Mood identification

Abstract

Computers are becoming more and more intelligent every day, providing users with personalised content and catering for their every need. One further step to enhance this service would be to actually recognise the current user's mood and act accordingly. Computer programs of the future will be able to recognise when the user is angry and provide a response that will attempt to better their mood by means of multimedia content and other mood stimulating activities. Even the demeanour of the program would change to accommodate the user's mood. That is for example, like humans, the computer would try to cheer the user up if it recognises that he/she is sad or upset. Before any of this is possible the computer program must first find a reliable way how to establish the user's mood. Our goal is to find out what has been achieved in this field and to present a computer program that is able to recognize a speaker's mood from both speech samples and the text equivalent of the sample. The first thing to consider is that moods are extremely hard to categorise and distinguishing between each category is also very hard, even for humans. To this extent our system will be able to recognize between five very well known mood categories which are Happy, Excited, Neutral, Angry, and Sad. We will show how certain categories, such as happy and excited, are hard to distinguish from one another, by using analysis of speech and text. We will discuss how certain features, found in speech samples, can be extracted and used to determine a person's mood. The person's speech rate, i.e. the speed at which the person talks, can be used to determine various moods. For example when a person talks slowly this may indicate that they are either sad or bored. To further elaborate on this, the energy level of the sample could be extracted. A low energy level would indicate that the speaker is sad or bored. Combining such features together one could come to certain conclusions that would aid in determining the speaker's mood. Again these methods are not fool proof, as it can be said that even when a person is in a neutral mood, he/she would still talk slowly and softly [...].