Skip to Main Content
Professor Elaine Biddiss (BME) uses non-pitched percussion instruments with researchers at Holland Bloorview Kids Rehabilitation Hospital. (photo by Neil Ta)

Researchers from the University of Toronto are laying the technological groundwork to transform early childhood music education.  

Professor Elaine Biddiss (BME), a senior scientist at the Bloorview Research Institute, worked in collaboration with Dr. Tilak Dutta at KITE-UHN, to lead a new study that addresses a gap in musical instrument classification of non-pitched percussion instruments. Their findings were published in a recent issue of PLOS one. 

Pitched instruments, which play specific notes or pitches, have been extensively studied and classified. But non-pitched percussion instruments, such as tambourines, maracas and castanets, present unique challenges in identification due to overlaps in frequency bands and variations in sound quality and play style.  

In response to this challenge, the research team developed a sophisticated musical instrument classifier capable of identifying these instruments with remarkable accuracy. 

A composite photo. On the left there is a smiling woman wearing a blazer with her hair pulled back; on the right a man wearing a white t-shirt smiles with his arms crossed.
Left to right: Professor Elaine Biddiss (BME) and Brandon Rufino (BME MASc 2T1), who is the first author of the new study. (photo courtesy of Holland Bloorview Kids Rehabilitation Hospital; photo by Qin Dai)

The researchers generated a comprehensive dataset comprising diverse instruments, including variations in brand, materials, construction and play styles. This dataset, which included over 369,000 samples recorded in-lab and 35,361 samples recorded in family homes, is the largest of its kind for non-pitched instruments. 

Utilizing advanced signal processing techniques paired with machine learning algorithms, the team optimized feature selection, windowing time, and model selection to develop an efficient classifier. From this data they were able to develop a model, achieving over 84% accuracy in lab settings and over 73% accuracy in home settings across all three instrument families. 

“This research represents a significant step forward in early childhood music education,” says Biddiss. “By leveraging cutting-edge technology, we’ve developed a tool that can accurately detect non-pitched percussion instruments, opening doors for more inclusive and engaging music learning experiences.” 

The development of a mixed-reality music application, capable of detecting children’s use of non-pitched percussion instruments, holds promise for enhancing early childhood music education and play. Moreover, the study emphasizes the importance of inclusive design practices, catering to participants with diverse physical and cognitive abilities. 

“Music is a powerful tool for learning and development, particularly in early childhood,” says Brandon Rufino (BME MASc 2T1), first author of the new study. 

“With this technology, we aim to make music education more accessible and enjoyable for all children, regardless of their background or abilities.” 

Media Contact

Fahad Pinto
Communications & Media Relations Strategist