Psychological nursing decision system for elderly chronic obstructive pulmonary disease patients based on emotion recognition

Abstract

The present application relates to the field of health care technology, in particular to a kind of psychological nursing decision system for old chronic obstructive pulmonary disease patient based on emotion recognition.The system includes obtaining patient daily voice signal and synchronous blood oxygen saturation, extracting sound characteristic value and calculating its sound abnormality degree;combining blood oxygen change and sound abnormality degree, assessing the blood oxygen fluctuation stability degree of each sampling time, generating initial pathological interference degree;by analyzing the correlation between different sound characteristic value initial pathological interference degree, correcting initial pathological interference degree, obtaining target pathological interference degree;based on the target pathological interference degree of all sound characteristic values at the same sampling time, screening out the analysis time with small pathological interference, using the voice signal of analysis time for accurate emotion recognition and individualized psychological nursing decision, effectively overcome the interference of chronic obstructive pulmonary disease physiological symptoms on emotion recognition, improve the accuracy of emotion state judgment for chronic obstructive pulmonary disease patients.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of health care, in particular to a psychological nursing decision system for elderly patients with chronic obstructive pulmonary disease based on emotion recognition. BACKGROUND

[0002] Chronic obstructive pulmonary disease (COPD) is a common chronic respiratory disease characterized by persistent airflow limitation. Negative emotions can reduce the patient's compliance with medication and rehabilitation training, forming a vicious cycle of "dyspnea-anxiety-symptom exacerbation", so psychological nursing for elderly patients with COPD is crucial.

[0003] Generally, the patient's current emotion is analyzed by recognizing the patient's speech emotion, pulse and other physiological characteristics, and psychological nursing is carried out based on the emotion. In this way, the physiological changes in speech caused by COPD are highly overlapped with the acoustic features related to emotion, which can easily lead to algorithm misjudgment. For example, due to decreased lung function, patients often "steal air" when speaking, resulting in an involuntary decrease in speech rate and frequent pauses and inhalation sounds interrupting the sentence, which is almost identical to the manifestations of depression and low mood (slow speech and frequent pauses). The algorithm is prone to misjudging such physiological manifestations of dyspnea as low mood. At the same time, due to airflow limitation, the patient's voice is often low, hoarse and weak, which is highly consistent with the speech characteristics in a state of depression or fatigue (low pitch and weak volume), further reducing the accuracy of emotion recognition. That is, the conventional emotion recognition mechanism for emotion recognition of COPD patients will be interfered by the pathology, and the accuracy is low. SUMMARY

[0004] In order to solve the technical problem that the conventional emotion recognition mechanism for emotion recognition of COPD patients will be interfered by the pathology, and the accuracy is low in the related art, the present application provides a psychological nursing decision system for elderly patients with chronic obstructive pulmonary disease based on emotion recognition, and the technical solution adopted is as follows:

[0005] The present application provides a psychological nursing decision system for elderly patients with chronic obstructive pulmonary disease based on emotion recognition, which comprises:

[0006] The acquisition module is used to acquire the speech signal of the patient's daily speech and the blood oxygen saturation at different sampling times within the same speech time range, and to extract different sound feature values from the speech signal; based on the standard deviation of each sampling time sound feature value, the sound abnormality degree of each sound feature value is determined;

[0007] The interference analysis module is used to determine the blood oxygen fluctuation stability degree of each sampling time according to the numerical change of the blood oxygen saturation at different sampling times; and to determine the initial pathological interference degree of the sound feature value affected by COPD by combining the blood oxygen saturation, the sound abnormality degree and the blood oxygen fluctuation stability degree at each sampling time.

[0008] The correction analysis module is configured to determine initial correlation of the two sound characteristic values by combining initial pathological interference degrees of the two sound characteristic values at different sampling moments; and correct the initial pathological interference degrees according to the initial correlation between the sound characteristic values to obtain target pathological interference degrees.

[0009] The decision module is configured to screen out unaffected analysis moments according to the target pathological interference degrees of all sound characteristic values at the same sampling moment, and make psychological care decisions based on voice signals of the analysis moments.

[0010] Further, the sound characteristic values include speech speed, tone and volume.

[0011] Further, the determination of the sound abnormality degree of each sound characteristic value based on the standard deviation of the sound characteristic value at each sampling moment comprises:

[0012] The absolute value of the difference between each sound characteristic value and a preset standard value is calculated as the sound abnormality degree of the corresponding sound characteristic value.

[0013] Further, the determination of the blood oxygen fluctuation stability degree of each sampling moment according to the numerical change of the blood oxygen saturation at different sampling moments comprises:

[0014] Any sampling moment is taken as a target moment, and the nearest preset number of sampling moments from the target moment are taken as adjacent sampling moments of the target moment.

[0015] The standard deviation of the blood oxygen saturation of the target moment and the adjacent sampling moments is calculated, and the inverse of the standard deviation is normalized and taken as the blood oxygen fluctuation stability degree of the target moment.

[0016] Further, the determination of the initial pathological interference degree of the sound characteristic value affected by the chronic obstructive pulmonary disease by combining the blood oxygen saturation, the sound abnormality degree and the blood oxygen fluctuation stability degree of each sampling moment comprises:

[0017] The inverse of the blood oxygen saturation is normalized and taken as a blood oxygen influence index;

[0018] The sound abnormality degree is normalized and taken as a sound influence index;

[0019] The weighted sum value of the blood oxygen influence index and the sound influence index is calculated, and the product value of the weighted sum value and the blood oxygen fluctuation stability degree is normalized as the initial pathological interference degree.

[0020] Further, the determination of the initial correlation of the two sound characteristic values by combining the initial pathological interference degrees of the two sound characteristic values at different sampling moments comprises:

[0021] According to the sampling time, the initial pathological interference degrees of each sound characteristic value are sorted to obtain an interference sequence;

[0022] Based on the Pearson correlation algorithm, the Pearson correlation coefficients of the interference sequences of any two sound characteristic values are calculated, and the absolute values of the Pearson correlation coefficients are taken as initial correlations.

[0023] Further, the initial pathological interference degrees are corrected according to the initial correlations between the sound characteristic values to obtain target pathological interference degrees, including:

[0024] The mean value of the initial correlations of any sound characteristic value with all other sound characteristic values is calculated as an interference correction coefficient;

[0025] The product value of the interference correction coefficient and the initial pathological interference degree is calculated, and the sum value of the product value and the initial pathological interference degree is normalized as the target pathological interference degree.

[0026] Further, the unaffected analysis time is screened according to the target pathological interference degrees of all sound characteristic values at the same sampling time, including:

[0027] The slow lung influence coefficient of the sampling time is determined according to the target pathological interference degrees of all sound characteristic values at the same sampling time.

[0028] The sampling time with the slow lung influence coefficient less than a preset influence threshold is taken as the analysis time.

[0029] Further, the slow lung influence coefficient of the sampling time is determined according to the target pathological interference degrees of all sound characteristic values at the same sampling time, including:

[0030] The target pathological interference degrees of all sound characteristic values at the same sampling time are weighted and averaged to obtain the slow lung influence coefficient.

[0031] Further, the psychological care decision is made based on the speech signals at the analysis time, including:

[0032] The speech signals at all analysis times are grouped into different speech segments according to the time sequence.

[0033] The speech segments are input into a pre-trained emotion recognition model to realize emotion state analysis and obtain target emotion analysis results; and psychological care decisions are made based on the target emotion analysis results.

[0034] The present application has the following advantages:

[0035] The embodiment of the present application effectively solves the problem of emotional misjudgment caused by pathological interference when the existing voice emotion recognition technology is applied to patients with chronic obstructive pulmonary disease. Specifically, the voice signal and the synchronous blood oxygen saturation data of the patient are obtained, and various sound characteristic values are extracted from the voice signal and the synchronous blood oxygen saturation data; then the blood oxygen fluctuation stability degree of each sampling moment is evaluated combined with the blood oxygen saturation change, and the initial pathological interference degree of each sound characteristic value is calculated; further, the initial pathological interference degree is corrected by analyzing the initial correlation between different sound characteristic values, and a more accurate target pathological interference degree is obtained; finally, the analysis moment with the smallest pathological interference is selected, and psychological nursing decision is made based on the voice signal of these moments. This technical means can effectively distinguish the physiological voice changes caused by chronic obstructive pulmonary disease (such as slow speech speed, low volume, etc.) from the acoustic characteristics corresponding to the true emotional state, significantly reduce the emotional recognition error caused by the high overlap between disease symptoms and emotional performance, thereby improving the accuracy of judging the emotional state of patients with chronic obstructive pulmonary disease, and laying a reliable foundation for subsequent accurate and personalized psychological nursing intervention. BRIEF DESCRIPTION OF DRAWINGS

[0036] In order to more clearly illustrate the technical solutions and advantages of the embodiments of the present application or the prior art, the following will briefly introduce the drawings needed to be used in the embodiment or the prior art description. Obviously, the drawings in the following description are only some embodiments of the present application, and those skilled in the art can obtain other drawings according to these drawings without creative labor.

[0037] Figure 1 A structure diagram of a psychological nursing decision system for elderly patients with chronic obstructive pulmonary disease based on emotion recognition is provided for an embodiment of the present application. DETAILED DESCRIPTION

[0038] In order to further illustrate the technical means and effects adopted by the present application to achieve the predetermined object, the following will combine the drawings and the preferred embodiments to specifically describe the specific implementation, structure, features and effects of a psychological nursing decision system for elderly patients with chronic obstructive pulmonary disease based on emotion recognition according to the present application. In the following description, different "one embodiment" or "another embodiment" do not necessarily refer to the same embodiment. In addition, the specific features, structures or characteristics in one or more embodiments can be combined in any suitable form.

[0039] 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 the present application belongs.

[0040] The specific scheme of a psychological nursing decision system for elderly patients with chronic obstructive pulmonary disease based on emotion recognition provided by the present application will be specifically described below with reference to the drawings.

[0041] Please refer to Figure 1 It shows a structure diagram of a psychological nursing decision system for elderly patients with chronic obstructive pulmonary disease based on emotion recognition provided by an embodiment of the application, and the system comprises:

[0042] The acquisition module 101 is configured to acquire a voice signal of daily voice of a patient and blood oxygen saturation at different sampling time ranges of the voice time range, extract different sound feature values from the voice signal, and determine a sound abnormality degree of each sound feature value based on a standard deviation of the sound feature value at each sampling time.

[0043] The embodiment of the application is applicable to elderly patients with chronic obstructive pulmonary disease in a home rehabilitation period, and is used for daily psychological state monitoring in a non-acute attack period.

[0044] The voice signal of the patient can be collected based on a portable device carried by the patient with chronic obstructive pulmonary disease or a fixed collection device.

[0045] After the voice signal is collected, the system is preprocessed (such as noise reduction, frame division, etc.), and an initial sound feature set is extracted, including:

[0046] 1. Speech rate: The number of pronunciation phonemes or the number of words in a unit of time is counted, and the formula is "speech rate = total number of phonemes / speech effective time length".

[0047] 2. Pitch: The base frequency sequence can be directly output by the librosa.pyin() function through the autocorrelation method or the YIN algorithm.

[0048] 3. Volume: The short-time energy corresponding to the voice signal is calculated, the square sum or amplitude of each frame signal is calculated after the voice is divided into frames, and the mean and peak values of the short-time energy are the volume-related features.

[0049] Of course, in other embodiments of the application, other sound features such as pause frequency can also be extracted, and the actual detection requirements are adjusted, and the limitation is not made.

[0050] The extracted original features are standardized (normalized to the [0, 1] interval) to remove the dimension influence, and the sound feature values are obtained.

[0051] In one embodiment of the present application, the normalization processing can be specifically, for example, maximum-minimum value normalization processing, and the normalization in the subsequent steps can all adopt maximum-minimum value normalization processing, and in other embodiments of the present application, other normalization methods can be selected according to the specific range of values, which will not be described here.

[0052] It should be noted that, in order to facilitate operation, all index data involved in the operation in the embodiments of the present application are subjected to data preprocessing, thereby canceling the dimensional influence. The means for canceling the dimensional influence are well-known technical means to those skilled in the art, which are not limited here.

[0053] Synchronously, the blood oxygen saturation is collected through a medical-grade wearable device (such as a smart bracelet), and the system accurately aligns and binds the sound characteristic value and the blood oxygen saturation data at the same sampling time through a time stamp to form a data sample for subsequent analysis.

[0054] The collection of the voice signal and the blood oxygen saturation in the embodiments of the present application is authorized by the relevant user, the collection process does not violate the relevant laws and regulations, and does not violate the public order and good customs.

[0055] Since different sound characteristic values have different situations when analyzing the influence on chronic obstructive pulmonary disease, in order to unify the analysis, in the embodiments of the present application, a standard value of each sound characteristic value under standard conditions is preset, and then standard difference analysis is realized, so as to obtain the sound abnormality degree of each sound characteristic value at each sampling time. The sound abnormality degree represents the abnormality level of the sound itself.

[0056] Further, in some embodiments of the present application, based on the standard difference of the sound characteristic value at each sampling time, the sound abnormality degree of each sound characteristic value is determined, including: calculating the absolute value of the difference between each sound characteristic value and the preset standard value as the sound abnormality degree of the corresponding sound characteristic value.

[0057] The preset standard value is the numerical value of each sound characteristic value under normal conditions. Since the sound characteristics of each patient are inconsistent, the patient's voice can be analyzed under normal conditions in advance to determine the preset standard value of the patient's sound characteristic value, or the preset standard value can be obtained according to the analysis of the group characteristics. Since the sound characteristic value is a standardized parameter, each preset standard value also needs to be subjected to the same normalization processing. For example, the preset standard value of the volume is 50 decibels, which can be set to 0.6 after the above data preprocessing.

[0058] The preset standard value is taken from the average level (baseline) of the patient's historical long-term statistical data or the universal baseline of the healthy population of the same age group. The system has a "baseline calibration" function, which automatically updates the baseline when the patient self-reports good condition and normal blood oxygen.

[0059] The preset values (such as 0.6) given in the embodiments of the present application are empirical values obtained in a typical hardware configuration and test scenario, and are intended to facilitate understanding of the present application. In actual applications, those skilled in the art can adjust, calibrate or optimize these parameters according to the specific hardware performance, scene complexity and data characteristics, which do not constitute a limitation of the present application.

[0060] It can be understood that the greater the value of the sound abnormality degree, the greater the absolute value of the difference between the sound feature value and the preset standard value, that is, the more abnormal the sound feature value itself is, such as abnormal increase or decrease in pitch. Whether it is abnormal increase or abnormal decrease, it is likely to be affected by emotion or COPD, therefore, it is necessary to analyze the interference of COPD, and the specific analysis process is described in subsequent embodiments.

[0061] The interference analysis module 102 is configured to determine the blood oxygen fluctuation stability degree of each sampling time according to the numerical change of the blood oxygen saturation at different sampling times; and determine the initial pathological interference degree of the sound feature value affected by COPD in combination with the blood oxygen saturation, the sound abnormality degree and the blood oxygen fluctuation stability degree of each sampling time.

[0062] The essence of COPD is airflow limitation and gas exchange disorder, which directly leads to a decrease in the oxygen content in the blood when producing pathological effects. Therefore, low blood oxygen saturation is an objective and quantitative indicator of the severity of COPD pathology. The lower the blood oxygen saturation at a time, the greater the influence of COPD on the patient's physiological state at that time.

[0063] However, determining the influence of COPD at each historical time based on only low blood oxygen saturation is one-sided, because it can be caused by temporary non-pathological factors, such as the patient holding his breath or the device having a transient measurement error. Therefore, the selected COPD state must be stable, rather than a sudden or abnormal transient condition. A stable low blood oxygen level reflects the patient's persistent pathological state, and the voice features (such as shortness of breath and slow speech) at this time are more likely to be caused purely by COPD itself.

[0064] Therefore, the embodiments of the present application need to analyze the persistent state, and further, in some embodiments of the present application, the blood oxygen fluctuation stability degree of each sampling time is determined according to the numerical change of the blood oxygen saturation at different sampling times, including: taking any sampling time as a target time, and taking the nearest preset number of sampling times from the target time as adjacent sampling times of the target time; calculating the standard deviation of the blood oxygen saturation of the target time and the adjacent sampling times, and normalizing the inverse of the standard deviation as the blood oxygen fluctuation stability degree of the target time.

[0065] The preset quantity is a quantity value preset for continuous analysis, and the specific sampling time can be 0.1 second, that is, the sound signal and the blood oxygen saturation are acquired every 0.1 second, and in the continuous analysis, the preset quantity can be determined as 30, that is, the nearest 30 sampling time points to the target time point are taken as the adjacent sampling time points of the target time point, and the nearest time sequence can be forward in time sequence and backward in time sequence, and the overall analysis is performed. Of course, the preset quantity value can be adaptively adjusted according to the sampling interval or the detection accuracy, and no limitation is made thereto.

[0066] The blood oxygen fluctuation stability degree represents the influence degree of the numerical fluctuation.

[0067] If the blood oxygen saturation fluctuates sharply in a short time (that is, the "gap is too large"), this can be caused by acute exacerbation, inaccurate measurement or other confounding factors, and the voice feature change at this time no longer has typical pathological representativeness, and if used for analysis, noise will be introduced, and the accuracy of subsequent processing will be reduced.

[0068] Therefore, in the embodiment of the application, the standard deviation is used to analyze the sharp fluctuation, the standard deviation of the blood oxygen saturation at the target time and the adjacent sampling time is calculated, the inverse of the standard deviation is normalized and taken as the blood oxygen fluctuation stability degree of the target time.

[0069] The greater the value of the standard deviation, the more sharply the blood oxygen saturation fluctuates in a short time, which indicates that the current state can be in a non-steady state (such as acute exacerbation, measurement noise, etc.), and the voice feature extracted at this time can not represent the typical pathological-voice correlation of the stable period of chronic obstructive pulmonary disease, so it should be excluded or weightedly reduced when constructing a specific model, that is, it does not have typical pathological representativeness of chronic obstructive pulmonary disease, and the lower the blood oxygen fluctuation stability degree.

[0070] Therefore, the inverse of the standard deviation is normalized and taken as the blood oxygen fluctuation stability degree of the target time. The normalization can be, for example, maximum-minimum value normalization, and the maximum value and the minimum value can be set according to experience, or the standard deviation values of all sampling time points can be normalized according to the actual scene, and no limitation is made thereto.

[0071] Traditionally, if all historical data is directly used to fit the relationship between the sound features and the blood oxygen saturation, the model will be disturbed by those moments of unclear COPD availability (such as moments of normal blood oxygen or moments of severe fluctuation), resulting in ambiguous proportional relationship. Now for a historical moment determined to be greater in COPD availability (i.e. low and stable blood oxygen saturation), it can be considered that the sound features recorded at that time (such as slow speech, low pitch, weak volume, etc.) are mainly caused by the clear pathological state of COPD, and the higher the value of the stability of blood oxygen fluctuation is. The stability of blood oxygen fluctuation can be used as a feature weight to realize pathological interference analysis in combination with the blood oxygen saturation and the degree of sound abnormality.

[0072] Further, in some embodiments of the present application, in combination with the blood oxygen saturation, the degree of sound abnormality and the stability of blood oxygen fluctuation at each sampling moment, the initial pathological interference degree of the sound feature value affected by COPD is determined, including: the reciprocal of the blood oxygen saturation is normalized as a blood oxygen influence index; the degree of sound abnormality is normalized as a sound influence index; the weighted sum value of the blood oxygen influence index and the sound influence index is calculated, and the product value of the weighted sum value and the stability of blood oxygen fluctuation is normalized as the initial pathological interference degree.

[0073] It can be understood that the smaller the blood oxygen saturation value is, the more likely it is to be affected by the pathological interference of COPD, and the greater the degree of sound abnormality is, the more it needs to be focused on, therefore, the reciprocal of the blood oxygen saturation is normalized as a blood oxygen influence index; the degree of sound abnormality is normalized as a sound influence index. The blood oxygen influence index and the sound influence index are two different dimension analysis parameters, in the embodiments of the present application, feature fusion is realized in the form of weighting.

[0074] Specifically, the blood oxygen influence index can be assigned a weight of 0.7, and the sound influence index can be assigned a weight of 0.3, then the product value of the blood oxygen influence index and 0.7 is calculated, and the product value of the sound influence index and 0.3 is calculated, and the sum value of the two product values is taken as the weighted sum value.

[0075] The specific numerical values (such as 0.7, 0.3, etc.) given in the embodiments of the present application are empirical values obtained under a typical hardware configuration and test scene, and are intended to facilitate understanding of the present application. In actual application, those skilled in the art can adjust, calibrate or optimize these parameters according to the specific hardware performance, scene complexity and data characteristics, which do not constitute a limitation of the present application.

[0076] In the embodiments of the present application, the product value of the weighted sum value and the stability of blood oxygen fluctuation is normalized as the initial pathological interference degree, and the initial pathological interference degree is given to reflect the influence degree of COPD on the sound features.

[0077] The correction analysis module 103 is configured to determine initial correlation of the two sound characteristic values according to initial pathological interference degrees of the two sound characteristic values at different sampling time points; and correct the initial pathological interference degrees according to the initial correlation between the sound characteristic values to obtain target pathological interference degrees.

[0078] The initial pathological interference degrees obtained by the analysis above are obtained according to a single sound characteristic value, and the different sound characteristic values also have correlation. If the analysis is performed without considering the correlation, the deviation of the overall result will be large. Therefore, the influence of the correlation of different sound characteristic values needs to be analyzed.

[0079] The different sound characteristic values have certain correlation. The higher the correlation is, the greater the probability that the change of the voice characteristic is caused by the slow lung pathology is, and the less the probability that the change of the voice characteristic is caused by other accidental factors (such as emotional fluctuation and environmental noise) is. Therefore, it can be preliminarily determined that the slow lung has high correlation.

[0080] Further, in some embodiments of the present application, the initial correlation of the two sound characteristic values is determined according to the initial pathological interference degrees of the two sound characteristic values at different sampling time points, including: sorting the initial pathological interference degrees of each sound characteristic value according to the sampling time points to obtain an interference sequence; calculating a Pearson correlation coefficient of the interference sequences of any two sound characteristic values based on a Pearson correlation algorithm, and taking an absolute value of the Pearson correlation coefficient as the initial correlation.

[0081] It should be noted that the Pearson correlation coefficient is a calculation index known to those skilled in the art, and the value of the Pearson correlation coefficient ranges from -1 to 1. The closer to 0, the lower the correlation is. The closer to -1, the more negative the correlation is. The closer to 1, the more positive the correlation is. In the embodiments of the present application, both positive and negative correlations represent correlation. Therefore, the absolute value of the Pearson correlation coefficient is taken as the initial correlation.

[0082] After the initial pathological interference degrees of any two sound characteristic values are sorted and the initial correlation thereof is determined based on the Pearson correlation algorithm, in order to further improve the accuracy of the pathological interference evaluation, the initial pathological interference degrees of the sound characteristic values need to be corrected by using the initial correlation. This is because different sound characteristic values may exhibit a cooperative or coupled interference mode under the influence of the slow lung. If only the initial interference degree of a single characteristic is used for judgment, it is still difficult to effectively separate the common pathological influence and the real emotional signal. By introducing the correlation information between the characteristics, the interference components dominated by the common pathological factors can be identified, so that the initial pathological interference degrees are weighted and corrected, so that the degrees of the characteristics affected by the disease are more truly reflected.

[0083] The correction process helps to more accurately distinguish the physiological voice changes caused by chronic obstructive pulmonary disease from the emotion-related acoustic features, provides a reliable basis for subsequent screening of analysis moments not affected by pathological interference, and thus significantly improves the emotion recognition accuracy and intervention effectiveness of psychological care decision-making.

[0084] Further, in some embodiments of the present application, the initial pathological interference degree is corrected according to the initial correlation between the sound feature values to obtain a target pathological interference degree, including: calculating the mean value of the initial correlation of any sound feature value with all other sound feature values as an interference correction coefficient; calculating the product value of the interference correction coefficient and the initial pathological interference degree, and normalizing the sum value of the product value and the initial pathological interference degree as the target pathological interference degree.

[0085] For any sound feature value, first, the mean value of the initial correlation of the sound feature value with all other sound feature values is calculated, and the value obtained by the mean value is defined as the interference correction coefficient of the sound feature value, which represents the degree of influence of common pathological factors.

[0086] Subsequently, the interference correction coefficient is multiplied by the initial pathological interference degree of the corresponding sound feature value to obtain a product term reflecting the cooperative interference intensity; the product term is added to the original initial pathological interference degree to form a corrected interference degree sum; finally, the sum is normalized by maximum and minimum to obtain the final target pathological interference degree.

[0087] By introducing the weighting correction mechanism driven by the correlation between features, the target pathological interference degree can more accurately reflect the true interference level of each sound feature value under the pathological action of chronic obstructive pulmonary disease, providing a highly reliable basis for subsequent screening of speech analysis moments not affected by pathological interference.

[0088] The decision module 104 is configured to screen the unaffected analysis moment according to the target pathological interference degree of all sound feature values at the same sampling moment, and make psychological care decisions based on the speech signal of the analysis moment.

[0089] To further ensure that the speech data relied on by emotion recognition is not disturbed by disease physiological factors, the overall degree of interference by chronic obstructive pulmonary disease at the sampling moment is comprehensively evaluated, and the truly "clean" analysis moment is selected accordingly.

[0090] Further, in some embodiments of the present application, the target pathological interference degree of all sound feature values at the same sampling moment is used to screen the unaffected analysis moment, including: determining a chronic obstructive pulmonary disease influence coefficient of the sampling moment according to the target pathological interference degree of all sound feature values at the same sampling moment; and taking the sampling moment with a chronic obstructive pulmonary disease influence coefficient less than a preset influence threshold as the analysis moment.

[0091] After obtaining the target pathological interference degrees of all sound feature values at each sampling moment, the degree of the whole being affected by the chronic obstructive pulmonary disease at each sampling moment is further evaluated based on the target pathological interference degrees. Specifically, the system aggregates (for example, takes the mean, weighted addition, or maximum value, etc.) the target pathological interference degrees corresponding to all sound feature values at the same sampling moment, thereby calculating the chronic obstructive pulmonary disease influence coefficient at the sampling moment.

[0092] Further, in some embodiments of the present application, the chronic obstructive pulmonary disease influence coefficient at the sampling moment is determined according to the target pathological interference degrees of all sound feature values at the same sampling moment, including: weightedly averaging the target pathological interference degrees of all sound feature values at the same sampling moment, and normalizing to obtain the chronic obstructive pulmonary disease influence coefficient.

[0093] Among them, the weights of different kinds of sound feature values corresponding to the influence of chronic obstructive pulmonary disease may not be consistent, therefore, the target pathological interference degrees of different sound feature values are weightedly averaged and normalized to obtain the chronic obstructive pulmonary disease influence coefficient.

[0094] For example, the sound feature values include three dimensions of speech rate, pitch, and volume, the weight of the speech rate dimension is set to 0.2, the weight of the pitch dimension is set to 0.5, and the weight of the volume dimension is set to 0.3, then the product value of the target pathological interference degree of the speech rate dimension and 0.2, the product value of the target pathological interference degree of the pitch dimension and 0.5, and the product value of the target pathological interference degree of the volume dimension and 0.3 are calculated, and the mean value of the three product values is normalized as the chronic obstructive pulmonary disease influence coefficient.

[0095] It needs to be further explained that the specific weight values (such as 0.2, 0.3, 0.5, etc.) given in the embodiments of the present application are empirical values obtained under typical hardware configurations and test scenarios, and are intended to facilitate understanding of the present application. In actual application, those skilled in the art can adjust, calibrate or optimize these parameters according to the specific hardware performance, scene complexity and data characteristics, which do not constitute a limitation of the present application.

[0096] The chronic obstructive pulmonary disease influence coefficient is used to quantify the intensity of the whole being affected by the disease physiological interference at the moment; then, the chronic obstructive pulmonary disease influence coefficient is compared with a preset influence threshold value, only when the coefficient is less than the threshold value, it is determined that the speech signal at the sampling moment is basically not significantly interfered by the chronic obstructive pulmonary disease pathological factors, and it is screened as an "analysis moment" for subsequent emotion recognition and psychological care decision.

[0097] Among them, the preset influence threshold value is a threshold value of the chronic obstructive pulmonary disease influence coefficient, and the preset influence threshold value in the embodiments of the present application can be specifically, for example, 0.5, that is, when the chronic obstructive pulmonary disease influence coefficient is less than 0.5, the corresponding sampling moment is taken as the analysis moment.

[0098] Further, in some embodiments of the present application, the psychological care decision is made based on the voice signal at the analysis moment, including: grouping all voice signals at the analysis moment into different voice segments in time sequence; inputting the voice segments into a pre-trained emotion recognition model to realize emotion state analysis and obtain a target emotion analysis result; and making a psychological care decision based on the target emotion analysis result.

[0099] The voice signal at the analysis moment is input into a pre-trained emotion recognition model. The model (such as a convolutional neural network CNN) has been trained on a large amount of annotated healthy population voice data, and has learned to associate various voice feature patterns with specific emotional states (such as calm, anxiety, depression, and anger). When the voice signal at the analysis moment is input into the model, the model will perform nonlinear calculation and feature abstraction layer by layer, and finally output a probability distribution through a classifier (such as Softmax), indicating the possibility of the patient being in each emotional state. The system takes the emotion with the highest probability as the main recognition result. This part of the process is the existing neural network model recognition process, which will not be further limited and described.

[0100] In the process of specific psychological care decision, a nursing decision knowledge base can be pre-built in the system to implement standardized intervention strategies corresponding to different emotional states. For example, when "high anxiety" is recognized, the knowledge base may trigger the following decision-making process: first, the system will judge the intensity and duration of anxiety. If it is acute and high-intensity anxiety, the decision-making suggestions may include "immediately pushing the breathing relaxation training audio guide to the patient end" for on-site intervention, and at the same time "generating an alarm to notify the responsible nurse for emergency telephone counseling".

[0101] The embodiment of the present application effectively solves the problem of emotional misjudgment caused by pathological interference when the existing voice emotion recognition technology is applied to patients with chronic obstructive pulmonary disease. Specifically, by obtaining the patient's daily voice signal and synchronous blood oxygen saturation data, and extracting a plurality of voice feature values therefrom; then, the blood oxygen fluctuation stability degree at each sampling moment is evaluated in combination with the blood oxygen saturation change, and the initial pathological interference degree of each voice feature value is calculated accordingly; further, the initial pathological interference degree is corrected by analyzing the initial correlation between different voice feature values to obtain a more accurate target pathological interference degree; finally, the analysis moment with the smallest pathological interference is selected, and psychological care decision is made based on the voice signal at these moments. This technical means can effectively distinguish between physiological voice changes (such as slow speech and low volume) caused by chronic obstructive pulmonary disease and acoustic features corresponding to true emotional states, significantly reducing the emotion recognition error caused by the high overlap between disease symptoms and emotional expressions, thereby improving the accuracy of emotional state judgment for patients with chronic obstructive pulmonary disease and laying a reliable foundation for subsequent accurate and personalized psychological care intervention.

[0102] It is to be noted that the sequential order of the above-described embodiments of the present application only for the purpose of description, but not the advantages and disadvantages of the embodiments. The processes depicted in the drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multi-task processing and parallel processing are also possible or can be advantageous.

[0103] Each of the embodiments in the specification is described in a progressive manner, and the same or similar parts between the embodiments can be referred to each other. Each embodiment focuses on the difference from other embodiments.

Claims

1. An emotion recognition-based psychological care decision system for elderly patients with chronic obstructive pulmonary disease, characterized by, The system comprises: An acquisition module is configured to acquire a voice signal of a patient's daily voice and blood oxygen saturation at different sampling time points in a same time range as the voice, extract different voice feature values from the voice signal, and determine a voice abnormality degree of each voice feature value based on a standard difference of the voice feature value at each sampling time point; An interference analysis module is configured to determine a blood oxygen fluctuation stability degree at each sampling time point according to a numerical change of the blood oxygen saturation at different sampling time points, determine an initial pathological interference degree of the voice feature value affected by the COPD by combining the blood oxygen saturation, the voice abnormality degree, and the blood oxygen fluctuation stability degree at each sampling time point; A correction analysis module is configured to determine an initial correlation of two voice feature values by combining the initial pathological interference degrees of the two voice feature values at different sampling time points, correct the initial pathological interference degree according to the initial correlation between the voice feature values, and obtain a target pathological interference degree; A decision module is configured to filter an unaffected analysis time point according to the target pathological interference degrees of all voice feature values at the same sampling time point, and make a psychological care decision based on a voice signal at the analysis time point. The method for determining the voice abnormality degree comprises: Calculating an absolute value of a difference between each voice feature value and a preset standard value as the voice abnormality degree of the corresponding voice feature value.

2. The psychological care decision system for the elderly COPD patients based on emotion recognition according to claim 1, wherein, The voice feature values include a speech rate, a pitch, and a volume.

3. The psychological care decision system for the elderly COPD patients based on emotion recognition according to claim 1, wherein, The method for determining the blood oxygen fluctuation stability degree at each sampling time point according to the numerical change of the blood oxygen saturation at different sampling time points comprises: Taking any sampling time point as a target time point and a preset number of sampling time points closest to the target time point as adjacent sampling time points of the target time point; Calculating a standard deviation of the blood oxygen saturation at the target time point and the adjacent sampling time points, normalizing an inverse of the standard deviation, and taking the normalized inverse as the blood oxygen fluctuation stability degree at the target time point.

4. The psychological care decision system for the elderly COPD patients based on emotion recognition according to claim 1, wherein, The method for determining the initial pathological interference degree of the voice feature value affected by the COPD by combining the blood oxygen saturation, the voice abnormality degree, and the blood oxygen fluctuation stability degree at each sampling time point comprises: Normalizing an inverse of the blood oxygen saturation as a blood oxygen influence index; Normalizing the voice abnormality degree as a voice influence index; Calculating a weighted sum of the blood oxygen influence index and the voice influence index, normalizing a product of the weighted sum and the blood oxygen fluctuation stability degree, and taking the normalized product as the initial pathological interference degree.

5. The emotion recognition based psychological care decision system for elderly COPD patients as claimed in claim 1 wherein, The method for determining the initial correlation of two voice feature values by combining the initial pathological interference degrees of the two voice feature values at different sampling time points comprises: Sorting the initial pathological interference degrees of each voice feature value according to the sampling time points to obtain an interference sequence; Calculating a Pearson correlation coefficient of the interference sequence of any two voice feature values based on a Pearson correlation algorithm, and taking an absolute value of the Pearson correlation coefficient as the initial correlation.

6. The emotion recognition based psychological care decision system for elderly COPD patients as claimed in claim 1 wherein, The method for correcting the initial pathological interference degree according to the initial correlation between the voice feature values to obtain the target pathological interference degree comprises: Calculating a mean value of the initial correlations of any voice feature value and all other voice feature values as an interference correction coefficient; The product value of the interference correction coefficient and the initial pathological interference degree is calculated, and the sum value of the product value and the initial pathological interference degree is normalized as the target pathological interference degree.

7. The emotion recognition based psychological care decision system for elderly COPD patients as claimed in claim 1 wherein, The target pathological interference degree according to all sound characteristic values at the same sampling time is used to screen the unaffected analysis time, and the target pathological interference degree according to all sound characteristic values at the same sampling time is used to screen the unaffected analysis time. The target pathological interference degree according to all sound characteristic values at the same sampling time is used to determine the slow lung influence coefficient of the sampling time. The sampling time with the slow lung influence coefficient less than the preset influence threshold is used as the analysis time.

8. The emotion recognition based psychological care decision system for elderly COPD patients as claimed in claim 7 wherein, The target pathological interference degree according to all sound characteristic values at the same sampling time is used to determine the slow lung influence coefficient of the sampling time. The target pathological interference degrees of all sound characteristic values at the same sampling time are weighted and averaged, and the slow lung influence coefficient is obtained through normalization processing.

9. The emotion recognition based psychological care decision system for elderly COPD patients of claim 1, wherein, The psychological nursing decision is made based on the voice signal of the analysis time, and the psychological nursing decision is made based on the voice signal of the analysis time. The voice signals of all analysis times are combined into different voice segments according to the time sequence. The voice segment is input into a pre-trained emotion recognition model to realize emotion state analysis and obtain a target emotion analysis result, and psychological nursing decision is made based on the target emotion analysis result.

Images