A Literature Review on Emotion Recognition in Speech.

In our age, we are bombarded with multimedia content daily. Although, face-to-face communication always outgrows the potential factors of healthy assessment of our peers through recorded content or live media interaction, (be it text, video, images, speech) new approaches to render us able to understand and discern between emotions of our peers on multimedia content are getting more and more popular and more complex. Two robust topics in this regard are generally named as sentiment analysis and emotion detection. The advent and exponential growth of social networks and for instance, the employment of speech bots have made it a necessity to particularly address the problem of healthy emotion recognition outside face-to-face, everyday conversations or interactions. Machines’ capability to perform the set of tasks through Machine Learning approaches, namely consisting of detecting, expressing, and understanding emotions is collectively known as, as in humans, emotional intelligence. Different modes of input as human behavior like those taken from audio, image, video sources and signal interpretations processed through Electro-encephalography (EEG), related brain wave measurements are used in emotion recognition. My study aim is intended to be the examination and review of recent study approaches in Emotion Detection in Speech, possibly establishing links or differences between recent study publishes because each study paper focuses on a single or set of Machine Learning approaches which are employed in Emotion Detection in Speech. This paper tries to examine various relevant research involving methods of Machine Learning which were studied and tested under this research respective to Speech Emotion Recognition (SER). Effectiveness of the involved methods and databases are discussed while commenting on the studies and expressed in the form of their findings. Improvements throughout these studies are, though not chronologically, compared using simple tables which show independent accuracies of several Machine Learning classifier combinations. [ABSTRACT FROM AUTHOR]