Image-based facial emotion recognition using convolutional neural network on emognition dataset

Scientific Reports, Journal Year: 2024, Volume and Issue: 14(1)

Published: June 23, 2024

Detecting emotions from facial images is difficult because expressions can vary significantly. Previous research on using deep learning models to classify has been carried out various datasets that contain a limited range of expressions. This study expands the use for emotion recognition (FER) based Emognition dataset includes ten target emotions: amusement, awe, enthusiasm, liking, surprise, anger, disgust, fear, sadness, and neutral. A series data preprocessing was convert video into augment data. proposes Convolutional Neural Network (CNN) built through two approaches, which are transfer (fine-tuned) with pre-trained Inception-V3 MobileNet-V2 building scratch Taguchi method find robust combination hyperparameters setting. The proposed model demonstrated favorable performance over experimental processes an accuracy average F1-score 96% 0.95, respectively, test

Language: Английский

---

Multimodal emotion recognition by fusing complementary patterns from central to peripheral neurophysiological signals across feature domains

Engineering Applications of Artificial Intelligence, Journal Year: 2025, Volume and Issue: 143, P. 110004 - 110004

Published: Jan. 8, 2025

Language: Английский

---

A multimodal shared network with a cross-modal distribution constraint for continuous emotion recognition

Engineering Applications of Artificial Intelligence, Journal Year: 2024, Volume and Issue: 133, P. 108413 - 108413

Published: April 12, 2024

Language: Английский

---

Multimodal Emotion Recognition Using Feature Fusion: An LLM-Based Approach

IEEE Access, Journal Year: 2024, Volume and Issue: 12, P. 108052 - 108071

Published: Jan. 1, 2024

Multimodal emotion recognition is a developing field that analyzes emotions through various channels, mainly audio, video, and text. However, existing state-of-the-art systems focus on two to three modalities at the most, utilize traditional techniques, fail consider emotional interplay, lack scope add more modalities, aren't efficient in predicting accurately. This research proposes novel approach using rule-based convert non-verbal cues text, inspired by limited prior attempt lacked proper benchmarking. It achieves multimodal utilizing distilRoBERTa, large language model fine-tuned with combined textual representation of audio (such as loudness, spectral flux, MFCCs, pitch stability, emphasis) visual features (action units) extracted from videos. evaluated datasets RAVDESS BAUM-1. high accuracy (93.18% 93.69% BAUM-1) both datasets, performing par SOTA (state-of-the-art) systems, if not slightly better. Furthermore, highlights potential for incorporating additional transforming them into text refine further pre-trained models, giving rise comprehensive recognition.

Language: Английский

---

A sturdy CNN-based model for ambience recognition of acoustic communication

Neural Computing and Applications, Journal Year: 2025, Volume and Issue: unknown

Published: Feb. 3, 2025

Language: Английский

---

Leveraging Symmetry and Addressing Asymmetry Challenges for Improved Convolutional Neural Network-Based Facial Emotion Recognition

Symmetry, Journal Year: 2025, Volume and Issue: 17(3), P. 397 - 397

Published: March 6, 2025

This study introduces a custom-designed CNN architecture that extracts robust, multi-level facial features and incorporates preprocessing techniques to correct or reduce asymmetry before classification. The innovative characteristics of this research lie in its integrated approach overcoming challenges enhancing CNN-based emotion recognition. is completed by well-known data augmentation strategies—using methods such as vertical flipping shuffling—that generate symmetric variations images, effectively balancing the dataset improving recognition accuracy. Additionally, Loss Weight parameter used fine-tune training, thereby optimizing performance across diverse unbalanced classes. Collectively, all these contribute an efficient, real-time system outperforms traditional models offers practical benefits for various applications while also addressing inherent detection. Our experimental results demonstrate superior compared other methods, marking step forward ranging from human–computer interaction immersive technologies acknowledging privacy ethical considerations.

Language: Английский

---

A disentanglement mamba network with a temporally slack reconstruction mechanism for multimodal continuous emotion recognition

Multimedia Systems, Journal Year: 2025, Volume and Issue: 31(2)

Published: March 23, 2025

Language: Английский

---

Predicting Activity in Brain Areas Associated with Emotion Processing Using Multimodal Behavioral Signals

Multimodal Technologies and Interaction, Journal Year: 2025, Volume and Issue: 9(4), P. 31 - 31

Published: March 31, 2025

Artificial agents are expected to increasingly interact with humans and demonstrate multimodal adaptive emotional responses. Such social integration requires both perception production mechanisms, thus enabling a more realistic approach alignment than existing systems. Indeed, emotion recognition methods rely on behavioral signals, predominantly facial expressions, as well non-invasive brain recordings, such Electroencephalograms (EEGs) functional Magnetic Resonance Imaging (fMRI), identify humans’ emotions, but accurate labeling remains challenge. This paper introduces novel examining how physiological signals can be used predict activity in emotion-related regions of the brain. To this end, we propose deep learning network that processes two categories recorded alongside during conversations: (video audio) one signal (blood pulse). Our enables (1) prediction from these inputs, (2) assessment our model’s performance depending nature interlocutor (human or robot) region interest. Results proposed architecture outperforms models anterior insula hypothalamus regions, for interactions human robot. An ablation study evaluating subsets input modalities indicates local was reduced when omitted. However, they also revealed data pulse) achieve similar levels predictions alone compared full model, further underscoring importance somatic markers central nervous system’s processing emotions.

Language: Английский

---

CNN in Neural Networks for Image-based Face Emotion Identification on Recognition Datasets

Research Square (Research Square), Journal Year: 2025, Volume and Issue: unknown

Published: April 15, 2025

Because facial expressions can vary greatly, it be challenging to identify emotions from face photographs. Prior studies on the use of deep learning models for image emotion classification have been conducted a variety datasets with restricted range expressions. The Recognition dataset, which contains ten target emotions—amusement, awe, enthusiasm, liking, surprise, anger, contempt, fear, sorrow, and neutral—is used in this work extend application recognition (FER). To transform video data into photos enhance data, number preparation steps were taken. This paper suggests two methods creating Convolutional Neural Network (CNN) models: transfer (fine-tuning) pre-trained Inception V3 Mobile Net V2 starting scratch using Taguchi technique determine. In order establish reliable combination hyperparameter settings, study developing (fine-tuned) V2, building technique. With an accuracy average F1-score 96% 0.95, respectively, test suggested model showed good performance across experimental procedures.

Language: Английский

---

Multimodal Emotion Recognition based on Face and Speech using Deep Convolution Neural Network and Long Short Term Memory

Circuits Systems and Signal Processing, Journal Year: 2025, Volume and Issue: unknown

Published: April 25, 2025

Language: Английский

---