Real-time Rhythmic Pattern Detection from Audio via Template Matching

ABSTRACT

Real-time Rhythmic Pattern Detection from live audio signals is possible using an audio programming environment such as Pure Data, but has been largely neglected. Lower level features such as amplitude, frequency, and tempo are commonly used as data for live processing, however, specific rhythmic pattern detection has not yet attained this same proliferation. We seek to prototype and implement a system for real-time rhythmic pattern detection from a live audio signal via template matching. Our approach is discussed, evaluating several alternative methods of preprocessing to the input audio signal and their effects on obtaining a distance measure between template and input signal. The system is further evaluated for robustness against near matches, and an online rhythmic pattern detector is implemented. Using the results from these tests, an outline for creating a real-time pattern detector in Pure Data is presented and future directions for this project are highlighted.

View Full Article