Formant analysis for automatic speech recognition (ASR)
September 2022
Final Project @ University of Artificial Intelligence, Wilmington, DE
I'm a music technology researcher, as well as an international concert organist and the main Organist at the St. John the Evangelist Anglican Church.
My current research interests are in the field of music information retrieval, artificial intelligence and computer simulation of expressive organ performance.
My doctoral project at the McGill University, "Temporal elasticity as the main tool for expressive performance of German late Romantic organ music," was funded by the prestigious FRQSC grant and CIRMMT Award.
I have an extensive concert repertoire, from Renaissance till Avantgarde, with the main focus on the German late Romantic organ music. Among others, I worked together with Basel symphony orchestra (Switzerland), symphony orchestra and choir of the "Helicon"-Theatre (Moscow, Russia), brass player's ensemble under the direction of Wolfgang Bauer (Stuttgart, Germany), jazz orchestra "VolgaOrchestra" (Volgograd, Russia), and baroque ensemble "Alta Capella" (Moscow, Russia). My concert activity led me through the Russian cities, then to Germany, Switzerland, Spain, Italy, Poland, the USA and Canada.
My music degrees
My research background
My IT Certificates
St. John the Evangelist
Church, London, Canada
(with Susan Millar Boldissar, cello)
Final Project @ University of Artificial Intelligence, Wilmington, DE
Project 1: Orchestral groups and music instruments recognition with Deep Neural Networks
Project 2: Reconstruction of audio files from the latent space with Variational Autoencoders (model based on Mel-frequency spectrograms)
Instead of analyzing the music scores manually, it is possible to extract the required features from MusicXML file and build a Machine Learning model to define the relative strength for each phrasing level.
The experiment is over! Click here to see the full description and results.
Abstract: The present study introduces a mathematical model for the symmetric phrasing scheme of German musicologist Hugo Riemann (1849–1919). The model is used to create an artificially expressive timing pattern in Max Reger’s (1873–1916) organ Prelude op. 135a/1 and is evaluated analytically against professional human interpretation.
View abstract and presentation.
The main idea of this project to compare the professional performance of Max Reger (1873-1916) organ music with the computer simulated performance built in accordance with the Hugo Riemann (1849-1919) phrasing principles (s. Riemann, Hugo. 1884. Musikalische Dynamik und Agogik, Lehrbuch der musikalischen Phrasierung auf Grund einer Revision der Lehre von der musikalischen Metrik und Rhythmik. Hamburg, Germany, p. 8). Several models are made in order to imitate different expression levels.
View abstract
Final project for the McGill graduate course "Digital Sound Synthesis and Audio Processing". The goal of the project was to provide a formantic analysis of the organ stop Vox Humana, compare its formantic structure to the human voice and resynthesize it using the "CHANT" methodology (Rodet, Xavier, Yves Potard and Jean-Baptiste Barriere. 1984. "The CHANT Project: From the Synthesis of the Singing Voice to Synthesis in General." Computer Music Journal 8(3): 15-31).
DownloadFinal project for the McGill graduate course "Music Information Acquisition, Preservation, and Retrieval". Evaluation of different beat tracking systems for temporal analysis of Mendelssohn's Organ Sonata op. 65 №3.
© Yulia Draginda