rapid eye movement sound types; the sleep pattern matching score includes a frequency matching score, an energy matching score, and a time matching score; the sleep monitoring adjustment configuration includes a fluctuation amplitude adjustment parameter, a sampling frequency adjustment parameter, and a sensitivity adjustment parameter; the abnormal sound identification record includes a crying feature record, a breathing pattern record, and a laryngeal sound change record; and the sleep stability assessment result includes a stability level, an abnormality occurrence frequency, and an abnormality duration.

3. The method for monitoring infant sleep status based on sound analysis according to claim 1, characterized in that: The steps for obtaining the sleep cycle sound category are specifically as follows: S111: Based on the smart crib, the microphone sensor inside the crib collects the baby's sound signals in real time, records the sound frequency data and volume change rate, analyzes the time series information, and then extracts the signal amplitude, frequency range and change trend characteristics to generate sound time series feature values; S112: Based on the sound time series feature value, the sound signal of each time period is classified, the sound signal is divided into fixed time windows, the sound amplitude within the window is collected, and the formula is used: Calculate the average sound energy EF of the current window i , and then distinguish deep sleep, light sleep and rapid eye movement period to generate sleep cycle sound categories, where T is the number of sampling points, AF t is the instantaneous voltage amplitude of the sound signal collected at time t.

4. The method for monitoring infant sleep status based on sound analysis according to claim 1, characterized in that: The steps for obtaining the sleep pattern matching score are specifically as follows: S211: Based on the sleep cycle sound category, perform frequency domain and time domain analysis on the sound data, calculate the frequency distribution, energy characteristics and time domain characteristics in each time period, and establish a sound feature parameter set; S212: Based on the sound feature parameter set, using the sleep reference sound in the crib audio database, the formula: Calculate the sound matching degree MR between the current sound data and the sleep state sample s , where Pq i is the energy value of the current sound data in the i-th frequency segment, Rq i is the energy value of the sleep reference sound in the database at the i-th frequency segment, N q is the total number of frequency segments; S213: Compare the sound matching degree with the sound data according to the reference benchmark of the sleep pattern, calculate the ratio in each matching degree interval, determine the category of the current sound data, and obtain the sleep pattern matching score.

5. The method for monitoring infant sleep status based on sound analysis according to claim 1, characterized in that: The steps for obtaining the sleep monitoring adjustment configuration are specifically as follows: S311: Based on the sleep pattern matching score, monitor the infant's current sleep status in real time, analyze the sound fluctuation amplitude and mutation frequency in the current sleep stage, calculate its time series change trend, and determine whether the sound fluctuation amplitude exceeds the stable range, and obtain the sound fluctuation deviation; S312: Based on the sound fluctuation deviation, in case of exceeding the stable range, adjusting the sound monitoring sensitivity of the crib, controlling the sampling frequency of the sound data, and matching the different sleep stages in combination with the sound feature intervals of the different sleep stages; S313: Based on the differentiated sleep stages, the sound fluctuation amplitudes of the differentiated sleep stages are classified according to the adjusted sound monitoring sensitivity and sampling frequency, using the formula: Calculate the monitoring adjustment index SY m , get the sleep monitoring adjustment configuration, where Vy i Represents the sound fluctuation amplitude at the i-th moment, Vy i-1 Represents the sound fluctuation amplitude at the i-1th moment, n y Represents the number of sampling points, Fy s Represents the adjusted sound detection sensitivity, Fy b Represents baseline sound detection sensitivity.

6. The method for monitoring infant sleep status based on sound analysis according to claim 1, characterized in that: The steps for obtaining the abnormal sound recognition record are specifically as follows: S411: According to the sleep monitoring configuration, the infant's crying characteristics, breathing sound pattern, and throat sound change rate are collected, and the collected sound data is subjected to feature separation to extract the time distribution, intensity fluctuation, and frequency change of the sound. The short-term energy values and peak time intervals of multiple types of sounds are calculated to obtain a sound feature parameter set. S412: Calling the sound feature parameter set, comparing it with the sound fluctuation range of the infant in the normal sleeping state, screening abnormal audio segments, detecting their duration and sudden change rate, and analyzing the distribution of abnormal audio segments on the time axis to obtain abnormal sound feature vectors; S413: Based on the abnormal sound feature vector, calculate the features of each abnormal sound category using the formula: Get abnormal sound recognition record Ao r , among which, So i Represents the intensity of the i-th abnormal sound segment, So norm Represents the average intensity of normal sleep sounds, To i Represents the duration of the i-th abnormal sound segment, To max Represents the longest duration of all abnormal sound segments, Fo i Represents the frequency of the i-th abnormal sound segment, Fo norm represents the average frequency of normal sleep sounds, n A Represents the total number of abnormal sound segments.

7. The method for monitoring infant sleep status based on sound analysis according to claim 1, characterized in that: The steps for obtaining the sleep stability evaluation result are specifically as follows: S511: Based on the abnormal sound recognition record, calculate the frequency and duration of abnormal sounds during the monitoring period, count the total number of abnormal sound events during the time period, and accumulate the duration of abnormal sound events to obtain abnormal sound statistical parameters; S512: Based on the abnormal sound statistical parameters and the sleep pattern matching score, the abnormal sound frequency and duration in each time period are compared, using the formula: Calculate the matching degree deviation value Sz d , and analyze the deviation between abnormal sound distribution and sleep pattern, evaluate whether the baby's current sleep state is abnormal, and obtain the sleep stability assessment result, among which Fz i Represents the number of abnormal sounds in the i-th time period, Tz i represents the duration of the i-th time period, Pz i represents the sleep pattern matching score corresponding to the i-th time period, n z Represents the total number of time periods.

8. A baby sleep status monitoring system based on sound analysis, characterized in that: The method for monitoring infant sleep status based on sound analysis according to any one of claims 1 to 7 is implemented, wherein the system comprises: The sound collection and classification module is based on the smart crib. It uses the microphone sensor inside the crib to collect the baby's sound signals in real time, record the sound frequency and volume change rate, and analyze the timing information to identify deep sleep, light sleep, and rapid eye movement sounds, and obtain the sleep cycle sound categories. The sound pattern matching module analyzes the frequency distribution, energy characteristics, and time domain characteristics of the sound data based on the sleep cycle sound category, compares the sound data with the standard sleep pattern, determines the category to which the current sound data belongs, and obtains a sleep pattern matching score; The monitoring and adjustment module monitors the current infant's sleep status in real time based on the sleep pattern matching score, determines whether the sound fluctuation amplitude exceeds the stable range, controls the sampling frequency change of the sound data, matches the differentiated sleep stages, and obtains the sleep monitoring adjustment configuration; The abnormal sound detection module adjusts the configuration according to the sleep monitoring, calculates the time distribution and intensity fluctuation of the sound, compares it with the normal sleep sound of the infant, detects continuous or sudden abnormal sounds, and obtains abnormal sound identification records; The sleep stability assessment module counts the frequency and duration of abnormal sounds based on the abnormal sound recognition records, combines the sleep pattern matching score, analyzes the stability level of the current sleep state, assesses whether the infant's current sleep state is abnormal, and obtains a sleep stability assessment result.

Images