Research and Application Analysis on Key Problems of Automatic Speech Recognition for Dysarthria

At present, automatic speech recognition technology has developed rapidly, and the recognition accuracy of Automatic Speech Recognition (ASR) based on deep learning has been very high. Therefore, the speech recognition problem of patients with dysarthria has been paid more attention in recent years. However, due to the particularity of dysarthria patients and the strong variability of their speech, the relevant available datasets are very scarce, and it is difficult to adapt to the current general recognition model. In order to promote the development and progress of this field, based on a large number of literature research, this paper summarizes the construction of data sets, the solution of speech variability, and the key issues and development trends of multi-feature speech data fusion. At the same time, it lists the current application of automatic speech recognition for dysarticulation in some fields. It is hoped that the need of social communication and intelligent life of patients with dysarthria who exist as a minority group can be solved as soon as possible.