Abstract

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.

中文翻译：

---

言语行为的谓语症状和生物特征

摘要

本文考虑了使用各种基于统计数据预测语音的方法和算法进行谓词分析的问题，以及使用神经网络学习系统进行语音识别的问题。它介绍了文本和语音行为的概率元素，在创建语音识别算法和发布语音改进建议时必须考虑这些因素。提出了一种句子分析算法，将语音信号转换为符号和单词的字符串。分析了现代语音识别系统的工作原理。提出了一种使用数学算法处理语音短语并评估信号电平的技术，该技术可以确定语音设备各个特性的影响程度。