Deep Psychological Affect Network for Emotion Recognition
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional emotion recognition systems face challenges in accurately understanding and measuring human emotions due to the complexity of neural mechanisms, noise from artifacts, low signal-to-noise ratio, inter-subject and intra-subject variability, and the multifaceted nature of emotions, which limits their ability to reliably automate emotional dynamics.
Innovation Solution
A deep psychological affect network (DPAN) is generated using a deep learning network and a temporal margin-based loss function, which learns physiological signals such as brain wave and heartbeat signals to estimate emotions by formulating emotion recognition as a spectrum-time sequence classification problem, improving the performance of emotion recognition through the use of a Convolutional Long Short-Term Memory (LSTM) network and a temporal margin-based classification loss function.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional biometric sensors and feature-based approaches are used for emotion recognition, then the system can be implemented with existing technology, but the recognition accuracy is limited due to noise, low signal-to-noise ratio, and inter-subject variability
Solution Approach 1:
The patent transforms the emotion recognition approach by changing from traditional feature extraction to deep learning-based parameter learning. The DPAN model learns optimal parameters directly from raw physiological signals, adapting to inter-subject variability through data-driven parameter optimization rather than fixed feature engineering.
Solution Approach 2:
The patent combines multiple physiological signal types (EEG, ECG, EMG, GSR) into a composite multi-modal input for the deep learning model. This composite approach leverages complementary information from different biological sources to improve recognition accuracy and robustness against individual variability.
2Measurement precision
If deep learning networks are used to improve emotion recognition performance, then recognition accuracy improves, but the complexity of the system increases
Solution Approach 1:
The patent segments the complex emotion recognition task into distinct computational components within the DPAN architecture: convolutional layers for local feature detection, LSTM layers for temporal dependency modeling, and fully connected layers for classification. This segmentation makes the complex model more manageable and interpretable.
Solution Approach 2:
The patent employs dynamic computational elements including LSTM networks that adaptively adjust hidden states based on temporal context, and dropout layers that dynamically deactivate neurons during training to prevent overfitting. These dynamic mechanisms enable the model to handle variable emotional expressions while controlling complexity.
3Measurement precision
If temporal margin-based loss function is applied to handle temporal changes in emotional states, then the model captures temporal dynamics better, but the computational burden increases
Solution Approach 1:
The patent applies preliminary action by pre-processing physiological signals through normalization and artifact removal before feeding them to the deep learning model. This preliminary preparation reduces the computational burden during the main training phase while maintaining temporal accuracy.
Solution Approach 2:
The patent maintains continuous useful action through the LSTM architecture that processes temporal sequences in a continuous manner, preserving temporal relationships without requiring intensive re-computation at each time step. The temporal margin loss function continuously guides the learning process to maintain temporal discrimination.
Data Source
AI summary
Disclosed are a method and a system for estimating human emotions using a deep psychological affect network for human emotion recognition. According to an embodiment of the present disclosure, a method for estimating emotion includes obtaining a physiological signal of a user, learning a network, which receives the obtained physiological signal, by using a temporal margin-based classification loss function considering a temporal margin, when the learning is in progress along a time axis, and estimating an emotion of the user through the learning of the network using the temporal margin-based classification loss function.