Emotion recognition is an important research area in speech recognition. The features of the emotions will affect the recognition efficiency of the speech recognition systems. Various techniques are used in identifying the emotions. In this paper a novel methodology for identification of emotions generated from speech signals has been addressed. This system is proposed using Skew Gaussian mixture model. The proposed model has been experimented over a gender independent emotion database. In order to extract the features from the speech signals cepstral coefficients are used. The developed model is tested using real-time speech data set and also using the standard and data set of Berlin. This model is evaluated in the presence of noise and without noise the efficiency of the model is evaluated and is presented by using confusion matrix.