Speech emotion recognition is a very important speech technology. an extensive research is made by using different speech information and signal for human emotion recognition. We develop a speech-based emotion classification method using SVM by using standard EMA database. In order to achieve a high emotion classification accuracy we have used SVM with kernel functions, From result obtained by using different kernels functions . From result we conclude that RBF Kernel function in which we got 94.96%, 96.02%, 98.96%, 98.76% accuracy results for Angry, Happy, Neutral, Sad emotions respectively using energy, formant and MFCC features. Our result shows that classification accuracy will be improve using kernel functions.