Comparison of Isolated and Continuous Text Models for Voice Based Attendance System
Arvind Kumar, Mahesh Chandra, Kumari Neelam, Sounik Kiran Kumar Dash, D.K. Singh · 2018 International Conference on Current Trends towards Converging Technologies (ICCTCT) · 2018
In this paper, a comparative study has been made between isolated and continuous text models for developing a voice based attendance system. A GUI application has been designed for testing a two layer authentication system for attendance system, the second layer being speech biometrics. Users were pre-enrolled in the system before using the application. For both the speech models, different combinations of feature vectors and classifiers were experimented. Various statistical parameters like mean and variance were studied for threshold evaluation. Final models were developed with MFCC feature and GMM classifier. Overall accuracy of the system for isolated text model and continuous text model is reported as 62.67% and 53.6% respectively.