The application of the additive model in the feature extraction of speech recognition

Considering the requirements of recognition rate, computational complexity and storage expenses, an additive model of the feature matrix is presented that is based on the analysis of the function factor on acoustic characteristics and its behavior. On this basis, the feature characteristic algorithm is introduced to estimate the feature characteristic matrix reflecting the essence of speech from observed feature matrices. When the feature characteristic matrix is employed as the final recognition template, experimental results demonstrate the effectiveness of the new method.