32 results about "Expression Feature" patented technology

This invention discloses a method and system for pain threshold determination based on multimodal automatic laser stimulation and animal behavior analysis. The method includes: using an RVC-Pose convolutional neural network model to detect key points on the animal's foot and outputting the foot's spatial coordinates in real time; positioning a laser spot on the foot and outputting continuously adjustable laser stimulation according to a preset gradient; extracting facial key points and micro-expression features using a Light-Face facial recognition algorithm, and / or reconstructing the animal's three-dimensional skeleton using a multi-view geometric reconstruction algorithm and extracting pain-related behavioral features; inputting the multimodal behavioral data into a multi-parameter fusion judgment model to automatically identify pain responses, adaptively adjusting the stimulation intensity under gradient stimulation mode, terminating stimulation and recording the pain threshold when an effective pain response is detected. This invention achieves fully automatic closed-loop control for pain threshold determination, solving the problems of inaccurate positioning, imprecise stimulation, and subjective judgment in existing technologies, significantly improving the objectivity and accuracy of pain assessment.

This application discloses an expression recognition method and apparatus, belonging to the field of artificial intelligence technology. The method includes: extracting features from a first set of video frames corresponding to a first video to obtain a first image feature set corresponding to the first set of video frames, wherein all video frames in the first set of video frames include facial images; inputting the first image feature set into a first expression recognition model, and performing denoising processing on the image features in the image feature set through a denoising module in the first expression recognition model to obtain a first expression feature set, wherein the first expression feature set includes expression features of facial images in video frames and expression change features of facial images between every two video frames; and reconstructing expressions based on the first expression feature set through the first expression recognition model to recognize facial expressions in the first video.

This invention proposes a text detection method generated by a large language model, comprising: segmenting the input text into multiple basic text units; extracting the rhetorical relationship types between the multiple basic text units and constructing a multi-relationship graph; extracting the first semantic expression features of the text segment corresponding to the first node and the second semantic expression features corresponding to the second node through a pre-trained language model; inputting the multi-relationship graph into a multi-relationship graph neural network model for feature learning, updating the first and second semantic expression features; and performing graph readout processing on the learned and updated first and second semantic expression features to generate probability distributions of different fine-grained categories corresponding to the input text. This method achieves fine-grained text detection.

This application discloses a method, apparatus, device, medium, and program product for identifying student psychological states, aiming to address the problems in existing technologies such as the difficulty in real-time, comprehensive, and dynamic identification of students' psychological states, and the lack of effective early detection and warning capabilities. The method includes: acquiring a multi-dimensional behavioral feature sequence of the target student, which includes at least an expression feature sequence, a behavioral trajectory feature sequence, and a voice emotion feature sequence; inputting the multi-dimensional behavioral feature sequence into a trained psychological state identification model to output psychological state representation information for characterizing the target student's psychological state; and identifying the target student's psychological state based on the psychological state representation information.

This invention relates to the field of health data analysis technology, and particularly to a method and application for analyzing the correlation between eating micro-expressions and food satisfaction based on deep learning. The method includes the following steps: S1, capturing facial video images of users during the eating process using a camera device, and extracting eating facial feature sequences from them; S2, separating chewing actions and facial muscle movements from the eating facial feature sequences based on the spatial displacement differences of facial key points, and obtaining target micro-expression feature vectors. In this invention, by cross-evaluating the captured eating micro-expression emotional state with the user's physiological health indicators, nutritional threshold constraints and weight corrections are applied to the initial food preferences generated based on emotions. This allows for strict control of the intake of core risk nutrients while catering to the user's personal taste satisfaction, ultimately generating personalized recommended recipes that balance emotional experience and medical health standards.

The application relates to the technical field of artificial intelligence, affective computing and computer-assisted teaching, in particular to a sign language emotion recognition and teaching feedback method based on machine learning, which collects learner sign language videos and extracts face, hand and posture holographic key points; action semantic features and emotion expression features are obtained in parallel by using a double-branch time sequence coding network; the sign language content and the emotion state containing continuous values of valence-arousal-dominance and discrete categories are obtained by a semantic recognition subnetwork and an emotion recognition subnetwork respectively; the recognition result is compared with standard semantics and emotion labels, and emotion intensity, naturalness, semantic matching degree scores and comprehensive quality scores are calculated; emotion correction instructions, reinforcement learning training sequence planning and key point trajectory visualization feedback are generated based on the scores; and teaching strategies are iteratively improved through closed-loop optimization and emotion resonance models. The application realizes synchronous recognition and quantitative evaluation of sign language semantics and emotions, and provides an adaptive feedback means for sign language emotion expression ability training.

This invention relates to the field of artificial intelligence technology and discloses a method for generating instant empathetic responses for robots based on micro-expression triggers. The method includes collecting facial micro-expression data of the interacting object, generating standardized micro-expression data through preprocessing, generating feature vector data based on micro-expression feature extraction, generating classification data by combining emotional state analysis, generating trigger signals for empathetic responses, generating empathetic response text data using natural language generation technology, and outputting instant empathetic responses through speech synthesis. Simultaneously, the micro-expression analysis model is optimized based on interaction feedback data. This invention ensures the reliability of the data foundation for generating empathetic response trigger signals, guarantees the real-time performance and reliability of interactive responses, and improves the friendliness and overall effect of robot interaction.

The application provides an image processing method and device, electronic equipment and computer readable storage medium; the method comprises: acquiring image data of a facial expression; extracting a first expression feature and a first identity feature from the image data; performing feature splicing on the first expression feature and the first identity feature to obtain a first joint feature, and extracting joint information between the first expression feature and the first identity feature from the first joint feature; estimating mutual information between the first expression feature and the first identity feature, adjusting the first expression feature to a second expression feature with the mutual information minimized as the target; and predicting a classification result of the facial expression based on the second expression feature and the joint information; in this way, expression and identity collaborative modeling is realized.

The application relates to the technical field of artificial intelligence and psychological interviews, in particular to an emotional conversation intelligent interview method and system, which comprises the following steps: collecting voice, text and micro-expression features through a multi-modal sensing unit; constructing a semantic-emotion coupling feature model, monitoring the mapping relationship between semantic logic and emotional fluctuation and calculating a conflict coefficient; when the semantic feedback and the emotional valence are negatively correlated, triggering a contradiction seeking logic; using a heterogeneous graph converter to construct interview multi-dimensional features into a heterogeneous graph, generating a context vector through an attention mechanism; generating a guiding prompt word with an emotional regulation factor based on the vector, guiding the interviewee to release real psychological motivation. The application can accurately identify the real intention hidden under the disguise, eliminate individual expression differences, balance deep mining and empathy injection, and realize self-adaptive and high-precision collection of interview information.

This application relates to the field of computer vision technology and discloses a method and system for detecting pain expressions. The method includes: acquiring a facial video stream, segmenting muscle regions and extracting texture features after facial key point localization and inter-frame alignment; calculating motion energy based on these features, extracting enhanced micro-expression temporal segments and multi-scale spatiotemporal features, constructing a muscle dynamics model and completing state estimation; generating an adaptive candidate spatiotemporal window set through spatiotemporal attention weighted fusion; performing temporal modeling for dynamic feature fusion, and selecting the optimal window by fusing micro-expression features; and outputting the final pain level after double consistency verification and loop closure optimization. This application overcomes the limitations of static images, accurately captures facial dynamics and temporal changes, strengthens feature correlation, and significantly improves detection accuracy.

An expression generation device includes a memory and processing circuitry configured to acquire, as processing targets, a face feature related to a face image of a person and a brainwave of the person, estimate an expression feature of the acquired brainwave by using a first model obtained by performing learning on a relationship between a brainwave and an expression feature related to an expression image, which is an image representing an expression, and generate a face image with an expression by using the face feature and the estimated expression feature.

This invention discloses a cross-domain emotion recognition method based on cue learning, belonging to the field of multimodal emotion recognition technology. In the feature alignment stage, this invention introduces a spatial-channel collaborative attention module within the CLIP multimodal framework to enhance the capture of micro-expression features in low-light blurred regions, achieving effective alignment of emotional features in the visible and low-light domains. In the recognition and inference stage, this method employs a semantically guided contextual cue learning approach, fully utilizing the knowledge of the pre-trained CLIP model to construct a text representation more adapted to the target domain features for each emotion category. A visual-text dual-path collaborative optimization framework is designed, achieving effective alignment and robust recognition of emotional features in the visible and low-light domains through cross-modal contrastive learning and domain adversarial loss. This invention significantly improves the model's generalization ability and recognition performance in cross-illumination domain scenarios.

This invention discloses a live streaming content compliance risk control method and system based on screening rules, relating to the field of streaming media content risk control technology. The proposed solution includes acquiring a real-time live stream, performing temporal slicing on the live stream, and performing visual object recognition and visual expression recognition on the video frames in the current slice to obtain object feature fragments, expression feature fragments, and their temporal position information. This application changes the existing approach of treating subsequent information as merely supplementary information by first associating and clustering object feature fragments and expression feature fragments appearing in stages across time units into matching fragment groups, and then concatenating them according to temporal position information to form a risk chain to be closed. This solves the technical defect that risk information cannot be closed and identified for a long time when it appears in stages across time units, and achieves continuous closed tracking of the same live streaming display chain, thereby reducing continuous missed judgments due to insufficient evidence in a single time unit.

The application discloses an ECA lightweight facial expression recognition method based on edge cloud cooperation, relates to the technical field of computer vision and artificial intelligence, and comprises the following steps: S1, training a general model in the cloud; S2, edge model training: S21, dividing an expression data set into a training set and a test set, and preprocessing the data set; S22, loading and freezing the first four convolutional layers of the cloud pre-training, and only fine-tuning the subsequent network layers in the training process; S23, introducing an ECA attention mechanism on the basis of transfer learning, highlighting key expression regions and suppressing background interference; S24, adopting deep separable convolution and H-swish activation functions in the subsequent layers, and reducing the parameter quantity and the calculation amount; and S25, using a loss function FocalLoss to solve the uneven quantity problem among expression categories in the data set. Finally, expression feature extraction and classification are independently completed at the edge, and lightweight and efficient expression recognition is realized.

The application provides an autism emotion ability evaluation method based on electroencephalogram and visual emotion features, acquires electroencephalogram signals and facial video data of an autism subject, respectively extracts electroencephalogram emotion features and visual emotion features, the electroencephalogram features include frequency band energy calculation and brain region function connection analysis, the visual features include facial key point expression features and emotion arousal degree sequences. Through an emotion arousal degree screening mechanism, effective emotion segments are screened, and multi-modal features are input into a structured Prompt reasoning model for semantic analysis to generate an emotion ability analysis report of the subject. According to the analysis result, scores of the subject in emotion recognition ability, emotion understanding ability and emotion regulation ability and other dimensions are output, and objective and quantitative evaluation of the emotion ability of an autism individual is realized. Through the cooperative processing of electroencephalogram and visual information, the evaluation accuracy and reliability can be improved, and scientific basis and application value are provided for early screening, intervention assistance and individualized rehabilitation.

The application discloses a multi-modal emotion recognition method based on confidence dynamic gating, acquires voice and video data in a natural interaction scene, obtains a text sequence through voice recognition, and establishes a speech-level multi-modal sample index; forms a window-level multi-modal sample through preprocessing, and respectively extracts text semantic features, voice acoustic features and visual dynamic expression features; calculates three-modal confidence based on each modal prediction probability distribution, determines an effective modal set in combination with an effective step mask and a modal loss or abnormal mask; generates initial fusion weights under the constraint of the effective modal set, and obtains fused emotion features through gating fusion; outputs a window-level emotion recognition result, aggregates the result into a speech-level and a conversation-level result after stability correction, thereby reducing low-reliability modal and invalid modal interference and reducing continuous recognition result abnormal jumps.

An audio synthesis method, model training method, apparatus, device, medium, and product are disclosed. The method includes: converting reference audio into a first audio with a specified timbre, wherein the timbre of the reference audio differs from the specified timbre, and the specified timbre is selected from a variety of candidate timbres; obtaining first text, extracting semantic features from the first text, and converting the first text into a first phoneme, wherein the first text describes the content of the audio to be synthesized; extracting first timbre features and first expression features from the first audio, wherein the first expression features include at least one of a first prosodic feature or a first emotional feature; concatenating the first text semantic features, the first phoneme, and the first expression features to predict the audio content, thereby obtaining a first audio content representation; and fusing the first audio content representation, the first phoneme, and the first timbre features to obtain a target audio with the specified timbre. This method can improve the controllability of timbre and the accuracy of audio synthesis.

This application provides a training method, apparatus, medium, and electronic device for a driver expression recognition model. The method includes: acquiring multiple batches of image training sets, and training the driver expression recognition model using these batches to reach a preset training batch size. The driver expression recognition model of this application, based on the structural characteristics of a Transformer encoder, is combined with a ResNet18 residual network, and a center loss function is introduced to improve the distribution of expression features. The driver expression recognition model strengthens the correlation between long-distance feature information in expression images, enabling it to extract discriminative feature information. Training parameters are updated by combining the Softmax cross-entropy loss function with the center loss function. The introduction of the center loss function improves the distribution of expression features, reducing the internal spacing of expressions of the same category and thus increasing the distance between different categories of expression features, making it easier for the network to distinguish facial expression features and improving recognition accuracy.

The application provides a high-risk operation personnel safety state intelligent monitoring method and system, relates to the high-risk operation monitoring technical field, and comprises collecting facial video streams, behavior data and physiological signals; the facial video streams are pretreated and time sequence feature extraction is performed to obtain a time sequence expression feature vector sequence; the time sequence features are time-aligned and standardized to obtain target time sequence features; the target time sequence features are input into a multi-modal fusion analysis model to obtain a quantized and time-varying personnel state dynamic risk assessment value; when the personnel state dynamic risk assessment value continuously exceeds a preset risk threshold for a certain length of time or a specific mode of sharp rise occurs, a graded early warning signal is generated according to a preset rule, and early warning information, associated data segments and psychological analysis conclusions of the to-be-monitored personnel are pushed to a safety supervision terminal. The application solves the problem of low monitoring effect of the existing high-risk operation personnel state monitoring technology through the above scheme.

A face drive generation method based on neural radiance field includes the following steps: S1, extracting feature points of a face region from image frames of a video, calculating expression features and posture features of a corresponding three-dimensional face model according to the feature points, and extracting speech features from a speech part of the video; S2, for each position in a three-dimensional space where the three-dimensional face is located, predicting a mixed shape vector for controlling the three-dimensional face, a posture correction coefficient, and a linear mixed skinning weight through a neural network; S3, based on a deformation technology of the three-dimensional face model, calculating a deformed position of each current position according to the posture features and the speech features obtained in step S1 and the parameters predicted in step S2; S4, predicting a flow field according to each calculated position, aggregating feature information of adjacent time through a multi-frame set manner, and optimizing a face reconstruction loss by using time aggregation, so as to constrain smooth transformation of a dynamic neural radiance field in time. The method improves the accuracy of expression deformation, and makes the generated video more smooth and consistent.

The application discloses a disease auxiliary diagnosis method and system based on a large language model, relates to the field of medical artificial intelligence, and comprises the following steps: analyzing and extracting past case records, processing the past case records by using an expression mapping unit to obtain a symptom-diagnosis result sample set, building a symptom-sign correlation framework, supplementing the symptom-sign correlation framework based on a pathological feature set and a diagnosis strategy recommendation set to obtain a symptom-sign correlation graph, collecting current text information, processing the current text information based on the expression mapping unit to obtain intermediate expression features, screening standard features based on the current selection situation to obtain a symptom set, inputting the symptom set into the symptom-sign correlation graph to obtain a diagnosis recommendation strategy, and collecting survival strategies or replacement strategies to feed back into the symptom-sign correlation graph for optimization. The application effectively assists doctors in improving diagnosis efficiency and reducing misdiagnosis probability, and provides standardized and accurate technical support for clinical diagnosis and treatment.

This invention relates to the field of oil and gas worker safety monitoring technology, specifically to a method and system for monitoring oil and gas worker fatigue based on multimodal information fusion. The method includes: acquiring facial images, full-body images, and heart rate data of oil and gas miners; performing multiple enhancements on the facial and limb images: adaptive illumination equalization, dust defogging, and motion deblurring processing to obtain three enhanced images, which are then weighted and fused to obtain enhanced facial and full-body images; extracting fatigue expression feature values ​​and facial dynamic feature values ​​from the enhanced facial image, and extracting limb feature values ​​from the enhanced full-body image, and fusing the three to obtain a fatigue quantification value; and correcting the fatigue quantification value based on heart rate data to determine the fatigue level, thereby achieving monitoring and graded early warning of miner fatigue. This invention improves the accuracy of miner safety monitoring.

The invention discloses a semantic-loss-free face desensitization method based on expression feature positioning. The semantic-loss-free face desensitization method comprises the steps that a training set and a test set are collected; constructing a double-branch feature learning network which comprises a visual feature learning branch and a facial action unit feature learning branch, and extracting visual features and facial action unit features; constructing a loss constraint system which comprises consistency loss and facial action unit regularization loss and is used for jointly optimizing the network; training the double-branch feature learning network by using the training set, and optimizing through a loss constraint system; inputting the image to be desensitized into the double-branch feature learning network to generate an attention map; based on the attention map, generating a mask used for identifying the privacy area, and performing pixel-level confusion and geometric deformation desensitization processing on the privacy area; and evaluating the expression recognition precision and privacy protection effect of the desensitized image on the test set. According to the invention, face privacy protection and face desensitization are realized under the condition of keeping image semantic information.

The application discloses a kind of old person health monitoring method and system based on expression emotion calculation in the field of computer vision, method includes real-time collection old person's face image;The face image collected is input into the old person health monitoring model based on expression emotion calculation constructed, and the physiological health signal and the psychological health signal of old person based on face image are output.Based on the old person health monitoring model based on expression emotion calculation, the key expression area is positioned based on the proposed regional navigation multi-task training module, face expression and face posture are jointly trained, expression feature extraction is completed, spatiotemporal motion enhancement network is trained using the extracted expression feature, multi-pose and dynamic face expression recognition is completed, and the expression information obtained is converted into physiological health signal and psychological health signal.The application can identify the expression information of old person facial posture real-time change in natural scene, improve recognition accuracy, and further monitor the health signal of old person.

This invention belongs to the field of emotion recognition and medical auxiliary diagnosis and treatment technology, and discloses an emotion recognition system applied to cosmetic body dysmorphic disorder. The system includes an image acquisition module that acquires and de-identifies patient facial videos during pre-operative, intra-operative, and post-operative follow-up stages; a face detection module that locates the face region using a YOLO network adapted to mask occlusion, supine posture, and complex lighting; an expression feature extraction module that extracts multi-level features in parallel using ResNet101-SE, VGG16, and EfficientNet-B0; a feature fusion and discrimination module that normalizes the feature vectors and inputs them into a gradient boosting decision tree classifier, adjusting the output emotion recognition result based on class weights; a risk assessment module that correlates the recognition result with BDD-YBOCS, SAS, and SDS scale scores, generating a risk warning when a threshold is exceeded; and a result processing module that stores data in a time series and interacts with a medical information system to establish a continuous mental health record in the electronic medical record.

This disclosure provides a spatial domain identification method based on expression topology-driven diffusion autoencoder, belonging to the field of data identification technology. Specifically, it includes: acquiring and preprocessing spatial transcriptome data to obtain a processed expression matrix; constructing an expression adjacency matrix based on the gene expression similarity between measurement units in the expression matrix; constructing a diffusion autoencoder network containing an encoder and a decoder; inputting the expression matrix into the encoder, encoding it through graph convolutional units and expression diffusion units to obtain a low-dimensional latent representation; inputting the latent representation into the decoder to reconstruct the expression feature representation; constructing and optimizing a joint objective function, and training the diffusion autoencoder network accordingly; inputting the spatial transcriptome data to be identified into the trained diffusion autoencoder network, outputting the latent representation, and performing cluster analysis to divide the measurement units into different spatial domains. This disclosed method improves the consistency and stability of the identification results.

The present application relates to the field of artificial intelligence and network information security technology, and particularly relates to a rumor identification method based on multi-source information fusion. The method comprises: obtaining a text to be identified; obtaining content audit original labels returned by a plurality of third-party content audit platforms; mapping heterogeneous labels of different platforms to a unified clue category system to obtain a clue set based on platform content audit; using a rumor clue supplement vocabulary to identify language expression features of the text to be identified to obtain a clue set based on language features; calculating fusion scores of various clues according to the clue set based on platform content audit and the clue set based on language features, and generating a rumor clue representation vector according to a preset threshold; fusing a text representation of the text to be identified with the rumor clue representation vector, and inputting the fusion result into a rumor identification model to obtain a rumor identification result. The present application can unify mapping, supplement and fusion of multi-platform content audit labels and language expression features, reduce the influence of single information source label loss, label heterogeneity and semantic inconsistency on the identification result, and help to improve the accuracy and stability of rumor identification in the public opinion dissemination scenario.

This invention discloses a semi-automatic annotation method and system for student facial expression data based on multimodal fusion, comprising: S1 generating multidimensional expressions based on a semantic conditional diffusion model, and simultaneously outputting discrete expression classifications and continuous VAD emotion dimension labels through text prompts in an educational setting; S2 obtaining synthetic expression images and their corresponding 68 facial key point coordinates based on S1, performing classroom scene-adaptive style transfer, and using a key point-constrained adversarial generative network to preserve expression features and adapt to classroom lighting and viewing angle; S3 identifying the classification probability of discrete expression categories and the three-dimensional predicted value of the continuous VAD emotion dimension, using hybrid uncertainty-driven active learning, and combining classification entropy and regression variance to select high-value samples; S4 performing identity decoupling and privacy anonymization on the selected high-value samples, reconstructing the face through a 3D deformation model and specifically blurring the eyebrow region to achieve anonymity protection. This invention can ensure privacy compliance, reduce annotation costs, and improve model accuracy and robustness.