Research of Promising Methods of Digital Processing in Speech Recognition Systems

Currently, speech recognition systems occupy a large part of a person's life. Many people use voice assistants and voice input. Speech recognition systems can also make life easier for people with disabilities. In business, too, a lot is paid to these systems. Now there are smart speakers from many well-known technology giants, these same companies produce their own product — voice assistants: Alice from Yandex, Siri from Apple and Google Assistant from Google. In business, speech recognition systems can also make life easier for consultants. Many stores and banks put a voice assistant on the hotline, which can answer the client's most basic questions and help with the choice of a consultant, and then switch to the right specialist. Speech processing applications, such as speech enhancement and speaker identification, are based on evaluating the relevant parameters of the speech signal. These parameters often need to be estimated based on noise observations, since speech signals are rarely obtained in «pure» acoustic environments in the real world. As a result, algorithms for estimating the parameters used must be resistant to environmental factors, such as additive noise and reverberation.