Speak-Correct: A Computerized Interface for the Analysis of Mispronounced Errors

摘要

Any natural language may have dozens of accents. Even though theequivalent phonemic formation of the word, if it is properly called in differentaccents, humans do have audio signals that are distinct from one another. Amongthe most common issues with speech, the processing is discrepancies in pronunciation,accent, and enunciation. This research study examines the issues ofdetecting, fixing, and summarising accent defects of average Arabic individualsin English-speaking speech. The article then discusses the key approaches andstructure that will be utilized to address both accent flaws and pronunciationissues. The proposed SpeakCorrect computerized interface employs a cuttingedgespeech recognition system and analyses pronunciation errors with a speechdecoder. As a result, some of the most essential types of changes in pronunciationthat are significant for speech recognition are performed, and accent defects definingsuch differences are presented. Consequently, the suggested techniqueincreases the Speaker’s accuracy. SpeakCorrect uses 100 h of phoneticallyprepared individuals to construct a pronunciation instruction repository. Theseprerecorded sets are used to train Hidden Markov Models (HMM) as well asweighted graph systems. Their speeches are quite clear and might be considerednatural. The proposed interface is optimized for use with an integrated phoneticpronounced dataset, as well as for analyzing and identifying speech faults inSaudi and Egyptian dialects. The proposed interface detects, analyses, andassists English learners in correcting utterance faults, overcoming problems,and improving their pronunciations.