Multiresolution alignment for multiple unsynchronized audio sequences using Sequential Monte Carlo samplers

摘要

With proliferation of smart devices such as smart phones, it is common that an event is recorded by multiple individuals creating several audio and video perspectives. Such user generated content is mostly unorganized (not synchronized). In this work, we consider the problem of aligning of multiple unsynchronized audio sequences and propose a multiresolution alignment algorithm using Sequential Monte Carlo samplers in a course to fine structure. The proposed method is evaluated with a real-life dataset from Jiku Mobile Video Datasets and has proven to be competitive with the baseline fingerprinting based alignment methods, with the proper choice of parameters.