Vinoodhini D, Ajai Ram and Arockia Xavier Annie R, Anna University, India

Audio verification is a key biometric authentication method used to confirm an individual's identity based on their voice. This research addresses challenges such as dynamic acoustic conditions (e.g., background noise, reverberation, and microphone variability) and diverse vocal traits to enhance speaker verification robustness. Existing approaches are ineffective in practical situations where security demands necessitate reliable performance under unpredictable environments. Leveraging the DF-ResNet architecture, which integrates a transformation module with depth-first search, our approach optimizes voice feature extraction and analysis .The model was tested on real-world datasets simulating environments like crowded public spaces, quiet offices, and reverberant halls. Its ability to increase accuracy while preserving low computational complexity is demonstrated by experimental results, which makes it a workable option for contemporary biometric identification systems.

Biometric Authentication, Depth First Resnet, Transformation Module, Speaker verification.