Overview of the latest research related to smart speakers

Abstract

This article provides an overview of the latest research related to smart speakers - speech recognition, speech synthesis, context recognition, artificial neural networks used. Various tasks related to smart speakers are discussed - which neural networks are effective in speech synthesis, how to make synthesized speech more natural, how to detect and classify errors in automatic speech recognition systems, how to protect personal voice assistants from recording / playback and other acoustic attacks, speech recognition using convolutional neural network, how to use dynamic time transformation (DTW) algorithm with training for speech recognition, using voice assistant to set up smart home systems, recursive algorithm to solve the problem of echo cancellation, reverberation and noise reduction, processes based on different algorithms in every voice assistant and different search engines lead to huge differences in the answers to questions, a problem of integration with smart devices using different data transfer protocols.