Robust Emotion Recognition using Pitch Synchronous and Sub-syllabic Spectral Features - Robust Emotion Recognition Using Spectral and Prosodic Features - page 25

(a)

(b)

(c)

(d)

(e)

(f)

(g)

20

(h)

0

−20

0 1,000 2,000 3,000 4,000

Frequency (Hz)

Fig. 2.2 Spectra for the steady region of the syllable /tha/ from the utterance thallidhandrulanu
gauravincha valenu . a Anger, b Disgust, c Fear, d Happy, e Neutral, f Sadness, g Sarcastic, and
h Surprise

2.3.4 Extraction of Sub-syllabic Spectral Features

In the context of Indian languages, a syllable can be viewed as the combination of
consonants and a vowel, in the form C m VC n , where C is the consonant and V is the
vowel such that, m

3. In the syllable, the vowel is treated as the
nucleus and consonants may or may not be present. In the context of Indian languages,
most common syllable forms are CV, CCV, CCVC, and CVC. Among these forms,
more than 90 % of the syllables are of the type CV. In a CV unit, the speech signal to
the left of the VOP (before VOP) is a consonant region and to the right of it is the vowel
region [ 8 ]. Vowel onset point as a junction point between the consonant and vowel of a
CV unit. At this point the characteristics of the consonant segment are terminated and
the characteristics of vowel are originated. Hence, it is important to extract features
around this crucial point [ 9 ]. After determining the VOP, 40 ms to the left of the VOP

, n ≥ 0 and m , n ≤