Personality Traits from Speech Signal using Cross- Corpus Technique

TitlePersonality Traits from Speech Signal using Cross- Corpus Technique
Publication TypeConference Proceedings
Year of Conference2018
AuthorsVijay, N., S.. Tripathi, and S.. Lalitha
Conference NameIEEE International Conference on Computational Intelligence and Computing Research
KeywordsDept. of Electronics and communication Engineering.

The focus of this work is to detect the psychological emotional state of a human and also determine the personality trait of the person using speech samples. Cross-corpus technique has been employed for validation. Various Spectral features of speech along with Domain-Adaptive Least square Regression (DaLSR) and Auto-encoder classifier are considered. Voice samples from the publicly available database of Berlin and Enterface are used. The work is extended to classify the personality of a person into introvert or extrovert using detected psychological state. Using cross-corpus technique an improvement of 15% is obtained for psychological state classification compared to the existing reported work. The technique of personality classification is an initial attempt and needs to be improved for better recognition.