Put sounds in intervals, measure energies, identify words, and parse sentences to recognize speech {speech recognition}. Segment acoustic waveform into ten-millisecond intervals {segment, sound} {sound segment}.
energy
In intervals, measure several frequency-band energies from 200 Hertz to 6000 Hertz. Linear predictive coding coefficients, zero-crossing rates, glottal frequency, and total energy are measures.
word
Energy sustained over several intervals indicates word. Interval number is word duration. Compressing longer words and elongating shorter words can standardize word duration.
pause
Low energy in one or more intervals indicates pause between words.
identification
Compare word frequency-band energy spectrum to stored word templates to identify word.
parsing
Use word syntactic and semantic information to parse sentence.
Systems {language understanding system} can detect strings using features and then classify strings and substrings by syntax.
Outline of Knowledge Database Home Page
Description of Outline of Knowledge Database
Date Modified: 2022.0225