Introduction - Robust Emotion Recognition Using Spectral and Prosodic Features - page 3

The emotions generally represent a synthesis of subjective experience, expressive
behavior, and neurochemical activity. There are more than 300 crisply identified
emotions by researchers [ 3 , 4 ]. However generally, all of them are not experienced
in day-to-day life. In this regard, most researchers agree on the principles of Palette
theory that quotes, any emotion is the composition of six primary emotions as any
color is the combination of 3 primary colors [ 5 ]. Anger, disgust, fear, happiness, sad-
ness and surprise are considered as the primary or basic emotions bymost researchers
[ 6 ]. These are also referred as archetypal emotions [ 7 ].
In psychology, expression of emotions is considered as a response to stimuli that
involves characteristic physiological changes [ 8 , 9 ]. According to physiology, an
emotion is defined as a disruption in the homeostatic baseline [ 8 ]. Based on these
changes, the properties of emotions can be explained in a three-dimensional space.
Essential dimensions of emotional states are captured by the features of activation
(arousal; measured as an intensity), affect (valence or pleasure; measured as posi-
tive or negative feeling after emotion perception) and power (control; measured as
dominance or submissiveness in emotion expression). According to the physiology
of emotion production mechanisms, it has been found that the nervous system is
stimulated by the expression of high arousal emotions like anger, happiness and fear.
This phenomenon causes an increased heart beat rate, higher blood pressure, changes
in respiration pattern, greater sub-glottal air pressure in the lungs and dryness of the
mouth. The resulting speech is correspondingly louder, faster and characterized with
strong high-frequency energy, higher average pitch, and wider pitch range [ 10 ]. On
the other hand, for the low arousal emotions like sadness, the nervous system is stim-
ulated causing the decrease in heart beat rate, blood pressure, leading to increased
salivation, slow and low-pitched speech with little high-frequency energy. Thus,
acoustic features such as pitch, energy, timing, voice quality, and articulation of the
speech signal highly correlate with the underlying emotions [ 11 ]. However, distin-
guishing emotions without any ambiguity, using any one of the above mentioned
dimensions, is a difficult task. For example, both anger and happiness have high
activation but they convey different affect (valence or pleasure information). This
difference is characterized by using both activation and valence dimensions. Hence,
emotion is the complex psychophysiological experience of an individual's state of
mind as interacting with biochemical (internal) and environmental (external) influ-
ences. In humans, emotion fundamentally involves physiological arousal, expressive
behaviors, and conscious experience. Therefore the three aspects of emotions are
psychological (what one is thinking), physiological (what one's body is doing), and
expressive (how one reacts) in nature [ 12 ].
1.3 Emotion from Speech Signal Perspective
Psychological and physiological changes caused due to emotional experience lead
to certain actions. Speech is one of the important outcomes of the emotional state of
human beings. A speech signal is produced from the contribution of the vocal tract