A variety of classifiers, including Long Short-Term Memory (LSTM), Recurrent Neural Network (RNN), Bidirectional LSTM, Gated Recurrent Unit, and Bidirectional GRU, were evaluated for their effectiveness in accurately identifying the dataset, obtained from Kaggle.com, comprising 16,000 instances. The study explores the preprocessing steps, including data collection and labeling, audio sampling, noise reduction, audio normalization, segmentation, feature extraction (MFCC and spectrograms), data augmentation, and data splitting. The models achieved high accuracy levels of 100% for the BiGRU model and 99.82%, for the LSTM model. The conclusion emphasizes that effective preprocessing and feature extraction are critical to the model's performance. By employing methods like noise reduction, normalization, and advanced feature extraction (MFCC and spectrograms), the quality of the provided data is significantly enhanced. The findings contribute to security and personalized user experiences. This research examines how deep learning models can be applied to detect speakers.