A Stratified Approach for Spoken Word and Accent Recognition: Validation & Analysis

摘要

Speech translation is a process of both speech recognition and equivalent phonemic to word translation. Accent is a pattern which differentiates the pronunciation and acoustic features based on the specific language group. The recognition and classification of words with different accent is also challenging problem in speech recognition research. There are many issues which limit the performance of such systems since a word spoken by different persons can have different acoustic properties due to variation in physiological characteristics, emotional status, and cultural background. Speech recognition is a process of identifying phonemes from the speech segment which is affected by the accent of the speaker. Some of the English words spoken by South Indian people whose native language being other than English like Telugu, Kannada, Tamil, Malayalam and Marathi, etc., will have a typical accent pronounced under the influence of their mother-tongue are incorrectly recognized by most of the translating systems. In this paper, an illustrative attempt has been made to effectively increase the efficiency of the system adopting a robust speech and accent recognition methodology. The focus is on automatically identifying the dialect or accent of a speaker given a sample of their speech, and demonstrates how Syllable MFCC, HMM and FO-ANN algorithm, a stratified approach through MATLAB tool can be employed to improve Automatic Speech and Accent Recognition (ASAR).