Electronic Device Emotion Recognition Weight Model Update
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for emotion recognition using neural network models on electronic devices consume significant resources and raise privacy concerns due to the need for relearning models based on user feedback, especially on lightweight devices and when models exist on external servers.
Innovation Solution
An electronic device applies feedback information to a weight model to provide personalized emotion information, minimizing resource consumption and network traffic by updating the weight model locally, thus reducing the need for frequent relearning and external data transmission.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If neural network models are relearned for personalization based on user feedback, then emotion recognition accuracy is improved, but resource consumption increases
Solution Approach 1:
The patent extracts only the essential feedback information from user responses and applies it to update the weight model, rather than performing complete relearning. This selective extraction of useful information allows personalization while minimizing computational resource consumption.
Solution Approach 2:
The patent changes the approach from relearning the entire neural network model to updating only the weight model parameters with feedback information. This parameter-level update maintains personalization benefits while dramatically reducing the computational resources required.
2Measurement precision
If neural network models are relearned for personalization, then emotion recognition accuracy is improved, but processing time increases
Solution Approach 1:
The patent performs preliminary actions by pre-training the neural network model and pre-establishing the weight model before actual emotion recognition tasks. During operation, only the weight model needs updating based on feedback, which is much faster than complete relearning, thus reducing processing time while maintaining accuracy.
Solution Approach 2:
The patent extracts only the necessary feedback information and applies it selectively to the weight model rather than performing comprehensive relearning. This extraction approach significantly reduces the time required for personalization updates.
3Measurement precision
If feedback information is transmitted to external servers for model updates, then personalization accuracy is improved, but network resource consumption and privacy risks increase
Solution Approach 1:
The patent implements self-service by enabling the electronic device to update its own weight model using locally stored feedback information. This eliminates the need to transmit sensitive user data to external servers, thereby maintaining personalization accuracy while removing privacy risks associated with network transmission.
Solution Approach 2:
The patent introduces the weight model as an intermediary that stores and processes feedback information locally. This intermediary allows the system to maintain personalization capabilities without directly transmitting raw user data to external servers, thus mitigating privacy concerns.
4Measurement precision
If multiple neural network models are used for different multimedia modalities, then emotion recognition accuracy is improved, but device complexity increases
Solution Approach 1:
The patent applies universality by using a single weight model that can process feedback information from multiple multimedia modalities (audio, visual, linguistic). This unified approach maintains the ability to recognize emotions across different data types while avoiding the complexity of maintaining separate models for each modality.
Data Source
Figure 1~2
Figure 3
Figure 4
AI summary
Emotion information is obtained by an electronic device in order to improve communication between a person and the electronic device. Multimedia data is obtained regarding a person, predicted values for the person are obtained by applying the multimedia data to neural network models, and emotion information of the person is obtained by applying the predicted values to a weight model. Then, feedback information is obtained from the person with respect to the first emotion information of the person. Finally, the weight model is updated by using the feedback information. Subsequently, when multimedia data are again obtained regarding the person, new predicted values for the person are obtained by applying later multimedia data the plurality of neural network models, and emotion information of the person is again obtained, but this time using the weight model updated using the feedback information.