Hello Everyone. My name is Patrick Wande. I am a Computer Science Major at Earlham College with a strong interest in machine learning advancement and optimizations.
Abstract
Facial recognition technology is gaining attention for its applications in security, surveillance, and identity verification. Optimizing facial recognition algorithms is essential for improving efficiency and accuracy. However, while offering promising applications in security and identity verification, facial recognition has raised concerns regarding its potential for bias, particularly with respect to racial and ethnic groups. This paper presents a research investigation into the possibility of racial bias in the Principal Component Analysis (PCA) optimization of a promising facial recognition technique, Elastic Bunch Graph Matching (EBGM).
EBGM is a facial recognition technique acknowledged for its ability to effectively capture facial structural information for robust recognition. It’s Principal Component Analysis optimization reduced the dimensionality of the data that is inputted into the EBGM model.
This project utilizes the FERET database, a diverse dataset comprising images of individuals from various racial and ethnic backgrounds. The EBGM algorithm is trained using PCA optimization on this dataset, with a focus on evaluating its accuracy in recognizing non-Caucasian faces compared to Caucasian faces. Key aspects of the research include experimental design to assess the performance of the facial recognition model across different racial groups, employing appropriate evaluation metrics such as recognition accuracy and false positive/negative rates. Statistical analysis is conducted to determine if there are significant differences in the model’s performance based on race. The findings of this research contribute to the understanding of potential racial biases in facial recognition algorithms optimized using PCA. Insights gained from this study can inform the development of strategies to mitigate bias and enhance the fairness and inclusivity of facial recognition systems.
Keywords: Facial recognition, EBGM, PCA, Optimization, Comparative Analysis, Racial Bias
Research Proposal (CS 388)
Below is the full proposal document for the discussed research:
Software Architecture Diagram
