Transparent Peer Review By Scholar9
Facial Expression Recognition Using Deep Learning Algorithm
Abstract
Facial emotion recognition extracts the human emotions from the images. As such, it requires an algorithm to understand and model the relationships between faces and facial expressions, and to recognize human emotions. Recently, deep learning models are extensively utilized enhance the facial emotion recognition rate. However, the deep learning models suffer from the overfitting issue. Moreover, deep learning models perform poorly for images which have poor visibility and noise. Therefore, in this dissertation, a novel deep learning based facial emotion recognition tool is proposed. Initially, a joint trilateral filter is applied to the obtained dataset to remove the noise. Thereafter, contrast-limited adaptive histogram equalization (CLAHE) is applied to the filtered images to improve the visibility of images. Finally, a deep convolutional neural network is trained. Nadam optimizer is also utilized to optimize the cost function of deep convolutional neural networks. Experiments are achieved by using the benchmark dataset and competitive human emotion recognition models. Comparative analysis demonstrates that the proposed facial emotion recognition model performs considerably better compared to the competitive models.
Aravind Ayyagari Reviewer
24 Sep 2024 05:23 PM
Approved
Relevance and Originality
The research addresses a key issue in computer vision by focusing on facial emotion recognition, essential for applications like mental health assessment and human-computer interaction. By employing advanced deep learning techniques, the study presents an innovative method to improve recognition rates, contributing to both academia and practical applications. The originality is highlighted by integrating noise reduction and visibility enhancement methods, marking a significant advancement over existing solutions.
Methodology
The methodology is well-structured, starting with a joint trilateral filter to remove noise, followed by contrast-limited adaptive histogram equalization (CLAHE) for image visibility enhancement. These preprocessing steps are vital for data preparation. Training a deep convolutional neural network with the Nadam optimizer adds rigor to the approach. However, more details about the dataset’s size, diversity, and collection would enhance this section's robustness.
Validity & Reliability
The findings' validity is strengthened by using benchmark datasets, allowing credible comparisons with existing models. However, addressing potential biases in the dataset and training process would improve reliability. Information on how the model performs across different demographic groups would further enhance its generalizability and robustness in real-world applications.
Clarity and Structure
The article is organized logically, guiding the reader through problem identification, methodology, and results. However, some complex sentences could be simplified for better readability. Clear definitions of technical terms would improve comprehension, especially for less familiar readers. A more straightforward presentation of ideas would enhance the research's clarity.
Result Analysis
The results are effectively analyzed, comparing the proposed model with existing systems, showcasing its advantages. However, specific performance metrics, like accuracy and F1 score, would provide clearer insight into efficacy. Including visual aids, such as confusion matrices or ROC curves, would further enhance result analysis, making it easier for readers to interpret the findings.
IJ Publication Publisher
Done Sir
Aravind Ayyagari Reviewer