An ASR model for Individuals with Hearing Impairment using Hidden Markov Model

Automatic Speech Recognition (ASR) technology for normal speech recognition has gained popularity among researchers since its early days. However, applying ASR to recognize speech by persons with hearing disability is an area in which not much research has been conducted so far. This research paper presents a fresh perspective to enhance the recognition of disordered speech in persons with hearing disabilities using Hidden Markov Models (HMMs). These models have been widely utilized in speech recognition tasks, but their application to the specific context of disordered speech by hearing-impaired speech is a novel attempt. The research utilizes a comprehensive dataset of disordered speech samples from hearing-impaired individuals having varying degrees of hearing impairment. HMMs were used to model the underlying acoustic and phonetic patterns in this unique speech data. Various experimental analyses were done using data collected from Normal and Hearing Impaired speakers. The whole experimentation process involves training and fine-tuning the HMMs using the collected dataset and evaluating their performance in speech recognition. The results of this study indicate that HMM-based models provide both relevant and promising results in recognizing disordered speech, even in the presence of varying degrees of hearing impairment. This could be better utilized for the development of assistive technologies that can aid hearing-impaired individuals to communicate with normal speakers in a better way. The research paper also sheds light on the potential of Hidden Markov Models as a valuable tool for recognizing and transcribing disordered speech of persons with hearing disability.