A gait-emotion joint recognition method based on heterogeneous space neural architecture search
By using a heterospatial neural architecture search method, the 3D human posture joints are transformed to different coordinate systems, a joint search strategy is constructed, and the neural architecture is automatically built. This solves the problem of ignoring angle information in existing technologies and improves the accuracy and generalization ability of gait emotion recognition.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- THE 28TH RES INST OF CHINA ELECTRONICS TECH GROUP CORP
- Filing Date
- 2024-08-06
- Publication Date
- 2026-07-10
AI Technical Summary
Existing deep learning-based gait emotion recognition technologies ignore the angular information between human joints. Limited by the limitations of hand-designed models, they cannot fully realize the potential of the model structure and need to be redesigned or adjusted when faced with new tasks or data.
A heterogeneous neural architecture search method is adopted. By transforming the 3D human pose joints to Cartesian and spherical coordinate systems, a joint search strategy is constructed to automatically build a neural architecture in the space of position and angle information, and a Softmax classifier is used for emotion recognition.
It improves the accuracy and generalization ability of gait emotion recognition, fully explores the potential of model structure, and avoids the shortcomings of manually designed models.
Smart Images
Figure CN119091485B_ABST
Abstract
Description
Technical Field
[0001] This invention relates to a gait emotion joint recognition method, and more particularly to a gait emotion joint recognition method based on heterospatial neural architecture search. Background Technology
[0002] This section provides only background information relevant to this disclosure and is not necessarily prior art.
[0003] Gait emotion recognition is a research field based on computer vision and machine learning techniques, aiming to infer an individual's emotional state by analyzing their gait patterns. Research shows that emotional state can influence a person's biometrics, such as posture, stride length, walking speed, and postural variations. This provides a theoretical basis for gait emotion recognition, as emotional states can manifest unique characteristics in gait. With the continuous advancement of artificial intelligence, deep learning techniques, such as convolutional neural networks (CNNs) and recurrent neural networks (RNNs), have been widely used in gait emotion recognition. These deep learning methods can automatically extract and learn features, improving the accuracy and robustness of emotion recognition. Gait emotion recognition has broad application prospects, including health monitoring, autonomous driving, virtual reality, and emotional intelligence assistive tools. The needs of different fields have driven the continuous development and application of gait emotion recognition technology.
[0004] Current deep learning-based gait emotion recognition primarily relies on manually designed deep learning models to extract real-time gait information features and then uses these features for emotion recognition. Currently, obtaining human gait emotion features mainly involves extracting features from human joints in a Cartesian coordinate system using deep learning models, thus neglecting the importance of angular information between these joints. Furthermore, manually designed deep learning models are heavily limited by the designer's creativity and experience, failing to fully utilize the potential advantages of the model structure. Moreover, when faced with new tasks or data, manually designed models may require redesign or adjustment, thus limiting their effectiveness in gait emotion recognition.
[0005] It should be noted that the information disclosed in the background section above is only used 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
[0006] Purpose of the invention: The technical problem to be solved by the present invention is to provide a gait emotion joint recognition method based on heterospatial neural architecture search, which addresses the shortcomings of the existing technology.
[0007] To address the aforementioned technical problems, this invention discloses a gait and emotion joint recognition method based on heterospatial neural architecture search, comprising the following steps:
[0008] Step 1: Extract the 3D human pose key points from video data of different emotions;
[0009] Step 2: The extracted 3D human pose joints are spatially transformed into Cartesian coordinate system and spherical coordinate system respectively. Based on the similarity minimization strategy, the position information and angle information of the joints in different coordinate systems are spatially mapped.
[0010] Step 3: Construct neural architecture search spaces and joint search strategies under different spaces. Using the joint search strategy, search for the optimal backbone network in the neural architecture search spaces under different spaces respectively. This backbone network is used to extract spatiotemporal emotional features in the location information space and angle information space, and train the above-mentioned optimal backbone network respectively.
[0011] Step 4: Using the trained optimal backbone network, extract human temporal gait emotion features in different spaces and input them into the Softmax classifier to obtain scores for different emotions.
[0012] Step 5: The joint scoring module performs joint scoring on the different emotion scores in different spaces of the input, thus completing the joint recognition of gait emotion based on heterospatial neural architecture search.
[0013] Furthermore, the three-dimensional human posture joints mentioned in step 1 include: root joints, spinal joints, cervical joints, head joints, left shoulder joints, left elbow joints, left hand joints, right shoulder joints, right elbow joints, right hand joints, left hip joints, left knee joints, left foot joints, right hip joints, right knee joints, and right foot joints, etc.
[0014] Furthermore, the spatial transformation described in step 2 specifically includes:
[0015] Transform the 3D human pose joints extracted in Cartesian coordinates from step 1 into spherical coordinates:
[0016]
[0017]
[0018]
[0019] In this system, any point in the rectangular coordinate system is (x, y, z), and the corresponding point in the transformed spherical coordinate system is...
[0020] Furthermore, obtaining the optimal backbone network as described in step 3 specifically includes the following steps:
[0021] Step 3-1: Construct neural architecture search spaces under different spaces, wherein the different spaces are: position information space and angle information space, and the neural architecture search space includes sub-modules for constructing the backbone network;
[0022] Step 3-2, construct the joint spatial search strategy, specifically as follows:
[0023]
[0024] Where G represents the joint spatial search strategy, maximize means to maximize, ACC represents the joint gait emotion recognition accuracy, and t c T represents the feature extraction speed of the backbone network constructed by each sub-module in the neural architecture search space under the location information space. c t represents the predefined feature extraction speed. s T represents the feature extraction speed of the backbone network constructed by each sub-module in the neural architecture search space under this information space. s This represents the predefined feature extraction speed, where w1 and w2 are weighting factors, as detailed below:
[0025]
[0026]
[0027] Where α1, β1, α2, and β2 are constants;
[0028] Step 3-3: Based on the joint search strategy constructed in Step 3-2, search for the optimal combination of each sub-module in the neural architecture search space under the position information space and angle information space, and use it as the optimal backbone network.
[0029] Furthermore, the sub-modules for constructing the backbone network mentioned in step 3-1 specifically include: a 3*3 convolution sub-module, a 5*5 convolution sub-module, a 3*3 max pooling sub-module, a 3*3 average pooling sub-module, an STGCN sub-module, a channel attention sub-module, and a spatial attention sub-module.
[0030] Furthermore, step 3-3, which involves searching for the optimal combination of each sub-module in the neural architecture search space under the position information space and angle information space, specifically includes:
[0031] Step 3-3-1: Use the controller to search for the initial backbone network in the position information space and the angle information space respectively. The backbone network is a random combination of each sub-module in the neural architecture search space.
[0032] Step 3-3-2: Train and evaluate the initial backbone network using the gait emotion dataset to obtain the joint gait emotion recognition accuracy and the feature extraction speed of the initial backbone network in each space.
[0033] Step 3-3-3: Perform calculations according to the joint spatial search strategy described in step 3-2, and feed the results back to the controller;
[0034] Step 3-3-4: The controller re-optimizes the new backbone network in the search position information space and angle information space, that is, the sub-modules in the neural architecture search space are randomly recombined to obtain a new backbone network;
[0035] Step 3-3-5: Using the new backbone network as the initial backbone network, repeat steps 3-3-2 to 3-3-4 iteratively until the preset number of iterations, to obtain the optimal backbone network in the position information space and angle information space respectively.
[0036] Furthermore, the training of the optimal backbone network described in step 3 specifically includes:
[0037] The optimal backbone network is trained using a gait sentiment dataset containing emotion labels.
[0038] Furthermore, step 4, which involves obtaining scores for different emotions, specifically includes:
[0039] Step 4-1: Using the optimal backbone network trained in Step 3, extract sentiment features from the location information space and the angle information space, respectively.
[0040] Step 4-2: Use the Softmax classifier to classify and score the emotional features extracted in different spaces in Step 4-1 to obtain scores for different emotions.
[0041] Furthermore, the joint scoring described in step 5 specifically includes:
[0042] The joint scoring module performs a joint scoring on the scores of different emotions obtained in step 4, resulting in a joint score for different emotions based on different spaces. The emotion with the highest score is then determined as the final result of the emotion recognition.
[0043] Furthermore, the joint scoring module described in step 5 is as follows:
[0044] F i =S i +C i
[0045] Where i represents the emotion category, S iC represents the scores of different emotions in the angular information space. i F represents the score of different emotions in the location information space. i This indicates the combined score output by the combined scoring module.
[0046] Beneficial effects:
[0047] This invention makes full use of the position and angle information of the three-dimensional human body joints in space, and avoids the shortcomings of manually designing deep learning models. It fully explores the potential of the model structure and effectively improves the accuracy and generalization of the gait emotion recognition model. Attached Figure Description
[0048] The present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments, and the advantages of the present invention in the above and / or other aspects will become clearer.
[0049] Figure 1 This is a schematic diagram of the overall process of the present invention.
[0050] Figure 2 This is a flowchart illustrating the joint search of backbone networks in different spaces according to the present invention.
[0051] Figure 3 This is a schematic diagram showing the result of gait skeleton recognition outputting the corresponding emotion as frustration. Detailed Implementation
[0052] This invention utilizes raw 3D pose joint data to map it to 3D pose joint representations in different spaces, and employs a joint search strategy to complete the backbone network search in different spaces, automatically constructing a gait emotion feature extraction network, and based on this, completing gait emotion recognition. The technical solution is as follows:
[0053] A gait emotion joint recognition method based on heterospatial neural architecture search specifically includes the following steps:
[0054] Step S1: Extract the 3D human pose joints from video data of different emotions;
[0055] Step S2: The extracted 3D human pose joints are spatially transformed into angular coordinate system and spherical coordinate system respectively, and the position information and angle information of the joints in different coordinate systems are spatially mapped based on the similarity minimization strategy.
[0056] Step S3: Construct neural architecture search spaces and joint search strategies under different spaces, use the joint search strategy to jointly search for the backbone network that extracts spatiotemporal emotional features from the location information space and angle information space, and complete the training based on the gait emotion dataset;
[0057] Step S4: Use the backbone network to extract human temporal gait emotion features in different spaces, and output scores for different emotions by using the Softmax classifier for the emotion features in different spaces.
[0058] Step S5: The joint scoring module performs joint scoring on the different emotion scores in different spaces of the input, thereby completing gait emotion recognition.
[0059] Preferably, the different emotional video data in step S1 includes emotions such as happiness, frustration, and neutrality.
[0060] Preferably, the three-dimensional human posture joints extracted in step S1 include root joints, spinal joints, cervical joints, head joints, left shoulder joints, left elbow joints, left hand joints, right shoulder joints, right elbow joints, right hand joints, left hip joints, left knee joints, left foot joints, right hip joints, right knee joints, and right foot joints.
[0061] Preferably, the spatial transformation in step S2 includes transformation to a position information space and an angle information space. The 3D human posture joints extracted from the emotion video belong to a Cartesian coordinate system, while the 3D human posture joints in the spherical coordinate system are transformed using a transformation formula between Cartesian and spherical coordinate systems. To avoid overlap in information spaces, a similarity minimization strategy is used to complete the mapping of the position information space and the angle information space.
[0062] Preferably, step S3 involves constructing neural architecture search spaces and joint search strategies under different spaces, and jointly searching for backbone networks from which features are extracted in different spaces. This specifically includes the following steps:
[0063] Construct a search space that includes 3*3 convolution, 5*5 convolution, 3*3 max pooling, 3*3 average pooling, STGCN module, channel attention module, and spatial attention module;
[0064] A joint spatial search strategy is constructed, which mainly consists of joint gait emotion recognition accuracy, backbone network feature extraction speed in the location information space, and backbone network feature extraction speed in the angle information space. Specifically, it can be represented as:
[0065]
[0066] Where ACC represents the joint gait emotion recognition accuracy, t c T represents the backbone network feature extraction speed in the location information space. c t represents the predefined feature extraction speed. s T represents the feature extraction speed of the backbone network in this information space. sThis represents the predefined feature extraction speed, where w1 and w2 are weighting factors, specifically defined as:
[0067]
[0068]
[0069] Where α1, β1, α2, and β2 are constants.
[0070] The optimal backbone network in the location information space and angle information space is found based on the joint search strategy; the entire search process is as follows: Figure 2 As shown, specifically, the controller first searches for backbone networks in the location information space and angle information space in the search space. Then, it uses the gait emotion dataset to train and evaluate the model to obtain the joint gait emotion recognition accuracy and the feature extraction speed of the backbone network in each information space. Based on the search strategy, the controller is fed back to the controller, prompting the controller to re-optimize the search for backbone networks in the location information space and angle information space. This process is repeated multiple times to obtain the optimal backbone network for extracting features in different spaces.
[0071] The backbone network was trained using a gait emotion dataset that included labels for happy, depressed, and neutral emotions.
[0072] Preferably, step S4 utilizes a backbone network to extract temporal gait emotion features of the human body in different spaces, and outputs scores for different emotions using a Softmax classifier for the emotion features in different spaces. Specifically, this involves using a trained backbone network to extract emotion features of three-dimensional human posture joints in both positional and angular information spaces, and then using a Softmax classifier to classify and score the emotion features in different spaces.
[0073] Preferably, step S5, which involves jointly scoring the different emotion scores in different spaces of the input through the joint scoring module, refers to jointly scoring the different emotion scores in different spaces of the Softmax output in step S4, thereby obtaining the joint scores of different emotions based on different spaces, thus completing gait emotion recognition.
[0074] The principle of this invention is:
[0075] The neural architecture search technology is used to automatically construct the backbone network for extracting three-dimensional human posture features in Cartesian and spherical coordinate systems. The Softmax classifier is used to output emotion classification scores, and the classification scores are then input into the joint scoring module to output individual emotional states, thus completing gait emotion recognition.
[0076] Example 1:
[0077] like Figure 1 As shown, a gait emotion joint recognition method based on heterogeneous spatial neural architecture search specifically includes the following steps:
[0078] Step S1: Use the VideoPose3D 3D human pose estimation algorithm to extract 3D human pose joints from video data of different emotions. The video data mainly includes happy, depressed, and neutral emotions. The 3D human pose joints include root joints, spinal joints, neck joints, head joints, left shoulder joints, left elbow joints, left hand joints, right shoulder joints, right elbow joints, right hand joints, left hip joints, left knee joints, left foot joints, right hip joints, right knee joints, and right foot joints.
[0079] Step S2: Perform spatial transformation on the extracted 3D human posture joints, converting them to both position information space and angle information space. Since the human posture joints extracted in step S1 are in a Cartesian coordinate system, it is only necessary to transform the coordinates of the 3D human joints in the Cartesian coordinate system to the spherical coordinate system according to the spatial transformation formula. The Cartesian coordinate system (x,y,z) and the spherical coordinate system... The space transformation formula is as follows:
[0080]
[0081]
[0082]
[0083] Furthermore, based on a similarity minimization strategy (using the cosine similarity calculation formula to calculate the feature similarity in the positional information space mapping and the angle information space, minimizing this similarity value during training, thereby enabling features to be mapped to different spaces), key points in different coordinate systems are mapped to positional information space and angle information space.
[0084] Step S3: Construct neural architecture search spaces and joint search strategies in different spaces. Utilize the joint search strategy to jointly search for backbone networks that extract spatiotemporal emotion features from the location information space and angle information space, and complete training based on the gait emotion dataset. The search space includes 3*3 convolutions, 5*5 convolutions, 3*3 max pooling, 3*3 average pooling, STGCN modules, channel attention modules, and spatial attention modules. The joint search strategy mainly consists of the joint gait emotion recognition accuracy, the feature extraction speed of the backbone network in the location information space, and the feature extraction speed of the backbone network in the angle information space, specifically expressed as:
[0085]
[0086] Where ACC represents the joint gait emotion recognition accuracy, t c T represents the backbone network feature extraction speed in the Cartesian location information space. c t represents the predefined feature extraction speed. s T represents the backbone network feature extraction speed in the angular information space. s This represents the predefined feature extraction speed, where w1 and w2 are weighting factors, specifically defined as:
[0087]
[0088]
[0089] Where α1, β1, α2, and β2 are constants.
[0090] The optimal feature extraction backbone network in the location information space and angle information space is searched according to the joint search strategy. The specific search process is as follows: Figure 2 As shown, the details are as follows:
[0091] The controller first searches for backbone networks in the location and angle information spaces within the search space. Then, it trains and evaluates the model using a gait sentiment dataset to obtain the joint gait sentiment recognition accuracy and the feature extraction speed of the backbone networks in each information space. This feedback, based on the search strategy, is fed back to the controller, prompting it to re-optimize the search for backbone networks in the location and angle information spaces. This process is iterated multiple times to obtain the optimal backbone networks for extracting features in different spaces. Finally, the backbone network is trained using a gait sentiment dataset containing labels for happy, frustrated, and neutral emotions.
[0092] Step S4: Use the backbone network trained in step S3 to extract three-dimensional human emotion features in the location information space and angle information space respectively, and then use the Softmax classifier to classify and score the emotion features in different spaces.
[0093] Step S5: The joint scoring module performs joint scoring on the different emotion scores in different spaces of the input, thereby completing gait emotion recognition. The joint scoring module can be represented as follows:
[0094] F i =S i +C i
[0095] Where i represents the category, S i C represents the scores of different sentiment categories in the angular information. i F represents the scores of different sentiment categories in the location information space. i This represents the scores for different sentiment categories output by the joint scoring module.
[0096] Example 2:
[0097] In one embodiment, the proposed gait emotion joint recognition method based on heterospatial neural architecture search and a gait emotion dataset are used to search for backbone networks in the information space and angular information space, and the searched backbone network is trained. Using the trained backbone network, temporal gait emotion features of the human body in different spaces are extracted and input into a Softmax classifier to obtain scores for different emotions. Finally, a joint scoring module is used to jointly score the scores of different emotions in different spaces, thereby completing gait emotion recognition. Figure 3 The diagram shown is a schematic representation of the result in the current embodiment where the gait skeleton recognition outputs the corresponding emotion as frustration.
[0098] In its specific implementation, this application provides a computer storage medium and a corresponding data processing unit. The computer storage medium is capable of storing a computer program, which, when executed by the data processing unit, can run the invention's content regarding a gait and emotion joint recognition method based on heterospatial neural architecture search, as well as some or all of the steps in various embodiments. The storage medium can be a magnetic disk, optical disk, read-only memory (ROM), or random access memory (RAM), etc.
[0099] Those skilled in the art will clearly understand that the technical solutions in the embodiments of the present invention can be implemented using computer programs and their corresponding general-purpose hardware platforms. Based on this understanding, the technical solutions in the embodiments of the present invention, or the parts that contribute to the prior art, can be embodied in the form of computer programs, i.e., software products. These computer program software products can be stored in a storage medium and include several instructions to cause a device containing a data processing unit (which may be a personal computer, server, microcontroller, MCU, or network device, etc.) to execute the methods described in various embodiments or certain parts of the embodiments of the present invention.
[0100] This invention provides an idea and method for gait and emotion joint recognition based on heterospatial neural architecture search. Many methods and approaches exist for implementing this technical solution; the above description is merely a preferred embodiment of the invention. It should be noted that those skilled in the art can make various improvements and modifications without departing from the principles of this invention, and these improvements and modifications should also be considered within the scope of protection of this invention. All components not explicitly stated in this embodiment can be implemented using existing technologies.
Claims
1. A gait emotion joint recognition method based on heterospatial neural architecture search, characterized in that, Includes the following steps: Step 1: Extract the 3D human pose key points from video data of different emotions; Step 2: The extracted 3D human pose joints are spatially transformed into Cartesian coordinate system and spherical coordinate system respectively. Based on the similarity minimization strategy, the position information and angle information of the joints in different coordinate systems are spatially mapped. Step 3: Construct neural architecture search spaces and joint search strategies under different spaces. Using the joint search strategy, search for the optimal backbone network in the neural architecture search spaces under different spaces respectively. This backbone network is used to extract spatiotemporal emotional features in the location information space and angle information space, and train the above-mentioned optimal backbone network respectively. Step 4: Using the trained optimal backbone network, extract human temporal gait emotion features in different spaces and input them into the Softmax classifier to obtain scores for different emotions. Step 5: The joint scoring module performs joint scoring on the different emotion scores in different spaces of the input, thereby completing the joint recognition of gait emotion based on heterospatial neural architecture search. The process of obtaining the optimal backbone network in step 3 specifically includes the following steps: Step 3-1: Construct neural architecture search spaces under different spaces, wherein the different spaces are: position information space and angle information space, and the neural architecture search space includes sub-modules for constructing the backbone network; Step 3-2, construct the joint spatial search strategy, specifically as follows: ； in, This indicates a joint space search strategy. To maximize, This represents the accuracy of joint gait emotion recognition. The feature extraction speed of the backbone network constructed by each sub-module in the neural architecture search space under the location information space represents the feature extraction speed of the backbone network. This represents the predefined feature extraction speed. This represents the feature extraction speed of the backbone network constructed by each sub-module in the neural architecture search space under this information space. This represents the predefined feature extraction speed. and These are weighting factors, as detailed below: ； in, It is a constant; Step 3-3: Based on the joint search strategy constructed in Step 3-2, search for the optimal combination of each sub-module in the neural architecture search space under the position information space and angle information space, and use it as the optimal backbone network.
2. The gait emotion joint recognition method based on heterospatial neural architecture search according to claim 1, characterized in that, The three-dimensional human posture joints mentioned in step 1 include: root joint, spine joint, neck joint, head joint, left shoulder joint, left elbow joint, left hand joint, right shoulder joint, right elbow joint, right hand joint, left hip joint, left knee joint, left foot joint, right hip joint, right knee joint, and right foot joint.
3. The gait emotion joint recognition method based on heterospatial neural architecture search according to claim 1, characterized in that, The spatial transformation described in step 2 specifically includes: Transform the 3D human pose joints extracted in Cartesian coordinates from step 1 into spherical coordinates: ； ； ； Wherein, any point in the rectangular coordinate system is The corresponding point in the transformed spherical coordinate system is .
4. The gait emotion joint recognition method based on heterospatial neural architecture search according to claim 3, characterized in that, The sub-modules for constructing the backbone network mentioned in step 3-1 specifically include: a 3*3 convolution sub-module, a 5*5 convolution sub-module, a 3*3 max pooling sub-module, a 3*3 average pooling sub-module, an STGCN sub-module, a channel attention sub-module, and a spatial attention sub-module.
5. The gait emotion joint recognition method based on heterospatial neural architecture search according to claim 4, characterized in that, Step 3-3, which describes searching for the optimal combination of each sub-module in the neural architecture search space under the position information space and angle information space, specifically includes: Step 3-3-1: Use the controller to search for the initial backbone network in the position information space and the angle information space respectively. The backbone network is a random combination of each sub-module in the neural architecture search space. Step 3-3-2: Train and evaluate the initial backbone network using the gait emotion dataset to obtain the joint gait emotion recognition accuracy and the feature extraction speed of the initial backbone network in each space. Step 3-3-3: Perform calculations according to the joint spatial search strategy described in step 3-2, and feed the results back to the controller; Step 3-3-4: The controller re-optimizes the new backbone network in the search position information space and angle information space, that is, the sub-modules in the neural architecture search space are randomly recombined to obtain a new backbone network; Step 3-3-5: Using the new backbone network as the initial backbone network, repeat steps 3-3-2 to 3-3-4 iteratively until the preset number of iterations, to obtain the optimal backbone network in the position information space and angle information space respectively.
6. The gait emotion joint recognition method based on heterospatial neural architecture search according to claim 1, characterized in that, Step 3, training the optimal backbone network, specifically includes: The optimal backbone network is trained using a gait sentiment dataset containing emotion labels.
7. The gait emotion joint recognition method based on heterospatial neural architecture search according to claim 1, characterized in that, Step 4, which describes obtaining scores for different emotions, specifically includes: Step 4-1: Using the optimal backbone network trained in Step 3, extract sentiment features from the location information space and the angle information space, respectively. Step 4-2: Use the Softmax classifier to classify and score the emotional features extracted in different spaces in Step 4-1 to obtain scores for different emotions.
8. The gait emotion joint recognition method based on heterospatial neural architecture search according to claim 1, characterized in that, The joint scoring described in step 5 specifically includes: The joint scoring module performs a joint scoring on the scores of different emotions obtained in step 4, resulting in a joint score for different emotions based on different spaces. The emotion with the highest score is then determined as the final result of the emotion recognition.
9. The gait emotion joint recognition method based on heterospatial neural architecture search according to claim 8, characterized in that, The joint scoring module mentioned in step 5 is as follows: ； in, Indicates the category of emotion. The scores represent different emotions within the angular information space. This represents the scores for different emotions within the location information space. This indicates the combined score output by the combined scoring module.