Hybrid Deep Convolutional Neural Network based Speaker Recognition for Noisy Speech Environments

Speaker recognition depends on identifying the speaker using particular segments of the sound stream. A single speech characteristic only reveals the speaker's identity partially. Current advances in machine learning have considerably enhanced automatic voice recognition and localization systems. Nevertheless, this advantage comes at the expense of requiring complicated models and calculations. Additional microphone arrays will be used, as well as practice data. This study introduces a novel deep convolutional neural network-based end-to-end hybrid identification and localization model (HDCNN). HDCNN are employing a cutting-edge data augmentation strategy. This model can recognize both single- and multi-speaker arrangements and show which speaker is active with outstanding accuracy. HDCNN, a hybrid machine-learning algorithm. The final outcomes of proposed HDCNN model show greatest performance with an accuracy of 98.33%, which is higher than existing model's performance metrics.