Application of Wavelet Karhunen-Loeve Transform to Text Independent Speaker Identification

S.-Y. Lung (Taiwan)


Speaker recognition, Wavelet transform, Karhunen-Loevetransform


A new speaker feature extracted from wavelet Karhunen Loeve transform (KLT) for speaker recognition is described. The wavelet KLT is a two stages representation of the features. Feature extraction first the data in the spatial domain using the wavelet transform and afterwards using the KLT in spectral domain. Our results have shown that this wavelet KLT feature introduced better performance than the cepstrum and -cepstrum with respect to the percentages of recognition.

Important Links:

Go Back