Identifying Periodic Signal Patterns in Audio Streams

Abstract

We develop a novel and efficient method for identifying periodic signal patterns in audio streams. For this purpose we introduce the concept of a similarity function that measures the degree of equivalency of audio samples. By aggregating the measurements in the form of a so-called similarity matrix, we can thoroughly visualize the similarity of every pair of samples of an audio signal. This visualization (similarity map) is subsequently used to identify the existence of periodic patterns. Audio compression and stream reduction are two applications of our proposed method. Specifically, it can be used in light-weight stream reduction algorithms that benefit battery-powered networks such as sensor networks.