A hardware preprocessor for use in speech recognition: Speech Input Device SID3
Abstract
A device which reduces the amount of data sent to the computer for speech recognition, by extracting from the speech signal the information that conveys the meaning of the speech, all other data being discarded is presented. The design includes signal to noise ratios as low as 10 dB, public telephone frequency bandwidth and unconstrained speech. It produces continuously at its output 64 bits of digital information, which represents the way 16 speech parameters vary. The parameters cover speech quality, voice pitch, resonant frequency, level of resonance and unvoiced spectrum color. The receiving computer must have supporting software containing recognition algorithms adapted to SID3 parameters.
- Publication:
-
NASA STI/Recon Technical Report N
- Pub Date:
- May 1983
- Bibcode:
- 1983STIN...8331923R
- Keywords:
-
- Data Reduction;
- Preprocessing;
- Signal Encoding;
- Speech Recognition;
- Voice Data Processing;
- Bandwidth;
- Computer Programming;
- Frequency Analyzers;
- Noise Spectra;
- Resonant Frequencies;
- Communications and Radar