Abstract: The speaker recognition is an important research direction of speech recognition. The feature extraction is a key technology in the speaker recognition system. This paper proposes an efficient scheme for Mel Frequency Cepstral Coefficients （MFCC） computation. It exploits the triangular structure of the frequency response function of the MFCC filters. The simulation results show that the scheme proposed in this paper reduced the access time by 83.6% in single frame memory and the proposed scheme gained better latency with less memory.

Key words: LabVIEW, speech recognition, MFCC, pretreatment