Face emotion recognition based on Gabor wavelet and particle swarm optimization algorithm

• Published in: Multimedia systems Vol. 31; no. 6; p. 435
• Main Authors: Xu, Yankui, Cheng, Lina
• Format: Journal Article
• Language: English
• Published: Berlin/Heidelberg Springer Berlin Heidelberg 01.12.2025; Springer Nature B.V
• Subjects: Access control; Accuracy; Algorithms; Artificial intelligence; Biometric identification; Biometrics; Computer Communication Networks; Computer Graphics; Computer Science; Convergence; Cryptology; Data Storage Representation; Efficiency; Emotion recognition; Emotional factors; Emotions; Facial recognition technology; Feature extraction; Filter banks; Fourier transforms; Image filters; Machine learning; Methods; Morlet wavelet; Multimedia Information Systems; Neural networks; Operating Systems; Optimization algorithms; Particle swarm optimization; Pattern recognition; Regular Paper; Semantics; Signal processing; Support vector machines; Voice recognition; Wavelet transforms; Feature extraction; Wavelet basis function; Particle swarm optimization; Gabor wavelet; Facial emotion recognition
• ISSN: 0942-4962; 1432-1882
• DOI: 10.1007/s00530-025-01976-2

Emotion is an important part of human behavior, which plays an important role in thinking, decision-making and social communication. By analyzing facial expressions, we can track and identify people’s emotional states, so as to better understand their inner feelings. Therefore, a face emotion recognition method based on Gabor wavelet and particle swarm optimization algorithm is proposed. Using Gabor wavelet algorithm to extract facial emotional image features, constructing filter banks of different scales and directions, combining one-dimensional and two-dimensional wavelets, comprehensively capturing the subtle changes of the face in different emotional states, and introducing FFT solving method to improve calculation speed. The 2DPCA + LBP algorithm is used to perform dimensionality reduction on the extracted features, effectively reducing the feature dimensions while retaining sufficient recognition information. Using particle swarm optimization algorithm to optimize the reduced dimensional features, gradually converging to the vicinity of the optimal solution through iterative process, and obtaining the optimized facial emotion feature vector. Introducing an improved deep residual network model for facial emotion recognition, solving the gradient vanishing problem by introducing residual block structure, enabling the network to learn feature representations at a deeper level, and using L2-SVM classifier for classification, improving the accuracy of recognition. Experimental results show that this method has strong ability of feature extraction and dimension reduction of facial emotion image, and can effectively obtain facial emotion feature vector through particle swarm optimization algorithm. And it has high accuracy in facial emotion recognition and good application effects.