The present idea is to convert a research tool of the ERC-funded project (269670) into a working single-word Automatic Speech Recognition (ASR) demonstration system to attract speech technology companies to use this method in commercial ASR systems. The difference to existing ASR systems is that no training is needed; a wide range of speakers and dialects are covered; it can handle non-standard, non-ideal speech; it is less susceptible to interference from noise; and the system can be easily adapted to new languages. Our planned ASR system [the Flexible Speech Recognition system (FlexSR)] has evolved from a linguistic model of phonological representation, which is the focus of the present ERC grant.
Call for proposal
See other projects for this call