This paper proposes a new method for speaker identification, which based on the non-uniformly distributed speaker information in frequency bands. In order to discard the linguistic information effectively, in this study, we adopt an improved Fisherfs F-ratio called the phoneme mean F-ratio to measure the dependences between frequency components and individual characteristics. Then we adopt an adaptive frequency filter to extract more discriminative feature. The experiment shows that the recognition rate using the proposed feature is increased by 0.62% compared with the F-ratio feature, and increased by 3.46% compared with the MFCC feature. The results confirmed that emphasizing the features from highly individual dependent frequency bands is valid for improving speaker recognition performance.
Index Terms: speaker identification, frequency warping, F-ratio