Abstract
Biometric is used extensively in the present times in authentication systems with the aim of enhancing security and convertibility. A biometric system used commonly is the voice identification system, which uses the distinct audio features of an individual's voice to identify a person. There are various applications that use this system, for example healthcare services, mobile banking, voice dialing and workforce management. At high noise voice records, there are certain limitations of classical schemes in voice identification applications. It is believed that neural network approach can be used effectively to handle classical schemes problems that emerge in voice identification applications. This study performs a comparison between deep neural network (DNN) and Gaussian Mixture Model (GMM) by determining the accuracy under distinct noise circumstance (i.e., low noise, no noise, heavy noise). The findings depict that there is high accuracy of voice identification system based DNN with distinct noise circumstance, while high noisy voice records are not identified by the GMM model.