8th International Conference on Spoken Language Processing

Jeju Island, Korea
October 4-8, 2004

Unscented Kalman Filtering of Line Spectral Frequencies

Andrew Errity (1), John McKenna (1), Stephen Isard (2)

(1) Dublin City University, Ireland
(2) University of Edinburgh, Scotland

We propose a new method for estimating Line Spectral Frequency (LSF) trajectories which uses unscented Kalman filtering (UKF). This method is based upon an iterative Expectation Maximisation (EM) approach in which LSF estimates are generated during a forward pass and then smoothed during a backward pass. The EM approach also provides re-estimated Kalman filter parameters for further forward-backward passes that improve estimation. This approach exploits the non-independence of neighbouring spectra. We estimate LSFs as they have good interpolation and quantization properties. This allows us to estimate LSF trajectories that are guaranteed to result in stable filters. We analyse noisy synthetic speech using this technique. The results compare favourably with other methods.

Full Paper

Bibliographic reference.  Errity, Andrew / McKenna, John / Isard, Stephen (2004): "Unscented kalman filtering of line spectral frequencies", In INTERSPEECH-2004, 2697-2700.