ABSTRACT
This chapter describes the approach to voice control of Internet of Things (IoT) objects, which are part of the “Smart House” system, providing high accuracy of speech recognition by set of special means of artificial intelligence. Analysis of deep neural network adaptation algorithms has shown that deep neural network adaptation using i-vectors significantly increases recognition accuracy by providing the deep neural network with additional phonogram information. Thus, the proposed method is based on the assumption that the features extracted from a deep neural network with a narrow neck, adapted by i-vectors. Improving the accuracy of the command recognition through the use of approaches to building language models that allow effective consideration of the distant semantic context, as well as morphological, syntactic, and semantic information. And, first of all, increase the speed of the command recognition of the “Smart House” system.
