Robust Emotion Recognition using Sentence, Word and Syllable Level Prosodic Features - Robust Emotion Recognition Using Spectral and Prosodic Features - page 64

90

Initial words
Middle words
Final words

80

70

60

50

40

30

20

10

0

Anger Disgust Fear Happy Neutral Sad Sarcastic Surprise

Emotions

Fig. 3.6 Comparison of emotion recognition performance using global + local prosodic features
extracted from initial , middle and final words

the words than the entire duration of words. In general, final syllables contribute
heavily toward emotion recognition. Average emotion recognition due to only final
syllables is about 61%, whereas it is 51 and 46% respectively for the groups of ini-
tial and middle syllables. The comparison of recognition of different emotions using
utterance-wise initial, middle, and final syllables is given in Fig. 3.7 . From the figure,
it is observed that, final syllables have more emotion discriminative information, in
case of most of the emotions except anger and happiness. These are high-arousal
emotions, and hence, their discrimination is better in the case of initial syllables.
Middle syllables do not contribute much toward emotion recognition than the initial
and final syllables. The overall emotion recognition performance due to the combi-
nation of initial, middle, and final syllables is given in Table 3.13 , This is comparable
to the results of word and utterance level studies (See Tables 3.9 and 3.11 ).

3.5.3.2 Region-Wise Syllable-Level Emotion Recognition

At the word-level emotion recognition analysis, we have studied emotion discrimina-
tive characteristics from the set of initial, middle and final words. Within these groups
of words, there may be some additional emotion discriminative information present
at the syllable level. Therefore, to capture emotion-specific information from the
syllables within the group of words (initial, middle and final), the syllables of these
words are divided into initial, middle and final syllables. The syllables within each
region of words (initial, middle and final words) are grouped into initial, middle, and
final syllables, based on their positions in the words. The details of syllable groups
are given in Table 3.5 . The block diagram of the ERSs developed using region-wise
syllable-level prosodic features is similar to the ERS developed using word level fea-
tures. Here, the ERS in each block has the same structure as shown in Fig. 3.3 .Overall