On sequential Monte Carlo sampling methods for Bayesian filtering
Statistics and Computing
Nonlinear filtering for speaker tracking in noisy and reverberant environments
ICASSP '01 Proceedings of the Acoustics, Speech, and Signal Processing, 2001. on IEEE International Conference - Volume 05
EURASIP Journal on Applied Signal Processing
Particle filter with integrated voice activity detection for acoustic source tracking
EURASIP Journal on Applied Signal Processing
A tutorial on particle filters for online nonlinear/non-GaussianBayesian tracking
IEEE Transactions on Signal Processing
IEEE Transactions on Audio, Speech, and Language Processing - Special issue on processing reverberant speech: methodologies and applications
Directional acoustic source orientation estimation using only two microphones
Digital Signal Processing
Hi-index | 0.08 |
The source height estimation (SHE) based hybrid particle filtering (PF) algorithm was originally presented by the authors in EUSIPCO 2008. The accuracy of the source height estimation was proposed to be used as a direct judgement of whether the current acoustic sensor array's observation is originated from the true source state or from the reverberation. According to the judgement, the hybrid PF algorithm adaptively allocates the particles using either the bootstrap (BS) algorithm or the sequential importance sampling (SIS) algorithm to effectively tackle the reverberation effects. In this paper, a comprehensive analysis is conducted on the work reported in that conference paper and additional novelties are included too. The newly designed quad-stage statistical (QSS) source height estimation algorithm (QSS-SHE) creates more hypotheses in the effective evaluation of the accuracy of the height estimation. Based on the QSS evaluations, the more precisely defined statistical weights (SW) based hybrid structure is introduced. At every time instance, particles sampled using both the BS and the SIS algorithms are mixed and updated according to their associated SW. The resulted hybrid particle filtering (SW-PF) algorithm demonstrates more robust tracking performance as compared to the hybrid PF algorithm originally proposed in the conference paper.