Real Time Percussive Technique Recognition

Real-time percussive technique recognition aims to automatically identify and classify different percussive actions within audio, such as drum strokes or guitar body percussion. Current research focuses on developing robust algorithms, often employing convolutional neural networks (CNNs), sometimes combined with variational autoencoders (VAEs), to analyze audio features like Mel spectrograms and to learn embeddings representing distinct percussive techniques. This field is significant for advancing music information retrieval, enabling real-time interaction with musical instruments and potentially improving the analysis of musical performance and the creation of new musical instruments.