A Novel Multiple Music Recognition Technique Based on Deep Learning

Abstract

Multiple Music Recognition is an attractive technology were most of the researches are concentrating on recent times. The introduction and Deep Learning (DL) and Artificial Intelligence (AI) technology created lots of new opportunities in this area. The primary drawbacks which are currently present in this technology are reaching maximum accuracy is the creation of multiple music. To overcome that drawback in this article a new model is developed namely, CNN and SelfSupervised Learning with BoYaTCN (CNSSBO) model which is the combination of the model like Convolutional Neural Networks (CNN) and self-supervised learning with the BoYaTCN model. Primarily, it begins with dataset preparation and preprocessing, which includes Mel-spectrogram generation, data processing and augmentation, noise injection, pre-training, and fine-tuning stages to enhance data quality. The CNSSBO model integrates CNN architectures, YAMNet for sound detection, an MLP-CNN model, and the BoYaTCN model to effectively capture the audio signals. Additionally, loss function analysis is incorporated to optimize model performance. The simulation is implemented in Python, and the effectiveness of the CNSSBO model is evaluated using performance metrics such as accuracy, precision, recall, and F1-score. The results demonstrate the capability of the hybrid framework in improving music recognition performance.