Speech consists of acoustic pressure waves created bythe voluntary movements of anatomical structures in the humanspeech production system. These waveforms are broadly classifiedinto voiced and unvoiced speech. Voiced sounds (vowels forexample), produce quasi-periodic pulses of air which areacoustically filtered as they propagate through the vocal tract.The main distinction between vowels and consonants is that vowelsresonate in the throat. Formants are exactly the resonantfrequencies of a vocal tract when pronouncing a vowel. In thispaper we attempt to carry out Vowel Recognition through FormantAnalysis in Serbian language, wherein we detect which of the fiveSerbian vowels is spoken by the Speaker. Here we describe astandard approach for classification of vowels in continuousspeech based on three formants: F1, F2 and F3. We haveinvestigated the correlations between formants in each vowel anddeveloped the algorithm to reduce the overlap of different vowelsin F1-F2 and F2-F3 planes.
展开▼
