The Predicative Symptoms and Biometry of Speech Behavior

This paper considers the issues of predicative analytics using various methods and algorithms for predicting speech based on statistical data, as well as speech recognition using neural network learning systems. It presents the probabilistic elements of text and speech behavior, which must be taken into account in creating a speech-recognition algorithm and issuing recommendations for speech improvement. A sentence analysis algorithm is proposed to convert a spoken acoustic signal into a string of symbols and words. The principles of operation of modern speech recognition systems are analyzed. A technique is proposed for processing speech phrases using mathematical algorithms with an assessment of signal levels, which makes it possible to determine the degree of influence of individual characteristics of the speech apparatus. An assessment of the severity of abnormal symptoms in subjects with speech impairments has been carried out and a method for quantitative assessment of the size of the deviation of symptoms in these subjects has been proposed.