Speaker Independent Phoneme Recognition Based on Fisher Weight Map

We have already proposed a new feature extraction method based on higher-order local auto-correlation and Fisher weight map (FWM) at Interspeech2006. This paper shows effectiveness of the proposed FWM in speaker dependent and speaker independent phoneme recognition. Widely used MFCC (Mel-frequency cepstrum coefficient) features lack temporal dynamics. To solve this problem, local auto-correlation features are computed and accumulated by weighting high scores on the discriminative areas. This score map is called Fisher weight map. From the speaker dependent phoneme recognition, the proposed FWM showed 79.5% recognition rate, by 5.0 points higher than the result by MFCC. Furhermore by combing FWM with MFCC and DeltaMFCC, the recognition rate improved to 88.3%. In the speaker independent phoneme recognition, it showed 84.2% recognition rate, by 11.0 points higher than the result by MFCC. By combining FWM with MFCC and DeltaMFCC, the reecognition rate improved to 89.0%.