Investigating the performance of realistic neural networks in speaker recognition

In this study, the performance of an auditory system-inspired automatic speaker recognition system was investigated. In this system, the information processing, analysis and classification abilities of the ear, auditory nerves and the brain are simulated with a very realistic approach and their performance is investigated in automatic speaker recognition. The findings highlight the significant potential of the proposed approach in classifying auditory signals. Accordingly, compared to the state-of-the-art methods presented in the literature, the proposed approach can fulfill the task of automatic speaker recognition with high success rate. In addition, according to the findings of the hybrid approach proposed in the study, it has been shown that the sequential use of deep neural networks and realistic neural networks has a performance-enhancing effect on classification problems. As a result, it has been revealed that using bio-inspired systems in classification can achieve higher performance than state-of-the-art methods.