Audio stream restoration using speech recognition methods and algorithms

A study has been carried out in the field of noise suppression in a speech signal. Identification of speech from a noisy signal occurs with the help of hidden Markov models of isolated words. A method for cleaning a speech signal from extraneous noise using a Wiener filter based on reference words is proposed as an alternative to spectral subtraction and Wiener filtering methods using a noise sample. The results obtained are reflected in the software implementation of the studied filtration method.