Improving the Robustness of Phoneme Classification Using Hybrid Features
This is the latest version of this eprint.
This work is concerned with improving the robustness of phoneme classification to additive noise with hybrid features using support vector machines (SVMs). In particular, the cepstral features are combined with local energy features of acoustic waveform segments to form a hybrid representation. The local energy features are taken into account separately in the SVM kernel, and a simple subtraction method allows them to be adapted effectively in noise. This hybrid representation with mean and variance normalization of the cepstral features contributes significantly to the robustness of phoneme classification and narrows the performance gap to the ideal baseline of classifiers trained under matched noise conditions. Further improvements are obtained by extending the multiclass prediction method from standard discrete error-correcting codes to adaptive continuous codes.
Available Versions of this Item