Attention level approximation of a conversation in human-robot vocal interaction using prosodic features of speech

Abstract

With the widespread use of robots and integration of robots into the daily human life, the communication abilities of the robots are highlighted in order to ensure smooth interaction. Domestic robots with conversation capabilities are getting an increasing demand due to the fact that modern lifestyles are driving elderly community into isolation. One of the requirements of a conversation system is to ensure that all the people involved are interested in carrying out the conversation. Humans use various physical parameters in order to determine the interest level of the speaker to ensure the conversation is interesting for both parties. The interest level estimation can be used by the robot as a form of feedback to evaluate the effectiveness of the conversation system. In order to evaluate the attention level of the user a system using the prosodic features of speech is proposed. Prosodic features can be associated with the auditory and acoustic measures. These features are important in indicating emotions and attitudes of the speaker. By using the proposed system the robot will be able to determine the interest level using the responses given by the user to the robot by analyzing the prosodic parameters of the vocal response.