In this paper we investigate Linear Discriminant Analysis (LDA) for the TI connected digit recognition task (TI task) and the Wall Street Journal large vocabulary recognition task (WSJ task). In addition to previous variants of LDA implementations, we avoided the explicit incorporation of derivatives in the acoustic vector. Instead a sliding window without derivatives was used. This large-sized vector was then taken to extract the features by an LDA transformation. Tests for this feature generation were performed both for Laplacian and Gaussian densities.
Bibliographic reference. Beulen, K. / Welling, L. / Ney, Hermann (1995): "Experiments with linear feature extraction in speech recognition", In EUROSPEECH-1995, 1415-1418.