Robust Emotion Recognition using Pitch Synchronous and Sub-syllabic Spectral Features - Robust Emotion Recognition Using Spectral and Prosodic Features - page 34

1

Extract LPCC
features

LPCC +
formant
features

4

Parameters directly
from entire speech
signal

Extract formant
specific
features

3

MFCC +
formant
features

5

Extract MFCC
features

2

Extract LPCC
features

6

LPCC +
formant
features

8

Extract formant
specific
features

Parameters from
vowel region of a
CV unit

Vowel region
identification

MFCC +
formant
features

9

Extract MFCC
features

7

10

Extract LPCC
features

LPCC +
formant
features

12

Speech
Signa l

Consonant
region
identification

Extract formant
specific
features

Parameters from
consonant region of
a CV unit

VOP detection

MFCC +
formant
features

13

Extract MFCC
features

11

14

Extract LPCC
features LPCC +

formant
features

16

Transition
region
identification

Extract formant
specific
features

Parameters from
transition region of
a CV unit

MFCC +
formant
features

17

Extract MFCC
features

15

18

Extract LPCC
features LPCC +

formant
features

20

Pitch
synchronous
analysis

Extract formant
specific
features

Parameters from
within a pitch period
of a speech signal

Pitch period
identification

MFCC +
formant
features

21

Extract MFCC
features

19

Fig. 2.9 Block diagram indicating various emotion recognition systems developed using the pro-
posed spectral features, extracted from the entire speech utterance, sub-syllabic segments (vowel,
consonant, and CV transition), and pitch synchronous analysis