Recognizing emotions at better accuracy is very challenging task. Therefore, in recent time, the human-machine interaction technology has gained so much success for recognizing the emotional states depending on physiological signals. The human emotional states can be detected by using facial expressions, but sometimes the accurate results are not achieved. Therefore in proposed work, the emotions are recognized using Electroencephalogram (EEG) which work on the basis of brain signal. Here, the human emotional states data is collected using DEAP Dataset and Artificial Neural Network (ANN) is used as classifier. Five time domain features namely correlation, average, variance, kurtosis and skewness are calculated for three frequency bands theta, alpha and beta. The data for two emotional dimensions valence and arousal is taken from DEAP Dataset. The proposed work gives better recognition results for valence and arousal dimensions which are 85.60 % and 87.36 % respectively. So we get the success in achieving significant accuracy.