Multimodal blind source separation for moving sources

A Multimodal Approach to Blind Source Separation of Moving Sources

“…The sampling importance resampling particle filter (SIR-PF) is also suitable for this case as used in our work [53]. In the second experiment, both speakers are simultaneously moving and their motion is more complicated as they cross over.…”

“…Independent SIR-PF for each speaker is the most optimal choice and is already used in our work [53]. Results show no significant difference (based on Euclidean error) but MCMC-PF reduces the computational complexity in multispeaker tracking.…”

“…In [53], we used the random walk model; another model which is shown to work well, to represent the time-varying location of a speaker in a typical room [53], [54], is the Langevin model [55], also used in [31], [42], and [44]. The motion of the speakers in each coordinate is assumed to be independent in this state model.…”

“…As we know the above PI based on [3] is insensitive to permutation. We therefore evaluated the permutation on the basis of the criterion, i.e., for permutation free FDCBSS, used in our works [8], [53]. …”

Sampling Adaptive Learning Algorithm for Mobile Blind Source Separation

“…The core of BSS is its ability to extract independent components from an observed mixture signal, without requiring a prior knowledge. Such flexibility has made BSS popular in many applications [5][6][7] especially in mobile intelligence [8,9].…”