Mobile and Personal Speech Assistant for the Recognition of Disordered Speech

摘要

Recently, Assistive Technologies tend to exploit speech-based interfaces as a means of communication between humans and machines. While they perform very well for normal speech, their efficacy is very limited for people suffering from a variety of speech disorders. Moreover, in the systems targeted for disordered speech, the recognition performance is highly diminished by the environmental factors related to the disease. This limits the practical applicability of these solutions. To overcome this problem, we propose a Mobile and Personal Speech Assistant (mPASS) - a platform providing the users with a set of tools, which enable to intuitively create their own speech recognition system corresponding to their needs and capabilities. Our long term vision is that handicapped users, without computer science and artificial intelligence knowledge, will use this platform to design at home their own speech recognition system, tailored to the domain, vocabulary, and language they find most useful. As a result, a personalized speech recognizer will be created, which can be used with diversified speech-based applications.