Application of support vector machines to recognize speech patterns of numeric digits
Gracieth C. Batista, Washington Luis Santos Silva · 2015
An application of Support Vector Machine (SVM) in recognition of speech patterns is presented on this paper. In other words, numeric digits of the Brazilian Portuguese language are pre-processed by Discrete Cosine Transform (DCT) with mel-cepstral coefficients and transformed in two-dimensional matrices. Such matrices have the mean and variance of speech signals acquired by the pre-processing. Besides, after that, these matrices are applied as input of the SVM training algorithm to find the best pattern of each digit. In a third step, these patterns are compared to other independent voice bank and thus, the recognition is completed.