Words phonetic decoding method with the suppression of background noise

Abstract

A new algorithm for the suppression of background noise in systems of recognition of voice commands is proposed based on phonetic decoding of words. The principle of the algorithm employs frequencydomain correction of quasi-stationary fragments of the signal of voice command using selective estimation of noise variance in pauses between syllables. The desired effect lies in a several-fold increase in the signal-to-noise ratio at the input of the decision unit. The theoretical results are proven with the aid of a physical experiment.