A survey of open-source speech recognition software for voice actuated control

Abstract

Nowadays, speech recognition technologies are actively developing and find application in various fields, such as controlling a computer using voice, dictating texts, and human-computer interaction or communicating with a computer on an intellectual level.Non-contact and natural for human ways of interacting with a computer, based on automatic recognition and synthesis of speech, text, gesture and tactile information, as well as paralinguistic information, including non-verbal aspects of speech and text information, are especially relevant.Much attention is paid to creating an accessible environment for people with disabilities and disabilities.An important means of ensuring accessibility and improving the quality of life, social interaction, and integration into society for people with disabilities are computer facilities and specialized information systems.In this paper we presented a series of experiments and analyzed the opensource software for voice actuated control.Simon is a software for Internet surfing, mailing, managing multimedia applications that can be adapted to the needs of older people.To work with the acoustic model, we need to either have our own model or load it from the library.The speech recognition was evaluated using two criteria, they are WER and Latency.The analysis of the software for recognition of continuous speech showed that at present there is no universal system for recognizing continuous speech that would be capable of self-learning, would be speaker-independent, resistant to noise, and have a low error rate.The considered software solutions at the moment are not universal and accurate, the speech recognition error greatly depends on the presence of extraneous medium and high-frequency noise, as well as on the microphone quality.