Dinoj Surendran

Can voice quality improve mandarin tone recognition

We investigate several measures of voice quality (VQ) to improve tone recognition in Mandarin Chinese. We find that band energy measures such as spectral balance (Sluijter and van Heuven, 1996) work better than measures based on glottal flow estimation and harmonic-formant differences. We also determine a set of bands and measures that improve tone classification accuracy on broadcast news speech to 64.1% from 60.4% when added to a traditional pitch-duration-intensity set of features. Most improvement is for the neutral tone, for which the F score increases from 0.345 to 0.619.

Visualizing High Dimensional Datasets Using Partiview

A standard method of visualizing high-dimensional data is reducing its dimensionality to two or three using some algorithm, and then creating a scatterplot with data represented by labelled and/or colored dots. Two problems with this approach are (1) dots do not represent data well, (2) reducing to just three dimensions does not make full use of several dimensionality-reduction algorithms. We demonstrate how Partiview can be used to solve these problems, in the context of handwriting recognition and image retrieval.