Signal Processing for Language Sanitization: Detection and Censorship of Obscene Words in Speech Recordings

In this paper, we have given a method for automatically detecting and censoring obscene and profane words from digital speech signals. Our method involves denoising the speech signal using spectral subtraction, converting speech into text, and examining it for the existence of obscene words with the help of a pre-defined lexicon. If an obscene word is present, speech and text registration is employed to identify the start and end times of the unwanted word in the actual speech. We insert a beep sound in place of the obscene word, providing an effective and socially acceptable method of censoring obscene and profane words from digital speech. The algorithms have been tested with the real speech signal and the results are promising. Our proposed method offers an effective approach to sanitizing speech signals by removing obscene and profane language in a socially acceptable manner.