A novel speech recognition method for student management system

Speech recognition is one of the most important technologies in speech application. This paper proposes a key word detection method for continuous speech in noisy environment. In the proposed method, we extract the widely used energy, zero crossing, entropy and MFCCs to generate an audio feature set. Moreover, we have also used a robust endpoint detection algorithm which makes the feature modify its parameter by adapting to the strength of background noise. Then HMMs are used for the classifiers. Experiments were made under different types of noises and the results show that this method is more accurate and more anti-noise than traditional methods. Moreover, we used this method in a student management system to recognize some key words.