Measuring the Accuracy of A Voiceprint Analysis System Designed by Applying the Euclidean Distance Function and Genetic Algorithm

Creative Commons License

Barakat A., Naqra A., Hussian A. R.

SSRG International Journal of Electrical and Electronics Engineering, cilt.11, sa.3, ss.220-230, 2024 (Scopus)

Özet

This research aims to measure the accuracy of the work of the voiceprint analysis system. The system comprises three stages: (i) recording voice, deleting noise, and extracting the voiceprint, (ii) establishing the database, and (iii) comparing the data and decision-making process. The process of deleting the noise and extracting the voiceprint, in which noise deletion is the biggest challenge, is in the first stage. Next, the voice is analyzed by applying the MFCC algorithm, and then a statistical equation is utilized to extract the voiceprint. Creating a database in which the voiceprint samples are saved and comparing and making a decision by applying the Euclidean distance function and the genetic algorithm, respectively. The test results showed speakers’ recognition ratios among the user groups (10, 20, 30, 40), by applying the Euclidean distance function, are (93%, 89.5%, 82.83%, and 73.37%) respectively. The distinction was improved by adding the genetic algorithm to the Euclidean distance function for the same number of users. The results were as follows (94%, 90.75%, 83.83%, and 74.87%), respectively. The average time for voice analysis and voiceprint extraction was (3.183, 3.174, 3.171, and 3.169 sec.); the average time for testing (0.00807, 0.00808, 0.0082, and 0.0258 sec.) by applying the Euclidean distance function; and the average time for testing (0.00615, 0.023711, 0.020747, and 0.022438 sec.) by applying the Euclidean distance function and the genetic algorithm, and thus speeding up the testing and decision-making process is achieved.