Arabic Automatic Speech Recognition for Speakers With Speech Disorders: A Comprehensive Review

Automatic speech recognition (ASR) aims to convert human speech signal into text, where the quality of the speech signal is essential to assure correct recognition. However, some people suffer speech disorders (SD) of various kinds such as distortion and substitution, which may degrade the recognition of the speech signal in ASR domain. Based on our literature investigation, most of the available ASR systems handle normal/regular speech with no consideration for SD and disabilities. This paper provides a comprehensive review on Arabic ASR for speakers suffering SD. In this paper, classifications of SD are provided. In addition, research attempts on Arabic ASR for speakers suffering SD are also highlighted in order to identify the benchmark datasets, tools, algorithms, and research gaps. To the best of our knowledge, all previous research attempts on Arabic ASR for speakers suffering SD focused on the Arabic letter /r/ in limited isolated words mode. This shows that there is a need to consider other types of SD in ASR domain for Arabic language, and consider continuous speech mode. In addition, Arabic speech datasets are very limited for SD research, which can be treated as an important research gap to be considered.