Revolutionizing Vocal Track Extraction: Innovative Hybrid Neural Network Approaches with Deep Clustering, U-net, and UH-net Models

Deep neural networks have become a cornerstone in various recognition and classification tasks due to their ability to learn complex patterns from raw data. This paper explores the potential application of neural networks in the domain of vocal extraction. We investigate the utilization of neural network architectures, specifically the deep clustering model based on recurrent neural networks (RNNs) and the U-net model based on convolutional neural networks (CNNs), for the task of vocal track extraction. Additionally, we propose a novel hybrid approach that incorporates a pretrained RNN model to enhance the performance of the U-net model in vocal track extraction.