Extraction of feature parameters for spoken word recognition
Abstract
Extraction of local peaks in spectra of voiced signals for a speech recognition system is discussed. Since there are differences of slopes of speech spectra between members, an adaptive correction for the slope is required. A simple method of compensating a slope using a linear phase digital filter is given and it is compared with the correcting method by a least mean square fitting.
- Publication:
-
In its Res. Rept. of the Faculty of Eng
- Pub Date:
- December 1984
- Bibcode:
- 1984rrfe.nasa...57T
- Keywords:
-
- Signal Processing;
- Speech Recognition;
- Weighting Functions;
- Amplitudes;
- Digital Filters;
- Least Squares Method;
- Slopes;
- Speech;
- Communications and Radar