Self-aware face emotion accelerated recognition algorithm: a novel neural network acceleration algorithm of emotion...

Area of Science:

• Computer Science
• Artificial Intelligence
• Human-Computer Interaction

Background:

• Growing need for accurate and efficient human emotion recognition in various applications.
• Importance of understanding student psychology in academic settings for timely intervention.
• Limitations of existing Convolutional Neural Networks (CNNs) in real-time emotion recognition tasks.

Purpose of the Study:

• To propose a novel algorithm, Self-Aware Face Emotion Accelerated Recognition Algorithm (SFEARA), for improved facial emotion recognition.
• To enhance the computational efficiency and reduce the energy consumption of emotion recognition models.
• To develop a model suitable for real-time analysis, particularly for international students' emotional states.

Main Methods:

• Development of the SFEARA algorithm, which differentiates between critical and non-critical regions of input data.
• Implementation of high-precision computation for critical regions and low-precision computation for non-critical regions during inference.
• Integration of results from different computational precision levels to achieve final emotion recognition.

Main Results:

• SFEARA demonstrated 1.3x to 1.6x higher computational efficiency compared to conventional CNNs.
• The algorithm achieved 30% to 40% lower energy consumption in emotion recognition tasks.
• Experimental data confirmed SFEARA's suitability for real-time emotion recognition scenarios.

Conclusions:

• SFEARA offers a significant improvement in efficiency and energy consumption for facial emotion recognition.
• The algorithm's design is well-suited for real-time applications requiring rapid and accurate emotional state assessment.
• This research contributes to the advancement of human-computer interaction through more effective emotion recognition technologies.