Design and development of a gender and language recognition system
Abstract
Gender and Language recognition is an important topic because of its wide range of applications for computer-based speech pronunciation, speech translation, and visual impairment. If people could speak to their computer systems in the same way they speak to another person; this would encourage more people to embrace technology. Recognizing speech signals as input transforms the computer from a tool to an assistant enabling a hands-free interaction with a computer, expanding the range of possibilities for human-computer usage and improvements in human-human communication. A gender and language recognition system which takes input speech signals extracts features and recognizes the gender and language of a speaker based on the features, was developed. The gender detection algorithm uses the feature extraction data for the pattern matching technique using the Euclidean distance to recognize gender. The language detection algorithm implemented the Mahalanobis distance as the pattern matching technique to recognize features of given language-specific-phrase. This system has been verified for twenty-five test subjects, more than 2300 feature samples, in two different test environments. The test finds the average recognition rate for correct gender recognition was 85%-92% for English, 79%-88% for Spanish, 78%-93% for German, 82-91% for Farsi and 88-93% for Arabic for all test phrases, test subjects and labs. During the testing and evaluation of the language recognition system, it was observed that the average recognition rate for accurate language recognition was 79%-87%, with high average language recognition of 99% for at least one test phrase. German’s average language recognition rate for was 82%-88%, with a high of 94%. For Arabic, the overall rate of recognition was 77%-83%, with 90% average recognition rate for at least one test phrase. Spanish had 65%-75%, with 82% average recognition rate for at least one test phrase. Farsi had 71%-80% average language recognition rate, with high average 89%.
Subject Area
Electrical engineering|Computer science
Recommended Citation
LaTisha Roberts,
"Design and development of a gender and language recognition system"
(2008).
ETD Collection for Tennessee State University.
Paper AAI1461698.
https://digitalscholarship.tnstate.edu/dissertations/AAI1461698