Measures of distance between speech signals in the frequency domain
Abstract
The energy density spectra and the time behavior of the autoregressive model of voice signals are studied. The first is a complete representation of the voice signal in the frequency domain, derived from the application of the Fourier transform to the sampled signal. The second is based on a voice signal generation model in which the channel is represented by poles only (resonances). A symmetrical version of the Itakura distance is the best compromise between conflicting requirements.
- Publication:
-
NASA STI/Recon Technical Report N
- Pub Date:
- February 1984
- Bibcode:
- 1984STIN...8510267D
- Keywords:
-
- Autocorrelation;
- Spectral Energy Distribution;
- Speech Recognition;
- Voice Data Processing;
- Fourier Analysis;
- Sampling;
- Signal Analysis;
- Time Functions;
- Verbal Communication;
- Communications and Radar