A dynamic filtering based cardiopulmonary sound preprocessing system

Abstract

The present application relates to the technical field of signal processing, in particular to a kind of cardiopulmonary sound preprocessing system based on dynamic filtering, system includes signal segmentation optimization module, noise suppression and spectrum reconstruction module, signal quality evaluation module, dynamic filtering adjustment module, abnormal signal tracing module.In the present application, the correlation of the time distribution, spectral change and amplitude fluctuation of the signal is extracted and analyzed, the local feature difference of the cardiopulmonary sound signal can be accurately described, the sensitivity and pertinence of signal processing are improved, the hierarchical identification and targeted suppression of different types of noise are realized by comparing the energy distribution of frequency band and the amplitude difference of noise, the clarity and continuity of the target signal are further enhanced, the time evolution law of abnormal signal segment is used to identify the low quality period and make dynamic response, effectively avoid the interference residue and misjudgment phenomenon in cardiopulmonary sound signal processing, enhance the adaptability and stability of the system in complex background.

Description

Technical Field

[0001] This invention relates to the field of signal processing technology, and in particular to a cardiopulmonary sound preprocessing system based on dynamic filtering. Background Technology

[0002] Signal processing technology is an important branch of information and electronic engineering, primarily researching how to analyze, transform, extract, and optimize continuous or discrete signals to achieve effective signal acquisition and processing in scenarios such as communication, biomedicine, acoustic detection, and image recognition. Core aspects of this technology include signal acquisition, noise suppression, filtering transformation, feature extraction, and signal reconstruction. By calculating and optimizing the time and frequency domain characteristics of signals, effective identification and differentiation of target signals can be achieved. Common methods in this field involve various mathematical and computational techniques such as dynamic filtering, adaptive filtering, and spectral analysis to improve signal quality and usability. Traditional cardiopulmonary sound preprocessing systems refer to systems that filter and reduce noise in the raw cardiopulmonary sound signals acquired from sensors to remove environmental noise and interference from human physiological signals. Traditional systems use fixed-parameter filtering methods, such as signal smoothing based on bandpass filtering or threshold-limited noise reduction methods. These methods rely on preset filtering coefficients to enhance signals and suppress noise in specific frequency bands. These approaches often depend on static filters operating within a fixed frequency range, setting filtering boundaries based on the frequency distribution characteristics of cardiopulmonary sounds to perform preliminary signal cleaning and preparation.

[0003] Existing technologies primarily rely on static filters for signal enhancement and noise suppression within fixed frequency bands. These methods are ill-suited to complex changes in the signal environment and perform poorly under dynamic noise interference. Traditional systems, using static parameter settings, are easily limited by the frequency range and neglect the time-varying characteristics and spectral distribution changes of the signal. This results in poor filtering performance when processing variable signals. Especially when facing interference from various noise sources, they cannot adjust appropriately according to real-time changes, leading to signal quality degradation and an inability to fully eliminate all unnecessary noise, thus affecting the accuracy of subsequent analysis and diagnosis. Summary of the Invention

[0004] The purpose of this invention is to address the shortcomings of existing technologies by proposing a cardiopulmonary sound preprocessing system based on dynamic filtering.

[0005] To achieve the above objectives, the present invention adopts the following technical solution: a cardiopulmonary sound preprocessing system based on dynamic filtering includes:

[0006] The signal segmentation optimization module acquires the temporal distribution characteristics, spectral intensity changes, and amplitude fluctuation range of the cardiopulmonary sound signal, analyzes the spectral overlap relationship between signal segments, determines the signal segmentation boundary and spectral coverage length based on the spectral distribution density and amplitude response order, and performs spectral segmentation and amplitude mapping to generate a cardiopulmonary sound segmentation feature spectrum.

[0007] Based on the segmented feature spectrum of the cardiopulmonary sounds, the noise suppression and spectrum reconstruction module selects the noise energy proportion and amplitude fluctuation value of key frequency bands, matches and compares the degree of overlap between noise frequency and spectrum, identifies the key level of noise suppression, and obtains a set of frequency band energy weight evaluation indicators.

[0008] The signal quality assessment module extracts the time distribution and duration of abnormal signal segments based on the frequency band energy weight assessment index set, determines the distribution stage of signal quality troughs, classifies the time trends of similar abnormal signal segments, and generates a dynamic evolution sequence of cardiopulmonary sound quality.

[0009] The dynamic filtering adjustment module calls the dynamic evolution sequence of cardiopulmonary sound quality, combines the amplitude fluctuation value and response time delay within the jurisdiction, filters the target signal segment within the trend segment, and revises the filtering parameter value segment by segment to obtain a segmented dynamic correction table of filtering parameters.

[0010] As a further aspect of the present invention, the cardiopulmonary sound segmentation feature spectrum includes signal segmentation boundary classification, spectrum coverage path structure, and amplitude fluctuation distribution; the frequency band energy weight evaluation index set includes frequency band level label, frequency band occupancy coefficient, and priority scoring interval; the cardiopulmonary sound quality dynamic evolution sequence includes abnormal time distribution, abnormal duration, and trend classification results; and the segmented filtering parameter dynamic correction table includes a target signal list, parameter revision range, and delay correlation factor.

[0011] As a further aspect of the present invention, the signal segmentation optimization module includes:

[0012] The signal distribution and spectrum recognition submodule acquires the time distribution characteristics, spectrum intensity changes, and amplitude fluctuation range data of the cardiopulmonary sound signal. It aligns the three signals by timestamp, extracts signal abrupt change points, locates spectrum intensity and amplitude changes, identifies response differences, and generates signal operation response deviation.

[0013] The spectrum hysteresis analysis submodule extracts the signal delay segment and spectrum response sequence based on the signal operation response deviation, counts the delay length and response sequence, determines the spectrum dependence direction, measures the number of signal trigger sequences within the hysteresis segment, and generates the signal spectrum hysteresis structure quantity.

[0014] The spectrum linkage structure generation submodule extracts the amplitude fluctuation change point and the signal spectrum linkage period based on the signal spectrum hysteresis structure quantity, compares the spectrum response frequency and time delay, identifies the linkage number and average delay ratio, analyzes the dependency relationship and spectrum mapping, and generates a segmented feature spectrum of cardiopulmonary sounds.

[0015] As a further aspect of the present invention, the noise suppression and spectrum reconstruction module includes:

[0016] The amplitude fluctuation extraction submodule extracts the amplitude fluctuation values ​​of key frequency bands based on the segmented feature spectrum of the heart and lung sounds, filters sequences whose fluctuations exceed the amplitude reference value, calculates the fluctuation anomaly ratio parameter, analyzes the proportion of abnormal fluctuations, and generates an abnormal amplitude fluctuation frequency ratio.

[0017] The frequency dependency matching submodule calls the abnormal amplitude fluctuation frequency ratio, combines the frequency band task frequency and the number of dependency paths, identifies the frequency dependency comparison item, adjusts the frequency band abnormal offset according to the proportion of tasks on the dependency path, and obtains the fluctuation coupling priority index.

[0018] The priority level determination submodule extracts the numerical range corresponding to the frequency band based on the fluctuation coupling priority index, sets the frequency band level division interval group, assigns the level identifier according to the index value falling into the interval, sorts and determines the noise suppression level, and obtains the frequency band energy weight evaluation index set.

[0019] As a further aspect of the present invention, the signal quality assessment module includes:

[0020] The abnormal record screening submodule collects the trigger and end times of abnormal signals based on the frequency band energy weight evaluation index set, identifies the time intervals between adjacent abnormalities, filters records that exceed the abnormal interval benchmark value, and generates an abnormal time interval sequence.

[0021] The abnormal period segmentation submodule calls the abnormal time interval sequence to count the number of consecutive occurrences of the same type of abnormal event. Based on the abnormal duration, interval time, and signal load index, the following formula is used:

[0022]

[0023] Calculate the periodic distribution trend value of the anomaly and establish the periodic distribution stages of the anomaly;

[0024] Where P represents the abnormal periodic distribution trend value, and D t T represents the duration of the anomaly at time t, μ represents the average duration of the anomaly, and T represents the duration of the anomaly at time t. t The time-related index represents time t, and N represents the total number of samples.

[0025] The periodic trend classification submodule calls the abnormal periodic distribution stage, identifies abnormal events of the same periodic trend, identifies differences in trend fluctuations, classifies and archives them according to the difference threshold, and generates a dynamic evolution sequence of cardiopulmonary sound quality.

[0026] As a further aspect of the present invention, the dynamic filtering adjustment module includes:

[0027] The trend sequence recognition submodule calls the dynamic evolution sequence of cardiopulmonary sound quality, extracts incremental change values, determines the trend upward range based on the cumulative increase and fluctuation amplitude, marks key signal segments, and generates an automated signal trend anomaly recognition list;

[0028] The response characteristic extraction submodule calls the signal number in the automated signal trend anomaly identification list, identifies the amplitude fluctuation value and response delay within the segment, and analyzes the signal response differences by combining the delay distribution and fluctuation frequency to obtain the automated signal response characteristic difference value.

[0029] The parameter revision submodule identifies the parameter revision item structure based on the difference value of the automated signal response characteristics and the corresponding period of the trend segment, analyzes the parameter offset degree and frequency distribution amplitude, identifies the parameter revision offset degree, and adjusts the filtering frequency and time distribution in combination with the original parameter structure differences to obtain a segmented filtering parameter dynamic correction table.

[0030] As a further aspect of the present invention, the step of judging the trend upward interval based on the cumulative increase and fluctuation amplitude means that when the incremental change value of the dynamic evolution sequence of cardiopulmonary sound quality is higher than the first preset increase threshold and the fluctuation amplitude of the incremental change value is less than the second preset fluctuation threshold within a continuous preset time period, it is judged as a trend upward interval.

[0031] The term "marking key signal segment" refers to marking a segment as a key signal segment when, within an upward trend range, the number of local peaks in the incremental change value exceeds a preset peak number threshold, or the rate of change of the incremental change value exceeds a preset rate threshold.

[0032] As a further aspect of the present invention, the system also includes an abnormal signal tracing module:

[0033] The abnormal signal tracing module collects the response difference and fluctuation characteristics of the high-frequency filtered signal based on the segmented filtering parameter dynamic correction table, determines whether the fluctuation characteristics have the same origin, identifies the abnormal attribution node, and forms a cardiopulmonary sound abnormal feature tracing map.

[0034] The source atlas of abnormal cardiopulmonary sounds includes fluctuation homology feature groups, abnormal node localization results, and source path structure diagram.

[0035] As a further aspect of the present invention, the abnormal signal tracing module includes:

[0036] The response difference extraction submodule, based on the segmented filtering parameter dynamic correction table, checks the response time and change sequence of high-frequency signals, identifies the synchronization deviation between response time difference and rate of change, filters time points and signal numbers where the deviation exceeds the stable range, and generates an automated signal response offset list.

[0037] The wave origin discrimination submodule extracts the wave sequence based on the automated signal response offset list, compares the wave amplitude and direction at the offset time point, identifies continuous and consistent wave segments and records the intervals that overlap with the offset time point, and generates a response wave origin segment table.

[0038] The abnormal attribution identification submodule extracts the node path of the signal number according to the response fluctuation homogeneous segment table, tracks the node response sequence and signal transmission in the overlapping segment, identifies the abnormal response frequency of the signal source node, and forms a source map of abnormal cardiopulmonary sounds.

[0039] Compared with the prior art, the advantages and positive effects of the present invention are as follows:

[0040] This invention effectively improves signal processing accuracy through in-depth analysis of signal temporal distribution characteristics, spectral changes, and amplitude fluctuations. It enables real-time adjustment of filtering parameters in dynamic environments to optimize signal quality. By analyzing the spectral overlap between signal segments, it can more accurately identify and classify different types of noise, ensuring effective noise suppression. Based on a frequency band energy weight evaluation index set, it precisely extracts the dynamic evolution trend of abnormal signal segments, thereby optimizing the dynamic adjustment of cardiopulmonary sound quality and reducing the risk of misjudgment and missed judgment. Dynamic adjustment of filtering parameters further enhances the flexibility and real-time performance of signal processing, improves signal clarity and reliability, effectively addresses changes in different environments, and enhances the system's adaptability to complex cardiopulmonary sounds. Attached Figure Description

[0041] Figure 1 This is a system flowchart of the present invention;

[0042] Figure 2 This is a flowchart of the signal segmentation optimization module in this invention;

[0043] Figure 3 This is a flowchart of the noise suppression and spectrum reconstruction module in this invention;

[0044] Figure 4 This is a flowchart of the signal quality assessment module in this invention;

[0045] Figure 5 This is a flowchart of the dynamic filtering adjustment module in this invention;

[0046] Figure 6 This is a flowchart of the abnormal signal tracing module in this invention. Detailed Implementation

[0047] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0048] In the description of this invention, it should be understood that the terms "length," "width," "upper," "lower," "front," "rear," "left," "right," "vertical," "horizontal," "top," "bottom," "inner," and "outer," etc., indicating orientation or positional relationships, are based on the orientation or positional relationships shown in the accompanying drawings and are only for the convenience of describing the invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation, and therefore should not be construed as a limitation of the invention. Furthermore, in the description of this invention, "a plurality of" means two or more, unless otherwise explicitly specified.

[0049] Please see Figure 1 A cardiopulmonary sound preprocessing system based on dynamic filtering includes:

[0050] The signal segmentation optimization module acquires the temporal distribution characteristics, spectral intensity changes, and amplitude fluctuation range of the cardiopulmonary sound signal, analyzes the spectral overlap relationship between signal segments, determines the signal segmentation boundary and spectral coverage length based on the spectral distribution density and amplitude response order, and performs spectral segmentation and amplitude mapping to generate a cardiopulmonary sound segmentation feature spectrum.

[0051] The noise suppression and spectrum reconstruction module is based on the segmented feature spectrum of cardiopulmonary sounds. It selects the noise energy proportion and amplitude fluctuation value of key frequency bands, matches and compares the degree of overlap between noise frequency and spectrum, identifies the key level of noise suppression, and obtains a set of frequency band energy weight evaluation indicators.

[0052] The signal quality assessment module extracts the temporal distribution and duration of abnormal signal segments based on the frequency band energy weight assessment index set, determines the distribution stage of signal quality troughs, classifies the temporal trends of similar abnormal signal segments, and generates a dynamic evolution sequence of cardiopulmonary sound quality.

[0053] The dynamic filtering adjustment module calls the dynamic evolution sequence of cardiopulmonary sound quality, combines the amplitude fluctuation value and response time delay within the jurisdiction, filters the target signal segment within the trend segment, and revises the filtering parameter value segment by segment to obtain the segmented dynamic correction table of filtering parameters.

[0054] The abnormal signal tracing module is based on a segmented filter parameter dynamic correction table. It collects the response difference and fluctuation characteristics of high-frequency filtered signals, determines whether the fluctuation characteristics have the same origin, identifies abnormal attribution nodes, and forms a tracing map of abnormal cardiopulmonary sounds.

[0055] The cardiopulmonary sound segmentation feature spectrum includes signal segment boundary classification, spectrum coverage path structure, and amplitude fluctuation distribution. The frequency band energy weight evaluation index set includes frequency band level labels, frequency band occupancy coefficient, and priority scoring interval. The cardiopulmonary sound quality dynamic evolution sequence includes abnormal time distribution, abnormal duration, and trend classification results. The segmented filtering parameter dynamic correction table includes a target signal list, parameter revision range, and delay correlation factor. The cardiopulmonary sound abnormality feature tracing map includes fluctuation homology feature groups, abnormal node location results, and tracing path structure diagram.

[0056] Please see Figure 2 The signal segmentation optimization module includes:

[0057] The signal distribution and spectrum recognition submodule acquires the time distribution characteristics, spectrum intensity changes, and amplitude fluctuation range data of the cardiopulmonary sound signal. It aligns the three signals by timestamp, extracts signal abrupt change points, locates spectrum intensity and amplitude changes, identifies response differences, and generates signal operation response deviation.

[0058] Acquire and process heart and lung sound signals to obtain temporal distribution characteristics. For example, acquire signals for 20 seconds, sample at 4000 Hz, and divide the signal into frames with a 100-millisecond window. Calculate the energy value, zero-crossing rate, and short-time average amplitude for each frame to form a temporal distribution feature sequence. Perform a Fast Fourier Transform on each frame to calculate the spectral intensity of a specific frequency band (heart sound band 20-150 Hz, lung sound band 150-1000 Hz). Compare the spectral intensity between adjacent frames to obtain the spectral intensity change. For example, a change of 0.8 units in one frame to 1.2 units in the next, a change of 0.4 units. Within each frame, measure the difference between the maximum and minimum amplitude of the signal to obtain amplitude fluctuation range data. For example, the amplitude range is -0.5 to 0.5, and the fluctuation range is 1.0. Utilize the heart and lung sound acquisition... Timestamps are used to align the time distribution feature sequence, spectral intensity change sequence, and amplitude fluctuation range sequence with time. Signal abrupt change thresholds are set, for example, when the energy change between adjacent frames exceeds 0.3 units or the short-term average amplitude change exceeds 0.2 units, signal abrupt change points are identified; for example, when the energy changes from 0.1 to 0.4, abrupt change points are marked. The spectral intensity and amplitude changes within the time window where the abrupt change point is located are located; for example, when the spectral intensity changes from 1.0 to 2.5 and the amplitude fluctuation changes from 0.8 to 1.5. The observed spectral intensity and amplitude changes are compared with the steady-state baseline value or the average value of the previous time window to identify response differences; for example, if the baseline spectral intensity is 1.5 and the observed value is 2.5, the response difference is 1.0. All response differences identified at the abrupt change point are integrated.

[0059] Table 1: Examples of Signal Operation Response Deviation

[0060]

[0061] As shown in Table 1, this table records the difference in spectral intensity and amplitude fluctuation response for different signal numbers at different timestamps, which constitutes the signal operation response deviation.

[0062] The spectrum hysteresis analysis submodule extracts the signal delay segment and spectrum response sequence based on the signal operation response deviation, counts the delay length and response sequence, determines the spectrum dependence direction, measures the number of signal trigger sequences within the hysteresis segment, and generates the signal spectrum hysteresis structure quantity.

[0063] Based on the signal response deviation, the frequency bands where responses occur are identified by setting a response difference threshold (e.g., 0.6 units). This allows for the extraction of signal delay segments. For example, if the HS_001 response difference is 1.0 at 1.300 seconds, exceeding the 0.6 threshold, the response start point is marked. If a frequency band responds within 200 milliseconds, it is identified as a signal delay segment. Within the identified signal delay segments, the time points at which different frequency bands (low-frequency heart sounds 20-150 Hz, mid-frequency lung sounds 150-400 Hz, and high-frequency lung sounds 400-1000 Hz) reach the response threshold are compared to determine the spectral response order. For example, if low-frequency heart sounds t=1.300 seconds, mid-frequency lung sounds t=1.350 seconds, and high-frequency lung sounds t=1.400 seconds, the response order is low-frequency heart sounds → mid-frequency lung sounds → high-frequency lung sounds. The duration of each delay segment is then recorded as... The delay length is set to 100 milliseconds, and the response sequence of each frequency band is recorded. The consistency of the response sequence of the mid-frequency bands in a large number of delay segments is analyzed. For example, 80% of the delay segments show a "low frequency → mid frequency → high frequency" pattern, and the direction of spectral dependence is determined to be from low frequency to high frequency. For each hysteresis segment, the number of times the signal triggering sequence (e.g., the cardiac frequency band triggers before the pulmonary frequency band) occurs is counted. For example, in 100 hysteresis segments, "the cardiac frequency band triggers first, and the pulmonary frequency band triggers later" occurs 75 times. The start time, delay length, spectral response sequence, determined spectral dependence direction, and number of signal triggering sequences of each delay segment are summarized to generate the signal spectral hysteresis structure quantity. For example, 50 hysteresis segments, 30 low-frequency to high-frequency dependencies, and 20 high-frequency to low-frequency dependencies are recorded and identified in 20 seconds, with triggering times of 180 and 120 times, respectively.

[0064] The spectrum linkage structure generation submodule extracts the amplitude fluctuation change point and the signal spectrum linkage period based on the signal spectrum hysteresis structure quantity, compares the frequency and time delay of the spectrum response, identifies the linkage number and average delay ratio, analyzes the dependency relationship and spectrum mapping, and generates a segmented feature spectrum of cardiopulmonary sounds.

[0065] Based on the signal spectral hysteresis structure and the original amplitude fluctuation range data, the point where the amplitude fluctuation exceeds a preset threshold (e.g., 0.5) when spectral hysteresis occurs is identified. For example, during the spectral hysteresis period from t=1.300 seconds to t=1.400 seconds, the amplitude fluctuation increases from 1.0 to 1.8, and t=1.300 seconds is marked as the amplitude fluctuation change point. According to the spectral response sequence and delay length, the signal spectral linkage period is extracted, for example, the linkage between low-frequency heart sounds and high-frequency lung sounds from t=1.300 seconds to t=1.400 seconds. The frequency of occurrence of a specific spectral response combination (e.g., low-frequency heart sounds and high-frequency lung sounds) is counted as the spectral response frequency. For example, if it occurs 15 times in a 20-second record, the average time difference between the response of each frequency band in the linkage combination reaching the threshold is calculated as the linkage delay. Low-frequency heart sounds precede high-frequency lung sounds by 100 milliseconds; the number of linkages is the frequency of spectral response divided by the total analysis period (20 seconds), and the average delay ratio is the average delay divided by the minimum physiological response delay (e.g., 20 milliseconds); a linkage matrix is ​​constructed to analyze the linkage frequency, average delay ratio, and spectral dependence direction, clarifying the inter-band dependence relationship. For example, low-frequency heart sounds always precede high-frequency lung sounds, and there is a dependence relationship from low-frequency heart sounds to high-frequency lung sounds in the linkage frequency. A spectral mapping is constructed. For example, low-frequency heart sound band activity triggers high-frequency lung sound band linkage within 100 milliseconds in 80% of cases, with a linkage effect strength of 0.7; the amplitude fluctuation change point, spectral linkage period, number of linkages, average delay ratio, dependence relationship, and spectral mapping are integrated to generate a segmented feature spectrum of heart and lung sounds.

[0066] Please see Figure 3 The noise suppression and spectrum reconstruction module includes:

[0067] The amplitude fluctuation extraction submodule extracts amplitude fluctuation values ​​in key frequency bands based on the segmented feature spectrum of heart and lung sounds, and filters sequences whose fluctuations exceed the amplitude benchmark value using the following formula:

[0068]

[0069] Calculate the fluctuation anomaly ratio parameter, analyze the proportion of abnormal fluctuations, and generate the abnormal amplitude fluctuation frequency ratio;

[0070] Where S represents the fluctuation anomaly ratio parameter, A i B represents the amplitude fluctuation value of the i-th data point. i represents the amplitude reference value of the i-th data point, and n represents the total number of data points;

[0071] Based on the segmented feature spectrum of cardiopulmonary sounds, key frequency bands crucial for cardiopulmonary sound diagnosis are identified. For example, 20-150 Hz is defined as the key frequency band for heart sounds, and 150-1000 Hz as the key frequency band for lung sounds. Within each key frequency band, the amplitude fluctuation value within each time window (e.g., 100 milliseconds) is extracted from the feature spectrum. For example, within a 100-millisecond time window of a certain heart sound band, the amplitude fluctuation value is 0.75 units. Then, an amplitude reference value is set, which can be determined by comparing large... Statistical analysis is performed on normal heart and lung sound signals, using the average amplitude fluctuation within a specific frequency band or the upper limit of the 95% confidence interval to determine the amplitude. For example, the baseline amplitude value for the heart sound band is set to 0.6 units, and the baseline amplitude value for the lung sound band is set to 0.8 units. Then, sequences whose amplitude fluctuation values ​​exceed the baseline amplitude value of their corresponding frequency band are selected. For example, if the amplitude fluctuation value of a certain time window within the heart sound band is 0.75, exceeding the baseline value of 0.6, it is included in the abnormal fluctuation sequence. Subsequently, the formula is used... The fluctuation anomaly ratio parameter is calculated. In this formula, S represents the fluctuation anomaly ratio parameter, which quantifies the overall deviation between the actual amplitude fluctuation and the benchmark value. This formula reflects the total difference by summing the absolute values ​​of the differences between the actual amplitude fluctuation values ​​and the benchmark value of all data points in the numerator. The difference is then normalized by root mean square in the denominator, making the S value less sensitive to individual extreme outliers and more robustly reflecting the overall fluctuation anomaly. Where A... i This represents the amplitude fluctuation value of the i-th data point. For example, the amplitude fluctuation value of the heart sound segment extracted from the segmented feature spectrum of heart and lung sounds at a certain time point is 0.75 units, B. i The amplitude reference value represents the i-th data point. For example, the amplitude reference value for the cardiac frequency band is 0.6 units. n represents the total number of data points. For example, in a 2-second analysis window, one data point is taken every 100 milliseconds, then n=20. For example, suppose that in a sequence containing 5 time points (n=5), the amplitude fluctuation values ​​A are [0.75, 0.80, 0.65, 0.70, 0.90], and the corresponding reference value B is 0.60.

[0072] The calculation process is as follows:

[0073] ;

[0074] ;

[0075] ;

[0076] ;

[0077] ;

[0078] molecular:

[0079]

[0080] Denominator: ;

[0081] ;

[0082] The S-value quantifies the overall deviation of the amplitude fluctuation in the time series from the benchmark. An abnormal threshold for the S-value is set. For example, when the S-value is greater than 1.5, the series is considered to have abnormal fluctuations. The number of times the S-value exceeds 1.5 within the total analysis time (e.g., 10 seconds) is counted, and its proportion of the total number of calculations is calculated as the percentage of abnormal fluctuations. For example, if 100 S-values ​​are calculated within 10 seconds, and 25 of them exceed 1.5, then the percentage of abnormal fluctuations is 25%, generating the abnormal amplitude fluctuation frequency ratio.

[0083] The frequency dependency matching submodule calls the abnormal amplitude fluctuation frequency ratio, combines the frequency band task frequency and the number of dependency paths, identifies frequency dependency comparison items, adjusts the frequency band abnormal offset according to the proportion of tasks on the dependency path, and obtains the fluctuation coupling priority index.

[0084] The algorithm retrieves the frequency ratio of abnormal amplitude fluctuations, for example, 25% for the cardiac frequency band and 18% for the lung frequency band; it combines this with the frequency of the frequency band task, for example, a task frequency of 1.25 Hz (75 bpm) for the cardiac frequency band and 0.25 Hz (15 bpm) for the lung frequency band task; and it combines this with the number of dependent paths, for example, abnormal fluctuations in the cardiac frequency band affect two high-frequency lung sound sub-bands, with a dependent path count of 2; it matches the frequency ratio of abnormal amplitude fluctuations, the frequency band task frequency, and the number of dependent paths to identify frequency dependency comparison items; and it sets a method for calculating the task proportion adjustment coefficient on dependent paths, for example, abnormal fluctuations in the cardiac frequency band cause linkage in the lung frequency band, affecting the respiratory rhythm task. The proportion is 0.7 for bronchial sound tasks and 0.3 for the original abnormal offset, which is the product of the ratio of abnormal amplitude fluctuation frequency and the frequency of frequency band tasks, for example, 25% × 1.25 = 0.3125. The proportion of critical tasks is 0.8, and the adjusted abnormal offset is 0.3125 × 0.8 = 0.25. Taking into account the adjusted frequency band abnormal offset, frequency band task frequency, number of dependent paths, and other factors, a fluctuation coupling priority index is generated. For example, the weighted sum is: 0.6 × (adjusted abnormal offset) + 0.3 × (frequency band task frequency) + 0.1 × (number of dependent paths). For the cardiac sound band: 0.6 × 0.25 + 0.3 × 1.25 + 0.1 × 2 = 0.725.

[0085] The priority level determination submodule is based on the fluctuation coupling priority index, extracts the numerical range corresponding to the frequency band, sets the frequency band level division interval group, assigns the level label according to the index value falling into the interval, sorts and determines the noise suppression level, and obtains the frequency band energy weight evaluation index set.

[0086] Based on the fluctuation coupling priority index, the numerical ranges of each frequency band are extracted. For example, heart sounds are 20-150 Hz and lung sounds are 150-1000 Hz. Level division intervals for the fluctuation coupling priority index are set. For example, [0, 0.4) represents low noise level, [0.4, 0.7) represents medium noise level, and [0.7, 1.0] represents high noise level. The threshold is determined based on extensive noise analysis of heart and lung sound signals and expert experience. For example, values ​​above 0.7 are associated with noise interference. The fluctuation coupling priority index value of each frequency band is compared with the divided intervals, and a level label is assigned. For example, the heart sound band value of 0.725 falls into the [0.7, 1.0] interval and is assigned "high noise level". Based on the assigned level labels, the frequency bands are sorted to determine the noise suppression level of each band. For example, the heart sound band is determined as "high intensity noise suppression". Based on the noise suppression level, a set of frequency band energy weight evaluation indicators is generated. For example, the frequency band weight for "high intensity noise suppression" is 0.1, for "medium intensity noise suppression" it is 0.5, and for "low intensity noise suppression" it is 0.9.

[0087] Please see Figure 4 The signal quality assessment module includes:

[0088] The abnormal record screening submodule collects the trigger and end times of abnormal signals based on the frequency band energy weight evaluation index set, identifies the time intervals between adjacent abnormalities, filters records that exceed the abnormal interval benchmark value, and generates an abnormal time interval sequence.

[0089] Based on the frequency band energy weighting evaluation index set (e.g., frequency bands with a weight less than 0.5 are abnormal frequency bands), the trigger time and end time of each abnormal signal are collected from the original or intermediate signals. For example, the start time of the abnormal heart sound band is t=1.00 seconds and the end time is t=1.30 seconds. The time interval between adjacent abnormal events is calculated. For example, the interval between the first abnormal segment (1.00 seconds to 1.30 seconds) and the second abnormal segment (2.50 seconds to 2.80 seconds) is 1.20 seconds. An abnormal interval baseline value (e.g., 0.5 seconds) is set. It is determined based on clinical experience and signal characteristics. This baseline value can be obtained from the statistical analysis of a large number of abnormal heart and lung sound events. For example, 85% of independent abnormal events have an interval greater than 0.5 seconds. Records with intervals exceeding the baseline value are filtered out. The filtered independent abnormal interval values ​​are arranged in chronological order to generate an abnormal time interval sequence, for example, [1.20 seconds, 3.50 seconds, 2.10 seconds].

[0090] The abnormal cycle segmentation submodule calls the abnormal time interval sequence to count the number of consecutive occurrences of the same type of abnormal event. Based on the duration of the abnormality, the interval time, and the signal load index, the following formula is used:

[0091]

[0092] Calculate the periodic distribution trend value of the anomaly and establish the periodic distribution stages of the anomaly;

[0093] Where P represents the abnormal periodic distribution trend value, and D t T represents the duration of the anomaly at time t, μ represents the average duration of the anomaly, and T represents the duration of the anomaly at time t. t The time-related index represents time t, and N represents the total number of samples.

[0094] The abnormal time interval sequence is invoked, for example, the sequence is [1.20 seconds, 3.50 seconds, 2.10 seconds]. Then, according to the type of abnormal signal (e.g., high-frequency hissing sound, low-frequency friction sound), the number of consecutive occurrences of a specific type of abnormal event within the observation period is counted. For example, if 5 consecutive abnormal events belonging to high-frequency hissing sound are detected, after obtaining the abnormal duration, abnormal time interval (obtained from the abnormal time interval sequence), and signal load index data, the periodic distribution trend value P of the abnormality is calculated. In this formula, P represents the periodic distribution trend value of the abnormality, which measures the dispersion of the duration of the abnormal event relative to the average value, and is weighted according to a time correlation index, thereby more comprehensively reflecting the periodicity or trend of the abnormality; where D t T represents the duration of the anomaly at time t. For example, if an anomaly lasted 0.3 seconds, μ represents the average duration of the anomaly. For example, by statistically analyzing the duration of all anomalies of the same type in the current batch, the average duration is calculated to be 0.25 seconds. t The time correlation index represents time t. This index can be quantified based on the time period of the abnormal event, its correlation with a specific physiological cycle (such as the cardiac cycle or respiratory cycle), and the time interval between events. For example, if the abnormality occurs at the peak of inspiration and is strongly correlated with the respiratory cycle, then T... t The value is relatively high, so it is set to 2.0. If it only occurs randomly between two heartbeats, then T... t The value is low, so we set it to 0.5. Here, we set T. t This is the ratio of the interval between adjacent anomalies to the average interval between anomalies, reflecting the degree of clustering or dispersion of events over time. For example, if the interval between adjacent anomalies is 1.20 seconds and the average interval between anomalies is 2.27 seconds, then... This time-related indicator uses the absolute value of the difference between the weighted duration of anomalies and the average duration to ensure that more frequent or time-concentrated anomalies have a greater impact on the calculation of the periodic distribution trend value. N represents the total number of samples, i.e., the total number of similar anomalies counted. For example, if a total of 5 similar anomalies are counted, then N=5. For example, suppose we have N=3 similar anomalies with durations D of [0.3 seconds, 0.4 seconds, 0.2 seconds], and the average duration is... The corresponding abnormal intervals were [1.20 seconds, 3.50 seconds, 2.10 seconds], and the average abnormal interval was [missing information]. seconds, then T t They are respectively ,Right now ;

[0095] The calculation process is as follows:

[0096] ;

[0097] ;

[0098] ;

[0099] Summation term: ;

[0100] ;

[0101] The result P=0.271 indicates that the periodic distribution trend value of the current batch of similar abnormal events is low, meaning that the duration of the abnormality is relatively stable and the temporal correlation is not high. This indicates that the abnormal events exhibit a random or weakly periodic distribution. For subsequent periodic trend classification, this value will serve as an important quantitative feature for comparison. Finally, based on the calculated periodic distribution trend value P of the abnormality, combined with relevant indicators such as the type and duration of the abnormal event, the abnormal periodic distribution stages are established. For example, P values ​​less than 0.3 can be classified as "weakly periodic random stage", P values ​​between 0.3 and 0.6 can be classified as "medium periodic fluctuation stage", and P values ​​greater than 0.6 can be classified as "strongly periodic stage", thus providing a structured basis for subsequent abnormality classification.

[0102] The advantage of the formula lies in the introduction of the time-related index T. t By taking its square root, the calculation of the periodic trend value P not only considers the dispersion of the duration of the anomaly, but also takes into account the clustering or dispersion effect of the anomaly event in the time dimension. This allows for a more accurate capture of the inherent periodicity and correlation of the anomaly event, which has important guiding significance for subsequent identification of different types of anomaly evolution patterns.

[0103] The periodic trend classification submodule calls the abnormal periodic distribution stage, identifies abnormal events of the same periodic trend, identifies differences in trend fluctuations, classifies and archives them according to the difference threshold, and generates a dynamic evolution sequence of cardiopulmonary sound quality.

[0104] The system invokes abnormal periodic distribution stages, such as "weak periodic random stage," "medium periodic fluctuation stage," and "strong periodicity stage," and compares the periodic distribution trend value P and abnormality type of different abnormal events to identify events with similar periodic trends. For example, two high-frequency hissing sounds with P values ​​of 0.271 and 0.290 respectively both belong to the "weak periodic random stage" and are identified as similar. Within similar abnormal events, the system identifies differences in trend fluctuations; for example, one abnormality's duration varies by 0.05 seconds, while the other varies by 0.10 seconds. A difference threshold is set, for example, a threshold between the duration variations of similar abnormalities. A difference exceeding 0.03 seconds or a difference in P-values ​​exceeding 0.05 is considered a discrepancy. The threshold is determined based on statistical analysis of a large amount of physiological and pathological cardiopulmonary sound data and expert review. For example, cluster analysis of 500 cardiopulmonary sounds found that when the difference in P-values ​​was less than 0.05, the underlying mechanisms were similar. Based on the difference threshold, similar abnormal events are classified and archived. For example, high-frequency wheezing sounds with a "weak periodic random phase" and a duration variation range of less than 0.03 seconds are archived as "high-frequency wheezing sounds - type A". The classification and archiving results are integrated and a dynamic evolution sequence of cardiopulmonary sound quality is generated according to time sequence and abnormality type.

[0105] Please see Figure 5 The dynamic filter adjustment module includes:

[0106] The trend sequence recognition submodule calls the dynamic evolution sequence of cardiopulmonary sound quality, extracts the incremental change value, judges the trend upward range based on the cumulative increase and fluctuation amplitude, marks key signal segments, and generates an automated list of abnormal signal trends.

[0107] The trend-upper range is determined based on the cumulative increase and fluctuation range. When the cumulative increase of the incremental change value of the dynamic evolution sequence of cardiopulmonary sound quality is higher than the first preset increase threshold and the fluctuation range of the incremental change value is less than the second preset fluctuation threshold within a continuous preset time period, it is determined to be a trend-upper range.

[0108] Marking a key signal segment refers to marking a segment as a key signal segment when the number of local peaks in the incremental change value exceeds a preset peak number threshold or the rate of change of the incremental change value exceeds a preset rate threshold within an upward trend range.

[0109] The system invokes a dynamic evolution sequence of cardiopulmonary sound quality, including quantitative assessment values ​​for cardiopulmonary sound quality, such as a quality score from 0 to 1. It extracts the incremental change in quality score between adjacent time points in the sequence; for example, a score of 0.70 at t=1.0 seconds and 0.72 at t=1.1 seconds, with an increment of 0.02. When the incremental change value, within a continuous time period (e.g., a 10-second analysis window), has a cumulative increase (the sum of all incremental change values) exceeding a first preset increase threshold (e.g., 0.15), and the fluctuation range of the incremental change value (the difference between the maximum and minimum values) is less than a second preset fluctuation threshold (e.g., 0.03), the system is effective. If an area is identified as having an upward trend, for example, within a 10-second window, the total increment is 0.18 (higher than 0.15) and the fluctuation range is 0.02 (lower than 0.03), it is determined to be an upward trend area. Within the upward trend area, the number of local peaks or the rate of change of the increment change value are analyzed. When the number of local peaks exceeds a preset peak number threshold (e.g., 3), or the rate of change exceeds a preset rate threshold (e.g., 0.005 units / second), it is marked as a key signal segment. The identified upward trend areas, marked key signal segments and signal numbers are compiled to generate an automated signal trend anomaly identification list.

[0110] The response characteristic extraction submodule calls the signal number in the list of automated signal trend anomaly identification, identifies the amplitude fluctuation value and response delay within the segment, and analyzes the signal response differences by combining the delay distribution and fluctuation frequency to obtain the automated signal response characteristic difference value.

[0111] The system retrieves signal numbers from the automated signal trend anomaly identification list, such as HS_001 within the T1-T2 time period. It identifies amplitude fluctuations within key signal segments, for example, HS_001 exhibiting an average amplitude fluctuation of 0.85 units in the heart rate band during the T1-T2 period. It also identifies response delays, such as a 50-millisecond delay from triggering to a stable state. Combining signal delay distribution and fluctuation frequency, the system analyzes signal response differences. Delay distribution is determined by statistically analyzing multiple delay times, calculating the average and standard deviation, for example, an average delay of 45 milliseconds and a standard deviation of 5 milliseconds. Fluctuation frequency is defined as amplitude fluctuations exceeding a specific threshold. The number of times, for example, 10 times in the T1-T2 segment, is compared with the real-time observed delay distribution and fluctuation frequency with normal physiological response models or historical benchmark data (e.g., normal average delay of 40 milliseconds and fluctuation frequency of 3 times). The differences are analyzed, for example, the current average delay is 50 milliseconds, which is higher than 40 milliseconds; the fluctuation frequency is 10 times, which is higher than 3 times. The difference analysis results are quantified by calculating the percentage deviation or absolute difference between the actual value and the benchmark value to obtain the difference value of the automated signal response characteristics, for example, the delay difference is 25% and the fluctuation frequency difference is 233%, which are integrated into a comprehensive index, for example, 0.45.

[0112] The parameter revision submodule identifies the structure of parameter revision items based on the difference values ​​of the automatic signal response characteristics and the corresponding period of the trend segment, analyzes the degree of parameter offset and frequency distribution amplitude, identifies the parameter revision offset, and adjusts the filtering frequency and time distribution in combination with the differences of the original parameter structure to obtain a segmented filtering parameter dynamic correction table.

[0113] Based on the difference value of the automated signal response characteristics (e.g., 0.45) and the trend segment period (e.g., 10-second duration, 1.2-second period), the difference value is matched with the parameter revision rule base to identify the revision item structure. For example, if the overall difference value is greater than 0.4 and the heart rate cycle is slower than 60 beats / minute, the rule base indicates that the low-frequency filter cutoff frequency and high-frequency filter gain should be adjusted. The revision item structure includes "low-pass cutoff frequency", "high-pass cutoff frequency", and "band gain". The current filter settings are compared with the ideal response characteristics to analyze the parameter offset and frequency distribution amplitude of each parameter revision item. For example, if the current low-pass cutoff frequency is 20 Hz and the ideal is 25 Hz, the offset is 5 Hz. The frequency distribution amplitude refers to the impact of the offset on the frequency range. For example, 5 Hz... Hertz adjustments affect filtering in the 20-30 Hz range. The degree of parameter revision offset is identified; for example, the low-pass cutoff frequency is adjusted upwards by 5 Hz, and the high-frequency gain is adjusted downwards by 0.1 times. Combining the differences in the original parameter structure (e.g., the original filter was over-smoothed), and comparing it with the current revision structure, the filter frequency and time distribution are adjusted. For example, the low-pass filter cutoff frequency is adjusted from 20 Hz to 25 Hz, and the high-pass filter cutoff frequency is adjusted from 150 Hz to 140 Hz. Based on the periodicity of trend segments, the filter parameter update frequency and signal time window are adjusted; for example, in a 1.2-second periodic trend segment, the filter parameters are updated every 0.5 seconds, applied to a 1-second signal frame. The adjusted parameter values ​​and application rules are compiled to generate a segmented dynamic correction table for filter parameters.

[0114] Please see Figure 6 The abnormal signal tracing module includes:

[0115] The response difference extraction submodule examines the response time and change sequence of high-frequency signals based on the segmented filtering parameter dynamic correction table, identifies the synchronization deviation between the response time difference and the rate of change, filters the time points and signal numbers whose deviations exceed the stable range, and generates an automated signal response offset list.

[0116] Based on the correction table of the recorded filter parameter adjustment scheme, the cutoff frequency of the high-pass filter is increased to 500 Hz under high-frequency noise conditions. The corrected filter parameters are applied to process cardiopulmonary sound signals to examine the response time and variation sequence of high-frequency signals (e.g., frequencies above 200 Hz). For example, monitoring the time it takes for the amplitude of a high-frequency signal to reach its peak after stimulation (e.g., coughing) and its decay sequence; comparing the response time and variation sequence of different high-frequency bands or time points to identify synchronization deviations in response time differences and rates of change. For example, two high-frequency sub-bands should theoretically respond synchronously, but in reality, the former is delayed by 50 milliseconds compared to the latter; differences exist in amplitude variation rates, such as 200% / second and 100% / second, which should theoretically be synchronous. A stable region is then established. For example, the stable interval for response time difference is ±20 milliseconds, and the stable interval for rate of change synchronization deviation is ±30%. The stable intervals are derived from statistical analysis of a large number of normal cardiopulmonary sounds' high-frequency responses. For example, the 95% confidence interval is ±18 milliseconds, and 20 milliseconds is a conservative threshold. Time points and signal numbers where the response time difference or rate of change synchronization deviation exceeds the stable interval are filtered out. For example, at t=3.50 seconds, the response time difference of LS_HF_001 is 60 milliseconds (exceeding ±20 milliseconds), and the rate of change synchronization deviation is 40% (exceeding ±30%). Record t=3.50 seconds and LS_HF_001. A list of high-frequency signals exhibiting asynchronous or abnormal changes at specific times is also recorded, generating an automated signal response offset list.

[0117] The wave origin discrimination submodule extracts wave sequences based on the automated signal response offset list, compares the wave amplitude and direction at the offset time points, identifies continuous and consistent wave segments and records the intervals that overlap with the offset time points, and generates a response wave origin segment table.

[0118] According to the automated signal response offset list, for example, LS_HF_001 has a response offset at t=3.50 seconds; for the offset signal number, extract the fluctuation sequence before and after the offset time point from the original signal, for example, the amplitude fluctuation sequence of LS_HF_001 between t=3.40 seconds and t=3.60 seconds, compare the fluctuation amplitude and fluctuation direction of different frequency bands or sensor signals at and near the offset time point, for example, at t=3.50 seconds, the amplitude of LS_HF_001 increases from 0.2 to 0.8, while the amplitude of LS_LF_002 increases from 0.1 to 0.5, and the directions are consistent, then they are from the same source; set a fluctuation amplitude consistency threshold (e.g. (If the difference does not exceed 20% and the direction is consistent), the threshold is derived from the amplitude fluctuation characteristics of common noise sources in the cardiopulmonary sound signal. It identifies continuous and consistent fluctuation segments. For example, from t=3.45 seconds to t=3.55 seconds, the amplitude fluctuations of LS_HF_001 and LS_LF_002 are consistent, with a difference of less than 20%. The segment and its overlapping interval with the offset time point are recorded. A table recording which frequency bands or signals exhibit homologous fluctuations within a specific time period is generated, creating a table of homologous fluctuation segments. For example, from t=3.45 seconds to t=3.55 seconds, LS_HF_001 and LS_LF_002 exhibit homologous fluctuations, overlapping at the offset point of t=3.50 seconds.

[0119] The abnormal attribution identification submodule extracts the node path of the signal number based on the response fluctuation homogeneous segment table, tracks the node response sequence and signal transmission in the overlapping segment, identifies the abnormal response frequency of the signal source node, and forms a source map of abnormal cardiopulmonary sounds.

[0120] Based on the response fluctuation homogeneous region table, for example, the homogeneous fluctuations of LS_HF_001 and LS_LF_002 within t=3.45 seconds to t=3.55 seconds, the node paths involving the signal numbers of the homogeneous fluctuations are extracted. For example, LS_HF_001 and LS_LF_002 both belong to the "left lower lobe" region. After "high-frequency enhancement filtering" and "low-frequency enhancement filtering" processing, the path is represented as "left lower lobe → high-frequency enhancement filtering → LS_HF_001". Within the homogeneous fluctuation region, the response sequence and signal transmission process of each node on the node path are traced. It is analyzed whether the signal change at the original acquisition point "left lower lobe" precedes the signal change after filtering. By analyzing the nodes... By analyzing the amplitude, frequency components, and time delay changes of the signal, we can determine the source of the signal and how it evolves into an anomaly. For example, if the original signal from the "left lower lobe" has a broadband instantaneous impact at t=3.45 seconds, it is earlier than the homologous fluctuations of the filtered LS_HF_001 and LS_LF_002. We can also count the frequency of abnormal responses of each signal source node (e.g., the cardiopulmonary sound acquisition sensor, the left lower lobe, and the friction rub generation point). For example, out of 50 homologous fluctuations, 35 can be traced back to the "left lower lobe" as the source, with an abnormal response frequency of 35 times. By integrating the signal source nodes, node paths, signal transmission processes, node response sequences, and the frequency of abnormal responses of the signal source nodes, we can form a source atlas of abnormal cardiopulmonary sound characteristics.

[0121] The above are merely preferred embodiments of the present invention and are not intended to limit the present invention in any other way. Any person skilled in the art may make changes or modifications to the above-disclosed technical content to create equivalent embodiments that can be applied to other fields. However, any simple modifications, equivalent changes, and modifications made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the protection scope of the present invention.

Claims

1. A dynamic filtering based cardiopulmonary sound preprocessing system, characterized in that, The system comprises: The signal segmentation optimization module acquires the time distribution characteristics, spectral intensity variation, and amplitude fluctuation range of the cardiopulmonary sound signal, analyzes the spectral overlap relationship between signal segments, judges the signal segmentation boundary and spectral coverage length according to the spectral distribution density and amplitude response order, and performs spectral segmentation and amplitude mapping to generate a cardiopulmonary sound segmentation characteristic spectrum graph; The noise suppression and spectral reconstruction module selects the noise energy proportion and amplitude fluctuation value of the key frequency band based on the cardiopulmonary sound segmentation characteristic spectrum graph, matches and compares the noise frequency and spectral overlap degree, identifies the noise suppression key level, and acquires a frequency band energy weight evaluation index set; The signal quality evaluation module extracts the time distribution and duration of abnormal signal segments according to the frequency band energy weight evaluation index set, judges the signal quality trough distribution stage, classifies the time trend of similar abnormal signal segments, and generates a cardiopulmonary sound quality dynamic evolution sequence; The dynamic filtering adjustment module calls the cardiopulmonary sound quality dynamic evolution sequence, combines the amplitude fluctuation value and response time delay in the jurisdiction section, filters the target signal segment in the trend section, revises the filtering parameter value in sections, and obtains a segmented filtering parameter dynamic correction table.

2. The dynamic filtering based cardiopulmonary sound preprocessing system of claim 1, wherein, The cardiopulmonary sound segmentation characteristic spectrum graph includes signal segmentation boundary classification, spectral coverage path structure, and amplitude fluctuation distribution, the frequency band energy weight evaluation index set includes frequency band level label, frequency band occupation coefficient, and priority score interval, the cardiopulmonary sound quality dynamic evolution sequence includes abnormal time distribution, abnormal duration, and trend classification result, and the segmented filtering parameter dynamic correction table includes a target signal list, parameter revision range, and delay correlation factor.

3. The dynamic filtering based cardiopulmonary sound preprocessing system of claim 1, wherein, The signal segmentation optimization module comprises: The signal distribution and spectral recognition submodule acquires the time distribution characteristics, spectral intensity variation, and amplitude fluctuation range data of the cardiopulmonary sound signal, aligns the three signals according to the time stamp, extracts signal mutation points, locates spectral intensity and amplitude variation, identifies response difference, and generates signal operation response deviation; The spectral lag analysis submodule extracts signal delay segments and spectral response order according to the signal operation response deviation, counts the delay length and response sequence, judges the spectral dependence direction, measures the number of signal trigger sequences in the lag section, and generates a signal spectral lag structure quantity; The spectral linkage structure generation submodule extracts amplitude fluctuation points and signal spectral linkage periods based on the signal spectral lag structure quantity, compares the spectral response frequency and time delay, identifies the linkage frequency and average delay ratio, analyzes the dependence relationship and spectral mapping, and generates a cardiopulmonary sound segmentation characteristic spectrum graph.

4. The dynamic filtering based cardiopulmonary sound preprocessing system of claim 3, wherein, The noise suppression and spectral reconstruction module comprises: The amplitude fluctuation extraction submodule extracts the amplitude fluctuation value of the key frequency band based on the cardiopulmonary sound segmentation characteristic spectrum graph, filters the sequence whose fluctuation exceeds the amplitude reference value, calculates the fluctuation abnormality ratio parameter, analyzes the abnormal fluctuation frequency ratio, and generates an abnormal amplitude fluctuation frequency ratio; The frequency dependence matching submodule calls the abnormal amplitude fluctuation frequency ratio, combines the frequency band task frequency and dependence path number, identifies the frequency dependence comparison item, adjusts the frequency band abnormal offset according to the task proportion on the dependence path, and obtains a fluctuation coupling priority index; The priority level determination sub-module extracts a numerical interval corresponding to the frequency band based on the fluctuation coupling priority index, sets a frequency band level division interval group, assigns a level identifier according to the index value falling into the interval, sorts to determine a noise suppression level, and obtains a frequency band energy weight evaluation index set.

5. The dynamic filtering based cardiopulmonary sound preprocessing system of claim 4, wherein, The signal quality evaluation module comprises: The abnormal record screening sub-module collects time points of triggering and ending of abnormal signals according to the frequency band energy weight evaluation index set, identifies time intervals between adjacent abnormalities, screens records exceeding an abnormal interval reference value, and generates an abnormal time interval sequence; The abnormal period division sub-module calls the abnormal time interval sequence, counts the continuous occurrence number of the same type of abnormal events, calculates the periodic distribution trend value of the abnormality according to the abnormal duration, interval time and signal load index, and establishes an abnormal period distribution stage; ， The cycle trend classification sub-module calls the abnormal period distribution stage, identifies abnormal events of the same cycle trend, identifies trend fluctuation differences, classifies and archives according to the difference threshold, and generates a heart and lung sound quality dynamic evolution sequence. Wherein, P represents the abnormal cycle distribution trend value, D t represents the abnormal duration at the t time, μ represents the average value of the abnormal duration, T t represents the time correlation index at the t time, N represents the total number of samples; The dynamic filtering adjustment module comprises:

6. The dynamic filtering based cardiopulmonary sound preprocessing system of claim 5, wherein, The trend sequence identification sub-module calls the heart and lung sound quality dynamic evolution sequence, extracts an incremental change value, judges a trend rising interval according to the cumulative amplitude and fluctuation amplitude, marks a key signal segment, and generates an automatic signal trend abnormality identification list; The response characteristic extraction sub-module calls the signal number in the automatic signal trend abnormality identification list, identifies the amplitude fluctuation value and response delay in the section, analyzes the signal response difference by combining the delay distribution and fluctuation frequency, and obtains an automatic signal response characteristic difference value; The parameter revision sub-module identifies the parameter revision item structure according to the automatic signal response characteristic difference value and the trend section corresponding period, analyzes the parameter offset degree and frequency distribution amplitude, identifies the parameter revision offset degree, adjusts the filtering frequency and time distribution by combining the original parameter structure difference, and obtains a segmented filtering parameter dynamic correction table. The trend rising interval is judged when the incremental change value of the heart and lung sound quality dynamic evolution sequence in the continuous preset time period, the cumulative amplitude of the incremental change value is higher than the first preset amplitude threshold, and the fluctuation amplitude of the incremental change value is less than the second preset fluctuation threshold.

7. The dynamic filtering based cardiopulmonary sound preprocessing system of claim 6, wherein, The key signal segment is marked when the number of local peak values of the incremental change value exceeds the preset peak value threshold, or the change rate of the incremental change value exceeds the preset rate threshold in the trend rising interval. The system further comprises an abnormal signal tracing module:

8. The dynamic filtering based cardiopulmonary sound preprocessing system of claim 1, wherein, The abnormal signal tracing module collects the response difference and fluctuation characteristics of the high-frequency filtered signal based on the segmented filtering parameter dynamic correction table, judges whether the fluctuation characteristics have homology, identifies abnormal attribution nodes, and forms a heart and lung sound abnormal feature tracing map; The heart and lung sound abnormal feature tracing map comprises a fluctuation homology feature group, an abnormal node positioning result, and a tracing path structure diagram. The abnormal signal tracing module comprises:

9. The dynamic filtering based cardiopulmonary sound preprocessing system of claim 8, wherein, ​ The response difference extraction submodule checks the high-frequency signal response time and change sequence based on the segmented filter parameter dynamic correction table, identifies the synchronization deviation of the response time difference and the change rate, screens the time points and signal numbers whose deviation exceeds the stable interval, and generates an automatic signal response deviation list; The fluctuation homology identification submodule extracts the fluctuation sequence according to the automatic signal response deviation list, compares the fluctuation amplitude and direction of the deviation time points, identifies the continuous consistent fluctuation section and records the interval overlapping with the deviation time point, and generates a response fluctuation homology section table; The abnormality attribution identification submodule extracts the node path of the signal number according to the response fluctuation homology section table, traces the node response sequence and signal transmission in the overlapping section, identifies the response abnormality frequency of the signal source node, and forms a heart and lung sound abnormality feature tracing atlas.

Images