Pustaka Ilmiah Universitas Padjadjaran

Abstrak

Analysis of Physical Parameters of Emotional Human Speeches Employing Principal Component Analysis

Human speeches can be recognized from their paralinguistic and linguistic point of view. The development of artificial-intelligence machines required the ability of the system to recognize the emotional of the speakers. The emotional condition of the speakers (sadness, happiness and anger) could be identified from the speeches intonation differences characterized by their different physical parameter such as: pitch, duration, intensity and formants. These physical parameters extracted from recorded human speeches in different emotional condition. In order to analyze their emotional condition from their physical parameters we employed Principal Component Analysis (PCA) to see the distribution of these parameters to indicate different emotional condition. PCA is the statistical method to see data distribution. The results of PCA1 vs. PCA2 distribution indicate unique characteristic emotional components for every emotional condition. Our method successfully differentiates the emotional condition base on human speeches.