An automatic evaluation method, system, device and medium for uterine contraction pressure of pregnant women based on acoustic signal features

Abstract

The present application relates to a kind of automatic evaluation method, system, equipment and medium of pregnant woman uterine contraction pressure based on acoustic signal feature, method includes: based on maternal heart sound template mutual information measure, signal decoupling processing is carried out to the multi-channel abdominal acoustic vibration signal collected by flexible acoustic sensor, rejects heart sound interference component and carries out multi-channel weighted fusion, generates pure uterine contraction acoustic signal;The pure signal is extracted with hilbert envelope and adaptive double threshold segmentation, detects uterine contraction event and generates uterine contraction event sequence;Based on the event sequence, uterine contraction frequency time sequence is calculated;Combining wavelet packet energy feature, recursive quantization analysis parameter and envelope peak amplitude, uterine contraction comprehensive intensity index is calculated and intensity grade is output;Finally, generate the monitoring curve chart containing uterine contraction waveform chart and quantization report.The present application effectively solves the problem that maternal heart sound interference is difficult to decouple and the accuracy of intensity evaluation is insufficient in the existing acoustic uterine contraction monitoring.

Description

Technical Field

[0001] This invention relates to the fields of biomedical signal processing and computer-aided diagnostic technology, specifically to an automatic assessment method, system, device, and medium for uterine contraction pressure in pregnant women based on acoustic signal characteristics. Background Technology

[0002] Uterine contractions are a core physiological event in the labor process, and accurate assessment of their frequency and intensity is of significant clinical importance for judging the progress of labor, identifying abnormal deliveries, and ensuring maternal and infant safety. Currently, the main methods for monitoring uterine contractions in clinical practice include palpation and external dynamometers. Palpation involves healthcare professionals estimating the occurrence and intensity of contractions by sensing changes in abdominal tension in the pregnant woman. While simple to operate, this method is highly dependent on the operator's clinical experience, resulting in poor consistency in judgment among different personnel, and it cannot provide continuous quantitative monitoring data. Dynamometers, on the other hand, use pressure sensors attached to the pregnant woman's abdomen to convert changes in abdominal tension caused by uterine contractions into pressure waveform signals, enabling continuous recording. However, this technology has significant limitations in practical application: the pressure sensor requires a certain pre-tension to the abdomen to ensure coupling, and prolonged wear can cause discomfort for the pregnant woman; for obese pregnant women, the subcutaneous fat layer significantly attenuates the pressure signal, leading to decreased signal quality; furthermore, maternal respiratory movements, fetal movements, and heart sounds can interfere with the uterine contraction pressure signal, affecting the accurate identification of the contraction waveform.

[0003] To address these issues, researchers have attempted to use acoustic sensors to collect low-frequency acoustic vibration signals generated during uterine contractions, aiming to achieve non-contact, highly comfortable uterine contraction monitoring. However, existing acoustic uterine contraction monitoring methods still face key technical bottlenecks: On the one hand, the acoustic vibration signals collected from the pregnant woman's abdomen are a mixture of multiple sound sources, including uterine contraction vibrations, maternal heart sounds, fetal heart sounds, fetal movements, and intestinal peristalsis. Among these, the energy of maternal heart sounds is often significantly higher than that of uterine contraction signals, causing contraction characteristics to be masked. Existing methods lack effective signal decoupling techniques to separate the pure acoustic component of uterine contractions from the mixed signals. On the other hand, the quantitative assessment of uterine contraction intensity mainly relies on the peak value or area integral of the time-domain waveform. These indicators are easily affected by baseline drift and residual noise and are difficult to reflect the inherent nonlinear dynamic characteristics of uterine contraction signals, resulting in insufficient accuracy in intensity assessment. Summary of the Invention

[0004] Based on this, the purpose of the present invention is to provide an automatic assessment method, system, device and medium for pregnant women's uterine contraction pressure based on acoustic signal characteristics, which can effectively decouple maternal heart sound interference and objectively quantify the frequency and intensity of uterine contractions from multi-dimensional features.

[0005] The objective of this invention is achieved through the following solution:

[0006] In a first aspect, the present invention provides an automatic assessment method for uterine contraction pressure in pregnant women based on acoustic signal characteristics, comprising the following steps:

[0007] S1: The multi-channel acoustic vibration signal of the pregnant woman's abdomen collected by the flexible acoustic sensor array is decoupled and combined with the mutual information measurement between the pre-stored maternal heart sound templates to remove the heart sound interference component and then the multi-channel signal after removing the interference component is weighted and fused to generate a pure uterine contraction acoustic signal.

[0008] S2: Hilbert envelope extraction and adaptive dual-threshold segmentation are performed on the pure uterine contraction acoustic signal to generate a smooth envelope curve. Based on the upward and downward crossing events of the smooth envelope curve, the start point, peak point and end point of each uterine contraction event are detected and marked to generate a uterine contraction event sequence.

[0009] S3: Based on the uterine contraction event sequence, using a preset duration as the sliding window length, count the number of uterine contraction event start points in each sliding window, calculate the uterine contraction frequency in each sliding window, and generate a frequency time series.

[0010] S4: Based on the uterine contraction event sequence, signal segments within each uterine contraction event window are extracted from the pure uterine contraction acoustic signal. Wavelet packet decomposition is performed on the signal segments to extract the wavelet packet energy features of the frequency band related to uterine contraction. Phase space reconstruction and recursive quantization analysis are performed on the signal segments to extract the recursion rate, determinism, and recursion entropy. Combined with the peak amplitude of the smooth envelope curve, the comprehensive uterine contraction intensity index is calculated, and the uterine contraction intensity level is output.

[0011] S5: Based on pure uterine contraction acoustic signals, smooth envelope curves, contraction event sequences, frequency time series, and intensity levels, generate a contraction monitoring curve that includes contraction waveforms and contraction event markers, as well as a quantitative report of contractions that includes the occurrence time, duration, frequency, and intensity level of each contraction event.

[0012] Secondly, the present invention provides an automatic assessment system for uterine contraction pressure in pregnant women based on acoustic signal characteristics, the system being configured with the following modules:

[0013] The uterine contraction acoustic signal purification module is used to decouple the multi-channel acoustic vibration signals of the pregnant woman's abdomen collected by a flexible acoustic sensor array, combine the mutual information measurement and recognition between pre-stored maternal heart sound templates, remove the heart sound interference components, and perform weighted fusion of the multi-channel signals after removing the interference components to generate a pure uterine contraction acoustic signal.

[0014] The uterine contraction event detection module is used to extract the Hilbert envelope and perform adaptive dual-threshold segmentation on the pure uterine contraction acoustic signal to generate a smooth envelope curve. Based on the upward and downward crossing events of the smooth envelope curve, the module detects and marks the start point, peak point and end point of each uterine contraction event to generate a uterine contraction event sequence.

[0015] The contraction frequency time series module is used to generate a frequency time series by counting the number of contraction event start points in each sliding window based on the contraction event sequence, using a preset duration as the sliding window length;

[0016] The contraction intensity assessment module is used to extract signal segments within each contraction event window from the pure acoustic signal of uterine contractions based on the contraction event sequence. The signal segments are decomposed by wavelet packet to extract the wavelet packet energy features of the frequency bands related to uterine contractions. The signal segments are then reconstructed in phase space and recursively quantized to extract the recursion rate, determinism, and recursion entropy. Combined with the peak amplitude of the smooth envelope curve, the comprehensive contraction intensity index is calculated and the contraction intensity level is output.

[0017] The monitoring report generation module is used to generate a uterine contraction monitoring curve containing uterine contraction waveforms and uterine contraction event markers, based on pure uterine contraction acoustic signals, smooth envelope curves, uterine contraction event sequences, frequency time series, and intensity levels, as well as a uterine contraction quantitative report containing the occurrence time, duration, frequency, and intensity level of each uterine contraction event.

[0018] Thirdly, this application provides a computer device, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement any of the above-mentioned automatic assessment methods for uterine contraction pressure of pregnant women based on acoustic signal characteristics.

[0019] Fourthly, this application provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements any of the above-mentioned methods for automatically assessing uterine contraction pressure in pregnant women based on acoustic signal characteristics.

[0020] In summary, the automatic assessment method for uterine contraction pressure in pregnant women based on acoustic signal characteristics provided in this application effectively eliminates maternal heart sound interference through signal decoupling processing, and can extract pure uterine contraction acoustic signals from mixed acoustic vibration signals. This solves the problem in existing technologies where heart sounds mask contraction characteristics, making it difficult to identify uterine contraction waveforms. Through Hilbert envelope extraction and adaptive dual-threshold segmentation, the starting point, peak point, and ending point of uterine contraction events can be accurately detected, enabling automated statistical analysis of uterine contraction frequency. This overcomes the shortcomings of traditional palpation methods, such as strong subjectivity, poor consistency, and susceptibility to interference from labor force gauge signals. Furthermore, by combining wavelet packet energy characteristics and recursive quantification analysis, the intensity of uterine contractions can be characterized from multiple dimensions, including time-frequency domain and nonlinear dynamics. This compensates for the insufficiency of the single envelope peak method in terms of sensitivity to noise and baseline drift, achieving a refined grading effect for uterine contraction intensity. The final generated uterine contraction monitoring curve and quantitative report can provide continuous, objective, and traceable labor monitoring data for clinical use, thereby significantly improving the accuracy and reliability of uterine contraction pressure assessment.

[0021] To better understand and implement this invention, the following detailed description is provided in conjunction with the accompanying drawings. Attached Figure Description

[0022] Figure 1 A flowchart illustrating an automatic assessment method for uterine contraction pressure in pregnant women based on acoustic signal characteristics, provided for an embodiment of this application;

[0023] Figure 2 This is a schematic diagram of the structure of an automatic assessment system for uterine contraction pressure in pregnant women based on acoustic signal characteristics, provided as another embodiment of this application. Detailed Implementation

[0024] To facilitate understanding of the present invention, a more complete description will be given below with reference to the accompanying drawings. Preferred embodiments of the invention are shown in the drawings. However, the invention can be implemented in many different forms and is not limited to the embodiments described herein. Rather, these embodiments are provided to provide a thorough and complete understanding of the disclosure of the invention.

[0025] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains. The terminology used herein in the description of the invention is for the purpose of describing particular embodiments only and is not intended to be limiting of the invention. The term "and / or" as used herein includes any and all combinations of one or more of the associated listed items.

[0026] In one embodiment, such as Figure 1As shown, an automatic assessment method for uterine contraction pressure in pregnant women based on acoustic signal characteristics is provided. This embodiment illustrates the application of this method to a terminal. It is understood that this method can also be applied to a server, or to a system including both a terminal and a server, and implemented through interaction between the terminal and the server. In this embodiment, the method includes the following steps:

[0027] S1: The multi-channel acoustic vibration signals of the pregnant woman's abdomen collected by the flexible acoustic sensor array are decoupled and processed. Combined with the mutual information measurement and recognition between the pre-stored maternal heart sound templates, the interference components of the heart sounds are removed and the multi-channel signals after removing the interference components are weighted and fused to generate a pure uterine contraction acoustic signal.

[0028] Specifically, the system deploys a flexible acoustic sensor array, which is fitted to the area corresponding to the uterus in the lower abdomen of the pregnant woman. Acoustic coupling between the sensors and the abdominal skin is achieved using a medical coupling agent. The output of the sensor array is connected to a host computer via a data acquisition card, enabling the synchronous acquisition and storage of multi-channel acoustic vibration signals. During acquisition, the system simultaneously records the pregnant woman's basic physiological information. The system pre-stores maternal heart sound templates corresponding to different combinations of physiological parameters, constructing a maternal heart sound template library. Before monitoring begins, the system acquires the acoustic vibration signals of the pregnant woman at rest, extracts the heart sound components, matches them with the heart sound templates in the template library, and updates them to obtain a personalized heart sound template adapted to the pregnant woman. The system calculates the mutual information value between the acquired multi-channel acoustic vibration signals and the personalized heart sound template; the mutual information value characterizes the degree of correlation between the two signals. The system sets a mutual information threshold, which is obtained through training with a large amount of clinical data. When the mutual information value between a signal segment in a certain channel signal and the heart sound template is greater than the set threshold, the system determines that the segment is a heart sound interference component. The system uses an adaptive filtering algorithm to remove the interference component from the original signal. During the filtering process, the system adjusts the filtering coefficient in real time.

[0029] After interference removal, the system calculates the signal-to-noise ratio (SNR) of each channel signal. The SNR is calculated as the ratio of effective signal power to noise power. The effective signal power is calculated by integrating the power within the characteristic frequency band of the uterine contraction signal, and the noise power is calculated by integrating the signal power during non-contraction periods. Using the SNR of each channel as a weight, the system normalizes the weighting coefficients and then employs a linear weighted fusion algorithm to fuse the multi-channel signals, generating a pure uterine contraction acoustic signal.

[0030] S2: Hilbert envelope extraction and adaptive dual-threshold segmentation are performed on the pure uterine contraction acoustic signal to generate a smooth envelope curve. Based on the upward and downward crossing events of the smooth envelope curve, the start point, peak point and end point of each uterine contraction event are detected and marked to generate a uterine contraction event sequence.

[0031] Specifically, the system performs a Hilbert transform on the pure uterine contraction acoustic signal to obtain the analytic signal, the magnitude of which is the instantaneous amplitude envelope. The system then uses Gaussian filtering to smooth the instantaneous amplitude envelope, resulting in a smoothed envelope curve that reflects the intensity of the contraction signal over time. The system uses statistical methods to determine adaptive dual thresholds, including an upper and lower threshold. The system extracts the baseline value of the smoothed envelope curve, which is the mean of the envelope curve during non-contraction periods. The system uses a sliding window to statistically analyze the mean envelope value during non-contraction periods and updates the baseline value in real time. The lower threshold is determined based on the baseline value and its standard deviation, as is the upper threshold. The baseline standard deviation is calculated statistically from the envelope curve during non-contraction periods. The dual thresholds are updated adaptively in real time at a fixed frequency.

[0032] The system detects the start, peak, and end points of uterine contractions based on a comparison of a smooth envelope curve and adaptive dual thresholds. When the smooth envelope curve first crosses the lower threshold, the system marks that point as the start of the contraction. When the envelope curve reaches its maximum value, the system marks that point as the peak, and the envelope value corresponding to the peak point represents the instantaneous intensity of the contraction. When the envelope curve first crosses the lower threshold and remains below it for a certain period, the system marks that point as the end of the contraction. The system sets a minimum duration threshold for contractions; envelope fluctuations below this threshold are considered noise and not marked as contractions. The system arranges each detected contraction in chronological order, forming a contraction sequence. Each element in the sequence contains the start time, peak time, end time, and peak amplitude of the contraction.

[0033] S3: Based on the uterine contraction event sequence, using a preset duration as the sliding window length, count the number of uterine contraction event start points within each sliding window, calculate the uterine contraction frequency within each sliding window, and generate a frequency time series.

[0034] Specifically, the system pre-sets the sliding window length and sliding step size based on clinical treatment guidelines. These parameters can be adaptively adjusted according to clinical needs. The sliding window starts from the first point when the system detects a uterine contraction event. The system controls the sliding window to slide sequentially until monitoring ends. For each sliding window, the system counts the number of uterine contraction event start points contained within the window, discarding duplicate start points. If the start point of a uterine contraction event is located in the overlapping area of ​​two adjacent windows, the system only counts that start point within the initial window. The system calculates the uterine contraction frequency based on the ratio of the number of uterine contraction events within the window to the window duration. The system uses the center time of each sliding window as a time node and the corresponding uterine contraction frequency as a numerical value, arranging them in chronological order to form a frequency time series. The sampling frequency of the time series is consistent with the sliding step size. For windows where no uterine contraction events are detected, the system records its uterine contraction frequency as 0. The generated frequency time series is updated in real time by the system, synchronously reflecting the dynamic changes in uterine contraction frequency and providing medical staff with data support related to uterine contraction frequency. The system uses this step to achieve real-time statistics and dynamic presentation of uterine contraction frequency, providing basic data for judging the progress of labor, ensuring the accuracy and real-time nature of frequency statistics, and avoiding the impact of random fluctuations in a single uterine contraction on frequency assessment.

[0035] S4: Based on the uterine contraction event sequence, signal segments within each uterine contraction event window are extracted from the pure uterine contraction acoustic signal. Wavelet packet decomposition is performed on the signal segments to extract the wavelet packet energy features of the frequency bands related to uterine contraction. Phase space reconstruction and recursive quantization analysis are performed on the signal segments to extract the recursion rate, determinism, and recursion entropy. Combined with the peak amplitude of the smooth envelope curve, the comprehensive uterine contraction intensity index is calculated, and the uterine contraction intensity level is output.

[0036] Specifically, based on the generated sequence of uterine contraction events, the system extracts signal segments corresponding to each contraction event from the pure acoustic signal of uterine contractions. The extraction range completely includes the rising edge, peak value, and falling edge of the contraction signal, avoiding feature loss due to signal truncation. Each extracted signal segment is numbered sequentially according to the contraction event sequence. The system performs wavelet packet decomposition on the signal segments, dividing the signal into multiple sub-frequency bands, with the uterine contraction signal concentrated in a specific low-frequency sub-band. After decomposition, the system calculates the sum of squares of the wavelet packet node coefficients within the target frequency band to obtain the wavelet packet energy feature, which is positively correlated with the intensity of the contraction signal. The system normalizes the wavelet packet energy feature to eliminate the influence of signal amplitude differences. Based on the correlation theorem, the system reconstructs the phase space of the signal segments. The reconstruction parameters are determined by a specific method, and the reconstructed phase space reflects the inherent nonlinear dynamic characteristics of the contraction signal.

[0037] Furthermore, the system performs recursive quantization analysis based on the reconstructed phase space, extracting three core nonlinear features: recursion rate, determinism, and recursion entropy. The recursion rate reflects the degree of recursion in the signal, determinism reflects the regularity of the signal, and recursion entropy reflects the complexity of the signal. The system normalizes all three nonlinear features. The system extracts the amplitude corresponding to the peak point of each uterine contraction event from the smooth envelope curve as a time-domain feature index. This index directly reflects the instantaneous intensity of the contraction, and the system normalizes the peak amplitude. The system calculates the comprehensive intensity index using a weighted summation method. The weight allocation is based on the influence of each feature on the intensity of the contraction, and has been validated using clinical data. The system classifies the contraction intensity levels according to the comprehensive intensity index, associates the comprehensive intensity index of each contraction event with its corresponding intensity level, and outputs the intensity level of each contraction event.

[0038] S5: Based on pure uterine contraction acoustic signals, smooth envelope curves, contraction event sequences, frequency time series, and intensity levels, generate a contraction monitoring curve that includes contraction waveforms and contraction event markers, as well as a quantitative report of contractions that includes the occurrence time, duration, frequency, and intensity level of each contraction event.

[0039] Specifically, the system uses professional data visualization tools to draw uterine contraction monitoring curves. The horizontal axis represents time, and the vertical axis is divided into upper and lower regions. The upper region plots the acoustic signal waveform of pure uterine contractions and the smooth envelope curve, using different colors to mark the start, peak, and end points of contraction events, visually displaying the waveform changes and key nodes of the contraction signal. The lower region plots the frequency time series curve and contraction intensity level markers, clearly reflecting the dynamic changes in contraction frequency and intensity. The time axis of the monitoring curve is consistent with the signal acquisition time, and necessary axis labels, legends, and units are added to the graph to ensure readability. The system generates a quantitative report on contractions, presented in text format, including basic monitoring information, detailed parameters for each contraction event, overall statistical information, and abnormal alerts.

[0040] The basic monitoring information includes the monitoring start time, monitoring duration, and basic information about the pregnant woman. Detailed parameters for each contraction event include occurrence time, duration, peak time, peak amplitude, comprehensive intensity index, and intensity level. Overall statistical information includes the total number of contraction events during the monitoring period, average contraction frequency, average comprehensive intensity index, and the number and percentage of contractions at different intensity levels. Abnormal alerts are used to indicate abnormal contraction frequency or intensity, providing warnings to medical staff. The system links and stores contraction monitoring curves with quantitative reports, which can be displayed on terminal devices or printed out, providing medical staff with objective and accurate quantitative evidence for judging labor progress and identifying abnormal deliveries, achieving automatic, continuous, and high-precision assessment of contraction pressure.

[0041] In summary, the automatic assessment method for uterine contraction pressure in pregnant women based on acoustic signal characteristics provided in this application effectively eliminates maternal heart sound interference through signal decoupling processing, and can extract pure uterine contraction acoustic signals from mixed acoustic vibration signals. This solves the problem in existing technologies where heart sounds mask contraction characteristics, making it difficult to identify uterine contraction waveforms. Through Hilbert envelope extraction and adaptive dual-threshold segmentation, the starting point, peak point, and ending point of uterine contraction events can be accurately detected, enabling automated statistical analysis of uterine contraction frequency. This overcomes the shortcomings of traditional palpation methods, such as strong subjectivity, poor consistency, and susceptibility to interference from labor force gauge signals. Furthermore, by combining wavelet packet energy characteristics and recursive quantification analysis, the intensity of uterine contractions can be characterized from multiple dimensions, including time-frequency domain and nonlinear dynamics. This compensates for the insufficiency of the single envelope peak method in terms of sensitivity to noise and baseline drift, achieving a refined grading effect for uterine contraction intensity. The final generated uterine contraction monitoring curve and quantitative report can provide continuous, objective, and traceable labor monitoring data for clinical use, thereby significantly improving the accuracy and reliability of uterine contraction pressure assessment.

[0042] In one embodiment, S1 of the automatic assessment method for uterine contraction pressure in pregnant women based on acoustic signal characteristics provided by the present invention specifically includes the following steps:

[0043] S11: Perform empirical mode decomposition on each single-channel signal in the multi-channel acoustic vibration signal of the pregnant woman's abdomen collected by the flexible acoustic sensor array, decompose it into multiple intrinsic mode function components and residuals, calculate the normalized mutual information between each intrinsic mode function component and the maternal heart sound template in combination with the pre-stored maternal heart sound template, mark the intrinsic mode function components with normalized mutual information greater than the preset interference judgment threshold as the dominant heart sound components, and generate the heart sound interference identification results for each channel.

[0044] Specifically, the system acquires multi-channel acoustic vibration signals from the abdomen of pregnant women collected by a flexible acoustic sensor array, and performs empirical mode decomposition (EMD) on each single-channel signal. EMD decomposes the signal according to its own time-scale characteristics without the need for pre-defined basis functions. Through this decomposition operation, the system decomposes each single-channel original signal into multiple intrinsic mode function (IMF) components and residuals. The IMF components reflect the fluctuation characteristics of the signal within different frequency ranges, while the residuals reflect the overall trend of the signal. The system calls a pre-stored maternal heart sound template, which is constructed based on maternal heart sound data under different physiological states and can cover the maternal heart sound characteristics of different individuals.

[0045] The system calculates the normalized mutual information between each intrinsic modal function component (IMF) obtained from the decomposition of each single channel and the maternal heart sound template. The normalized mutual information quantifies the correlation between the two signals, and its value range is within a fixed interval. The system calls a preset interference threshold, which is trained using a large amount of clinical data and is used to distinguish between heart sound interference-related components and uterine contraction signal-related components. The system compares the normalized mutual information corresponding to each IMF component with the preset interference threshold, marking IMF components with normalized mutual information greater than the preset threshold as dominant heart sound components. After marking all IMF components in each channel, the system integrates the marking results to generate the heart sound interference identification result for each channel, which clearly identifies which IMF components in each channel belong to the dominant heart sound components.

[0046] S12: Based on the heart sound interference identification results, remove the dominant heart sound component from the original signal of each channel and reconstruct the remaining intrinsic mode function components and residuals to generate the interference-free signal of each channel after removing heart sound interference.

[0047] Specifically, the system reads the heart sound interference identification results for each channel, identifying the specific intrinsic mode function (IMF) components marked as dominant heart sounds in each channel. Based on this identification result, the system performs interference removal on the original signal of each channel, removing all IMF components marked as dominant heart sounds from the original signal decomposition products of the corresponding channel. After the removal operation, each channel retains only the IMF components of the unmarked dominant heart sounds and the residuals obtained from the decomposition. The system then performs a reconstruction operation on the remaining IMF components and residuals of each channel. The reconstruction process follows the inverse process of empirical mode decomposition, superimposing the remaining IMF components and residuals to restore a continuous time-domain signal. This reconstruction operation can retain the signal components related to uterine contractions in the original signal while completely removing the signal components corresponding to heart sound interference. After sequentially performing the dominant heart sound component removal and signal reconstruction operations on all channels, the system generates a de-interference signal for each channel. The de-interference signal for each channel has eliminated the influence of heart sound interference and contains only acoustic vibration signal components related to uterine contractions.

[0048] S13: Estimate the signal-to-noise ratio (SNR) of the interference-removed signals for each channel, calculate the SNR estimate for each channel and normalize it to a weighted fusion weight, and then linearly sum the interference-removed signals of each channel according to the weighted fusion weight to generate a pure uterine contraction acoustic signal.

[0049] Specifically, the system acquires the de-interference signals from each channel and performs signal-to-noise ratio (SNR) estimation on each channel's de-interference signal. SNR estimation is achieved by analyzing the power ratio of the effective signal component to the noise component in the de-interference signal. The effective signal component is the acoustic vibration signal related to uterine contractions, while the noise component is the weak, irrelevant signal remaining after de-interference. The system obtains the estimated SNR value for each channel's de-interference signal, which reflects the purity of the corresponding channel's de-interference signal. The system performs normalization processing on the estimated SNR values ​​of all channels, converting them into weighted values ​​that meet the requirements of weighted fusion. The sum of the weighted values ​​of all channels satisfies a fixed condition. The system associates the de-interference signal of each channel with its corresponding normalized weight value and performs a linear weighted summation operation. Each channel's de-interference signal is first multiplied by its corresponding weight value, and then the weighted signals from all channels are superimposed. The linear weighted summation operation highlights the contribution of channels with higher SNR and suppresses the residual noise in channels with lower SNR. After the system completes the linear weighted summation of all channels, it generates a pure acoustic signal of uterine contraction. This signal has completely eliminated the interference of heart sounds and integrated the effective signal components of each channel, which can accurately reflect the acoustic vibration characteristics generated by uterine contraction, providing a reliable signal basis for subsequent detection and quantification of uterine contraction events.

[0050] In one embodiment, S2 of the automatic assessment method for uterine contraction pressure in pregnant women based on acoustic signal characteristics provided by the present invention specifically includes the following steps:

[0051] S21: Perform Hilbert transform on the pure uterine contraction acoustic signal, calculate the instantaneous envelope of the pure uterine contraction acoustic signal by combining the original signal and the transformed orthogonal components, and perform smoothing filtering on the instantaneous envelope to generate a smooth envelope curve.

[0052] Specifically, the system acquires a pure uterine contraction acoustic signal, performs a Hilbert transform on this signal, and obtains orthogonal components that are orthogonal to the original signal by integrating the pure uterine contraction acoustic signal. The system uses the pure uterine contraction acoustic signal as the original signal and the result of the Hilbert transform as the orthogonal components, and calculates the instantaneous envelope of the pure uterine contraction acoustic signal based on both. Preferably, the instantaneous envelope is calculated using the following formula:

[0053]

[0054] in, Represents the instantaneous envelope at time t. The raw signal representing the pure acoustic signal of uterine contractions. This represents the orthogonal components obtained after performing a Hilbert transform on the original signal. This represents the coupling coefficient between the orthogonal components and the original signal, used to enhance the instantaneous envelope's response to changes in the amplitude of the uterine contraction signal. After calculating the instantaneous envelope, the system performs smoothing filtering on it. The smoothing formula is as follows:

[0055]

[0056] in, This represents the smoothed envelope value at time t. Indicates the length of the smooth window. express The instantaneous envelope value at time t. This represents the weighting coefficients within the smoothing window. These weighting coefficients adapt to changes in the window position to achieve adaptive smoothing. Through the aforementioned smoothing filtering operation, the system eliminates high-frequency noise in the instantaneous envelope, generating a smooth envelope curve that accurately reflects the amplitude variation trend of the uterine contraction acoustic signal.

[0057] S22: Calculate the mean and standard deviation of the smooth envelope curve based on the statistical distribution characteristics of the smooth envelope curve, and set the high threshold and low threshold for uterine contraction detection according to the linear combination of the mean and standard deviation to generate an adaptive dual-threshold segmentation benchmark.

[0058] Specifically, the system acquires a smooth envelope curve, performs statistical distribution analysis on the curve, and extracts its statistical distribution characteristics. Based on these characteristics, the system calculates the mean and standard deviation of the smooth envelope curve. The mean reflects the overall amplitude level of the smooth envelope curve, while the standard deviation reflects the degree of amplitude fluctuation. Based on the calculated mean and standard deviation, the system sets high and low thresholds for uterine contraction detection using a linear combination. The thresholds are set using the following formula:

[0059]

[0060]

[0061] in, This indicates a high threshold for uterine contraction detection. This indicates a low threshold for uterine contraction detection. This represents the mean of the smoothed envelope curve. The standard deviation of the smooth envelope curve is represented by... and The proportionality coefficient representing the standard deviation and This represents the coupling coefficient between the mean and standard deviation, used to achieve adaptive adjustment of the threshold. The system calculates the high threshold and low threshold for uterine contraction detection using the above two formulas, and integrates them to form an adaptive dual-threshold segmentation benchmark. This benchmark can adapt to the amplitude characteristics of uterine contraction signals of different pregnant women, avoiding detection errors caused by fixed thresholds.

[0062] S23: Based on the adaptive dual-threshold segmentation benchmark, the smooth envelope curve is threshold scanned. When the smooth envelope curve crosses the low threshold of uterine contraction detection from below, it is marked as a candidate starting point. Within a predetermined time interval after the candidate starting point, it is detected whether the smooth envelope curve can rise above the high threshold of uterine contraction detection. If it can, the candidate starting point is confirmed as a valid uterine contraction starting point and the local maximum value is located as the peak point. When the smooth envelope curve crosses the low threshold of uterine contraction detection from above, it is marked as an ending point. If the smooth envelope curve fails to exceed the high threshold of uterine contraction detection after the candidate starting point, it is judged as noise interference and discarded. Adjacent uterine contraction events with a starting point interval less than a predetermined merging threshold are merged to generate a uterine contraction event sequence containing the starting point, peak point, and ending point information of each uterine contraction event.

[0063] Specifically, the system acquires an adaptive dual-threshold segmentation benchmark and performs a threshold scanning operation on the smooth envelope curve based on this benchmark. During the threshold scanning process, the system monitors the amplitude changes of the smooth envelope curve in real time. When the smooth envelope curve crosses the low threshold of uterine contraction detection from below, the system marks this moment as a candidate starting point. After marking the candidate starting point, the system continuously monitors the amplitude changes of the smooth envelope curve within a predetermined time interval after the candidate starting point to determine whether it can rise above the high threshold of uterine contraction detection. If the smooth envelope curve can exceed the high threshold of uterine contraction detection, the system confirms the candidate starting point as a valid uterine contraction starting point and continues to locate the local maximum value in the signal segment after the valid uterine contraction starting point, marking the moment corresponding to the local maximum value as the peak point.

[0064] When the smooth envelope curve crosses below the low threshold of contraction detection from above, the system marks this moment as the end point. If the smooth envelope curve fails to exceed the high threshold of contraction detection within a predetermined time interval after the candidate start point, the system determines that the signal change corresponding to that candidate start point is noise interference and discards it. The system filters all detected valid contraction events, calculates the start point interval between adjacent valid contraction events, and uses a merging determination formula:

[0065]

[0066] in, This indicates the merge decision value. and These represent the start times of two adjacent effective uterine contraction events, respectively. and These represent the durations of two consecutive effective uterine contractions. This represents a weighting coefficient indicating the duration. When the merging determination value is less than a predetermined merging threshold, the system merges the corresponding adjacent contraction events. After completing all the above operations, the system generates a contraction event sequence, which contains the start point, peak point, and end point information for each valid contraction event.

[0067] In one embodiment, S3 of the automatic assessment method for uterine contraction pressure in pregnant women based on acoustic signal characteristics provided by the present invention specifically includes the following steps:

[0068] S31: Using the starting point of the first contraction event in the contraction event sequence as the zero point of the time axis, and using the preset duration as the sliding window length and the preset step size as the window movement step size, multiple partially overlapping sliding windows are constructed sequentially along the time axis to generate a sliding window partitioning sequence.

[0069] Specifically, the system acquires a sequence of uterine contraction events, extracts the starting point of the first contraction event from the sequence, and sets this starting point as the zero point of the entire timeline. A unified time coordinate system is established based on this zero point to ensure consistency in the time base of all subsequent operations. The system calls the preset sliding window length and window movement step size, and constructs sliding windows sequentially along the timeline based on this time coordinate system. The construction of the sliding window uses the following boundary calculation formula:

[0070]

[0071] in, This represents the m-th sliding window. This indicates the moment corresponding to the zero point of the time axis. Indicates the sequence number of the sliding window. Indicates the window movement step size. This represents the length of the sliding window. The system calculates the start and end boundaries of each sliding window sequentially according to this formula. All sliding windows are arranged continuously along the time axis with some overlap; the degree of overlap is determined by the relationship between the window length and the movement step size. After completing the boundary calculation and construction of all sliding windows, the system arranges all sliding windows sequentially by their serial numbers, generating a sliding window partitioning sequence. This sequence clearly defines the serial number, start boundary, and end boundary information of each sliding window.

[0072] S32: Based on the sliding window partitioning sequence, the starting point time coordinates of each event in the uterine contraction event sequence are compared one by one with the time boundaries of each window in the sliding window partitioning sequence. The cumulative number of uterine contraction event starting points covered within the time boundary of each window is counted. The starting point count value sequence of each window is generated with the starting point falling between the starting and ending boundaries of the window as the counting condition.

[0073] Specifically, the system acquires the sliding window segmentation sequence and the uterine contraction event sequence, and correlates the two sequences with time coordinates to ensure that their time bases are consistent. The system extracts the start time coordinate of each uterine contraction event in the uterine contraction event sequence, and simultaneously extracts the time coordinates corresponding to the start and end boundaries of each sliding window in the sliding window segmentation sequence. For each sliding window, the system compares the start time coordinates of all uterine contraction events with the start and end boundaries of that window one by one to determine whether the start point of each uterine contraction event falls within the time boundary range of that window. The counting condition is set as follows: the time coordinate of the start point of the uterine contraction event is greater than or equal to the start boundary time coordinate of the window, and less than or equal to the end boundary time coordinate of the window. Uterine contraction event start points that meet this condition are counted in the cumulative count of the corresponding window. Preferably, the system uses the following formula to count the start points of uterine contractions:

[0074]

[0075] in, This represents the starting point count value of the m-th sliding window. This represents the total number of contractions in the contraction event sequence. Represents the time coordinate of the starting point of the i-th uterine contraction event. This represents the starting boundary time coordinate of the m-th sliding window. This represents the end boundary time coordinate of the m-th sliding window. This represents a unit step function, outputting 1 when the input value is greater than or equal to 0, and 0 otherwise. The system performs the above comparison and statistical operations sequentially on each sliding window, arranging the starting point count values ​​of all sliding windows according to their window numbers, and generating a sequence of starting point count values ​​for each window.

[0076] S33: Based on the sequence of starting point count values ​​for each window, convert the starting point count value of each window into the frequency of uterine contractions within the corresponding window time interval. With the end time of the window as the horizontal axis anchor point and the converted frequency value as the vertical axis value, construct a frequency-time mapping relationship to generate a frequency time series.

[0077] Specifically, the system obtains the sequence of starting point count values ​​for each window. For each sliding window, it converts the corresponding starting point count value into the frequency of contractions within the time interval of that window. The conversion formula is as follows:

[0078]

[0079] in, This represents the frequency of uterine contractions within the m-th sliding window time interval. This represents the starting point count value of the m-th sliding window. This represents a frequency conversion factor, used to convert count values ​​into frequency units that conform to clinical statistical standards. This represents the length of the time interval for the m-th sliding window. After the system completes the frequency conversion for each sliding window, it extracts the end time point of each sliding window, uses this end time point as the anchor point of the horizontal axis, and uses the converted contraction frequency value as the vertical axis value to establish the coordinate point corresponding to each sliding window. The system sequentially associates all coordinate points according to the sliding window number, constructing a correspondence between frequency and time, i.e., a frequency-time mapping relationship. This mapping relationship can intuitively reflect the changes in the frequency of contractions within different time intervals. The system presents this mapping relationship in sequence form, generating a frequency time series, which contains the horizontal axis value and the corresponding vertical axis value of each coordinate point.

[0080] In one embodiment, the present invention provides an automatic assessment method for uterine contraction pressure in pregnant women based on acoustic signal features. This method extracts signal segments within each contraction event window from a pure uterine contraction acoustic signal based on a contraction event sequence, performs wavelet packet decomposition on the signal segments, and extracts wavelet packet energy features of the uterine contraction-related frequency bands, including:

[0081] S41: Based on the start and end points in the uterine contraction event sequence, the signal segments corresponding to each uterine contraction event are sequentially extracted from the pure uterine contraction acoustic signal. The extraction range extends from the first preset offset before the start point to the second preset offset after the end point, generating a window signal segment sequence for each uterine contraction event.

[0082] Specifically, the system acquires a sequence of uterine contraction events and a pure acoustic signal of uterine contractions, and correlates the two with their time coordinates to ensure that the time information of the uterine contractions is consistent with the time axis of the acoustic signal. The system sequentially extracts the start and end points of each uterine contraction event from the sequence, obtaining the corresponding time coordinates. The system calls a first preset offset and a second preset offset to calculate the starting boundary of signal truncation based on the starting point time coordinate of each uterine contraction event. The starting boundary is the time corresponding to the first preset offset forward from the starting point time coordinate. Similarly, the system calculates the ending boundary of signal truncation based on the ending point time coordinate of each uterine contraction event. The ending boundary is the time corresponding to the second preset offset backward from the ending point time coordinate. The first preset offset ensures that the truncation range includes the signal transition portion before the start of the uterine contraction event, and the second preset offset ensures that the truncation range includes the signal attenuation portion after the end of the uterine contraction event. Following the calculated starting and ending boundaries, the system sequentially truncates signal segments corresponding to each uterine contraction event from the pure acoustic signal of uterine contractions, with each signal segment corresponding one-to-one with its corresponding uterine contraction event. The system arranges all captured signal segments sequentially according to the order of uterine contraction events in the sequence, generating a window signal segment sequence for each uterine contraction event.

[0083] S42: Perform multi-level wavelet packet decomposition on each window signal segment to decompose the signal into multiple sub-bands. Calculate the sum of squares of the wavelet packet coefficients in each sub-band as the energy value of the corresponding sub-band. Extract the energy values ​​of the fast waveband corresponding to the uterine contraction fast waveband and the slow waveband corresponding to the uterine contraction slow waveband to generate the energy distribution vector of each sub-band.

[0084] Specifically, the system acquires window signal segment sequences for each uterine contraction event and performs multi-level wavelet packet decomposition on each window signal segment in the sequence. Wavelet packet decomposition decomposes the window signal segment into multiple sub-bands through an iterative decomposition method. Each sub-band corresponds to a different frequency range, which can fully explore the characteristics of the signal in different frequency ranges. When performing multi-level wavelet packet decomposition, the system decomposes the window signal segment step by step according to the preset number of decomposition levels. Each level of decomposition divides the frequency band of the previous level into two sub-bands, ultimately obtaining multiple sub-bands covering different frequency ranges. After decomposition, the system extracts the wavelet packet coefficients corresponding to each sub-band and calculates the sum of squares of all wavelet packet coefficients in each sub-band. This sum of squares is the energy value of the corresponding sub-band. Based on the frequency characteristics of the uterine contraction signal, the system distinguishes between the fast wave frequency band and the slow wave frequency band of uterine contraction. It extracts the fast wave frequency band energy value corresponding to the fast wave frequency band of uterine contraction from all sub-bands, and simultaneously extracts the slow wave frequency band energy value corresponding to the slow wave frequency band of uterine contraction. The system arranges the energy values ​​of all sub-bands in order of frequency, constructs an energy distribution vector for each window signal segment, and contains the energy information of all sub-bands, fully reflecting the energy distribution characteristics of the window signal segment in different frequency ranges.

[0085] S43: Summing the energy values ​​of the fast wave band and the slow wave band, then dividing by the sum of the energy values ​​of all sub-bands, yields the relative energy proportion. This proportion is then multiplied by the logarithmic compression factor calculated based on the sum of the energy values ​​of all sub-bands to generate the wavelet packet energy characteristic index.

[0086] Specifically, the system acquires the energy distribution vector corresponding to each window signal segment, and extracts the fast wave band energy value, slow wave band energy value, and all sub-band energy values ​​from the energy distribution vector. The system first calculates the sum of the fast wave band energy value and the slow wave band energy value, then calculates the sum of all sub-band energy values. The sum of the fast wave band energy value and the slow wave band energy value is divided by the sum of all sub-band energy values ​​to obtain the relative energy proportion. The relative energy proportion reflects the proportion of uterine contraction-related band energy in the overall signal energy. The system calculates a logarithmic compression factor based on the sum of all sub-band energy values. The logarithmic compression factor is used to compress the energy values, avoiding feature distortion caused by excessively large energy values. The system multiplies the relative energy proportion by the logarithmic compression factor to generate a wavelet packet energy feature index. The calculation formula for the wavelet packet energy feature index is:

[0087]

[0088] in, The wavelet packet energy characteristic index The energy value of the sub-band corresponding to the fast wave frequency band of uterine contraction represents the intensity component of the high-frequency rhythmic contraction of uterine smooth muscle. The energy value of the sub-band corresponding to the slow wave frequency band of uterine contraction represents the intensity component of low-frequency tonic contraction of uterine smooth muscle. The sum of the energy values ​​of all sub-bands, i.e. , The sum of squares of the wavelet packet coefficients in the j-th sub-band is used to reflect the total energy level of the window signal segment. The system performs the above calculation operation for each window signal segment to obtain the wavelet packet energy characteristic index corresponding to each uterine contraction event.

[0089] In one embodiment, the present invention provides an automatic assessment method for uterine contraction pressure in pregnant women based on acoustic signal characteristics. This method involves phase space reconstruction and recursive quantization analysis of signal segments, extracting recursion rate, determinism, and recursive entropy, combining the peak amplitude of the smooth envelope curve, calculating a comprehensive uterine contraction intensity index, and outputting the uterine contraction intensity level. The method includes:

[0090] S44: Perform phase space reconstruction on the window signal segment of each uterine contraction event, calculate the optimal delay time using the mutual information method, calculate the optimal embedding dimension using the pseudo-nearest neighbor method, map the original time domain signal to a high-dimensional phase space, and generate a phase space reconstruction point set.

[0091] Specifically, the system acquires the sequence of window signal segments for each uterine contraction event and performs phase space reconstruction for each window signal segment. The core of phase space reconstruction is to map the one-dimensional time-domain signal to a high-dimensional phase space to uncover the nonlinear dynamic characteristics of the signal. The system uses the mutual information method to calculate the optimal delay time. Mutual information is used to quantify the autocorrelation of the signal at different delay times. The optimal delay time is calculated using the following formula:

[0092]

[0093] in, Indicates the optimal delay time. This indicates that the signal has a delay time of... Mutual information value at time, This represents the mutual information decay coefficient, used to balance the decreasing trend of mutual information with the reasonableness of the delay time. The system uses the pseudo-nearest neighbor method to calculate the optimal embedding dimension. The pseudo-nearest neighbor method determines the embedding dimension by judging the authenticity of neighboring points in the phase space. The optimal embedding dimension is calculated using the following formula:

[0094]

[0095] in, Indicates the optimal embedding dimension. This represents the proportion of pseudo-nearest neighbors when the embedding dimension is m. This represents the pseudo-nearest neighbor ratio threshold coefficient. After determining the optimal delay time and the optimal embedding dimension, the system treats each data point of the original time-domain signal as a component of the phase space point according to the phase space reconstruction rule. Combining the optimal delay time and the embedding dimension, the one-dimensional window signal segment is mapped to a high-dimensional phase space to generate a phase space reconstruction point set. This point set can completely preserve the nonlinear characteristics of the original signal.

[0096] S45: Based on the phase space reconstructed point set, a recursion radius threshold is set and the Euclidean distance between any two phase points is calculated. When the distance is less than the recursion radius threshold, the recursion state is marked and a symmetric recursion matrix is ​​constructed to generate a recursion graph representation of the uterine contraction signal.

[0097] Specifically, the system acquires the phase space reconstruction point set corresponding to the window signal segment, which can fully reflect the nonlinear dynamic characteristics of the uterine contraction signal. Based on the distribution characteristics of the phase space reconstruction point set, the system sets a recursion radius threshold, which is set using the following adaptive formula:

[0098]

[0099] in, For the recursive radius threshold, and All are threshold adjustment coefficients. This represents the maximum Euclidean distance between all phase points in the phase space reconstruction point set. This represents the minimum Euclidean distance between all phase points. The system traverses all phase point pairs in the phase space reconstruction point set, calculates the Euclidean distance between any two phase points, and compares each Euclidean distance with a recursion radius threshold. When the Euclidean distance between two phase points is less than the recursion radius threshold, the system marks the phase point pair as being in a recursive state; otherwise, it marks it as being in a non-recursive state. Based on the recursion state markings of all phase point pairs, the system constructs a symmetric recursion matrix. The dimension of the recursion matrix is ​​consistent with the total number of phase points in the phase space reconstruction point set, and the values ​​of the matrix elements correspond to the recursion state of the phase point pair. The system performs a visual transformation on the recursion matrix, generating a recursion graph representation of the uterine contraction signal. This recursion graph clearly presents the recursive relationships between phase points, providing data support for subsequent statistical extraction of recursive features.

[0100] S46: Based on the recursion matrix, the proportion of recursive points to the total number of points in the recursion graph representation is used as the recursion rate, the proportion of recursive points on the diagonal structure to all recursive points is used as the determinism, and the negative logarithmic weighted sum of the probabilities of the diagonal length distribution is used as the recursion entropy. The recursion rate, determinism, and recursion entropy are combined into a nonlinear feature parameter set for the uterine contraction signal.

[0101] Specifically, the system obtains a recursion matrix and performs feature statistics on the recursion graph representation of uterine contraction signals based on this matrix. The system counts the total number of all elements in the recursion matrix, which corresponds to the total number of phase point pairs in the phase space reconstruction point set. The system counts the number of elements marked as recursive states in the recursion matrix, and divides the number of recursive state elements by the total number of matrix elements to obtain the recursion rate. The formula for calculating the recursion rate is:

[0102]

[0103] in, For recursion rate, The number of elements in the recursive state. The total number of matrix elements. This is the recursion rate correction coefficient. The system counts the number of recursive points on the diagonal structure of the recursion matrix, divides this number by the total number of all recursive points, and obtains the determinism. The determinism is calculated using the following formula:

[0104]

[0105] in, For certainty, The number of recursive points on the diagonal structure. The average length of all diagonal structures. The maximum detected diagonal length. The system statistically analyzes the probability distribution of diagonal lengths in the recursive matrix, where the recursive entropy is calculated using the composite recursive entropy formula:

[0106]

[0107] in, For recursive entropy, Let l be the probability distribution of the diagonal length l in the recursive matrix. The minimum diagonal length threshold. The maximum diagonal length detected. The system integrates recursion rate, determinism, and recursive entropy to generate a set of nonlinear feature parameters for uterine contraction signals.

[0108] S47: Extract the peak amplitude of each contraction event from the smooth envelope curve based on the contraction event sequence, obtain the global maximum amplitude of the entire smooth envelope curve during the monitoring period, divide the peak amplitude of each event by the global maximum amplitude for normalization, and generate the normalized envelope peak parameter of each contraction event.

[0109] Specifically, the system acquires the contraction event sequence and the smoothed envelope curve, and correlates them with time coordinates to ensure that the time information of the contraction events is consistent with the time axis of the smoothed envelope curve. The system extracts the start and end points of each contraction event from the contraction event sequence, and locates the corresponding curve segment in the smoothed envelope curve based on this time range. Within each curve segment corresponding to a contraction event, the system extracts the maximum value of the curve; this maximum value is the peak amplitude of the corresponding contraction event. The system traverses the entire smoothed envelope curve and extracts the maximum amplitude of the smoothed envelope curve within the monitoring period; this amplitude is the global maximum amplitude. The system normalizes the peak amplitude of each contraction event using the following formula:

[0110]

[0111] in, To normalize the envelope peak parameter, The peak amplitude of a single uterine contraction event. To achieve the global maximum amplitude of the smooth envelope curve, This is a normalization correction coefficient used to avoid extreme value distortion in the normalized parameters. The system performs the above peak amplitude extraction and normalization process for each contraction event, arranges all normalized parameters in the order of contraction events, and generates a normalized envelope peak parameter sequence for each contraction event.

[0112] S48: The wavelet packet energy feature index and three recursive parameters in the nonlinear feature parameter set are used as feature inputs for intensity assessment. Combined with the normalized envelope peak parameter, a multi-feature linear weighted fusion strategy is used to calculate the comprehensive contraction intensity index of each contraction event. The comprehensive contraction intensity index is mapped to discrete contraction intensity levels according to the preset grading threshold range, and the output includes three levels of contraction intensity: weak, moderate and strong.

[0113] Specifically, the system acquires wavelet packet energy feature indices, a set of nonlinear feature parameters, and normalized envelope peak parameters, using these four types of features as feature inputs for assessing uterine contraction intensity. The system employs a multi-feature linear weighted fusion strategy to calculate the comprehensive uterine contraction intensity index for each contraction event. The fusion calculation uses the following formula:

[0114]

[0115] in, The overall intensity index of uterine contractions. The wavelet packet energy characteristic index For recursion rate, For certainty, For recursive entropy, To normalize the envelope peak parameter, , , , , These are the weighting coefficients for each feature, and the sum of all weighting coefficients is 1. The system calls a preset grading threshold range, which is used to map the continuous comprehensive contraction intensity index to discrete contraction intensity levels. The system compares the comprehensive contraction intensity index of each contraction event with the grading threshold range, and maps it to the corresponding contraction intensity level according to the range in which the index falls. Contraction intensity levels are divided into three levels: weak, moderate, and strong. After completing the level mapping for all contraction events, the system outputs results containing the corresponding intensity level for each contraction event, providing a quantitative basis for the objective assessment of contraction intensity.

[0116] In one embodiment, S5 of the automatic assessment method for uterine contraction pressure in pregnant women based on acoustic signal characteristics provided by the present invention specifically includes the following steps:

[0117] S51: With time as the horizontal axis, plot the waveform curve of the pure uterine contraction acoustic signal in the first plotting area, plot the waveform curve of the smooth envelope curve in the second plotting area, and mark the position marks of the start point, peak point and end point of each event in the contraction event sequence in the second plotting area to generate a contraction monitoring curve containing the contraction waveform and contraction event marks.

[0118] Specifically, the system acquires pure uterine contraction acoustic signals, smooth envelope curves, and uterine contraction event sequences, and processes these three data points using a unified time coordinate system to ensure consistency in the time reference of all data. The system establishes a plotting coordinate system with time as the horizontal axis and signal amplitude as the vertical axis, constructing a complete plotting space. Within this space, the system divides the plotting space into two independent plotting areas, arranged sequentially along the vertical axis, with the horizontal axes of the two areas perfectly aligned to ensure the uniformity of the time coordinates and the correspondence of the data. The system extracts all amplitude data of the pure uterine contraction acoustic signal and their corresponding time coordinates, associating the amplitude data with the time coordinates one-to-one. Based on the association results, the system plots the waveform curve of the pure uterine contraction acoustic signal in the first plotting area. The horizontal axis of the curve strictly corresponds to time, and the vertical axis strictly corresponds to the signal amplitude, fully presenting the change process of the pure uterine contraction acoustic signal over time.

[0119] Furthermore, the system extracts all amplitude data and corresponding time coordinates of the smooth envelope curve. Using the same horizontal axis reference, it plots the waveform of the smooth envelope curve in the second plotting area. The horizontal axis of the curve is consistent with that of the first plotting area, and the vertical axis corresponds to the amplitude of the smooth envelope, clearly showing the changing trend of the contraction signal amplitude. The system extracts the time coordinates and corresponding amplitudes of the start point, peak point, and end point of each contraction event from the contraction event sequence. Based on these coordinate data, it marks the location of each event in the second plotting area. Different types of marks use different representation forms to clearly distinguish the start point, peak point, and end point. After completing all curve plotting and marking, the system integrates the layout of the two plotting areas, unifies the coordinate axis labeling specifications, and generates a contraction monitoring curve chart that includes the contraction waveform and contraction event marks. This curve chart can fully present the waveform changes of the contraction signal and the key time nodes of each contraction event, providing support for intuitive observation of the contraction situation.

[0120] S52: Based on the contraction event sequence, frequency time series, and intensity level, generate quantitative parameter entries for each contraction event in chronological order. Each entry includes the occurrence time, duration, corresponding contraction frequency value, and intensity level of the event, which are then summarized into a contraction quantitative report.

[0121] Specifically, the system acquires uterine contraction event sequences, frequency time series, and intensity level data, and performs time coordinate correlation processing on these three data points to ensure accurate time correspondence and avoid parameter matching errors caused by time misalignment. The system sorts the uterine contraction event sequences chronologically, clarifying the order of occurrence of all contractions and laying the foundation for the orderly generation of quantitative parameter entries. For each contraction event, the system extracts the start time coordinate from the contraction event sequence and uses this time coordinate as the occurrence time of the contraction event. The system calculates the duration of the contraction event by subtracting the start time coordinate from the end time coordinate, thus fully reflecting the duration of each contraction event.

[0122] Furthermore, based on the occurrence time of the contraction event, the system locates the corresponding time interval in the frequency time series, extracts the contraction frequency value within that time interval, and uses this frequency value as the contraction frequency value corresponding to the contraction event, achieving precise matching between contraction events and frequency data. The system extracts the intensity level corresponding to the contraction event, integrating the four parameters—occurrence time, duration, contraction frequency value, and intensity level—to form quantitative parameter entries for the contraction event. Each entry contains only the core quantitative information of the corresponding contraction event. The system performs the above parameter extraction and entry generation operations for each contraction event, arranging all quantitative parameter entries sequentially according to the chronological order of the contraction events to ensure that the order of entries matches the order of contraction occurrence. The system summarizes and organizes all quantitative parameter entries, standardizing the entry format and parameter expression specifications, and integrates them to form a contraction quantitative report. The report fully includes the quantitative parameter information of all contraction events, clearly presenting the specific parameters of each contraction event in chronological order, providing medical staff with systematic and accurate quantitative data support for analyzing contraction conditions.

[0123] It should be understood that although the steps in the flowcharts of the embodiments described above are shown sequentially according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the flowcharts of the embodiments described above may include multiple steps or multiple stages. These steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these steps or stages is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the steps or stages of other steps.

[0124] Based on the same inventive concept, this application also provides an automatic assessment device for pregnant women's contraction pressure based on acoustic signal characteristics, for implementing the aforementioned automatic assessment method for pregnant women's contraction pressure based on acoustic signal characteristics. The solution provided by this device is similar to the implementation scheme described in the above method. Therefore, the specific limitations of one or more embodiments of the automatic assessment device for pregnant women's contraction pressure based on acoustic signal characteristics provided below can be found in the limitations of the automatic assessment method for pregnant women's contraction pressure based on acoustic signal characteristics described above, and will not be repeated here.

[0125] Preferably, such as Figure 2 As shown, this invention provides an automatic assessment system 600 for uterine contraction pressure in pregnant women based on acoustic signal characteristics. This system is configured with the following modules:

[0126] The uterine contraction acoustic signal purification module 610 is used to decouple the multi-channel acoustic vibration signals of the pregnant woman's abdomen collected by the flexible acoustic sensor array, combine the mutual information measurement and recognition between the pre-stored maternal heart sound templates, remove the heart sound interference components, and perform weighted fusion of the multi-channel signals after removing the interference components to generate a pure uterine contraction acoustic signal.

[0127] The uterine contraction event detection module 620 is used to extract the Hilbert envelope and perform adaptive dual-threshold segmentation on the pure uterine contraction acoustic signal to generate a smooth envelope curve. Based on the upward and downward crossing events of the smooth envelope curve, the module detects and marks the start point, peak point and end point of each uterine contraction event to generate a uterine contraction event sequence.

[0128] The contraction frequency time series module 630 is used to generate a frequency time series by counting the number of contraction event start points in each sliding window, using a preset duration as the sliding window length, based on the contraction event sequence and using a preset duration as the sliding window length.

[0129] The uterine contraction intensity assessment module 640 is used to extract signal segments within each uterine contraction event window from the pure acoustic signal of uterine contraction based on the uterine contraction event sequence, perform wavelet packet decomposition on the signal segments, extract the wavelet packet energy features of the frequency band related to uterine contraction, and perform phase space reconstruction and recursive quantization analysis on the signal segments to extract the recursion rate, determinism and recursion entropy. Combined with the peak amplitude of the smooth envelope curve, the module calculates the comprehensive uterine contraction intensity index and outputs the uterine contraction intensity level.

[0130] The monitoring report generation module 650 is used to generate a uterine contraction monitoring curve containing uterine contraction waveforms and uterine contraction event markers, based on pure uterine contraction acoustic signals, smooth envelope curves, uterine contraction event sequences, frequency time series, and intensity levels, as well as a uterine contraction quantitative report containing the occurrence time, duration, frequency, and intensity level of each uterine contraction event.

[0131] In one embodiment, this application also provides a computer device, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the above-described automatic assessment method for uterine contraction pressure in pregnant women based on acoustic signal characteristics.

[0132] In one embodiment, this application also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the above-described method for automatically assessing uterine contraction pressure in pregnant women based on acoustic signal characteristics.

[0133] In the description of this specification, references to terms such as "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of this application. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples. Moreover, without contradiction, those skilled in the art can combine and integrate the different embodiments or examples described in this specification, as well as the features of those different embodiments or examples.

[0134] For the device embodiments, since they basically correspond to the method embodiments, the relevant parts can be referred to in the description of the method embodiments. The device embodiments described above are merely illustrative. The components described as separate parts may or may not be physically separate, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this disclosure according to actual needs. Those skilled in the art can understand and implement this without creative effort.

[0135] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any person skilled in the art can easily conceive of various variations or substitutions within the technical scope disclosed in this application, and these should all be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. An automatic assessment method for uterine contraction pressure in pregnant women based on acoustic signal characteristics, characterized in that, Includes the following steps: S1: The multi-channel acoustic vibration signal of the pregnant woman's abdomen collected by the flexible acoustic sensor array is decoupled and combined with the mutual information measurement between the pre-stored maternal heart sound templates to remove the heart sound interference component and then the multi-channel signal after removing the interference component is weighted and fused to generate a pure uterine contraction acoustic signal. S2: Perform Hilbert envelope extraction and adaptive dual-threshold segmentation on the pure uterine contraction acoustic signal to generate a smooth envelope curve. Based on the upward and downward crossing events of the smooth envelope curve, detect and mark the start point, peak point and end point of each uterine contraction event to generate a uterine contraction event sequence. S3: Based on the uterine contraction event sequence, using a preset duration as the sliding window length, count the number of uterine contraction event start points in each sliding window, calculate the uterine contraction frequency in each sliding window, and generate a frequency time series. S4: Based on the uterine contraction event sequence, extract signal segments within each uterine contraction event window from the pure uterine contraction acoustic signal, perform wavelet packet decomposition on the signal segments, extract wavelet packet energy features of the uterine contraction-related frequency bands, and perform phase space reconstruction and recursive quantization analysis on the signal segments to extract recursion rate, determinism, and recursion entropy. Combined with the peak amplitude of the smooth envelope curve, calculate the comprehensive uterine contraction intensity index and output the uterine contraction intensity level. S5: Based on the pure uterine contraction acoustic signal, the smooth envelope curve, the contraction event sequence, the frequency time series, and the intensity level, generate a contraction monitoring curve including the contraction waveform and contraction event markers, and a contraction quantitative report including the occurrence time, duration, frequency, and intensity level of each contraction event.

2. The method according to claim 1, characterized in that, S1 includes: S11: Perform empirical mode decomposition on each single-channel signal in the multi-channel acoustic vibration signal of the pregnant woman's abdomen collected by the flexible acoustic sensor array, decompose it into multiple intrinsic mode function components and residuals, calculate the normalized mutual information between each intrinsic mode function component and the maternal heart sound template in combination with the pre-stored maternal heart sound template, mark the intrinsic mode function components with normalized mutual information greater than the preset interference judgment threshold as the dominant heart sound components, and generate the heart sound interference identification results for each channel; S12: Based on the heart sound interference identification result, remove the dominant heart sound component from the original signal of each channel and reconstruct the remaining intrinsic mode function components and residuals to generate the interference-free signal of each channel after removing the heart sound interference. S13: Perform signal-to-noise ratio estimation on the de-interference signal of each channel, calculate the estimated signal-to-noise ratio of each channel and normalize it into a weighted fusion weight, and perform linear weighted summation on the de-interference signal of each channel according to the weighted fusion weight to generate a pure uterine contraction acoustic signal.

3. The method according to claim 1, characterized in that, S2 includes: S21: Perform Hilbert transform on the pure uterine contraction acoustic signal, calculate the instantaneous envelope of the pure uterine contraction acoustic signal by combining the original signal and the transformed orthogonal components, and perform smoothing filtering on the instantaneous envelope to generate a smooth envelope curve; S22: Calculate the mean and standard deviation of the smooth envelope curve based on the statistical distribution characteristics of the smooth envelope curve, and set a high threshold and a low threshold for uterine contraction detection according to the linear combination of the mean and the standard deviation to generate an adaptive dual-threshold segmentation benchmark. S23: Based on the adaptive dual-threshold segmentation benchmark, the smooth envelope curve is subjected to threshold scanning. When the smooth envelope curve crosses the low threshold of uterine contraction detection from below, it is marked as a candidate starting point. After the candidate starting point, within a predetermined time interval, it is detected whether the smooth envelope curve can rise above the high threshold of uterine contraction detection. If it can, the candidate starting point is confirmed as a valid uterine contraction starting point, and the local maximum value is further located as the peak point. When the smooth envelope curve crosses the low threshold of uterine contraction detection from above, it is marked as an ending point. If the smooth envelope curve fails to exceed the high threshold of uterine contraction detection after the candidate starting point, it is determined to be noise interference and discarded. Adjacent uterine contraction events with a starting point interval less than a predetermined merging threshold are merged to generate a uterine contraction event sequence containing the starting point, peak point, and ending point information of each uterine contraction event.

4. The method according to claim 1, characterized in that, S3 includes: S31: Taking the starting point of the first contraction event in the contraction event sequence as the zero point of the time axis, using the preset duration as the sliding window length and the preset step size as the window movement step size, multiple partially overlapping sliding windows are constructed sequentially along the time axis to generate a sliding window partitioning sequence. S32: Based on the sliding window partitioning sequence, the starting point time coordinates of each event in the contraction event sequence are compared one by one with the time boundaries of each window in the sliding window partitioning sequence. The cumulative number of contraction event starting points covered within the time boundary of each window is counted. The starting point count value sequence of each window is generated with the starting point falling between the starting and ending boundaries of the window as the counting condition. S33: Based on the sequence of starting point count values ​​for each window, convert the starting point count value of each window into the frequency of uterine contractions within the corresponding window time interval. With the end time of the window as the horizontal axis anchor point and the converted frequency value as the vertical axis value, construct a frequency-time mapping relationship to generate a frequency time series.

5. The method according to claim 1, characterized in that, The step involves extracting signal segments within each contraction event window from the pure uterine contraction acoustic signal based on the contraction event sequence, performing wavelet packet decomposition on the signal segments, and extracting wavelet packet energy features of the uterine contraction-related frequency bands, including: S41: Based on the start and end points in the uterine contraction event sequence, sequentially extract the signal segments corresponding to each uterine contraction event from the pure uterine contraction acoustic signal. The extraction range extends from a first preset offset before the start point to a second preset offset after the end point, generating a window signal segment sequence for each uterine contraction event. S42: Perform multi-level wavelet packet decomposition on each of the window signal segments to decompose the signal into multiple sub-bands. Calculate the sum of squares of the wavelet packet coefficients in each sub-band as the energy value of the corresponding sub-band. Extract the fast waveband energy value corresponding to the uterine contraction fast waveband and the slow waveband energy value corresponding to the uterine contraction slow waveband to generate the energy distribution vector of each sub-band. S43: The relative energy ratio is obtained by summing the fast waveband energy value and the slow waveband energy value and dividing by the sum of all sub-band energy values. This ratio is then multiplied by a logarithmic compression factor calculated based on the sum of all sub-band energy values ​​to generate a wavelet packet energy characteristic index. The calculation formula for the wavelet packet energy characteristic index is as follows: in, The wavelet packet energy characteristic index The energy value of the sub-band corresponding to the fast wave frequency band of uterine contraction represents the intensity component of the high-frequency rhythmic contraction of uterine smooth muscle. The energy value of the sub-band corresponding to the slow wave frequency band of uterine contraction represents the intensity component of low-frequency tonic contraction of uterine smooth muscle. The sum of the energy values ​​of all sub-bands, i.e. , Let be the sum of squares of the wavelet packet coefficients of the j-th sub-band.

6. The method according to claim 5, characterized in that, The process of performing phase space reconstruction and recursive quantization analysis on the signal segment, extracting the recursion rate, determinism, and recursion entropy, and calculating the comprehensive uterine contraction intensity index by combining the peak amplitude of the smooth envelope curve, and outputting the uterine contraction intensity level includes: S44: Reconstruct the phase space for the window signal segment of each uterine contraction event, calculate the optimal delay time using the mutual information method, calculate the optimal embedding dimension using the pseudo-nearest neighbor method, map the original time domain signal to a high-dimensional phase space, and generate a phase space reconstruction point set; S45: Based on the reconstructed point set in the phase space, set a recursion radius threshold and calculate the Euclidean distance between any two phase points. When the distance is less than the recursion radius threshold, mark the recursion state and construct a symmetric recursion matrix to generate a recursion graph representation of the uterine contraction signal. S46: Based on the recursion matrix, the proportion of recursive points to the total number of points in the recursion graph representation is used as the recursion rate; the proportion of recursive points on the diagonal structure to all recursive points is used as the determinism; and the negative logarithmic weighted sum of the probabilities of the diagonal length distribution is used as the recursion entropy. The recursion rate, the determinism, and the recursion entropy are combined into a nonlinear feature parameter set for the uterine contraction signal, wherein the recursion entropy is calculated using the composite recursion entropy formula: in, For recursive entropy, Let l be the probability distribution of the diagonal length l in the recursive matrix. The minimum diagonal length threshold. The maximum diagonal length detected; S47: Based on the uterine contraction event sequence, extract the peak amplitude of each uterine contraction event from the smooth envelope curve, obtain the global maximum amplitude of the entire smooth envelope curve during the monitoring period, divide the peak amplitude of each event by the global maximum amplitude for normalization processing, and generate the normalized envelope peak parameter of each uterine contraction event. S48: The wavelet packet energy feature index and the three recursive parameters in the nonlinear feature parameter set are used as feature inputs for intensity evaluation. Combined with the normalized envelope peak parameter, a multi-feature linear weighted fusion strategy is used to calculate the comprehensive contraction intensity index of each contraction event. The comprehensive contraction intensity index is mapped to discrete contraction intensity levels according to a preset grading threshold range, and the contraction intensity levels containing three levels are output: weak, moderate, and strong.

7. The method according to any one of claims 1-6, characterized in that, S5 includes: S51: With time as the horizontal axis, plot the waveform curve of the pure uterine contraction acoustic signal in the first plotting area, plot the waveform curve of the smooth envelope curve in the second plotting area, and mark the position marks of the start point, peak point and end point of each event in the uterine contraction event sequence in the second plotting area to generate a uterine contraction monitoring curve containing the uterine contraction waveform and uterine contraction event marks. S52: Based on the contraction event sequence, the frequency time series, and the intensity level, generate quantitative parameter entries for each contraction event in chronological order. Each entry includes the occurrence time, duration, corresponding contraction frequency value, and intensity level of the event, and summarizes them into a contraction quantification report.

8. An automatic assessment system for uterine contraction pressure in pregnant women based on acoustic signal characteristics, characterized in that, The system includes: The uterine contraction acoustic signal purification module is used to decouple the multi-channel acoustic vibration signals of the pregnant woman's abdomen collected by a flexible acoustic sensor array, combine the mutual information measurement and recognition between pre-stored maternal heart sound templates, remove the heart sound interference components, and perform weighted fusion of the multi-channel signals after removing the interference components to generate a pure uterine contraction acoustic signal. The uterine contraction event detection module is used to extract the Hilbert envelope and perform adaptive dual-threshold segmentation on the pure uterine contraction acoustic signal to generate a smooth envelope curve. Based on the upward and downward crossing events of the smooth envelope curve, the module detects and marks the start point, peak point and end point of each uterine contraction event to generate a uterine contraction event sequence. The contraction frequency time series module is used to calculate the contraction frequency within each sliding window based on the contraction event sequence, using a preset duration as the sliding window length, to count the number of contraction event start points within each sliding window, and generate a frequency time series. The uterine contraction intensity assessment module is used to extract signal segments within each uterine contraction event window from the pure uterine contraction acoustic signal based on the uterine contraction event sequence, perform wavelet packet decomposition on the signal segments, extract wavelet packet energy features of the uterine contraction-related frequency bands, and perform phase space reconstruction and recursive quantization analysis on the signal segments to extract recursion rate, determinism, and recursion entropy. Combined with the peak amplitude of the smooth envelope curve, the module calculates the comprehensive uterine contraction intensity index and outputs the uterine contraction intensity level. The monitoring report generation module is used to generate a uterine contraction monitoring curve graph containing uterine contraction waveforms and uterine contraction event markers, and a uterine contraction quantitative report containing the occurrence time, duration, frequency and intensity level of each uterine contraction event, based on the pure uterine contraction acoustic signal, the smooth envelope curve, the uterine contraction event sequence, the frequency time series and the intensity level.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the method of any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the method of any one of claims 1 to 7.

Images