Identification of Malay Stop Consonants Based on MFCC &Rasta PLP Features Using K-NN Classifier for Cued Speech Application

Abstract

: Phonological studies suggest that phoneme awareness in an early age through cued speech is reliable to measure the literacy skills and provide a strong language foundation for deaf children. This paper proposed a phonemic-based recognition of Malay Phonemes according to stop voicing. Eight stop consonants /p b t d t∫ dʒ k g/ preceding /a/ vowel are selected to encode each combination as a hand shape at a specified position. Features are extracted by using Mel -frequency Cepstral Coefficients (MFCC), MFCC with delta coefficients, MFCC with delta to delta coefficients and Rasta PLP. Mean of the featured group samples were taken to reduce the frame dimensions of the extracted features. These dimensionalities reduced features were fed into the k-Nearest Neighbors (k-NN) classifier for the classification. K-fold cross validation method is used to test the reliability of the classifier results. Experimental results show that the best identification rate is 92.5% upon feature fusion sets of MFCC, MFCC with delta coefficients, MFCC with delta to delta coefficients and Rasta PLP., voltage divider circuit voltage waveforms and inverter output waveforms are shown using MATLAB Simulink software