Affect Recognition in Human Emotional Speech using Probabilistic Support Vector Machines

The problem of inferring human emotional state automatically from speech has become one the central problems in Man Machine Interaction (MMI). Though Support Vector Machines (SVMs) were used several worksfor emotion recognition speech, potential using probabilistic SVMs for this task is not explored. emphasis current work on how to use efficient emotions speech. Emotional corpuses two Dravidian languages- Telugu & Tamil- constructed assessing accuracy Probabilistic SVMs. Recognition proposed model analyzed both and Tamil compared with three existing works. Experimental results indicated that significantly better methods.