Recognition of Paralinguistic Information Using Prosodic Features Related to Intonation and Voice Quality

Download or read book Recognition of Paralinguistic Information Using Prosodic Features Related to Intonation and Voice Quality written by Carlos T. Ishi and published by . This book was released on 2008 with total page pages. Available in PDF, EPUB and Kindle. Book excerpt: We proposed and evaluated intonation and voice quality-related prosodic features for automatic recognition of paralinguistic information (intentions, attitudes and emotions) in dialogue speech. We showed that intonation-based prosodic features were effective to discriminate paralinguistic information items expressing some intentions or speech acts, such as affirm, deny, thinking, and ask for repetition, while voice quality features were effective for identifying part of paralinguistic information items expressing some emotion or attitude, such as surprised, disgusted and admired. Among the voice qualities, the detection of pressed voices were useful to identify disgusted or embarrassed (for "e", "un"), and admiration (for "he"), while the detection of harsh/whispery voices were useful to identify surprised/unexpected or suspicious/disgusted/blame/dissatisfied. Improvements in the detection of voice qualities (harshness, pressed voice in nasalized voices, and syllable offset aspiration noise) can still improve the detection rate of paralinguistic information items expressing emotions/attitudes. Future works will involve improvement of voice quality detection, investigations about how to deal with context information, and evaluation in a human-robot interaction scenario.