Speech Signal-Based Modelling of Basic Emotions to Analyse Compound Emotion: Anxiety

Abstract

True emotions are rapid and short-lived. In realtime, two or more emotions can exist simultaneously at a given instant. The emotions can be expressed sequentially or merge to form a new emotion altogether. Combination of two or more basic emotions is known as compound emotions. Anxiety, an emotion triggered as a response to a potential threat, can be defined as a compound emotion comprising anger, fear and sadness predominantly. Speech signal-based emotion recognition system is developed to identify the basic emotions in anxiety. Autoencoders are used to learn the features of the voice samples and neural network classifies the basic emotions and estimates the likelihood of those in anxiety.