Singing Technique

Singing technique research focuses on understanding and classifying the diverse ways vocalists modify their voices for expressive purposes, encompassing techniques ranging from traditional singing styles to extreme metal vocals. Current research employs deep learning models, including convolutional recurrent neural networks (CRNNs) and generative adversarial networks (GANs) with convolutional autoencoders (CAEs), to analyze audio features and automatically detect or convert between different singing techniques. These efforts are driven by the need to improve music information retrieval, create more realistic voice synthesis systems, and gain a deeper understanding of the perceptual and acoustic properties of singing.