System of Smart Home Based on Speech Recognition with Machine Learning

Abstract

This article designs and implements a system of smart home based on speech recognition with machine learning and Android. The STM32 microcontroller is combined with actuating components to control home devices by obtaining server control commands through the network. This article designs a smart home APP based on Android and Google Voice Search, optimizes the speech recognition algorithm by using the training results from Word2Vector model to improve the recall rate of the speech recognition of command samples. This APP displays environmental data according to the scenario for users, such as the temperature, humidity, and alert of smoke gas. The test results show that the recognition recall rate of the test data of control commands from users of different ages is increased to 94.4% with high stability.