Abstract—In this study, a method that created a model by emotion using the MFCC and HMM structure, which is widely used in voice recognition systems, was adopted, thereby including explicit information and adding energy and speech speed to the MFCC feature to propose an emotional recognition method that uses features for emotional recognition. In addition, the validity of the method was verified through experiments. Furthermore, an instantaneous calculation method of speech speed was proposed and the experiment verified that the recognition rate was improved by 2.5% using the proposed method. For the calculation of speech speed, a codebook size of 32, the instantaneous calculation method of speech speed, and a window size of 200ms were employed.