Method and system for evaluating the effect of psychological intervention for mothers of premature infants

Abstract

The present invention discloses a method and system for evaluating the effect of psychological intervention for premature infants' mothers, which relates to the technical field of psychological assessment. A system for evaluating the effect of psychological intervention for premature infants' mothers includes: a data sorting module and an effect evaluation module. By simultaneously analyzing the behavioral analysis data set and the facial analysis data set of the evaluated mother, the present invention can comprehensively evaluate the mother's psychological state from multiple perspectives; the behavioral analysis model and the facial analysis model respectively provide independent evaluations of behavioral performance and emotional responses, making the evaluation results more comprehensive and accurate, and can effectively identify the mother's emotional fluctuations, psychological changes and responses to premature infants; through multi-level and multi-dimensional behavioral analysis, the mother's emotions, psychological stress and behavioral patterns can be deeply identified, not only considering the mother's behavior mode, but also deeply exploring the internal mechanism of her emotions, so as to provide strong data support for psychological intervention.

Description

Technical Field

[0001] The present invention relates to the technical field of psychological assessment, and particularly to a method and system for evaluating the psychological intervention effect of premature infants' mothers. Background Art

[0002] After a premature infant is born, the mother often faces huge psychological pressure, including emotional problems such as anxiety, depression, and post-traumatic stress disorder. These problems not only affect the mother's physical and mental health, but may also affect the care of the infant and the parent-child relationship. Therefore, psychological intervention for premature infants' mothers is particularly important. Existing intervention methods usually help mothers relieve psychological burdens through means such as psychological counseling, group counseling, and cognitive behavioral therapy. However, the evaluation of intervention effects lacks standardization and scientificity, making it difficult to accurately understand the actual effects of intervention measures. Conducting psychological assessment only through psychological scales is too one-sided. Therefore, for the evaluation of the psychological intervention effect of premature infants' mothers, more comprehensive evaluation methods need to be introduced, especially the evaluation combined with mother-infant interaction behavior, which not only helps to reflect the mother's psychological state, but also can reveal the parent-child relationship between the mother and the premature infant and the quality of their interaction. Summary of the Invention

[0003] The present invention aims to provide a method and system for evaluating the psychological intervention effect of premature infants' mothers to conduct a more comprehensive psychological assessment of premature infants' mothers.

[0004] A method for evaluating the psychological intervention effect of premature infants' mothers includes the following steps:

[0005] Conduct psychological observation and evaluation on the mother to be evaluated after psychological intervention, and collect the behavior analysis data set and facial analysis data set of the mother to be evaluated and the corresponding premature infant during the psychological observation and evaluation; record the total time of the psychological observation and evaluation as T0; the behavior analysis data set includes behavior contact images P t ; the facial analysis data set includes facial expression images E t ; t = 1, 2,..., T0; where t represents the discrete time points within the total time T0;

[0006] Analyze the behavior analysis data set using a psychological assessment behavior analysis model to obtain a behavior assessment result; the psychological assessment behavior analysis model uses multi-level and multi-dimensional behavior analysis to achieve psychological assessment based on action analysis;

[0007] Analyze the facial analysis data set using a psychological assessment facial analysis model to obtain a facial psychological assessment result; the psychological assessment facial analysis model is trained and optimized by combining the YOLOv5 model and the TAM module to achieve psychological assessment based on facial emotion recognition;

[0008] Conduct a comprehensive evaluation based on the behavior assessment result and the facial psychological assessment result to obtain a psychological intervention assessment result.

[0009] As a preferred technical solution of the present invention, the psychological assessment behavior analysis model includes an image processing layer, a feature analysis layer, and an output layer;

[0010] The image processing layer is used to perform feature segmentation on the behavior contact image P t to obtain the behavior contact component P （b）（t） ; where b = b1, b2, b3, and b4; b1 is the left - hand movement, b2 is the right - hand movement, b3 is the upper - body movement; b4 is the lower - body movement;

[0011] The feature analysis layer is used to perform grouped feature analysis on the behavior contact component P （b）（t） to obtain the evaluation feature result G （b） ;

[0012] The output layer is used to perform feature fusion on all the evaluation feature results G （b） to obtain the behavior evaluation result.

[0013] As a preferred technical solution of the present invention, in the image processing layer, the preset human millimeter - wave radar point cloud matrix is used to perform spatial variation alignment on the behavior contact image P t to obtain the pre - processed behavior contact image P t ';

[0014] Obtain the body feature data of the mother to be evaluated to get the body feature data;

[0015] Based on the pre - processed behavior contact image P t ', perform motion trajectory extraction to obtain the motion trajectory sequence;

[0016] Based on the body feature data and the motion trajectory sequence, perform feature fusion on the pre - processed behavior contact image P t ' to obtain the enhanced behavior contact image R t ;

[0017] Perform body part feature output on the enhanced behavior contact image R t to obtain the behavior contact component P （b）（t） .

[0018] As a preferred technical solution of the present invention, in the feature analysis layer, select the behavior contact component P （b）（t-q） , the behavior contact component P （b）（t） and the behavior contact component P （b）（t+q） to form the behavior contact grouping component F q ;

[0019] Perform internal cross - analysis on the behavior contact grouping component F q to obtain the behavior emotion probability matrix Z q ;

[0020] Perform temporal emotion analysis on the behavioral contact grouping component F q to obtain the behavioral temporal emotion probability matrix J q ;

[0021] According to the behavioral emotion probability matrix Z q and the behavioral temporal emotion probability matrix J q , use the formula H q =GMU(Z q , J q ) for feature fusion to obtain the grouped behavior evaluation result H q ; GMU() represents the fusion matrix unit;

[0022] Randomly change the value of q, repeat the above steps several times, and obtain several groups of grouped behavior evaluation results H q ;

[0023] wherein, q satisfies the following conditions: t - q > 0; t + q < T0; q ∈ the preset time span range;

[0024] Combine all the grouped behavior evaluation results H q to obtain the evaluation feature result G （b） .

[0025] As a preferred technical solution of the present invention, the psychological evaluation facial analysis model is used to extract features from the facial expression image E t to obtain the facial psychological evaluation result;

[0026] Use the YOLOv5 model as the basic structure of the psychological evaluation facial analysis model; aiming at the characteristics of the psychological evaluation facial analysis model, introduce the TAM module to replace the bottleneck layer in the backbone network of the YOLOv5 model; stack several TAM modules to form the facial feature extraction unit in the psychological evaluation facial analysis model; at the same time, use the global pooling layer to output the final facial psychological evaluation result.

[0027] As a preferred technical solution of the present invention, the specific steps for comprehensive evaluation based on the behavioral evaluation result and the facial psychological evaluation result include:

[0028] Obtain historical psychological scale data; extract features from the historical psychological scale data to obtain the scale evaluation result;

[0029] Record the behavioral evaluation result as A1, the facial psychological evaluation result as A2, and the scale evaluation result as A3; combine A1, A2 and A3 to obtain the psychological intervention result set;

[0030] From the set of psychological intervention results, randomly select two elements to calculate the difference degree, and obtain the mutually exclusive difference degree of psychological intervention;

[0031] If all the mutually exclusive difference degrees of psychological intervention are less than the preset psychological intervention difference degree, then perform weighted fusion on the set of psychological intervention results to obtain the psychological intervention evaluation result;

[0032] Otherwise, extract the elements corresponding to the mutually exclusive difference degrees of psychological intervention that are greater than the preset psychological intervention difference degree to obtain abnormal elements; if the abnormal elements are A1 or A2, then conduct a psychological observation evaluation again; otherwise, perform weighted fusion based on A1 and A2 to obtain the psychological intervention evaluation result.

[0033] A psychological intervention effect evaluation system for preterm infants' mothers, comprising:

[0034] A data sorting module, which includes a data acquisition unit; the data acquisition unit is used to conduct a psychological observation evaluation on the mother to be evaluated after psychological intervention, and collect the behavioral analysis data set and facial analysis data set of the mother to be evaluated and the corresponding preterm infant during the psychological observation evaluation; record the total time of the psychological observation evaluation as T0; the behavioral analysis data set includes the behavioral contact image P t ; the facial analysis data set includes the facial expression image E t ; t = 1, 2,..., T0; where t represents the discrete time points within the total time T0;

[0035] An effect evaluation module, which includes an action recognition unit, a facial recognition unit, and a comprehensive analysis unit;

[0036] The action recognition unit is used to analyze the behavioral analysis data set by using the psychological evaluation behavior analysis model to obtain the behavioral evaluation result; the psychological evaluation behavior analysis model realizes psychological evaluation based on action analysis by using multi-level and multi-dimensional behavioral analysis;

[0037] The facial recognition unit is used to analyze the facial analysis data set by using the psychological evaluation facial analysis model to obtain the facial psychological evaluation result; the psychological evaluation facial analysis model is trained and optimized by combining the YOLOv5 model and the TAM module to realize psychological evaluation based on facial emotion recognition;

[0038] The comprehensive analysis unit is used to conduct a comprehensive evaluation based on the behavioral evaluation result and the facial psychological evaluation result to obtain the psychological intervention evaluation result.

[0039] The present invention has the following advantages:

[0040] By simultaneously analyzing the behavioral analysis dataset and the facial analysis dataset of the mother to be evaluated, the present invention can comprehensively evaluate the mother's psychological state from multiple perspectives; the behavioral analysis model and the facial analysis model respectively provide independent evaluations of behavioral performance and emotional responses, making the evaluation results more comprehensive and accurate. This multi-dimensional evaluation method can effectively identify the mother's emotional fluctuations, psychological changes, and responses to premature infants; through multi-level and multi-dimensional behavioral analysis, it is possible to more deeply identify the mother's emotions, psychological stress, and behavioral patterns. This multi-level analysis not only considers the mother's behavior style but also delves into the internal mechanism of her emotions, thus providing strong data support for psychological intervention.

[0041] In the feature analysis layer of the present invention, by combining the behavioral contact components at different time points, internal emotion probability output and temporal emotion analysis are performed, enabling the capture of the mother's emotional fluctuations from multiple angles and time dimensions and the detailed analysis of emotional changes during the behavioral interaction process; through this multi-dimensional emotion evaluation, the mother's psychological state can be more comprehensively understood; by randomly changing the value of the time span, the analysis time window can be dynamically adjusted. This flexible selection of the time span can adapt to the evaluation requirements in different situations and avoid the limitations that may be brought by a fixed time window. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] Figure 1 It is a schematic structural diagram of a system for evaluating the psychological intervention effect of a premature infant mother adopted in an embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0043] In order to enable those skilled in the art of the present technology to better understand the technical solutions in the present invention, the following will clearly and completely describe the technical solutions in the embodiments of the present invention with reference to the accompanying drawings in the embodiments of the present invention.

[0044] Embodiment 1, a method for evaluating the psychological intervention effect of a premature infant mother, includes the following steps:

[0045] Perform psychological observation and evaluation on the mother to be evaluated after psychological intervention, and collect the behavioral analysis dataset and the facial analysis dataset of the mother to be evaluated and the corresponding premature infant during the psychological observation and evaluation; record the total time of the psychological observation and evaluation as T0; the behavioral analysis dataset includes behavioral contact images P t ; the facial analysis dataset includes facial expression images E t ; t = 1, 2,..., T0; where t represents the discrete time points within the total time T0.

[0046] In the psychological observation and assessment, the camera is set to capture the contact process between the mother being evaluated and the premature infant without disturbing the mother's behavior. The camera is set at different angles to capture the mother's postures, movements, gestures, and details of interactions with the premature infant. At the same time, a facial recognition camera is installed to analyze the mother's facial expression responses, such as emotional expressions like smiling, frowning, and eye contact. In the psychological observation and assessment, the mother being evaluated needs to be informed of the purpose and use of this observation to ensure the availability of the data.

[0047] The discrete points represent specific time points within the total time T0. At these time points, evaluation data is collected, such as the mother's behavior and facial expression images. Each discrete point can represent a specific instant within a time period, usually used to describe the state or changes within a certain time period, rather than a continuous process.

[0048] The behavioral analysis dataset is analyzed using a psychological assessment behavior analysis model to obtain the behavior assessment results. The psychological assessment behavior analysis model uses multi-level and multi-dimensional behavioral analysis to achieve psychological assessment based on action analysis.

[0049] The psychological assessment behavior analysis model includes an image processing layer, a feature analysis layer, and an output layer.

[0050] The image processing layer is used to perform feature segmentation on the behavior contact image P t to obtain the behavior contact components P （b）（t） ; where b = b1, b2, b3, and b4; b1 is the left hand movement, b2 is the right hand movement, b3 is the upper body movement; b4 is the lower body movement.

[0051] The feature analysis layer is used to perform grouped feature analysis on the behavior contact components P （b）（t） to obtain the evaluation feature results G （b） ;

[0052] The output layer is used to perform feature fusion on all the evaluation feature results G （b） to obtain the behavior assessment results.

[0053] The image processing layer performs feature segmentation on the behavior contact image to separately identify the behaviors of each part of the mother's body, such as the left hand, right hand, upper body, and lower limbs. This method helps to more accurately identify and analyze the specific behavior manifestations of the mother when interacting with the premature infant, rather than simply analyzing the overall behavior; this refined processing can reveal the behavioral differences of the mother in different body parts and help to more comprehensively understand her psychological state; the feature analysis layer performs feature analysis on each component of the mother's behavior, such as the left hand, right hand, etc.; through the feature extraction of these subdivided behaviors, the model can identify specific behavior patterns and further judge the mother's emotions or psychological state; for example, the differences in the movements of the mother's left hand and right hand may reflect her emotional investment or mood fluctuations; based on the analysis of behavior characteristics at different time points, the model can dynamically evaluate the mother's emotional changes; for example, during the process of the mother's contact with the infant, her behavior may change over time, and this change may reflect her gradual adaptation to the role of motherhood or fluctuations in her emotional state. Real-time behavior tracking makes the psychological assessment closer to the mother's actual psychological state;

[0054] In the image processing layer, the preset human millimeter-wave radar point cloud matrix is used to perform spatial variation alignment on the behavior contact image P t to obtain the preprocessed behavior contact image P t ’;

[0055] Obtain the body feature data of the mother to be evaluated to get the body feature data;

[0056] Based on the preprocessed behavior contact image P t ’, perform motion trajectory extraction to obtain the motion trajectory sequence;

[0057] Based on the body feature data and the motion trajectory sequence, perform feature fusion on the preprocessed behavior contact image P t ’ to obtain the enhanced behavior contact image R t ;

[0058] Perform body part feature output on the enhanced behavior contact image R t to obtain the behavior contact component P （b）（t） ;

[0059] The image processing layer is trained based on the CNN model;

[0060] Align the behavior contact image spatially using a preset millimeter-wave radar point cloud matrix to ensure accurate registration of the image data in space. The beneficial effect of this step is to reduce the deviation of the image data caused by shooting angle or environmental changes, ensuring the accuracy of subsequent analysis; through this alignment, the interaction behavior between the mother and the baby can be more precisely docked with the body characteristics, thus laying a foundation for subsequent feature extraction and fusion; by obtaining the body characteristic data of the mother to be evaluated, detailed body information is provided for subsequent behavior analysis. The beneficial effect of this step is that it can identify the dynamic changes in various parts of the mother's body, such as hands, upper limbs, lower limbs, etc., and accurately capture the mother's body movement pattern by extracting the sequence of motion trajectory coordinates, which enables subsequent analysis to reflect the specific behaviors of the mother-baby interaction from a dynamic perspective and analyze the emotional responses and behavior patterns more precisely;

[0061] During the feature fusion process, based on the preprocessed image, the mother's body characteristic data, and the sequence of motion trajectory coordinates, enhance the quality of the behavior contact image, making the image clearer and more hierarchical when showing the mother's behavior; through feature fusion, the details in the behavior image are strengthened, which helps to improve the recognition accuracy of the mother's behavior. For example, distinguish the differences in her different limb movements such as the left hand and the right hand, and the emotional indication of the movements such as tension or relaxation. This enhanced image enables the system to more accurately reflect the subtle emotional changes of the mother during the interaction;

[0062] Select the behavior contact component P （b）（t-q） 、the behavior contact component P （b）（t） and the behavior contact component P （b）（t+q） to form the behavior contact grouping component F q ;

[0063] Conduct an internal cross-analysis on the behavior contact grouping component F q to obtain the behavior emotion probability matrix Z q ;

[0064] Conduct a temporal emotion analysis on the behavior contact grouping component F q to obtain the behavior temporal emotion probability matrix J q ;

[0065] According to the behavior emotion probability matrix Z q and the behavior temporal emotion probability matrix J q , use the formula H q =GMU(Z q , J q ) for feature fusion to obtain the grouped behavior evaluation result H q ; GMU() represents the fusion matrix unit;

[0066] Randomly change the value of q and repeat the above steps several times to obtain several groups of evaluation results H of grouping behaviors q ;

[0067] Among them, q satisfies the following conditions: t - q > 0; t + q < T0; q ∈ the preset time span range; the preset time span range is set by professional technicians according to the actual situation;

[0068] Combine all the evaluation results H of grouping behaviors q to obtain the evaluation feature result G （b） ;

[0069] At the feature analysis layer, by combining the behavior contact components at different time points, internal emotion probability output and temporal emotion analysis are carried out. This method can capture the mother's emotional fluctuations from multiple angles and time dimensions, and carefully analyze the emotional changes of the mother during the behavior interaction process. Through this multi-dimensional emotion assessment, the mother's psychological state can be more comprehensively understood. For example, emotions such as anxiety, relaxation, and joy; by randomly changing the value of the time span, the analysis time window can be dynamically adjusted. This flexible time span selection can adapt to the evaluation needs in different situations and avoid the limitations that may be brought by a fixed time window; for example, some behavior patterns may require a longer time span to be accurately captured, while other patterns may require a shorter time window. This dynamic adjustment improves the adaptability and accuracy of the model in different scenarios; in the temporal emotion analysis, by capturing the mother's emotional state at different moments, the evolution of the mother's emotions over time can be analyzed. This temporal analysis helps to reveal the gradual changes in the mother's emotions during a specific behavior process, especially at the moments when the emotions fluctuate greatly; for example, the process of the mother changing from anxiety to gradually relaxation, or from confusion to generating comfort. This temporal emotion analysis can effectively help to understand the mother's emotional response during psychological intervention; by repeating the change of the time span value multiple times and obtaining several groups of evaluation results of the behavior contact of the evaluated mothers in groups, the stability and robustness of the emotion analysis are ensured. The combination of multiple evaluation results can eliminate some accidental errors or inconsistencies, making the final evaluation result more reliable and accurate; multiple evaluation results can also provide more solid data support for subsequent interventions;

[0070] Use the psychological assessment facial analysis model to analyze the facial analysis data set to obtain the facial psychological assessment result; the psychological assessment facial analysis model is trained and optimized by combining the YOLOv5 model and the TAM module to achieve psychological assessment based on facial emotion recognition;

[0071] The psychological assessment facial analysis model is used to extract features from the facial expression image E t to obtain the facial psychological assessment result;

[0072] Use the YOLOv5 model as the basic structure of the psychological assessment facial analysis model; aiming at the characteristics of the psychological assessment facial analysis model, introduce the TAM module to replace the bottleneck layer in the backbone network of the YOLOv5 model; stack several TAM modules to form the facial feature extraction unit in the psychological assessment facial analysis model; at the same time, use the global pooling layer to output the final facial psychological assessment result;

[0073] The TAM module, namely the Time Adaptive Module, is a time modeling module used in deep learning models, especially suitable for processing time series data, such as video understanding, action recognition, time series prediction and other tasks. The core idea of TAM is to adaptively model the feature changes in the time dimension, enabling the network to better capture and utilize time series information;

[0074] The specific steps for training the psychological assessment facial analysis model include:

[0075] Collect several groups of facial emotion recognition training samples, combine several groups of facial emotion recognition training samples to obtain a facial emotion recognition training set; perform model training based on the facial emotion recognition training set to obtain an initial psychological assessment facial analysis model; if the initial psychological assessment facial analysis model passes the model evaluation, then use the initial psychological assessment facial analysis model as the psychological assessment facial analysis model in the psychological assessment facial analysis model; otherwise, continue to perform model training using the facial emotion recognition training set;

[0076] Combining the training and optimization of the YOLOv5 model and the TAM module enables the psychological assessment facial analysis model to accurately recognize the facial emotions of the evaluated mother. The object detection ability of YOLOv5 combined with the advantages of the TAM module in facial recognition can better identify facial expression details and improve the recognition accuracy of facial emotions, especially under complex and dynamic emotion changes; introducing the TAM module to replace the bottleneck layer in the backbone network of the YOLOv5 model and stacking multiple TAM modules to form the facial feature extraction unit effectively improves the expressiveness and processing speed of facial feature extraction; by training several groups of facial emotion recognition samples and continuously optimizing the training model, it can adaptively adjust the feature extraction method according to different emotion datasets, improving the sensitivity and accuracy to the mother's facial expressions; the global pooling layer in the psychological assessment facial analysis model effectively integrates the local features in the facial expression image into global information, helping the model to recognize and analyze facial emotions from a global perspective. This layer helps to avoid over-reliance on the emotion information of a single local area, thereby enhancing the overall emotion recognition ability and ensuring that the overall facial emotions of the mother are accurately captured;

[0077] Based on the behavior assessment result and the facial psychological assessment result, conduct a comprehensive assessment to obtain the psychological intervention assessment result;

[0078] The specific steps for comprehensive evaluation based on behavior evaluation results and facial psychological evaluation results include:

[0079] Obtain historical psychological scale data; perform feature extraction on the historical psychological scale data to obtain scale evaluation results;

[0080] Record the behavior evaluation result as A1, the facial psychological evaluation result as A2, and the scale evaluation result as A3; combine A1, A2, and A3 to obtain a set of psychological intervention results;

[0081] Select any two elements from the set of psychological intervention results to calculate the difference degree, and obtain the mutually exclusive difference degree of psychological intervention;

[0082] If all the mutually exclusive difference degrees of psychological intervention are less than the preset psychological intervention difference degree, then perform weighted fusion on the set of psychological intervention results to obtain the psychological intervention evaluation result; the preset psychological intervention difference degree is set by professional and technical personnel according to the actual situation;

[0083] Otherwise, extract the elements corresponding to the mutually exclusive difference degrees of psychological intervention that are greater than the preset psychological intervention difference degree to obtain abnormal elements; if the abnormal element is A1 or A2, perform a psychological observation evaluation again; otherwise, perform weighted fusion based on A1 and A2 to obtain the psychological intervention evaluation result;

[0084] By combining the mother's behavior, facial expressions, and scale data for comprehensive evaluation, it is possible to comprehensively understand the mother's psychological state from multiple perspectives. This multi-dimensional evaluation method helps to make up for the limitations that may be brought about by a single evaluation dimension, thereby improving the accuracy and reliability of the evaluation of the psychological intervention effect; obtaining and analyzing the historical psychological scale data of the mother to be evaluated can provide a background and comparison for the current evaluation of the psychological intervention effect. This method helps the system identify the long-term changes in the mother's psychological state, provides more comprehensive data support for the evaluation of the intervention effect, and helps to understand whether the intervention has effectively improved the mother's mental health; by calculating the difference degrees among behaviors, facial expressions, and scale results, it is possible to automatically identify the contradictions or inconsistencies between different evaluation results. This difference degree calculation mechanism can effectively identify abnormal evaluation results, thereby triggering a re-evaluation or adjustment of the intervention measures in a timely manner when problems are found. This automated detection greatly improves the efficiency and accuracy of the evaluation process; by performing weighted fusion on different evaluation results, it is possible to integrate the evaluation results of each dimension to form a more accurate evaluation of the psychological intervention effect. This weighted fusion method can give different weights according to the importance of different evaluation dimensions, thereby ensuring the comprehensiveness and accuracy of the evaluation results;

[0085] In this embodiment, a single psychological observation and assessment includes the following steps: The assessor needs to communicate with the mother to be assessed during the assessment preparation stage, explain the purpose and process of the psychological observation and assessment, and ensure that the mother understands and agrees to participate in the assessment; at this time, the assessor will also set up necessary equipment, including cameras, sensors, etc., for real-time recording of the behavior and facial expression data of the mother and the premature infant; at the same time, the assessor collects data at preset time points, and the system will start monitoring the behavior and facial expressions of the mother to be assessed; after the data collection is completed, it enters the data processing and analysis stage, and a psychological assessment behavior analysis model is used to process the behavior contact images of the mother, analyze her behavior patterns, action frequencies, and emotional responses; through the facial expression images, analyze the mother's facial expressions and judge her emotional changes; combine the behavior analysis results and the facial emotion analysis results to obtain a comprehensive psychological assessment result, and evaluate the emotional and behavioral changes of the mother after psychological intervention; the assessment results will be sorted out and presented to the assessor or psychological intervention experts. At this stage, the assessor judges the performance of the mother during the psychological intervention based on the results, understands the improvement of her mental health, and provides further intervention suggestions if necessary. The assessor will also communicate the assessment results to the mother and explain her mental state and emotional changes.

[0086] Embodiment 2. A system for evaluating the effect of psychological intervention for premature infant mothers. Refer to Figure 1 as shown, including:

[0087] A data sorting module, which includes a data acquisition unit; the data acquisition unit is used to conduct a psychological observation and assessment on the mother to be assessed after psychological intervention, and collect the behavior analysis data set and facial analysis data set of the mother to be assessed and the corresponding premature infant during the psychological observation and assessment; record the total time of the psychological observation and assessment as T0; the behavior analysis data set includes behavior contact images P t ; the facial analysis data set includes facial expression images E t ; t = 1, 2,..., T0; where t represents the discrete time points within the total time T0;

[0088] An effect evaluation module, which includes an action recognition unit, a facial recognition unit, and a comprehensive analysis unit;

[0089] The action recognition unit is used to analyze the behavior analysis data set by using a psychological assessment behavior analysis model to obtain a behavior assessment result; the psychological assessment behavior analysis model uses multi-level and multi-dimensional behavior analysis to achieve psychological assessment based on action analysis;

[0090] The facial recognition unit is used to analyze the facial analysis data set by using a psychological assessment facial analysis model to obtain a facial psychological assessment result; the psychological assessment facial analysis model is trained and optimized by combining the YOLOv5 model and the TAM module to achieve psychological assessment based on facial emotion recognition;

[0091] The comprehensive analysis unit is used to perform a comprehensive evaluation based on the behavior evaluation result and the facial psychological evaluation result to obtain the psychological intervention evaluation result.

[0092] It should be understood that those of ordinary skill in the art can make improvements or transformations according to the above description, and all such improvements and transformations shall fall within the protection scope of the appended claims of the present invention. The parts not described in detail in this specification belong to the prior art well-known to those of ordinary skill in the art.

Claims

1. A method for evaluating the effect of psychological intervention on mothers of premature infants, characterized in that: It includes the following steps: Conduct psychological observation and evaluation on the evaluated mother after psychological intervention. In the psychological observation and evaluation, collect the behavioral analysis data set and facial analysis data set of the evaluated mother and the corresponding premature infant; record the total time of the psychological observation and evaluation as T0; The behavior analysis dataset includes behavioral contact images P t ; The facial analysis dataset includes facial expression images E t ; t=1, 2, ..., T0; where t represents a discrete time point within the total time T0; Analyze the behavioral analysis data set using the psychological evaluation behavioral analysis model to obtain the behavioral evaluation result; the psychological evaluation behavioral analysis model realizes psychological evaluation based on action analysis through multi-level and multi-dimensional behavioral analysis; Analyze the facial analysis data set using the psychological evaluation facial analysis model to obtain the facial psychological evaluation result; the psychological evaluation facial analysis model is trained and optimized by combining the YOLOv5 model and the time adaptive module to realize psychological evaluation based on facial emotion recognition; Conduct a comprehensive evaluation based on the behavioral evaluation result and the facial psychological evaluation result to obtain the psychological intervention evaluation result; Among them, the psychological evaluation behavioral analysis model includes an image processing layer, a feature analysis layer and an output layer; The image processing layer is used to process the behavior contact image P t Perform feature segmentation to obtain the behavioral contact component P （b）（t） ; Among them, b=b1, b2, b3 and b4; b1 is the left hand movement, b2 is the right hand movement, b3 is the upper body movement; b4 is the lower limb movement; The feature analysis layer is used to analyze the behavioral contact component P （b）（t） Perform group feature analysis to obtain the evaluation feature result G （b） ; The output layer is used to convert all the evaluation feature results G （b） Perform feature fusion to obtain behavior evaluation results; Among them, the behavioral contact component P is selected in the feature analysis layer （b）（t-q） , behavioral contact component P （b）（t） and behavioral contact component P （b）（t+q） Composition behavior contact group component F q ; Behavioral contact grouping component F q Perform internal cross analysis to obtain the behavior emotion probability matrix Z q ; Behavioral contact grouping component F q Perform time series emotion analysis to obtain the behavior time series emotion probability matrix J q ; According to the behavior emotion probability matrix Z q and behavior time series emotion probability matrix J q , using the formula H q =GMU(Z q , J q ) to perform feature fusion and obtain the grouping behavior evaluation result H q ; GMU() represents fusion matrix unit; Randomly change the value of q, repeat the above steps several times, and obtain several groups of group behavior evaluation results H q ; Among them, q satisfies the following conditions: t - q > 0; t + q < T0; q ∈ the preset time span range; All group behavior evaluation results H q Combine and get the evaluation feature result G （b） .

2. A method for evaluating the effect of psychological intervention for mothers of premature infants according to claim 1, characterized in that: In the image processing layer, the preset human body millimeter wave radar point cloud matrix is ​​used to process the behavior contact image P t Perform spatial variation alignment to obtain the preprocessed behavioral contact image P t '; Obtain the physical characteristic data of the evaluated mother to get the physical characteristic data; Based on the preprocessed behavior contact image P t 'Extract motion trajectory and obtain motion trajectory sequence; Based on the body feature data and motion trajectory sequence, the preprocessed behavior contact image P t 'Perform feature fusion to obtain enhanced behavior contact image R t ; To enhance the behavior contact image R t Output the body part features to obtain the behavioral contact component P （b）（t） .

3. A method for evaluating the effect of psychological intervention for mothers of premature infants according to claim 2, characterized in that: Psychological evaluation facial analysis model for facial expression images E t Perform feature extraction to obtain facial psychological assessment results; Use the YOLOv5 model as the basic structure of the psychological evaluation facial analysis model; aiming at the characteristics of the psychological evaluation facial analysis model, introduce a time adaptive module to replace the bottleneck layer in the backbone network of the YOLOv5 model; stack several time adaptive modules to form the facial feature extraction unit in the psychological evaluation facial analysis model; at the same time, use the global pooling layer to output the final facial psychological evaluation result.

4. A method for evaluating the effect of psychological intervention for mothers of premature infants according to claim 3, characterized in that: The specific steps of the comprehensive evaluation based on the behavioral evaluation result and the facial psychological evaluation result include: Obtain the historical psychological scale data; extract the characteristics of the historical psychological scale data to get the scale evaluation result; Record the behavioral evaluation result as A1, the facial psychological evaluation result as A2, and the scale evaluation result as A3; combine A1, A2 and A3 to get the psychological intervention result set; Select any two elements from the psychological intervention result set to calculate the difference degree to obtain the psychological intervention mutual exclusion difference degree; If all the psychological intervention mutual exclusion difference degrees are less than the preset psychological intervention difference degree, then perform weighted fusion on the psychological intervention result set to obtain the psychological intervention evaluation result; Otherwise, extract the elements corresponding to the psychological intervention mutual exclusion difference degree greater than the preset psychological intervention difference degree to obtain abnormal elements; if the abnormal element is A1 or A2, conduct a psychological observation and evaluation again; otherwise, perform weighted fusion based on A1 and A2 to obtain the psychological intervention evaluation result.

5. A psychological intervention effect evaluation system for premature infant mothers, characterized in that: The system applies a method for evaluating the psychological intervention effect of a premature infant mother described in any one of claims 1 - 4 above, including: The data sorting module includes a data acquisition unit; the data acquisition unit is used to conduct psychological observation and evaluation on the evaluated mother after psychological intervention, and collect the behavior analysis data set and facial analysis data set of the evaluated mother and the corresponding premature infant in the psychological observation and evaluation; the total time of the psychological observation and evaluation is recorded as T0; the behavior analysis data set includes the behavior contact image P t ; The facial analysis dataset includes facial expression images E t ; t=1, 2, ..., T0; where t represents a discrete time point within the total time T0; An effect evaluation module, including an action recognition unit, a facial recognition unit and a comprehensive analysis unit; The action recognition unit is used to analyze the behavioral analysis data set using the psychological evaluation behavioral analysis model to obtain the behavioral evaluation result; the psychological evaluation behavioral analysis model realizes psychological evaluation based on action analysis through multi-level and multi-dimensional behavioral analysis; The facial recognition unit is used to analyze the facial analysis data set using the psychological evaluation facial analysis model to obtain the facial psychological evaluation results; the psychological evaluation facial analysis model is trained and optimized in combination with the YOLOv5 model and the time adaptive module to achieve psychological evaluation based on facial emotion recognition; The comprehensive analysis unit is used to conduct a comprehensive assessment based on the behavioral assessment results and the facial psychological assessment results to obtain the psychological intervention assessment results.

Images