INVESTIGATING THE FACE RECOGNITION ALGORITHM ON EMOTIONAL EXPRESSIONS FOR VISUALLY IMPAIRED EMPLOYERS

Rationale: More than 4 million Americans are either legally blind or visually impaired. Despite their visual limitations, visually impaired or blind individuals play very pivotal roles in running a business and demonstrating their leadership for blind communities.

Objective: The study was to optimize a facial recognition algorithm for enhanced communication with the business clients and the interviewers through emotional cues. The project assumed that a visually impaired employer conducted an interview and acquired emotional expression data even from a remote area.

Methods: Various emotional expressions of a candidate from a webcam were obtained, and then stored as a digital format. With a facial recognition algorithm, the data was converted to Local Binary Patterns (LBP), and the LBP patterns of trained data were compared by a Central Neural Network (CNN). The Haar algorithm from the "open cv library" was used for facial recognition, and the CNN technique was applied to the trained data set to increase the Haar algorithm precision.

Results: Once the emotion was recorded, the emotional status of the interviewee could be reported with moderate accuracy to the blind interviewer.

Conclusion: The emotional expressions by a candidate were correctly measured with an accuracy of about 55% by the algorithm tested in this study with the optimal distance ranging from 20.0 ft to 25.0 ft between webcam and interviewee’s face.