Coursera Spring Sale
40% Off Coursera Plus Annual!
Grab it
Learn about an efficient, open-source C++ algorithm for loop classification and tempo estimation in this 41-minute conference talk from ADC 2024. Discover the technical implementation behind Audacity's new tempo detection feature, presented by Matthieu Hodgkinson from the Muse Group's Audacity team. Explore how this "classical" algorithm achieves remarkable performance metrics, including an area under the ROC curve (AUC) of 0.93 and processing speeds over 2500 times faster than real-time on modern hardware. Understand the algorithm's unique approach to tempo detection, which evaluates the likelihood of different tatum count hypotheses by measuring onset distances and using weighted averages of onset strengths. Examine the evaluation framework that enables regression-free development and tuning, complete with computation time measurements and configurable false positive rate targeting. Gain insights into the algorithm's design philosophy, which specifically targets loop audio files to improve discriminant factors and achieve higher success rates while maintaining low false positive rates. Access the complete open-source implementation and evaluation framework, designed for easy reusability across different applications in audio processing and digital audio workstation development.
