Skip to main content
Figure 7 | BioMedical Engineering OnLine

Figure 7

From: Formant analysis in dysphonic patients and automatic Arabic digit speech recognition

Figure 7

Recognition accuracy (%) of different voicing conditions using 12 MFCC features. The figure shows comparison of the performance of Arabic digit recognition system using 12 MFCC features for normal voice and six other voicing disorders. For all Gaussian mixtures, normal voice performs the best, which is quite usual. Among the voicing disorder, nodules have the highest performance, which is 76% using four mixtures. The lowest performance is with polyp. The number of mixture has no specific contribution towards the recognition performance for disordered voice.

Back to article page