UBC Researchers Tackle Sound Quality Issues in Live Performances

At a recent launch event, Dr. Miles Thorogood unveiled the SPIRAL lab at UBC Okanagan, a facility funded by the Canada Foundation for Innovation. This innovative space focuses on simulating the creative process in sound design, aiming to develop advanced models and algorithms for new computational tools. Dr. Thorogood and his team are addressing a common issue faced by audiences during live performances: sound distortion caused by technical glitches.

According to Dr. Thorogood, listening to live performances over the internet can exacerbate sound quality problems. “Yikes,” he remarked, highlighting the frustrations that often accompany remote viewing. The researchers at the Sonic Production, Intelligence, Research, and Applications Lab are utilizing advanced machine learning techniques to create glitch-free music performances. This effort aims to enhance creative collaboration and improve public art experiences.

Exploring Audio Data Transmission

The SPIRAL team is investigating how audio data travels through networks. They are employing neural networks for Packet Loss Concealment, a technique designed to generate synthetic audio that compensates for data loss. This approach ensures that listeners remain unaware of any interruptions during remote performances, particularly when high-quality audio is transmitted over networks in noisy urban settings.

Dr. Thorogood, who teaches in the Faculty of Creative and Critical Studies, explains that audio packet loss often disrupts the flow of music. This is especially problematic when data packets fail to reach their destination. The User Datagram Protocol (UDP), which is commonly used for real-time communication, lacks error control. Consequently, lost or corrupt packets cannot be retransmitted, leading to noticeable glitches in sound.

The research conducted by Dr. Thorogood and undergraduate media studies student Yashvardhan Joshi was recently published in IEEE Access. Their findings illustrate the intersection of art and science, as they work to conceal glitches in complex soundscapes that merge both natural and electronic sounds. The goal is to create immersive public installations and musical experiences that feel uninterrupted.

Application of Research in Public Art

During the design phase for the Light Up Kelowna project, Dr. Thorogood’s team implemented a wireless multi-node audio-visual format. They identified instances of audio packet loss, which prompted further investigation into the underlying issues. An analysis revealed that the large distances between nodes and the noisy radio frequency environment contributed to the problem.

To address this, the researchers explored packet loss concealment techniques to improve network music performance. This research not only benefited their academic pursuits but also provided Yashvardhan with an opportunity to investigate how deep learning algorithms can create innovative audio effects. His work extends beyond sound, involving the development of light and sound technology for public art.

Dr. Thorogood emphasizes the significance of this research, noting its potential to enhance creative workflows in the music industry. The methodologies developed at SPIRAL are designed to streamline the creative process, making it easier for artists to produce high-quality sound in various settings.

For more information on the ongoing work at SPIRAL, visit the UBC Okanagan website.