A Mixed Parameter Method Based on MFCC and Fractal Dimension for Speech Recognition

We propose a speech recognition approach with mixed parameter in this paper, which combines the traditional MFCC and fractal feature as the feature parameter. MFCC has higher spectrum resolution at low frequency segment, while it cannot represent speech nonlinearity. Fractal dimension is used to quantitatively describe the chaos nonlinearity in speech air flow. Experimental results demonstrate this method is promising in improving speech recognition performance.