Towards robust phoneme classification: Augmentation of PLP models with acoustic waveforms
Matthew Ager, Zoran Cvetkovic, Peter Sollich and Bin Yu
In: EUSIPCO 2008, Lausanne(2008).
The robustness of classification of phoneme segments using
generative classifiers is investigated for the PLP and acoustic
waveform speech representations in the presence of white
Gaussian noise. We show a method to combine the strengths
of both representations, specifically the excellent classification
accuracy of PLP in quiet conditions with the additional
robustness of acoustic waveform classifiers. This is
achieved using a convex combination of their respective loglikelihoods.
Issues of noise modelling and time-invariance of
acoustic waveforms are also addressed with initial solutions
shown. The resulting combined classifier has greater accuracy
than PLP alone and is significantly more robust to the
presence of additive noise during testing.