Skip to Main Content
The mandarin chinese oral speeches always contain a great number of emotions, and the speech emotion recognition is a very important key step of natural language understanding. As embedded system is widly applied in every walk of life, the speech emotion recognition technology will be extensively used in the future. Since some characteristics of embedded system such as poor resource, computing capability and etc., it is a great challenge that not only satisfactory score but also well efficiency of the speech emotion recognition algorithm must be requested. The paper describes an improved DTW-based speech emotion recognition model and an elementary emotional knowledge-base for mandarin chinese oral, and shows a embedded system application with speaker-dependent connected-word and isolated-word speech emotion recognition abilities.