Chinese medicine internal medicine sentiment disease syndrome identification method and system fusing machine learning

Abstract

The application discloses a TCM internal medicine emotional disease syndrome identification method and system fusing machine learning, relates to the technical field of TCM intelligent diagnosis, and comprises the following steps: around the emotional mutation period, the patient diagnosis and treatment timeline and symptom performance line are rearranged synchronously, the syndrome feature boundary weakening area is identified, the feature contrast zone is constructed, and the critical feature overlap list is formed in the feature contrast zone; based on the critical feature overlap list, the tongue image, pulse image, signs and emotional description are contrasted and analyzed item by item, the feature mutual offset position is marked, and the difference anchor point zone is integrated. The application keeps the syndrome identification of the emotional mutation stage continuous and stable through the synchronous rearrangement of the timeline and the symptom line, the tracking of the difference anchor point zone, the coordination of the inverse phase guide line and the elastic threshold window, makes the syndrome evolution trajectory converge in the stable identification area through the phase traction window array and the shadow shunt gate dynamic regulation sensitivity and compliance, and improves the reliability and consistency of the identification result.

Description

Technical Field

[0001] This invention relates to the field of intelligent diagnostic technology in traditional Chinese medicine, specifically to a method and system for identifying emotional and mental disorders in internal medicine that integrates machine learning. Background Technology

[0002] The integration of machine learning in TCM internal medicine for the identification of emotional disorders refers to the process of structuring and organizing patients' emotional changes, chief symptoms, tongue and pulse examinations, physical signs, and follow-up information within the theoretical framework of TCM internal medicine, focusing on the diagnosis and treatment of diseases related to emotional disorders. Machine learning methods are then introduced to mine patterns and model correlations in this multi-source diagnostic data, establishing a mapping relationship between the characteristics of emotional disorders and diagnostic conclusions. This enables automated analysis and assisted judgment of syndrome types. This approach breaks through the traditional model of syndrome identification relying on the physician's personal experience, allowing the diagnostic process of emotional disorders to maintain the theoretical essence of TCM while improving consistency, stability, and repeatability through a data-driven approach, providing technical support for the standardized diagnosis and treatment of emotional disorders in TCM internal medicine.

[0003] Existing technologies have the following shortcomings: Under current conditions, when a patient's emotional state is in a critical transitional zone from one dominant emotion to another, related symptoms, emotional reactions, and physical signs exhibit highly overlapping and rapidly changing characteristics, leading to a gradual weakening of the boundaries between the multidimensional features used for syndrome identification. Since existing syndrome identification models typically rely on relatively stable feature distribution structures, at the critical stage of emotional evolution, the feature vectors corresponding to different syndrome types become significantly closer or even overlap in their internal representation spaces, making it difficult for the model to effectively distinguish between syndrome types and thus preventing the formation of a clear discriminative path during inference. This results in insufficient directional guidance for syndrome output, leading to fuzzy judgments or frequent switching, severely interfering with the accurate grasp of syndrome types in clinical diagnosis and treatment, and restricting the reliability and stability of syndrome identification for emotional disorders in practical applications.

[0004] The information disclosed in the background section is only intended to enhance the understanding of the background of this disclosure, and therefore may include information that does not constitute prior art known to those skilled in the art. Summary of the Invention

[0005] The purpose of this invention is to provide a method and system for identifying emotional and mental disorders in traditional Chinese medicine by integrating machine learning, so as to solve the problems in the background art mentioned above.

[0006] To achieve the above objectives, the present invention provides the following technical solution: a method for identifying syndromes of emotional disorders in traditional Chinese medicine internal medicine by integrating machine learning, comprising the following steps: Around the period of emotional change, the patient's diagnosis and treatment timeline and symptom manifestation line are rearranged synchronously to identify the weakened boundary of syndrome characteristics, construct characteristic control bands, and compile a list of critical feature overlap within the characteristic control bands; Based on the list of overlapping critical features, the descriptions of tongue appearance, pulse appearance, physical signs and emotions are compared and analyzed item by item. The locations where features cancel each other out are marked and integrated to form a divergence anchor zone to reflect the interactive conflict area between different syndrome features. By connecting subtle changes within the divergence anchor zone according to the chronological order of evolution, the direction of abrupt changes in emotional evolution is extracted, and a continuously trackable reverse-phase guide line is generated to characterize the evolutionary path of abrupt changes in emotional state. An elastic threshold window corresponding to the syndrome type is established around the reverse phase guide line, so that the elastic threshold window dynamically opens and closes with the rhythm of sudden change in emotion, and is combined according to the time phase law to form a phase traction window array; Based on the phase-traction window array, time slot rotation traction is implemented, and a shadow diversion gate mechanism is introduced to dynamically improve the model sensitivity and reduce the model compliance during the critical period of emotional evolution, so that the syndrome evolution trajectory gradually converges towards the stable recognition area, and outputs a clear diagnosis result of emotional syndrome.

[0007] Preferably, the steps for forming the critical feature overlap list are as follows: Establish a patient diagnosis and treatment timeline and symptom manifestation line around the period of emotional change. The diagnosis and treatment timeline is based on the emotional changes, and the symptom manifestation line is based on the patient's chief complaint, physician's observation and examination records. Based on the starting point of emotional change, the diagnosis and treatment timeline and symptom manifestation timeline are rearranged synchronously. Symptom nodes from different sources are rearranged according to their duration and trend of change, so that emotional reactions and symptom changes correspond in time. Based on the synchronous rearrangement results, the overlapping intervals of symptoms and signs during the period of emotional change are identified, the weakened areas of syndrome feature boundaries are determined, and the overlapping periods are marked on the time axis. A feature comparison band is constructed around the boundary weakening region. Symptoms, signs and emotional features in continuous time slices are compared and analyzed to form a feature set and integrate them in chronological order to generate a critical feature overlap list.

[0008] Preferably, when constructing the feature control band, the symptoms, signs and emotional reactions during the period of emotional change are arranged in chronological order with the timeline as the vertical axis and the symptom features as the horizontal axis. The frequency of occurrence and correlation strength of the features in each time slice are recorded. Items with high degree of feature overlap and obvious attribute intersection are grouped into the same feature set to form a critical feature overlap list with correlation.

[0009] Preferably, the formation process of the divergence anchor zone is as follows: Based on the list of overlapping critical features, the information on tongue appearance, pulse appearance, physical signs and emotional descriptions is organized and classified, and divided into tongue appearance feature group, pulse appearance feature group, physical sign feature group and emotional description group according to perception type. Based on the time distribution in the list of overlapping critical features, the correspondence between each feature group is analyzed item by item, and the co-occurrence and opposition of different features in the same time slice are recorded. Feature points that conflict between tongue appearance, pulse appearance, physical signs and emotional descriptions are marked and extracted to form feature cancellation positions and construct conflict feature chains. Using the timeline as the main sequence, feature points with similar conflict attributes and continuous relationships in the time slice are sequentially linked and integrated to form a divergence anchor zone, which is used to reflect the interactive conflict area between different syndrome features.

[0010] Preferably, in the process of forming the divergence anchor zone, the feature points with continuously changing conflict attributes are sequentially connected according to the time sequence, with the time axis as the sequential reference. During the connection process, the feature points are layered and integrated according to the change in conflict intensity, so that the divergence anchor zone has continuity in the time dimension and a hierarchical structure of conflict strength in the feature dimension, thereby improving the expression accuracy of the interactive conflict area.

[0011] Preferably, the steps for generating the reverse phase leader are as follows: Based on the structural distribution of the divergence anchor point zone, the subtle changes in each anchor point are continuously connected in the order of time evolution, and the anchor points in adjacent time slices are connected in the order of appearance to form a time change chain. The subtle change characteristics of the linked anchor point sequence are analyzed, and directional information reflecting the evolution trend of emotional state is extracted based on the fluctuation amplitude of conflict strength and the transfer direction of the dominant characteristic attribute. Using the abrupt change nodes as key connection points, connect each abrupt change node in chronological order to form a reverse phase guide line, and mark the abrupt change stages and the dominant changes in the characteristic in the path; By refining the temporal structure and integrating the path continuity of the reverse-phase guiding line, the evolutionary path of emotional mutation remains continuous in time and coherent in the characteristic dimension.

[0012] Preferably, the phase-traction window array generation steps are as follows: Using the reverse-phase guiding line as a dual reference line for changes in time and emotional state, an initial threshold window corresponding to the syndrome type is established, and the threshold boundary is limited by time position and characteristic fluctuation amplitude. Based on the time rhythm and abrupt change amplitude of the reverse phase guide line, the initial threshold window is set with elastic characteristics, so that the threshold window can dynamically expand and contract with the rhythm of emotional changes, forming a flexible boundary structure. Based on the time phase characteristics of the reverse-phase guide line, a phase correlation relationship is established between threshold windows, so that multiple elastic threshold windows form a rhythmic phase correspondence chain in time sequence; By combining time phase patterns, multiple elastic threshold windows are integrated into a phase traction window array, enabling the emotional change rhythm and the syndrome recognition process to maintain a time-synchronized response.

[0013] Preferably, the elastic threshold window is dynamically adjusted in the phase traction window array according to the change amplitude of the emotional shift rhythm. When the amplitude of the emotional shift increases, the threshold window boundary expands to cover the feature fluctuation range, and when the emotion tends to be calm, the threshold window boundary shrinks to limit the feature concentration range, thereby maintaining the continuity and stability of the syndrome discrimination.

[0014] Preferably, based on the phase-traction window array, time slot rotation traction is performed, and a shadow diversion gate mechanism is introduced to dynamically regulate the model sensitivity and compliance during the critical period of emotional evolution, so that the syndrome evolution trajectory converges towards the stable recognition area and the emotional syndrome judgment result is output. The steps are as follows: Based on the phase traction window array, the time slot rotation traction operation is performed. Driven by the time phase law, the traction window is activated and released in a cyclical manner in time sequence, so that the syndrome identification process is rhythmically and dynamically promoted in time. Based on time-slot rotation traction, a shadow diversion gate mechanism is introduced to establish a balanced channel between different emotional phases, so as to flexibly allocate multi-source feature information and adjust the feature flow balance. By focusing on the dynamic fluctuations of the critical period of emotional evolution, the sensitivity and compliance of the model are adjusted in stages to keep the model's response characteristics synchronized with the rhythm of emotional changes. The characteristic trajectories after rotation traction and diversion adjustment are integrated into the stable recognition area, so that the syndrome evolution trajectory gradually converges towards the stable area and outputs a clear diagnosis result of emotional and mental disease syndrome.

[0015] A TCM internal medicine emotional syndrome identification system integrating machine learning includes a symptom rearrangement module, a feature overlap parsing module, an anchor point tracking module, a threshold-driven module, and a dynamic convergence module. The time-symptom rearrangement module rearranges the patient's diagnosis and treatment timeline and symptom manifestation timeline around the period of emotional change, identifies the weakened boundary of syndrome characteristics, constructs a characteristic control band, and gathers a list of overlapping critical characteristics within the characteristic control band; The feature overlap analysis module, based on the critical feature overlap list, performs a comparative analysis on each description of tongue appearance, pulse appearance, physical signs and emotions, marks the positions where features cancel each other out, and integrates them to form a divergence anchor point band to reflect the interactive conflict area between different syndrome features. The anchor point tracking module connects the subtle changes within the divergent anchor point bands in chronological order, extracts the direction of sudden changes in emotional evolution, and generates a continuously trackable reverse-phase guide line to characterize the evolutionary path of sudden changes in emotional state. The threshold traction module establishes an elastic threshold window corresponding to the syndrome around the reverse phase guide line, so that the elastic threshold window dynamically opens and closes with the rhythm of sudden emotional changes, and is combined according to the time phase law to form a phase traction window array. The dynamic convergence module performs time slot rotation traction based on the phase traction window array and introduces a shadow diversion gate mechanism to dynamically improve the model sensitivity and reduce the model compliance during the critical period of emotional evolution, so that the syndrome evolution trajectory gradually converges towards the stable recognition area and outputs a clear diagnosis result of emotional syndrome.

[0016] The technical effects and advantages provided by the present invention in the above technical solution are as follows: This invention introduces synchronous rearrangement of timelines and symptom lines, along with dynamic tracking of divergence anchor points, during the emotional abrupt change phase. This enables the syndrome identification process to maintain continuous discriminative ability under conditions of overlapping multidimensional features and weakened boundaries. Through the synergistic effect of the inverse guiding line and the elastic threshold window, the directionality and rhythm of emotional state changes are quantified, thereby maintaining the stability and temporal consistency of syndrome identification during emotional fluctuations and avoiding identification bias caused by feature abrupt changes.

[0017] This invention constructs a phase-traction window array and introduces a shadow diversion gate mechanism, enabling the model to dynamically adjust its response sensitivity and compliance based on the amplitude of characteristic fluctuations during critical periods of emotional evolution, thus forming an adaptive diagnostic control mechanism. This approach allows the trajectory of syndrome evolution to automatically converge to a stable recognition region, achieving stable output of diagnostic conclusions under complex and ever-changing emotional states, thereby improving the reliability and clinical reproducibility of identifying emotional disorders. Attached Figure Description

[0018] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in this invention. For those skilled in the art, other drawings can be obtained based on these drawings.

[0019] Figure 1 This is a flowchart of the method for identifying syndromes of emotional disorders in traditional Chinese medicine internal medicine that integrates machine learning, as described in this invention.

[0020] Figure 2 This is a schematic diagram of the modules of the TCM internal medicine emotional syndrome identification system that integrates machine learning, as described in this invention. Detailed Implementation

[0021] Exemplary embodiments will now be described more fully with reference to the accompanying drawings. However, these exemplary embodiments can be implemented in many forms and should not be construed as limited to the examples set forth herein; rather, they are provided so that the description of this disclosure will be more complete and fully convey the concept of the exemplary embodiments to those skilled in the art.

[0022] This invention provides, for example Figure 1 The method for identifying syndromes of emotional disorders in TCM internal medicine, which integrates machine learning, includes the following steps: Around the period of emotional change, the patient's diagnosis and treatment timeline and symptom manifestation line are rearranged synchronously to identify the weakened boundary of syndrome characteristics, construct characteristic control bands, and compile a list of critical feature overlap within the characteristic control bands; Around the period of emotional abrupt change, the patient's diagnosis and treatment timeline and symptom manifestation timeline are synchronously rearranged to establish a mutual mapping relationship between the time and symptom dimensions. This identifies areas of weakened syndrome feature boundaries and constructs feature control bands accordingly. Within these bands, a list of overlapping critical features is compiled. The specific implementation process is as follows: After identifying the period of emotional abrupt change, a continuous treatment timeline is established based on the patient's emotional state records, descriptions of main symptoms, accompanying signs, and physician observations throughout the entire treatment cycle. This timeline uses emotional changes as its core theme and continuous time slices as the smallest unit, covering the entire process from the initial shift in emotional state to the sudden change in the dominant emotional type. Simultaneously with the timeline construction, symptom manifestation lines are extracted for each period. These symptom manifestation lines, based on patient complaints, physician observations, and auxiliary examination records, reflect the order and duration of various physiological and psychological symptoms over time. By synchronously constructing the treatment timeline and symptom manifestation lines, a horizontal correspondence can be established on the timeline, ensuring that each time slice corresponds to a specific set of symptoms, thus laying the foundation for subsequent synchronous rearrangement of data association.

[0023] After establishing the timeline and symptom presentation line, a synchronized rearrangement operation is performed on the multi-source symptom information within the period of emotional abrupt change. The purpose of synchronized rearrangement is to readjust the symptom presentations from different sources and with different rhythms in the original records onto a unified reference timeline, making the temporal correspondence between emotional reactions and symptom changes closer. To this end, the starting point of the emotional abrupt change is used as the rearrangement benchmark in the timeline, and symptom nodes in the symptom presentation line that are synchronized with the emotional change are moved closer to this benchmark, and the symptom order is rearranged according to the duration and trend of symptom changes. Through this rearrangement, the originally scattered and intertwined symptom evolution trajectories can be integrated into an ordered time sequence, thereby revealing the evolutionary pattern of symptoms during the process of emotional abrupt change more clearly. At this time, the coupling relationship between the diagnosis and treatment timeline and the symptom presentation line is strengthened, providing an observable time-symptom mapping basis for identifying areas where the syndrome boundary is weakened.

[0024] After completing the synchronous rearrangement, the characteristic manifestations during the emotional abrupt change period are analyzed based on the rearrangement results to identify the weakening regions of syndrome feature boundaries. The core of this step lies in observing the degree of overlap, the rate of change, and the intersection intervals of symptoms and signs in consecutive time slices. By comparing the frequency of occurrence, intensity, and co-occurrence patterns of symptom features in adjacent time slices, the critical intervals where emotional states transition from one dominant state to another can be identified. Within these intervals, some symptom features simultaneously exhibit multiple syndrome attributes. For example, symptom clusters with characteristics of both liver stagnation and heart and spleen disorders appear frequently in this stage, indicating that the boundaries between syndromes are weakening. These identified weakening boundary regions represent the areas of mutual penetration of different syndrome features and are key observation areas in the process of emotional abrupt change. By marking these regions on the timeline, the periods of feature overlap can be delineated on the time axis, providing a locational basis for constructing feature control bands.

[0025] After identifying the weakened regions of syndrome characteristics, feature control bands are constructed around these regions, and a list of overlapping critical features is compiled within these control bands. The feature control bands are comprehensive analysis bands established with the timeline as the vertical axis and symptom characteristics as the horizontal axis, used to compare the overlapping relationships of symptom characteristics in different time segments. First, the symptom, sign, and emotional response features within the aforementioned weakened regions are arranged chronologically and compared item by item with the main features of adjacent time segments to select those that appear consecutively in time and exhibit similar characteristics. Second, the frequency of occurrence and the strength of interrelationships of features within each time segment are recorded in the control bands to reveal the superposition patterns between features. Third, items with high degree of feature superposition and obvious attribute overlap are grouped into the same feature cluster to form a set of features with correlational significance. Finally, these feature sets are integrated chronologically to generate a list of overlapping critical features. This list systematically summarizes all elements with overlapping features during periods of emotional abrupt change, preserving the correspondence between symptom manifestations and temporal evolution, and presenting the co-occurrence patterns of different syndrome characteristics within the weakened region.

[0026] Based on the list of overlapping critical features, the descriptions of tongue appearance, pulse appearance, physical signs and emotions are compared and analyzed item by item. The locations where features cancel each other out are marked and integrated to form a divergence anchor zone to reflect the interactive conflict area between different syndrome features. Starting with a list of overlapping critical features, this study focuses on multi-source diagnostic and treatment information during periods of emotional abrupt change. Through layer-by-layer analysis and feature fusion, it reveals the dynamic conflicts and checks and balances between different syndrome features. The specific implementation steps are as follows: Based on the critical feature overlap list, the tongue appearance, pulse, physical signs, and emotional description information involved are systematically organized and categorized. This critical feature overlap list originates from the previous stage's synchronous rearrangement of timelines and symptom lines and the construction of feature control bands, containing overlapping entries and distribution information of various features during periods of emotional abrupt change. In this step, the features in the list are divided into tongue appearance feature group, pulse appearance feature group, physical sign feature group, and emotional description group according to perception type. Each group of features is arranged with time as the vertical axis and feature manifestation as the horizontal axis, forming a two-dimensional feature table that reflects the correlation between time changes and feature manifestations. Through this organization process, features of different dimensions are unified into a corresponding analytical framework, enabling tongue appearance, pulse, physical signs, and emotional descriptions to be compared vertically and correlated horizontally under the same time reference, laying a complete data structure foundation for subsequent comparative analysis.

[0027] After completing feature classification and aggregation, the correspondence between each feature group was analyzed item by item based on the time distribution recorded in the critical feature overlap list. Specifically, features within the same time slice in the tongue appearance group, pulse appearance group, physical signs group, and emotional description group were analyzed side-by-side to observe their consistency or differences in expression direction, intensity changes, and feature attributes. For example, when the tongue appearance feature is red, the pulse feature is wiry, the physical signs feature is flushed face, and the emotional description is irritability, these features show a mutually reinforcing relationship; conversely, when the tongue appearance feature is pale, the pulse feature is slow, the physical signs feature is fatigue, and the emotional description shows urgency and anxiety, it indicates a tendency for mutual cancellation or feature conflict. Through this comparative method, the co-occurrence or opposition of different physiological and emotional features within the critical interval of emotional abrupt change can be visually presented. The results of the comparative analysis were recorded in time series format and marked in the original feature table, making the feature interaction relationship of each time period clearly visible, thus providing a direct basis for identifying feature cancellation positions.

[0028] After completing the feature comparison analysis, feature points exhibiting conflict or mutual cancellation among tongue appearance, pulse, physical signs, and emotional descriptions were marked and extracted. The marking operation was performed on a time slice basis. When opposite directions of feature expression or mutually restrictive trends were detected within the same time slice, that time slice was marked as a feature cancellation location. For example, when tongue appearance changes towards a heat syndrome, while pulse and physical signs lean towards a deficiency-cold syndrome, that time slice was considered a potential syndrome conflict zone. At this stage, not only were specific conflict time locations marked, but also the feature items involved in the cancellation and their degree of expression were recorded to construct a complete conflict feature chain. By continuously comparing conflict locations across multiple time slices, cross-interference zones of multidimensional features during emotional evolution can be identified. This process further extends the static feature correspondence of the previous stage into a dynamic conflict trajectory, revealing the evolutionary law of mutual restriction and alternating dominance of different physiological and psychological features during emotional abrupt changes.

[0029] After marking and extracting the feature cancellation locations, the conflict information from each time slice is vertically integrated to form a divergence anchor band. During integration, using the time axis as the main sequence, all feature cancellation locations are sequentially linked together, grouping temporally adjacent, conflict-attribute-similar, or mutually continuous feature points into the same anchor chain. These chains are then combined to form the overall anchor band structure. Each divergence anchor represents a multi-feature interaction zone, where the feature conflict state reflects the dynamic game relationship between different symptom features. By arranging multiple anchor bands in chronological order, a conflict distribution trajectory spanning the entire emotional abrupt change stage can be formed, indicating the contraction and expansion trends of the symptom boundary during evolution. The formation of the divergence anchor band allows the scattered conflict information in the critical feature overlap list to be presented in a structured manner, providing a continuous reference benchmark for subsequent extraction of the direction of emotional evolution abrupt change.

[0030] By connecting subtle changes within the divergence anchor zone according to the chronological order of evolution, the direction of abrupt changes in emotional evolution is extracted, and a continuously trackable reverse-phase guide line is generated to characterize the evolutionary path of abrupt changes in emotional state. Based on the generated divergence anchor bands, and using time series as the dominant axis and multi-source feature conflicts as the core content, this method gradually extracts inverse guiding lines that reflect the trend of emotional abrupt changes through time series concatenation, trend identification, direction derivation, and continuous construction steps. This enables dynamic tracking and trajectory expression of emotional evolution over time. The specific implementation steps are as follows: Based on the structural distribution of the divergent anchor points, the subtle changes within each anchor point are continuously linked in chronological order. The divergent anchor points originate from the interactive conflict information between tongue appearance, pulse, physical signs, and emotional descriptions. Each anchor point records the location of feature cancellation, conflict intensity, and participating feature type. In this step, using the time axis as a vertical reference, the anchor points of adjacent time slices within the divergent anchor point zone are connected sequentially according to their appearance, forming a continuous chain of temporal changes. To ensure the continuity of the emotional abrupt change process, the direction of change, conflict degree, and feature participation of adjacent anchor points are compared. When adjacent anchor points are temporally continuous and have similar conflict attributes, they are grouped into the same change sequence; when a feature type shifts from one dominant to another, the node is recorded as a key abrupt change node. Through this linking process, the originally scattered divergent anchor point zones are integrated into a conflict trajectory with temporal extension, laying the foundation for subsequent identification of emotional abrupt change trends.

[0031] After completing the temporal chaining of the divergence anchor points, the subtle changes in the chained anchor point sequence are analyzed to extract directional information reflecting the trend of emotional state evolution. Specifically, the increase or decrease of feature changes within each anchor point, the fluctuation range of conflict intensity, and the direction of shift in the dominant attribute of the feature are examined sequentially along the time chain. For example, when in multiple consecutive anchor points, the emotional description feature gradually shifts from depression to irritability, the pulse changes from weak and thready to rapid and wiry, and the physical signs change from pale complexion to flushed complexion, it can be determined that a sudden shift in emotional state from introversion to extroversion is forming during this period. The extraction of such trends not only relies on the change of a single feature but also emphasizes the mutual checks and balances between multidimensional features. By comparing the sequence of subtle changes between anchor points and extracting trends, the directional changes in emotional evolution can be revealed, thus initially determining the temporal direction of the sudden shift. This step transforms the static conflict structure of the divergence anchor point band into a dynamic trend flow, making the emotional evolution process traceable and directional in the time dimension.

[0032] After clarifying the direction and trend of emotional shifts, a continuous temporal correlation is established around these shifts, and a reverse-phase guiding line representing the core path of emotional change is extracted. This step uses shift nodes as key connection points, connecting each shift node in chronological order to form a continuous path, and marking the key stages of the shift and the main content of the dominant changes in the characteristics within the path. The establishment of the reverse-phase guiding line is based on the opposite state changes in emotional evolution, such as internal repression-external expression or emptiness-activity. By comparing the characteristic manifestations before and after the shift node, the reverse trend of emotional change is determined. When the emotional description, tongue appearance, pulse appearance, and physical signs within a time slice form a reverse characteristic combination with the previous stage in terms of the direction of expression, that point is determined to be a reverse-phase turning point. After identifying multiple reverse-phase turning points in continuous time slices, these points are sequentially connected to form a continuously traceable reverse-phase guiding line. This guiding line runs through the entire emotional shift stage, recording the evolution of emotions from stability to fluctuation and then back to stability, forming the main framework of the emotional change path.

[0033] After generating the reverse-phase guideline, its temporal structure is refined and its path continuity is integrated to give the evolutionary path of emotional abrupt changes a complete expressive ability. This step uses the reverse-phase guideline as the central main line, and combines the conflict intensity and dominant feature type recorded in the divergence anchor point band to extend and correct the time slices on both sides of the guideline, incorporating anchor points with slight misalignment or feature delay into the influence range of the guideline. By extending the time slices, the short-term feature fluctuation intervals can be connected with the main direction of abrupt change, avoiding interruptions in the path of emotional change. Furthermore, transitional descriptions of emotional states are added at key nodes of the guideline to reflect the intermediate stage characteristics of the transformation of emotional states from one dominant to another, such as from depression to anger, or from thought to grief. The final reverse-phase guideline maintains continuity in the temporal dimension and coherence in the feature dimension, clearly depicting the relative movement trends and alternating relationships of various features during the emotional abrupt change, thus providing a structured directional basis for the subsequent threshold construction and dynamic guidance.

[0034] An elastic threshold window corresponding to the syndrome type is established around the reverse phase guide line, so that the elastic threshold window dynamically opens and closes with the rhythm of sudden change in emotion, and is combined according to the time phase law to form a phase traction window array; This step, based on the results of the reverse-phase guideline, uses the directional evolutionary trajectory during the abrupt change in emotional state as the core reference. Through four consecutive steps—threshold setting, flexible adjustment, phase correlation, and window array construction—it gradually forms a threshold control structure that can track emotional changes in time and adapt to their amplitude. This enables dynamic adaptation of the syndrome differentiation range and synchronous response of time and phase within the diagnostic information flow. The specific implementation process is as follows: Using the reverse-phase guideline as a dual reference line for changes in time and emotional state, an initial threshold window corresponding to the syndrome type is established. The reverse-phase guideline records the time sequence of abrupt changes in the direction of emotional evolution and the dominant emotional state, reflecting the inherent rhythm of the transformation of emotions from one dominant state to another. Based on this guideline, a corresponding syndrome type region is selected near each abrupt change node. According to the main syndrome type characteristics of emotional disorders in traditional Chinese medicine, the symptom characteristics, tongue characteristics, pulse characteristics, and physical signs related to the direction of emotional change are sequentially divided into several segments, with each syndrome type segment forming an initial threshold window. The boundary of this threshold window is jointly defined by the temporal position of the reverse-phase guideline and the amplitude of characteristic fluctuations. The inner side represents the range where the syndrome type characteristics tend to be stable, while the outer side represents the range where the emotional characteristics are still in a transitional or unstable stage. By establishing the initial threshold window, a structural boundary is provided for subsequent dynamic opening and closing regulation, enabling syndrome identification to have a stage-based hierarchical structure in time and a continuous correspondence in characteristics.

[0035] After forming the initial threshold window, the threshold window is flexibly configured based on the temporal rhythm and abrupt change amplitude of the reverse-phase guideline. The purpose of this step is to allow the threshold window to expand and contract appropriately with the rhythm of emotional changes, adapting to the actual differences in characteristic fluctuations at different stages. In practice, firstly, the rhythmic points of emotional abrupt changes in the reverse-phase guideline are identified, i.e., the time nodes when the emotional state accelerates, reverses, or stabilizes before oscillation. Then, using the rhythmic point as the center, the amplitude and duration of changes in emotional characteristics within the preceding and following time periods are measured, and the opening and closing range of the threshold window is adjusted according to this dynamic trend. When the amplitude of emotional abrupt changes is large and the symptom fluctuations are significant, the boundary of the threshold window automatically expands to cover a wider range of characteristic fluctuations, thereby preventing instability in syndrome differentiation due to excessive abrupt changes. When emotional changes tend to be mild and characteristics tend to stabilize, the boundary of the threshold window gradually contracts to limit the concentrated range of syndrome characteristics, ensuring stable convergence of the identification results. Through this flexible setting, the threshold window transforms from a fixed range into a dynamically responsive flexible boundary structure, enabling adaptive adjustment of syndrome differentiation during emotional abrupt changes.

[0036] After completing the elastic adjustment of the threshold windows, a phase correlation is established between the threshold windows based on the temporal phase characteristics of the inverse leading line. Emotional abrupt changes typically exhibit periodic rhythmic characteristics, with sequential responses among characteristic fluctuations. To capture this relationship, this step uses the time scale of the inverse leading line as the main line, arranging the elastic threshold windows within adjacent time intervals in chronological order, and forming a phase correspondence chain based on the difference between the positive and negative phases of the abrupt change direction. Each threshold window is assigned a phase identifier in the time series to indicate whether its corresponding emotional state is in the rising, reversing, or converging phase of the abrupt change rhythm. When the previous threshold window is in the emotional intensification phase, the next threshold window corresponds to the emotional relief or turning phase, forming a complementary phase relationship between the two threshold windows. Through this phase correlation, multiple elastic threshold windows no longer exist independently in the time dimension, but rather are mutually pulled in a rhythmic manner, thus forming a threshold sequence with an inherent coupling relationship that progresses over time. This structure maintains the continuity of the emotional abrupt change process and provides a temporal phase registration basis for the subsequent construction of the traction window array.

[0037] After establishing the phase correlation of threshold windows, multiple elastic threshold windows are integrated into a phase-traction window array by combining them according to temporal phase patterns. This step uses the time series as the vertical axis and the threshold window phase distribution as the horizontal structure, stacking threshold windows within the same rhythmic period in chronological order to form a traction window array with periodic response characteristics. Each row in the phase-traction window array represents the change process of emotional state within a sudden change cycle, and each column represents the response interval of different feature dimensions at the same time phase. When the emotional sudden change rhythm enters a rapid change phase, the connection density between adjacent threshold windows in the traction window array increases, the traction effect strengthens, and the dynamic adjustment speed of the syndrome features increases accordingly. When the emotional state enters a moderate or stable phase, the phase interval of the threshold windows in the traction window array increases, the traction effect weakens, and the syndrome discrimination tends to converge. Through this traction window array formed by combining according to phase patterns, the emotional change rhythm and the syndrome identification process form a dynamic mapping, enabling the identification model to maintain a time-synchronous response to sudden change trends at different emotional stages. The phase traction window array not only reflects the opening and closing patterns of characteristic boundaries under the rhythm of sudden emotional changes, but also provides a structural carrier that can be continuously followed for subsequent dynamic traction regulation.

[0038] Based on the phase-traction window array, time slot rotation traction is performed, and a shadow diversion gate mechanism is introduced to dynamically improve the model sensitivity and reduce the model compliance during the critical period of emotional evolution, so that the syndrome evolution trajectory gradually converges towards the stable recognition area, and outputs a clear diagnosis result of emotional syndrome. Based on a phase-driven window array and using the temporal phase law as the main thread, a dynamic balance control mechanism is established for periods of sudden emotional change through time slot rotation, diversion gate setting, dynamic regulation, and result convergence steps. This mechanism ensures that syndrome identification maintains the consistency of judgment and the focus of output even under changing emotional states. The specific implementation process is as follows: A time-slot rotation traction operation is performed based on a phase-traction window array. The phase-traction window array is formed by combining multiple elastic threshold windows according to a temporal phase pattern, with each threshold window corresponding to a specific phase interval in the emotional evolution process. Time-slot rotation traction involves cyclically activating and releasing the traction windows in chronological order, driven by the periodic opening and closing of each phase window within the time series. In specific implementation, adjacent phase-traction windows are divided into several time-slot units based on the rate of change of the emotional abrupt change rhythm, with each time-slot unit corresponding to a responsive traction cycle. When a time-slot unit is in a traction state, other phase windows remain in a standby state to prevent overlapping interference of feature responses. Through this rotation method, the syndrome identification process forms a rhythmic dynamic progression in time, that is, capturing feature changes concentratedly in one period and transitioning to a balanced and gradual release phase in the next period. This alternating traction structure helps maintain the consistency of model output during periods of frequent emotional abrupt changes and prevents judgment drift caused by abrupt changes in emotional features. This step, through the cyclical traction of time phase, enables the phase traction window array to have rhythm switching capabilities, providing the basic structure for subsequent diversion and regulation.

[0039] Building upon the time-slot rotation traction mechanism, a shadow diversion gate mechanism is introduced to regulate the balanced distribution of multidimensional feature flows during emotional upheavals. The shadow diversion gate is an auxiliary control structure referenced to time phase. Its function is to establish a balanced channel between different emotional phases, allowing for flexible distribution of multi-source feature information as it flows through the traction window array. When the rhythm of emotional upheaval changes rapidly and feature fluctuations are severe, the shadow diversion gate opens an additional slow-release pathway, temporarily guiding highly volatile features to the diversion zone to reduce feature interference within the main traction window. When the emotional state stabilizes, the diversion gate gradually closes, restoring the concentrated input of feature flows and returning the discrimination process to the main guidance path. The introduction of the shadow diversion gate enables the flow of feature information to have elastic adjustment capabilities, thus avoiding discrimination instability caused by excessive feature superposition during the upheaval phase. Through this mechanism, the identification of emotional syndromes can remain flexible and adaptable during the upheaval period, avoiding misjudgment due to oversensitivity or recognition delay due to slow response, achieving dynamic balance control of multi-source features.

[0040] After establishing the shadow diversion mechanism, the sensitivity and compliance of the model are adjusted in stages around the dynamic characteristic fluctuations during the critical period of emotional evolution. Sensitivity reflects the system's responsiveness to feature changes, while compliance reflects the system's continuation of the judgment trend of the previous stage. In the early stage of emotional abrupt change, feature changes are drastic and highly differentiated. At this time, increasing sensitivity can enhance the response to sudden features, enabling the model to capture the direction of emotional changes in a timely manner. However, when emotional abrupt change enters a stage of repeated fluctuations, excessively high sensitivity can easily cause frequent switching of the judgment path. Therefore, compliance is gradually reduced to give the system greater freedom of response, thereby maintaining judgment flexibility in a volatile environment. Subsequently, in the stage where the emotional state tends to be stable and feature fluctuations weaken, sensitivity is reduced and compliance is moderately increased to make the system tend to continue the judgment trajectory of the previous stage, thereby guiding the syndrome evolution results to gradually stabilize. Through this adjustment process, a dynamic balance between sensitivity and compliance is established at different stages, keeping the model's response characteristics synchronized with the rhythm of emotional abrupt change. This step uses the shadow diversion gate as an auxiliary fulcrum to dynamically adjust the response amplitude in the characteristic flow direction, thereby achieving adaptive balance in the process of identifying symptoms during abrupt changes.

[0041] After dynamic regulation of sensitivity and compliance, the feature trajectories, adjusted through rotational traction and diversion, are integrated into a stable identification zone to form the final convergence process of syndrome evolution. This step is time-phase-driven, longitudinally superimposing the syndrome change paths within multiple traction cycles on the time axis, and identifying intervals where feature fluctuations slow down and change directions converge as stable identification zones. As the proportion of feature flow passing through the shadow diversion gate gradually decreases, it indicates that the emotional state is transitioning from abrupt change to a stable phase, at which point the syndrome evolution trajectory gradually converges towards the stable zone. The convergence process is not a simple static pause, but rather a continuous advancement through time-slot rotational traction, causing the feature trajectory to gradually tighten over time, ultimately converging in the core area of ​​the stable identification zone. Within this area, the correlation features of tongue appearance, pulse appearance, physical signs, and emotional descriptions tend to be consistent, the boundaries of syndrome features are clear, and the system outputs a clear diagnosis of emotional disorders. Through this process, the entire identification chain gradually shifts from initial dynamic response and abrupt adaptation to balanced convergence, completing the entire process from multi-source feature conflict to stable diagnostic output.

[0042] This invention introduces synchronous rearrangement of timelines and symptom lines, along with dynamic tracking of divergence anchor points, during the emotional abrupt change phase. This enables the syndrome identification process to maintain continuous discriminative ability under conditions of overlapping multidimensional features and weakened boundaries. Through the synergistic effect of the inverse guiding line and the elastic threshold window, the directionality and rhythm of emotional state changes are quantified, thereby maintaining the stability and temporal consistency of syndrome identification during emotional fluctuations and avoiding identification bias caused by feature abrupt changes.

[0043] This invention constructs a phase-traction window array and introduces a shadow diversion gate mechanism, enabling the model to dynamically adjust its response sensitivity and compliance based on the amplitude of characteristic fluctuations during critical periods of emotional evolution, thus forming an adaptive diagnostic control mechanism. This approach allows the trajectory of syndrome evolution to automatically converge to a stable recognition region, achieving stable output of diagnostic conclusions under complex and ever-changing emotional states, thereby improving the reliability and clinical reproducibility of identifying emotional disorders.

[0044] This invention provides, for example Figure 2 The TCM internal medicine emotional syndrome identification system, which integrates machine learning, includes a symptom rearrangement module, a feature overlap parsing module, an anchor point tracking module, a threshold traction module, and a dynamic convergence module. The time-symptom rearrangement module rearranges the patient's diagnosis and treatment timeline and symptom manifestation timeline around the period of emotional change, identifies the weakened boundary of syndrome characteristics, constructs a characteristic control band, and gathers a list of overlapping critical characteristics within the characteristic control band; The feature overlap analysis module, based on the critical feature overlap list, performs a comparative analysis on each description of tongue appearance, pulse appearance, physical signs and emotions, marks the positions where features cancel each other out, and integrates them to form a divergence anchor point band to reflect the interactive conflict area between different syndrome features. The anchor point tracking module connects the subtle changes within the divergent anchor point bands in chronological order, extracts the direction of sudden changes in emotional evolution, and generates a continuously trackable reverse-phase guide line to characterize the evolutionary path of sudden changes in emotional state. The threshold traction module establishes an elastic threshold window corresponding to the syndrome around the reverse phase guide line, so that the elastic threshold window dynamically opens and closes with the rhythm of sudden emotional changes, and is combined according to the time phase law to form a phase traction window array. The dynamic convergence module performs time slot rotation traction based on the phase traction window array and introduces a shadow diversion gate mechanism to dynamically improve the model sensitivity and reduce the model compliance during the critical period of emotional evolution, so that the syndrome evolution trajectory gradually converges towards the stable recognition area and outputs a clear diagnosis result of emotional syndrome.

[0045] The method for identifying syndromes of emotional disorders in TCM internal medicine by incorporating machine learning provided in this embodiment of the invention is implemented through the aforementioned system for identifying syndromes of emotional disorders in TCM internal medicine by incorporating machine learning. For details of the specific methods and processes of the system for identifying syndromes of emotional disorders in TCM internal medicine by incorporating machine learning, please refer to the embodiments of the method for identifying syndromes of emotional disorders in TCM internal medicine by incorporating machine learning, which will not be repeated here.

[0046] The foregoing has only described certain exemplary embodiments of the present invention by way of illustration. Undoubtedly, those skilled in the art can modify the described embodiments in various ways without departing from the spirit and scope of the present invention. Therefore, the foregoing drawings and descriptions are illustrative in nature and should not be construed as limiting the scope of protection of the claims of the present invention.

Claims

1. A method for identifying syndromes of emotional disorders in Traditional Chinese Medicine internal medicine by integrating machine learning, characterized in that, Includes the following steps: Around the period of emotional change, the patient's diagnosis and treatment timeline and symptom manifestation line are rearranged synchronously to identify the weakened boundary of syndrome characteristics, construct characteristic control bands, and compile a list of critical feature overlap within the characteristic control bands; Based on the list of overlapping critical features, the descriptions of tongue appearance, pulse appearance, physical signs and emotions are compared and analyzed item by item. The positions where features cancel each other out are marked and integrated to form a divergence anchor point zone. By connecting the subtle changes within the divergence anchor zone according to the chronological evolution, the sudden shifts in emotional evolution are extracted to generate reverse phase guide lines; An elastic threshold window corresponding to the syndrome type is established around the reverse phase guide line, so that the elastic threshold window dynamically opens and closes with the rhythm of sudden emotional changes, and is combined according to the time phase law to form a phase traction window array; Based on the phase-traction window array, time slot rotation traction is implemented, and a shadow diversion gate mechanism is introduced to dynamically improve the model sensitivity and reduce the model compliance during the critical period of emotional evolution, so that the syndrome evolution trajectory gradually converges towards the stable recognition area, and outputs a clear diagnosis result of emotional syndrome.

2. The method for identifying TCM internal medicine emotional syndromes by integrating machine learning according to claim 1, characterized in that, The steps for forming the critical feature overlap list are as follows: Establish a patient diagnosis and treatment timeline and symptom manifestation line around the period of emotional change. The diagnosis and treatment timeline is based on the emotional changes, and the symptom manifestation line is based on the patient's chief complaint, physician's observation and examination records. Based on the starting point of emotional change, the diagnosis and treatment timeline and symptom manifestation timeline are rearranged synchronously. Symptom nodes from different sources are rearranged according to their duration and trend of change, so that emotional reactions and symptom changes correspond in time. Based on the synchronous rearrangement results, the overlapping intervals of symptoms and signs during the period of emotional change are identified, the weakened areas of syndrome feature boundaries are determined, and the overlapping periods are marked on the time axis. A feature comparison band is constructed around the boundary weakening area. Symptoms, signs and emotional features in continuous time slices are compared and analyzed to form a feature set and integrate them in chronological order to generate a critical feature overlap list.

3. The method for identifying syndromes of emotional disorders in traditional Chinese medicine internal medicine by incorporating machine learning according to claim 2, characterized in that, When constructing the feature control band, the timeline is used as the vertical axis and the symptom features are used as the horizontal axis. The symptoms, signs and emotional reactions during the period of emotional change are arranged in chronological order, and the frequency of occurrence and correlation strength of the features in each time slice are recorded. Items with high degree of feature overlap and obvious attribute intersection are grouped into the same feature set to form a critical feature overlap list.

4. The method for identifying syndromes of emotional disorders in traditional Chinese medicine internal medicine by incorporating machine learning according to claim 2, characterized in that, The formation process of the divergence anchor zone is as follows: Based on the list of overlapping critical features, the information on tongue appearance, pulse appearance, physical signs and emotional descriptions is organized and classified, and divided into tongue appearance feature group, pulse appearance feature group, physical sign feature group and emotional description group according to perception type. Based on the time distribution in the list of overlapping critical features, the correspondence between each feature group is analyzed item by item, and the co-occurrence and opposition of different features in the same time slice are recorded. Feature points that conflict between tongue appearance, pulse appearance, physical signs and emotional descriptions are marked and extracted to form feature cancellation positions and construct conflict feature chains. Using the timeline as the main sequence, feature points with similar conflict attributes and continuous relationships in the time slice are sequentially linked and integrated to form a divergence anchor zone, which is used to reflect the interactive conflict area between different syndrome features.

5. The method for identifying syndromes of emotional disorders in traditional Chinese medicine internal medicine by incorporating machine learning according to claim 4, characterized in that, In the formation of the divergence anchor zone, the feature points with continuously changing conflict attributes are sequentially connected according to the time sequence, based on the time axis. During the connection process, the feature points are layered and integrated according to the change in conflict intensity, so that the divergence anchor zone has continuity in the time dimension and a hierarchical structure of conflict strength in the feature dimension, thereby improving the expression accuracy of the interactive conflict area.

6. The method for identifying syndromes of emotional disorders in traditional Chinese medicine internal medicine by incorporating machine learning according to claim 4, characterized in that, The steps for generating a reverse-phase lead wire are as follows: Based on the structural distribution of the divergence anchor point zone, the subtle changes in each anchor point are continuously connected in the order of time evolution, and the anchor points in adjacent time slices are connected in the order of appearance to form a time change chain. The subtle change characteristics of the linked anchor point sequence are analyzed, and directional information reflecting the evolution trend of emotional state is extracted based on the fluctuation amplitude of conflict strength and the transfer direction of the dominant characteristic attribute. Using the abrupt change nodes as key connection points, connect each abrupt change node in chronological order to form a reverse phase guide line, and mark the abrupt change stage and the dominant characteristic changes in the path; By refining the temporal structure and integrating the path continuity of the reverse-phase guiding line, the evolutionary path of emotional mutation remains continuous in time and coherent in the characteristic dimension.

7. The method for identifying syndromes of emotional disorders in traditional Chinese medicine internal medicine by incorporating machine learning according to claim 6, characterized in that, The steps for generating a phase-traction window array are as follows: Using the reverse-phase guiding line as a dual reference line for changes in time and emotional state, an initial threshold window corresponding to the syndrome type is established, and the threshold boundary is limited by time position and characteristic fluctuation amplitude. Based on the time rhythm and abrupt change amplitude of the reverse phase guide line, the initial threshold window is set with elastic characteristics, so that the threshold window can dynamically expand and contract with the rhythm of emotional changes, forming a flexible boundary structure. Based on the time phase characteristics of the reverse-phase guide line, a phase correlation relationship is established between threshold windows, so that multiple elastic threshold windows form a rhythmic phase correspondence chain in time sequence; By combining time phase patterns, multiple elastic threshold windows are integrated into a phase traction window array, enabling the emotional change rhythm and the syndrome recognition process to maintain a time-synchronized response.

8. The method for identifying syndromes of emotional disorders in traditional Chinese medicine internal medicine by incorporating machine learning according to claim 7, characterized in that, The elastic threshold window dynamically adjusts within the phase traction window array according to the amplitude of changes in the emotional shift rhythm. When the amplitude of the emotional shift increases, the threshold window boundary expands to cover the characteristic fluctuation range, and when the emotional state tends to calm down, the threshold window boundary contracts to limit the range of characteristic concentration, thereby maintaining the continuity and stability of the syndrome differentiation.

9. The method for identifying syndromes of emotional disorders in traditional Chinese medicine internal medicine by incorporating machine learning according to claim 7, characterized in that, Based on the phase-traction window array, time slot rotation traction is implemented, and a shadow diversion gate mechanism is introduced. During the critical period of emotional evolution, the model sensitivity and compliance are dynamically adjusted to make the syndrome evolution trajectory converge towards the stable recognition area and output the emotional syndrome judgment result. The steps are as follows: Based on the phase traction window array, the time slot rotation traction operation is performed. Driven by the time phase law, the traction window is activated and released in a cyclical manner in time sequence, so that the syndrome identification process is rhythmically and dynamically promoted in time. Based on time-slot rotation traction, a shadow diversion gate mechanism is introduced to establish a balanced channel between different emotional phases, so as to flexibly allocate multi-source feature information and adjust the feature flow balance. By focusing on the dynamic fluctuations of the critical period of emotional evolution, the sensitivity and compliance of the model are adjusted in stages to keep the model's response characteristics synchronized with the rhythm of emotional changes. The characteristic trajectories after rotation traction and diversion adjustment are integrated into the stable recognition area, so that the syndrome evolution trajectory gradually converges towards the stable area and outputs a clear diagnosis result of emotional and mental disease syndrome.

10. A TCM internal medicine emotional syndrome identification system integrating machine learning, used to implement the TCM internal medicine emotional syndrome identification method integrating machine learning as described in any one of claims 1-9, comprising a symptom rearrangement module, a feature overlap parsing module, an anchor point tracking module, a threshold traction module, and a dynamic convergence module: The time-symptom rearrangement module rearranges the patient's diagnosis and treatment timeline and symptom manifestation timeline around the period of emotional change, identifies the weakened boundary of syndrome characteristics, constructs a characteristic control band, and gathers a list of overlapping critical characteristics within the characteristic control band; The feature overlap analysis module, based on the critical feature overlap list, performs a comparative analysis of tongue appearance, pulse appearance, physical signs and emotional descriptions item by item, marks the positions where features cancel each other out, and integrates them to form a divergence anchor point band. The anchor point tracking module connects the subtle changes within the divergent anchor point zone according to the time evolution sequence, extracts the sudden change direction of emotional evolution, and generates the reverse phase guide line; The threshold traction module establishes an elastic threshold window corresponding to the syndrome around the reverse phase guide line, so that the elastic threshold window dynamically opens and closes with the rhythm of sudden emotional changes, and is combined according to the time phase law to form a phase traction window array. The dynamic convergence module performs time slot rotation traction based on the phase traction window array and introduces a shadow diversion gate mechanism to dynamically improve the model sensitivity and reduce the model compliance during the critical period of emotional evolution, so that the syndrome evolution trajectory gradually converges towards the stable recognition area and outputs a clear diagnosis result of emotional syndrome.

Images